JP2019066909A - Object distribution estimation device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To estimate highly accurate information on the distribution of an object in a wide range from a photographed image of a space where congestion may occur in a wide field of view. An image acquisition means 30 acquires a photographed image of a space in which a predetermined object photographed by a photographing unit may exist. The density estimation means 50 and the single identification means 51 are recognition means for recognizing the number of objects photographed at arbitrary positions in the captured image, and a plurality of recognizers having different ranges of the number of recognizable objects can be recognized. It is a plurality of recognition means provided for each corresponding range of the number of objects used. The method map storage means 42 stores the position in the captured image in association with the accuracy guarantee recognition means capable of recognizing the number of objects at the position with an accuracy equal to or higher than a predetermined lower limit value among the plurality of recognition means. .. The distribution estimation means 52 generates distribution information of objects in space from the number of objects acquired in the captured image by the accuracy guarantee recognition means stored in the method map storage means 42. [Selection diagram] Fig. 2

Description

  The present invention relates to an object distribution estimation apparatus for estimating the distribution of an object from a photographed image in which a space in which a predetermined object such as a person can exist is photographed, and in particular, a photographed image in which a space where congestion may occur is photographed. The present invention relates to an object distribution estimation apparatus that estimates a distribution.

  In places where congestions may occur, such as marathons and events where parades and other events are held, it is required to provide a large number of security guards in areas where congestion occurs, for the purpose of preventing accidents and the like. Therefore, surveillance efficiency can be expected to be improved by arranging surveillance cameras at various places in the hall, estimating the distribution of persons from the photographed image, and displaying the distribution.

  One of the methods for detecting a person present in the event hall from the photographed image is a method of searching for the photographed image by a recognizer that has learned the features of the human image in advance.

  For example, in the object detection device described in Patent Document 1 below, an input image is generated using a classifier that has been learned in advance using multiple “person” image data and “non-person” image data as a recognizer. Detecting a person is described.

  Further, in the crowd analysis device described in Patent Document 2 below, the person density is estimated using a classifier that has been machine-learned for each person density using a learning image in which a crowd of people density exceeding the density lower limit is photographed in advance. By doing, it is described to determine the occurrence of the crowd.

  Hereinafter, a recognizer that recognizes each person is referred to as a single identifier, and a recognizer that estimates human density is referred to as a density estimator.

JP, 2011-186633, A Unexamined-Japanese-Patent No. 2017-068598

  By the way, since the event hall is generally large, it is required to widen the field of view of each monitoring camera and reduce the number of installed cameras from the viewpoint of installation and operation costs.

  However, the single classifier and the density estimator have different resolution ranges that can be recognized with high accuracy depending on the characteristics of the respective learning images. On the other hand, in the photographed image photographed in a wide field of view, the resolution of the image of the person according to the distance from the camera to the person is remarkably reduced.

  Therefore, there is a problem that it is difficult to provide highly accurate information if a recognizer is used without taking into consideration the characteristics of a photographed image photographed in a wide field of view.

  That is, while the single classifier can recognize the position of each person, the density estimator can not recognize the positions of each person. Also, in general, when the resolution of the image to be recognized (part of the photographed image) becomes lower than that of the learning image, the recognition accuracy decreases, and the learning of the single classifier uses the learning image occupied by the image of a single person. Since the learning of the density estimator is performed using the learning image of the field of view in which the images of a plurality of people can be taken while the recognition accuracy is reduced according to the distance from the camera to the person It is more likely that a single discriminator will occur.

  Therefore, the person distribution in the area near the camera in the captured image can provide more detailed information if it is recognized by the single-body discriminator, but if even the distribution of the person in the distant area is to be recognized by the single-body discriminator It will provide low information. On the other hand, although the distribution of people in the distant area can provide more accurate information if it is recognized by the density estimator, it is possible to recognize even the distribution of people in the area near the camera by the density estimator. It will fail to provide.

  The present invention has been made in view of the above problems, and provides an object distribution estimation apparatus capable of estimating, in a wide range, information with high accuracy regarding the distribution of objects from a photographed image obtained by photographing a space where congestion may occur with a wide view. The purpose is

  (1) The object distribution estimation apparatus according to the present invention is different from the image acquisition means for acquiring a photographed image of a space where a predetermined object photographed by the photographing unit may exist, and the number of photographed objects is different. A recognition unit that recognizes the number of objects photographed at an arbitrary position in the photographed image using a recognizer that learns in advance the features of each of a plurality of types of learning images, and the number of objects that can be recognized is mutually different The plurality of recognition means provided for each relevant range of the number of objects using a plurality of different recognition devices, the position in the photographed image, and the number of objects in advance at the position among the plurality of recognition means A storage unit that associates and stores an accuracy guarantee recognition unit that can be recognized with an accuracy equal to or higher than a defined lower limit, and the number of objects acquired in the photographed image by the accuracy guarantee recognition unit stored in the storage unit , And a distribution estimating means for generating distribution information of the object in the serial space.

  (2) In the object distribution estimation device according to (1), the lower limit value of the resolution of the image of the object which can be recognized with accuracy higher than the lower limit value by the recognition means is determined as the limit resolution of the recognition means. The storage means is one of the plurality of recognition means with respect to a position in the photographed image, one in which the resolution of the image of the object at the position is equal to or more than the limit resolution. It is possible to associate and store the configuration.

  (3) In the object distribution estimation device according to (2), the resolution is represented by the number of pixels of the image of the object, and the number of pixels corresponding to the limit resolution is present in the space as the limit number of pixels. A projection unit configured to project a model imitating the object on a shooting surface of the shooting unit to generate a projected image; and for a position in the shot image, the number of pixels of the projection image generated at the position and the number The image processing apparatus may further include selection means for comparing the limit pixel number of each of the plurality of recognition means and selecting the accuracy assurance recognition means to be associated with the position.

  (4) In the object distribution estimation device according to (2), the resolution is estimated based on a distance from the imaging unit to the object in the space, and the storage corresponding to the limit resolution is used as a limit distance. The means associates the position in the photographed image with one of the plurality of recognition means in which the distance for the object photographed at the position is equal to or less than the limit distance as the accuracy assurance recognition means. The configuration can be stored.

  (5) In the object distribution estimation apparatus according to (4), projection means for backprojecting a position in the photographed image on a horizontal plane at a height corresponding to the object in the space to obtain an object virtual position; The accuracy assurance recognition in which the distance between the virtual position of the object corresponding to the position and the photographing unit and the limit distance of each of the plurality of recognition means are compared with the position in the image and associated with the position. And a selection means for selecting a means.

  (6) Another object distribution estimation apparatus according to the present invention is an image acquisition unit for acquiring a photographed image of a space in which a predetermined object photographed by the photographing unit may exist, and the above-mentioned density at each predetermined density. A density estimation unit that recognizes the density of the object captured at an arbitrary position in the captured image using a density estimator that learns in advance features of density images obtained by capturing a space in which the object exists; A single identification means for recognizing the presence or absence of the object photographed at an arbitrary position in the photographed image using a single classifier which previously learned the features of the single image in which the object is photographed; a position in the photographed image And storage means for associating and storing the density estimation means and the single identification means among the accuracy assurance recognition means capable of recognizing the density or presence of the object at the position with an accuracy equal to or higher than a predetermined lower limit at a predetermined position. , From the density or the presence or absence of the object acquired in the captured image by the accuracy assurance recognizer stored in the storage unit, and a distribution estimating means for generating distribution information of the object in the space.

  According to the present invention, it is possible to estimate, in a wide range, highly accurate information on the distribution of an object from a photographed image obtained by photographing a space where congestion may occur with a wide view.

FIG. 1 is a block diagram showing a schematic configuration of an object distribution estimation device according to an embodiment of the present invention. It is a functional block diagram when the object distribution estimation apparatus which concerns on embodiment of this invention performs the distribution estimation process based on a picked-up image. It is a functional block diagram when the object distribution estimation apparatus which concerns on embodiment of this invention performs a method map creation process. It is the figure which represented typically the information of the single-piece | unit discriminator which the single-piece | unit discriminator storage means has memorize | stored. It is a schematic diagram which shows an example of a picked-up image and the method map corresponding to this. It is a schematic diagram which shows an example of the distribution image corresponding to the picked-up image of FIG. It is a schematic diagram which shows an example of the method map corresponding to the picked-up image of FIG. It is a schematic flowchart of operation | movement of the object distribution estimation apparatus based on embodiment of this invention. It is a flowchart of the outline of a method map creation process. It is a flowchart of the outline of distribution image generation processing.

  Hereinafter, an object distribution estimation apparatus 1 according to an embodiment of the present invention (hereinafter referred to as an embodiment) will be described based on the drawings.

[Configuration of Object Distribution Estimation Device 1]
FIG. 1 is a block diagram showing a schematic configuration of the object distribution estimation apparatus 1. As shown in FIG. The object distribution estimation device 1 includes a photographing unit 2, a communication unit 3, a storage unit 4, an image processing unit 5, an operation unit 6, a display control unit 7 and a display unit 8.

  The photographing unit 2 is a monitoring camera, and is connected to the image processing unit 5 via the communication unit 3, and photographing means for photographing a predetermined space where a predetermined object may be crowded at predetermined time intervals and outputting a photographed image It is. Hereinafter, the part which the imaging | photography part 2 image | photographs is called object space.

  For example, the photographing unit 2 is installed on a pole installed at the event hall with a view that overlooks the target space. The field of view may be fixed or may be changed in accordance with an external instruction via the communication unit 3. Also, for example, the photographing unit 2 photographs the target space at a frame period of 1⁄5 second to generate a color image. Instead of a color image, a monochrome image may be generated.

  The communication unit 3 is a communication circuit, one end of which is connected to the image processing unit 5 and the other end of which is the photographing unit 2 and the display control unit 7 via a communication cable such as a coaxial cable or LAN (Local Area Network) or the Internet. Connected with The communication unit 3 acquires a captured image from the imaging unit 2 and inputs the image to the image processing unit 5, and outputs the information input from the image processing unit 5 to the display control unit 7.

  The storage unit 4 is a memory device such as a read only memory (ROM) or a random access memory (RAM), and stores various programs and various data. The storage unit 4 is connected to the image processing unit 5 to input and output such information with the image processing unit 5.

  The image processing unit 5 is configured by an arithmetic device such as a central processing unit (CPU), a digital signal processor (DSP), and a micro control unit (MCU). The image processing unit 5 is connected to the storage unit 4 and operates as various processing means and control means by reading and executing a program from the storage unit 4 to store various data in the storage unit 4. read out. The image processing unit 5 is also connected to the imaging unit 2 and the display control unit 7 via the communication unit 3, and the person being imaged by analyzing the captured image acquired from the imaging unit 2 via the communication unit 3. And the distribution information such as the distribution image on which the estimation result is described is output to the display control unit 7 via the communication unit 3.

  The operation unit 6 is an input device for the display control unit 7 and includes a keyboard, a mouse, and the like. The operation unit 6 is connected to the display control unit 7, receives an instruction operation by the supervisor, and outputs the instruction operation to the display control unit 7.

  The display control unit 7 is configured by a PC (Personal Computer) or the like, and is a storage unit (not shown) configured by a memory device such as a ROM or a RAM, and an interface circuit with a communication network to which the communication unit 3 is connected. A control unit (not shown) including a communication unit (not shown) and an arithmetic device such as a CPU, an MCU, and an IC is provided. The display control unit 7 is connected to the communication unit 3 via the communication network, and is connected to the operation unit 6 and the display unit 8. The display control unit 7 receives information from the image processing unit 5 from the communication unit 3 and stores it, and receives an operation instruction from a monitoring person from the operation unit 6 and displays information corresponding to the operation instruction among the stored information Output to section 8.

  The display unit 8 is a display device such as a liquid crystal display or a CRT (Cathode Ray Tube) display, and is connected to the image processing unit 5 via the communication unit 3 and the display control unit 7, and information generated by the image processing unit 5 Is a display means for displaying The supervisor visually recognizes the displayed information to determine the occurrence of congestion, etc., and takes measures such as changing the staffing if necessary.

  In the present embodiment, the object distribution estimation apparatus 1 in which the number of the imaging unit 2 and the number of the image processing units 5 is one to one is exemplified. However, in another embodiment, the object distribution estimation device 1 of the imaging unit 2 and the image processing unit 5 The number may be many to one or many to many.

[Function of Object Distribution Estimation Device 1]
2 and 3 are functional block diagrams showing functions of the object distribution estimation apparatus 1. Among them, FIG. 2 is a functional block diagram when performing distribution estimation processing based on a photographed image. Further, FIG. 3 is a functional block diagram when the method map creation processing is performed prior to the distribution estimation processing.

(Distribution estimation process)
Among the functions of the object distribution estimation apparatus 1, the distribution estimation process will be described first. When the distribution estimation process is performed (FIG. 2), the communication unit 3 functions as the image acquisition unit 30 and the distribution information output unit 31 etc., and the storage unit 4 is the density estimator storage unit 40, the single identifier storage unit 41, the method map The image processing unit 5 functions as a storage unit 42, an object model storage unit 43 (not shown in FIG. 2), a camera parameter storage unit 44 (not shown in FIG. 2), etc. , Functions as distribution estimation means 52 and the like.

Among these, each of the density estimation means 50 and the simple identification means 51 uses a recognizer which has previously learned the features of each of a plurality of types of learning images different in the number of objects being photographed, and any position within the photographed image The density estimation means 50 and the simple identification means 51 are the recognition means for recognizing the number of objects that can be recognized using a plurality of recognizers that are different from each other in the range of the number of recognizable objects. Provide a plurality of recognition means provided for each. For example, the density estimation unit 50 may have a density of 0 person / m ^2, a density of more than 0 person / m ^{2 and not} more than 2 people / m ^2, a density of more than 2 people / m ^{2 and not} more than 4 people / m ² and 4 people Using a recognizer learned using a learning image with a density higher than 1 / m ² and using an area set at an arbitrary position in the captured image as a window, the density in the window is 0 people / m ² , 0 person / m ² two / ^{m 2} or less of a density higher than ^2, recognizes which of higher than two / ^m higher than ² Quadruple / ^{m 2} or less of the density and four / ^{m 2} density. Also, for example, the single identification unit 51 uses a recognizer learned using 0 and 1 learning images to recognize whether the inside of the window is 0 or 1 person. That is, the density estimation means 50 recognizes the density of the objects in the window as the number of objects, and the lower limit of the range is 0 person / m ² while the upper range extends to a value exceeding 4 people / m ² . On the other hand, the range of the number of objects that can be recognized by the single identification unit 51 in the window is 0 and 1.

  The image acquisition unit 30 sequentially acquires captured images from the imaging unit 2 which is an imaging unit, and sequentially outputs the acquired captured images to the density estimation unit 50 and the single identification unit 51.

  The density estimator storage means 40 is an estimated density calculation function which learns image features of each density image obtained by photographing a space in which an object (person) exists at the density for each predetermined density, and the feature amount of the image is calculated When input, an estimated value (estimated density) of the density of an object captured in the image is calculated, and information of an estimator (density estimator) which outputs the calculated estimated density is stored in advance. That is, parameters such as coefficients of the estimated density calculation function are stored in advance as information of the density estimator.

  The density estimation unit 50 extracts feature quantities (feature quantities for estimation) for density estimation from various places of the photographed image input from the image acquisition unit 30 and reads out the density estimator from the density estimator storage unit 40 and extracts The distribution (density distribution) of estimated density is estimated by inputting each of the estimated feature quantities to the density estimator, and the estimated density distribution is output to the distribution estimation means 52. Preferably, the density estimation means 50 further outputs the density distribution to the single identification means 51.

  The process of density estimation and the density estimator will be specifically described.

  The density estimation unit 50 sets a window (extraction window for estimation) at the position of each pixel of the captured image, and extracts the feature amount for estimation from the captured image in each extraction window for estimation. The feature amount for estimation is a feature of Gray Level Co-occurrence Matrix (GLCM).

  It is desirable that the regions in the target space captured by the estimation extraction windows have the same size. Therefore, preferably, the density estimation unit 50 performs processing in consideration of this point. In the process, the density estimation unit 50 uses a camera parameter storage unit 44 (not shown in FIG. 2).

  The camera parameter storage unit 44 stores camera parameters of the photographing unit 2. The camera parameters are information including external parameters such as the installation position and imaging direction of the imaging unit 2 in real space, focal length of the imaging unit 2, angle of view, lens distortion and other lens characteristics, and internal parameters such as the number of pixels of the imaging device. is there.

  The density estimation unit 50 reads the camera parameters of the imaging unit 2 from the camera parameter storage unit 44, and the area in the target space captured at an arbitrary pixel of the captured image becomes the same size by homography conversion using the camera parameters. As described above, the feature amount for estimation is extracted after the captured image is deformed.

  The density estimator can be realized by a classifier that identifies multiple classes of images, and can be a classification function learned by the multiclass Support Vector Machine (SVM) method.

The density is, for example, a "background" class in which there are no people, a "low density" class which is higher than 0 person / m ^{2 and not} more than 2 people / m ^{2, and not} less than 2 people / m ^{2 and not} more than 4 people / m ² It can be defined as four classes of "medium density" class and "high density" class higher than 4 persons / m ² .

  The estimated density is a value given in advance to each class, and is a value output as a result of distribution estimation. In the present embodiment, values corresponding to each class are described as “background”, “low density”, “medium density”, and “high density”.

  That is, the density estimator learns by applying the multiclass SVM method to the feature quantities of many density images belonging to each of the "background" class, the "low density" class, the "medium density" class, and the "high density" class. It is an identification function for identifying each class image from other classes. The parameters of the discriminant function derived by this learning are stored as a density estimator. The feature amount of the density image is the same as the feature amount for estimation, and is a GLCM feature.

  The density estimation means 50 acquires estimated density which is the output value by inputting each of the feature quantities for estimation extracted corresponding to each pixel to the density estimator. When the photographed image is deformed and the estimation feature amount is extracted, the density estimating unit 50 deforms the density distribution into the shape of the original photographed image by homography conversion using camera parameters.

  A group of estimated densities for each pixel of the captured image obtained in this manner is a density distribution.

  The density distribution outputted by the density estimating means 50 indicates the density of people in each part of the photographed image, but the density distribution does not indicate the position of each person. On the other hand, the single identification unit 51 is a unit that obtains more detailed information (the position of a person) by identifying the presence or absence of a person at each position of the captured image.

  The single identifier storage means 41 stores information of a identifier (single identifier) that has learned the image features of a single person (object).

  FIG. 4 is a diagram schematically representing the information of the single-piece discriminator stored in advance by the single-piece discriminator storage means 41. As shown in FIG.

  The single-body discriminator calculates and outputs an evaluation value (identification score) representing the likelihood that the image is a single person image (single-body image) when the image feature amount is input. It is represented by parameters such as a coefficient of the evaluation value calculation function and a threshold applied to the identification score.

  The single-piece discriminator can be a discriminator learned by applying the linear SVM method to the feature amount of a learning image consisting of a large number of single-piece images and a large number of unmanned images in which each person is only a person.

  When a linear SVM is used as a learning algorithm, the coefficient of the evaluation value calculation function is a weight vector. The weight vector is a weight for each element of the feature amount, and the value of the inner product of the feature amount of the input image and the weight vector represents the identification score. In learning, when the inner product of the weight vector and the feature amount is larger than 0, it is adjusted so as to be identified as a person, and when less than 0, it is identified as other than a person. Therefore, the threshold for identifying whether or not the input image is a single image is in principle 0, and the threshold can generally be set to 0. However, the threshold may be set to a value smaller than 0 in order to reduce an error in identifying a single image as not being a single image.

  Note that the feature amounts of the learning image are HOG (Histograms of Oriented Gradients) feature amounts.

  The single-piece classifier stored in the single-piece classifier storage unit 41 is a classifier that has learned image features of a small part of parts constituting a single object as the density is higher. The single discriminator storage means 41 is a single discriminator that is a single discriminator that has learned the image feature of a single person's whole body in association with a value representing a low density class, and is associated with a value representing a medium density class Upper-body discriminator 101, which is a simplex discriminator that has learned the image features of the upper 2/3 of a person, and a simplex discriminator that has learned the image characteristics of the upper 1/3 of a single person in association with a value representing a high density class The near-head identifier 102 is stored.

  The whole-body discriminator 100 is a single-piece discriminator learned using a single-body image of the whole body of a single person, and the upper-body discriminator 101 is a single-piece image of the upper two-thirds of a single person Is a single-body discriminator learned using an image obtained by cutting out the upper 2/3 of a single-body image taken, etc., and the near-head part discriminator 102 is a single-piece image in which the upper third of a single person is photographed ( It is a single-piece discriminator learned using an image etc. which cut out the upper 1/3 of the single-piece picture by which the whole human body was photoed.

  The single identification unit 51 extracts a single identification feature (identification feature) from each position of the captured image input from the image acquisition unit 30 and reads a single identification from the single identification storage unit 41. The presence or absence of an object is identified by inputting each of the extracted feature amounts for identification into a single identifier, and the presence or absence of an object at each position is output to the distribution estimation means 52.

  Specifically, the single identification unit 51 first sets a plurality of candidate positions at predetermined intervals in the captured image. The predetermined interval is one pixel, and the single identification unit 51 sequentially sets the position of each pixel of the captured image as a candidate position. The candidate position represents the center of gravity of a person's head.

  In addition, the single-piece identification means 51 sets a window defined in the shape of the upper third of a single person at each candidate position and refers to the density distribution input from the density estimation means 50 to estimate the inside of the window Aggregate the density. Then, the single identification unit 51 determines the largest estimated density at each candidate position as the density of the candidate position.

  Further, the single identification unit 51 sets an identification extraction window corresponding to the density of the candidate position at each candidate position, and extracts an identification feature amount from the photographed image in the identification extraction window. The extraction window for identification has a shape of a single image used for learning of a single classifier according to each density, and is a window having a size enlarged or reduced by a plurality of predetermined magnifications. That is, the extraction window for identification is a window defined in the shape of the whole body of a single person if the density of the candidate position is low, and if it is medium density, it is set in the shape of the upper 2/3 of the single person It is a window that is defined in the shape of the upper third of a single person if it is dense.

  In addition, the single identification unit 51 reads out from the single identification storage unit 41 a single identification corresponding to the density of the candidate position for each candidate position. That is, the simple identification unit 51 reads the whole-body identifier if the density of the candidate position is low, reads the upper-body identifier if the density is medium, and reads the near-head identifier if the density is high. Then, for each candidate position, the single identification unit 51 inputs the feature amount for identification extracted from the candidate position into the read single identifier, and acquires the identification score, which is the output value thereof, as the evaluation value of the candidate position. .

  Then, the single-piece identification unit 51 refers to the density for each candidate position and the evaluation value, and the candidate position for which the evaluation value satisfying the predetermined criteria is calculated is the position where the object is present, and the other candidate position is the object Decide that it is not a position.

  Specifically, the single identification unit 51 extracts a candidate position whose identification score is equal to or higher than the threshold corresponding thereto, and among the extracted candidate positions, one or more candidate positions having the same density and which are close to each other are selected. It is decided that the candidate positions summarized in one place are positions where people exist.

  The process of putting together the candidate positions is performed to cope with the fact that a high identification score is calculated for the same person in the vicinity as well as the position where the person is actually photographed. Specifically, for example, for each density, the single-piece identification unit 51 sequentially sets candidate positions for which identification scores higher than the threshold are calculated in the order of high identification score as a target position and a candidate whose identification score is lower than the target position. Set the position to the comparison position. Then, the single identification unit 51 deletes the information of the comparison position where the overlap between the identification extraction window set at the comparison position and the identification extraction window set at the target position among the comparison positions is larger than a predetermined ratio. A plurality of candidate positions are put together into one by doing.

  Then, the single identification unit 51 outputs information on the presence or absence of an object for each candidate position to the distribution estimation unit 52.

  The method map storage unit 42 sets the position in the photographed image and the number of objects photographed in the window when the window is set to the position among the plurality of recognition units to a predetermined lower limit value or more A technique map is stored that is associated with the accuracy assurance recognition means that can be recognized with the accuracy of.

  FIG. 5 is a schematic view showing an example of a photographed image and a method map corresponding thereto, and FIG. 5 (a) shows the photographed image 200, and FIG. 5 (b) shows a method map 250 corresponding to the photographed image 200. It is shown in tabular form. For each of the density estimation means 50 and the simple identification means 51, the limit (limit resolution) of the resolution of the human image that can be recognized with an accuracy higher than the lower limit was found by preliminary experiments using test images of various resolutions. ing. Further, “A” and “B” are given to the density estimation means 50 and the single identification means 51 in advance as codes (recognition means ID) for identifying them.

  In the example of FIG. 5, an area 201 in the photographed image 200 is an area where a person present far from the photographing unit 2 can be photographed. When the extraction window for estimation is set at an arbitrary position in the region 201, the resolution of the human image in the extraction window for estimation does not fall below the limit resolution of the density estimation unit 50, and therefore the density estimation result by the density estimation unit 50 is It can be obtained with an accuracy above the lower limit. When the extraction window for identification is set at an arbitrary position in the area 201, the resolution of the image of the person in the extraction window for identification falls below the limit resolution of the single identification means 51, and the identification result by the single identification means 51 is lower limit It can not be obtained with an accuracy higher than the value.

Corresponding to the above, in the method map 250, the code A given to the density estimation means 50 in association with the coordinates [..., (X ₂ , Y ₂ ), ...] of the pixel group in the area 201 is It is memorized.

  Further, in the example of FIG. 5, an area 202 in the photographed image 200 is an area where a person present in the vicinity of the photographing unit 2 can be photographed. When the extraction window for estimation is set at an arbitrary position in the region 202, the resolution of the human image in the extraction window for estimation does not fall below the limit resolution of the density estimation means 50, and therefore the density estimation result by the density estimation means 50 is It can be obtained with an accuracy above the lower limit. Further, when the extraction window for identification is set at an arbitrary position in the area 202, the resolution of the image of the person in the extraction window for identification does not fall below the limit resolution of the single identification unit 51. Is obtained with an accuracy equal to or higher than the lower limit value.

Corresponding to the above, in the method map 250, the code A given to the density estimation means 50 in association with the coordinates [..., (X ₃ , Y ₃ ), ...] of the pixel group in the region 202 The code B given to the single identification unit 51 is stored.

  The distribution estimation means 52 is an object in the target space from the number recognized by the accuracy assurance recognition means stored in association with the position in the method map storage means 42 among the number of objects at each position recognized by the plurality of recognition means. Distribution information is generated, and the generated distribution information is output to the distribution information output means 31.

  Specifically, the distribution estimation means 52 includes information on the density of the object at each position input from the density estimation means 50, information on the presence / absence of an object at each position input from the simplex identification means 51, and a method map storage means The distribution map is generated based on the number recognized by the accuracy assurance recognition means with reference to the method map stored in S.42.

  For example, the distribution estimation unit 52 projects an object model colored with a color according to the estimated density of the position to a pixel position that is associated with the code A and the code B in the method map and that an object is recognized to exist. Also, a color corresponding to the estimated density is set to the value of the pixel at the position associated with only the code A representing the density estimation means 50 in the method map to generate a distribution image. The color may be, for example, red if the estimated density is high, yellow if it is medium density, or green if it is low.

  In generating the distribution image, the distribution estimation means 52 uses the object model stored in the object model storage means 43. The object model storage means 43 stores an object model which approximates the shape of the object in advance. The object model is a three-dimensional model composed of three spheroids corresponding to the head, torso and legs of a standing person. By the way, let the center of gravity of the head be the representative position of a person. Note that the three-dimensional model can be simplified to represent the whole person as one spheroid, or more specifically, for example, a person's head, torso, both arms, and both legs may be separate spheroids. It can also be represented by

  The projection of the object model is basically the same as the projection means 53 of the method map generation process described later, using the object model read out from the object model storage means 43 and the camera parameters read out from the camera parameter storage means 44. To be done.

  FIG. 6 is a schematic view showing an example of a distribution image corresponding to the photographed image 200 of FIG. As described above, the area 202 in FIG. 5 is associated with the code A and the code B. In FIG. 6 corresponding to the area, an object whose head center of gravity is located at the pixel position where the object is recognized to exist The projected images 210 to 212 of the model are drawn in colors according to the estimated density of the pixel positions. Here, the hatching of each projection image represents a color according to the estimated density, and the hatching of the projection image 210 represents a color (red in the above example) corresponding to the high density of the estimated density. Similarly, the horizontal hatching of the projected image 211 and the hatching of the projected image 212 represent colors corresponding to the medium density and low density (yellow and green in the above example, respectively).

  As described above, the area 201 in FIG. 5 is associated with only the symbol A, and each pixel of the area 220 in FIG. 6 corresponding to the area is displayed in a color according to the estimated density. Specifically, the hatching of partial areas 221 to 223 in area 220 represents the color common to the above-described projected images 210 to 212, partial area 221 is high density, partial area 222 is medium density, partial area 223 Is an area estimated to be low density.

  The distribution estimation unit 52 may generate a distribution image by transparently combining the distribution image with the captured image. Alternatively, the distribution estimation unit 52 may generate a distribution image by transparently combining the distribution image with a projection image of a three-dimensional model that simulates the topography of a target space or a building.

  The distribution information output unit 31 transmits the distribution information input from the distribution estimation unit 52 to the display unit 8 via the display control unit 7 and causes the display unit 8 to display the distribution information.

(Method map creation process)
Subsequently, among the functions of the object distribution estimation apparatus 1, a method map creation process will be described. When performing a method map creation process (FIG. 3), the storage unit 4 functions as a method map storage unit 42, an object model storage unit 43, a camera parameter storage unit 44, a limit value storage unit 45, etc. It functions as the projection means 53 and the application method selection means 54 etc. The timing of creating the method map is when the method map is not stored yet, when the field of view is changed, or when the estimated desired area is changed. The estimated desired area is an area that the observer wants to cause the object distribution estimation device 1 to estimate the distribution of the object in the captured image. When the supervisor inputs the estimated desired area using the operation unit 6, the information on the area is input to the image processing unit 5 via the display control unit 7 and the communication unit 3.

  The projection means 53 refers to the estimated desired area input using the operation unit 6 and reads out the object model from the object model storage means 43 and the camera parameters from the camera parameter storage means 44, and each pixel in the estimated desired area The object model is projected to the position to generate a projection image, and the generated projection image for each pixel position is output to the application method selection unit.

  For example, the projection unit 53 derives a three-dimensional position corresponding to each pixel position in the estimated desired area in the horizontal plane 160 cm high in the virtual space simulating the target space, using the camera parameter. The height 160 cm is a numerical value predetermined as the average height of the head of a person in the target space, corresponding to the center of gravity at the representative position of the person as described above. Then, the projection unit 53 arranges the object model in the virtual space on the basis of the derived three-dimensional position, and projects the arranged object model on the imaging plane of the imaging unit 2 using the camera parameters.

  The application method selection means 54 compares the resolution of the image of the object at the pixel position with the limit resolution of each recognition means for each pixel position in the estimation desired area, and recognizes the recognition means whose resolution of the object image is at least the limit resolution. Select as accuracy assurance recognition means. That is, the application method selection unit 54 refers to the resolution of the projection image for each pixel position input from the projection unit 53 and the limit resolution of each of the plurality of recognition units stored in the limit value storage unit 45. The accuracy assurance recognition means at each pixel position in the estimation desired area is selected, and the correspondence between the pixel position and the accuracy assurance recognition means is stored in the method map storage means 42.

  Specifically, the resolution can be represented by the number of pixels of the image of the object. That is, the larger the number of pixels corresponding to the image of the object, the higher the resolution of the image.

  In this case, the number of pixels corresponding to the limit resolution can be defined as the limit pixel number, and the limit value storage unit 45 can store the limit pixel number as the limit resolution. In addition, the application method selection unit 54 counts the number of pixels of the projected image of the pixel position for each pixel position in the estimation desired area. The count value of the number of pixels is the resolution of the image of the object at the pixel position, and the application method selection means 54 compares the count value with the limit pixel number of each recognition means Is selected as the accuracy assurance recognition means.

FIG. 7 is a schematic view showing an example of a method map corresponding to the photographed image 200 of FIG. In this example, a method map is created for the estimated desired area 300 surrounded by an alternate long and short dash line. The projection means 53 sets the position of the pixel group in the estimated desired area 300 [..., (x ₁ , y ₁ ), ..., (x ₂ , y ₂ ), ..., (x ₃ , y ₃ ), ...] The model is projected to generate projected image groups [..., projected images 310, ..., projected images 320, ..., projected images 330, ...]. The application method selection means 54 derives resolution [..., r ₁ ,..., R ₂ ,..., R ₃ ,. The limit value storing unit 45, the resolution limit R _A, its resolution limit R _B in association with granted code B to a single identification means 51 is stored in association with the code A given to the density estimating means 50 ing. In this example, r ₁ <R _A and r ₁ <R _B for the resolution r ₁ of the projected image 310 at the pixel position (x ₁ , y ₁ ), and the projected image 320 at the pixel position (x ₂ , y ₂ ) Assuming that r ₂ > R _A and r ₂ <R _B for resolution r ₂ and r ₃ > R _A and r ₂ > R _B for resolution r ₃ of the projected image 330 at the pixel position (x ₃ , y ₃ ) Do. The application method selection means 54 compares the resolution [..., r ₁ , ..., r ₂ , ..., r ₃ , ...] with each of R _A and R _B and determines the accuracy of [..., (x ₁ , y ₁ ) There is no guarantee recognition means, ..., (x ₂ , y ₂ ) accuracy guarantee recognition means are the density estimation means 50, ..., (x ₃ , y ₃ ) accuracy guarantee recognition means are the density estimation means 50 and the simplex identification means 51 , ...] is determined. That is, while the application method selecting unit 54 determines that there is no accuracy assurance recognizing unit for each pixel position in the area 311 including the pixel position (x ₁ , y ₁ ) in the estimation desired area 300, the pixel position (x ₂ , Y ₂ ) is stored in the method map storage unit 42 for each pixel position in the area 321 including the pixel position (x ₂ , y ₂ ), and each pixel position in the area 331 including the pixel position (x ₃ , y ₃ ) The association with the code A and the code B is stored in the method map storage means 42.

[Operation of Object Distribution Estimation Device 1]
The operation of the object distribution estimation apparatus 1 will be described with reference to the flowcharts of FIGS. 8, 9 and 10.

  FIG. 8 is a schematic flow chart of the operation of the object distribution estimation device 1. When the object distribution estimation apparatus 1 starts operation, the imaging unit 2 installed at the event hall captures images of the monitoring space at predetermined time intervals and sequentially transfers the captured image to the image analysis center where the image processing unit 5 is installed. Send. The image processing unit 5 basically repeats the operation relating to the distribution estimation process of steps S5 to S10 in accordance with the flowchart of FIG. 8 each time a captured image is received. However, when it is necessary to create a method map, the method map creation process of steps S1 to S4 is performed prior to the distribution estimation process of steps S5 to S10.

  That is, when an instruction to change the estimation desired area is input from the operation unit 6 to the image processing unit 5 (in the case of “YES” in step S1), the image processing unit 5 performs method map generation processing (step S4). Further, when there is no method map as when the operation of the object distribution estimation device 1 starts, or when the field of view of the imaging unit 2 is changed (“YES” in step S2), the image processing unit 5 Then, the display unit 8 displays a request for the input of the estimation desired region, waits for the input of the estimation desired region from the operation unit 6 (in the case of “NO” in step S3), and the estimation desired region is input ( In the case of “YES” in step S3), method map creation processing S4 is performed. Then, in these cases, the image processing unit 5 performs the distribution estimation process of steps S5 to S10 after the method map creation process S4.

  On the other hand, if there is no change in the estimated desired area, and a method map already exists, and no change in the visual field occurs (“NO” in steps S1 and S2), the image processing unit 5 creates the method map The distribution estimation process of steps S5 to S10 is performed by omitting S4.

  FIG. 9 is a schematic flow chart of the method map creating process S4, and the method map creating process S4 will be described with reference to FIG.

  When the area desired for estimation is designated in step S1 or S3, the image processing unit 5 operates as the projection means 53, and sequentially sets each pixel in the area for estimation desired to the target position (step S40). The object model read out from the means 43 is projected on the photographing surface of the photographing unit 2 based on the camera parameters read out from the camera parameter storage means 44 (step S41). Next, the image processing unit 5 operates as the application method selection unit 54, and the application method selection unit 54 calculates the resolution of the projection image of the object generated by the projection unit 53 (step S42).

  The application method selection unit 54 compares the resolution of the projected image of the object at the position with the limit resolution of the recognition unit for each of the plurality of recognition units that recognize the number of objects at the target position. If it is a means for recognition of quality assurance, it is stored in the method map. Specifically, in the present embodiment, the recognition means is of two types using the density estimator and the single classifier respectively, and each of them is identified by the above-described recognition means ID and the symbols A and B.

  The application method selection unit 54 reads the limit resolution of the single unit discriminator stored in association with the code A from the limit value storage unit 45, and compares the resolution of the projection image with the limit resolution. Then, if the resolution of the projection image is equal to or higher than the limit resolution (in the case of "YES" in step S43), the single identifier is stored in the method map storage unit 42 as the accuracy assurance recognition unit of the target position (step S44). On the other hand, if the resolution of the projection image is less than the limit resolution (in the case of "NO" in step S43), the single identifier is not used as the accuracy assurance recognition means of the attention position, and step S44 is omitted.

  Subsequent to the determination of the accuracy assurance recognition means for the single unit discriminator in steps S43 and S44, the determination of the accuracy assurance recognition means for the density estimator is similarly performed in steps S45 and S46. That is, the application method selection means 54 reads the limit resolution of the density estimator stored in association with the code B from the limit value storage means 45 and compares it with the resolution of the projected image, and the resolution of the projected image is limited. If it is higher than the resolution (in the case of “YES” in step S45), the density estimator is stored in the method map storage unit 42 as the accuracy assurance recognition unit of the target position (step S46). If it is less than the resolution (in the case of "NO" in step S45), the density estimator is not used as the accuracy assurance recognition means of the position of interest, and step S46 is omitted.

  When the image processing unit 5 repeats the processing of steps S40 to S46 for all pixels in the estimation desired region (in the case of “NO” in step S47) and completes for all pixels (in the case of “YES” in step S47), The process proceeds to step S5 in FIG.

  In step S5, the communication unit 3 operates as the image acquisition unit 30, and waits for reception of a photographed image from the photographing unit 2. The image acquisition unit 30 that has acquired the photographed image outputs the photographed image to the image processing unit 5.

  The image processing unit 5 into which the photographed image is input operates as the density estimation unit 50, and estimates the density distribution from the photographed image (step S6). Specifically, the density estimation unit 50 extracts feature quantities for estimation of the target position with each pixel in the estimated desired area in the photographed image as the target position, and performs density estimation from the density estimator storage unit 40 of the storage unit 4 And the feature amount for estimation is input to the density estimator to obtain the estimated density at the position of interest.

  Further, the image processing unit 5 operates as the single identification unit 51, and identifies the presence or absence of an object from the photographed image (step S7). Specifically, the single-piece identification unit 51 extracts feature quantities for identification of the attention position with each pixel in the estimated desired area in the photographed image as the attention position, and the whole-body discriminator 100, the upper body discriminator 101, and the head Among the neighboring classifiers 102, the single classifiers corresponding to the estimated density of the target position obtained in step S6 are read out from the single classifier storage unit 41, and the classification features are input to the single classifiers to detect the presence or absence of an object. Identify

  The estimated density obtained in step S6 and the presence or absence of an object obtained in step S7 are input to the distribution estimating means 52 for each pixel in the estimated desired area.

  The distribution estimation means 52 reads out the method map stored in the method map storage means 42 (step S8), and generates a distribution image from the method map and the inputted estimated density and the presence or absence of an object (step S9).

  FIG. 10 is a schematic flowchart of the distribution image generation processing S9, and the distribution image generation processing S9 will be described with reference to FIG. The distribution estimation means 52 sequentially sets each pixel of the estimation desired area from the distant pixel to the target position (step S90). Then, with reference to the method map read out in step S8, it is checked whether the accuracy of the single identification unit is ensured for the target position (step S91).

  If the single identification unit is the accuracy assurance recognition unit (in the case of "YES" in step S91), distribution estimating unit 52 determines that the identification result in step S7 for the target position is an object ("in step S92" In the case of “YES”, the projection image of the object model at the target position is drawn (step S93). Here, in the example of the distribution map of the present embodiment described with reference to FIG. 5, the density estimation unit is also the accuracy assurance recognition unit at the pixel position where the single identification unit is the accuracy assurance recognition unit. Therefore, in step S93, as described with reference to FIG. 6, the projected image to be drawn is drawn in a color according to the estimated density in step S6. On the other hand, if the identification result in step S7 indicates no object ("NO" in step S92), step S93 is omitted.

  If the accuracy of the single identification unit is not guaranteed (in the case of "NO" in step S91), the distribution estimation unit 52 refers to the method map to check whether the accuracy of the density estimation unit is guaranteed for the target position (step S94).

  When the density estimation means is the accuracy assurance recognition means (in the case of “YES” in step S94), the distribution estimation means 52 adds a color according to the estimated density of step S6 to the pixels of the target position in the distribution image ( Step S95). On the other hand, when the accuracy of the density estimation means is not guaranteed ("NO" in step S94), step S95 is omitted.

  Distribution estimation means 52 repeats the processing of steps S90 to S95 for all pixels in the estimation desired region (in the case of “NO” in step S96), and when it is completed for all the pixels (in the case of “YES” in step S96), The process proceeds to step S10 in FIG.

  In addition, hidden surface elimination in the distribution image is performed by the paint overlapping method by setting the attention position in order from the distant pixel in step S90.

  In step S10, the communication unit 3 operates as the distribution information output unit 31, and outputs the distribution image generated by the distribution estimation unit 52 to the display control unit 7, and the display control unit 7 causes the display unit 8 to display the distribution image. .

[Modification]
(1) In the above embodiment, the number of pixels constituting the image of an object is defined as the resolution of the object image, but the size (that is, the number of pixels or the area) of the circumscribed rectangle of the image of the object, the height of the image of the object It is equivalent to define the size of a rectangle with a constant aspect ratio based on the width or the size of a rectangle with a constant aspect ratio based on the width of the image of the object as the resolution. This definition makes it easy to derive the resolution of the test image.

  (2) In the above embodiment and its modification, the number of pixels or the area is defined as the resolution of the object image, but when the imaging unit 2 and its imaging magnification are determined, the distance from the imaging unit 2 to the object and the resolution Since the relationship with is uniquely determined, the resolution can be estimated by the distance. That is, the distance from the imaging unit 2 to the object can be used as a value representing the resolution of the object image.

  In this case, the limit value storage unit 45 stores the distance (limit distance) corresponding to the limit resolution. The projection unit 53 derives a three-dimensional position corresponding to each pixel position in the estimated desired area, calculates the distance from the photographing unit 2 to the three-dimensional position, and the application method selection unit 54 calculates the distance calculated by the projection unit 53 And the limit distance are selected to select the accuracy assurance recognition means. That is, the application method selection unit 54 selects, as the accuracy assurance recognition unit, one of a plurality of recognition units for which the distance with respect to the object imaged at the position is equal to or less than the limit distance with respect to the position in the photographed image. . Then, the method map storage means 42 associates and stores the position in the photographed image and the accuracy assurance recognition means.

  For example, the projection unit 53 backprojects the position in the photographed image on a horizontal plane at a height according to the object in the target space to obtain the object virtual position, and the application method selecting unit 54 calculates the position in the photographed image. The distance between the virtual position of the object corresponding to the position and the photographing unit 2 is compared with the limit distance of each of the plurality of recognition means, and the accuracy assurance recognition means to be associated with the position is selected.

  (3) In the above-mentioned embodiment and its modification, distribution estimating means 52 generated a distribution image as distribution information, but distribution information is not limited to image representation. For example, the distribution estimation unit 52 sets the position of the pixel at which the single identification unit 51 recognizes the presence of the object in the photographed image among the pixels associated with the single identification unit 51 as the accuracy assurance recognition unit in the distribution map. Back projection onto the target space based on the camera parameters of the imaging unit 2 is used to calculate information on the three-dimensional position where each object exists. In addition, the distribution estimation unit 52 arranges the pixels for which the density estimation unit 50 estimates the density estimation unit 50 in the distribution map as the accuracy assurance recognition unit, into the areas for each same estimated density, Is backprojected onto the target space according to the camera parameters of the imaging unit 2 to calculate information of a three-dimensional area of each density. Thus, the distribution estimation means 52 may generate three-dimensional distribution information. In this case, distribution information may be generated only from the information of the single-body identification means 51 which is more detailed, with respect to the position where both the density estimation means 50 and the single-body identification means 51 are associated.

  (4) In the above embodiment and its modification, an example in which the object to be detected is a human is shown, but the invention is not limited to this, and an object to be detected may be a vehicle or an animal such as a cow or sheep. it can.

  (5) In the above embodiment and each variation, the density estimator learned by the multiclass SVM method is exemplified, but instead of the multiclass SVM method, a decision tree type random forest method, multiclass Adaboost Various density estimators such as a density estimator learned by (AdaBoost) method or multiclass logistic regression method can be used.

  Alternatively, it may be a density estimator using a discrimination type CNN (Convolutional Neural Network).

  (6) In the above-described embodiment and the respective variations thereof, classes of density other than the background estimated by the density estimator are three classes, but the classes may be divided more finely.

  (7) In the above-described embodiment and each of the modifications thereof, a density estimator that classifies into multiple classes is exemplified, but instead, a regression-type density estimator that regresses the value of density (estimated density) from feature amounts It can also be done. That is, a density estimator which learns parameters of a regression function for obtaining an estimated density from feature amounts by ridge regression method, support vector regression method, regression tree type random forest method, Gaussian process regression, or the like can do.

  Alternatively, it may be a density estimator using a regression type CNN.

  (8) In the above-described embodiment and each of the modifications, the GLCM feature is exemplified as the feature value to be learned by the density estimator and the feature value for estimation, but these may be local binary patterns (Local Binary) instead of the GLCM features. Pattern (LBP) can be a variety of feature quantities such as feature quantity, Haar-like feature quantity, HOG feature quantity, luminance pattern, or a combination of a plurality of GLCM features and these It can also be done.

  (9) In the above-described embodiment and each of the modifications, the density estimation unit 50 and the single identification unit 51 scan at intervals of one pixel. It is also possible to do it empty.

  (10) In the above embodiment and each of the modifications, the single classifiers learned by the linear SVM method are exemplified, but various learning methods conventionally known such as the AdaBoost method instead of the linear SVM method are exemplified. It can also be a single discriminator learned using. In addition, a pattern matching unit can be used instead of the classifier, and the classification score in that case is the inner product of the average pattern of the feature quantities extracted from the human learning image and the feature quantity of the input image, etc. The function can be a function having the score as an output value and the feature amount of the input image as an input value. Alternatively, a discrimination type CNN may be used as a single discriminator.

  (11) In the above embodiment and the respective variations thereof, the HOG feature has been exemplified as the feature to be learned by the single classifier. However, instead of the HOG feature, local binary pattern feature, Haar-like feature , And may be various feature amounts such as a luminance pattern, or may be a feature amount obtained by combining the HOG feature amount and a plurality of these.

  Reference Signs List 1 object distribution estimation device, 2 imaging unit, 3 communication unit, 4 storage unit, 5 image processing unit, 6 operation unit, 7 display control unit, 8 display unit, 30 image acquisition unit, 31 distribution information output unit, 40 density estimation Storage means 41 single discriminator storage means 42 method map storage means 43 object model storage means 44 camera parameter storage means 45 limit value storage means 50 density estimation means 51 single identification means 52 distribution estimation means 53 projection means, 54 application method selection means, 100 whole body discriminator, 101 upper body discriminator, 102 near head part discriminator, 200 photographed image, 250 method map, 300 estimated desired area.

Claims (6)

An image acquisition unit that acquires a photographed image of a space in which a predetermined object photographed by the photographing unit may exist;
A recognition unit that recognizes the number of objects photographed at an arbitrary position in the photographed image, using a recognizer that has previously learned the features of each of a plurality of types of learning images in which the number of photographed objects differs A plurality of recognition means provided for each range of the number of objects using a plurality of the recognizers having different ranges of the number of recognizable objects.
Storage means for associating and storing the position in the photographed image and the accuracy assurance recognition means capable of recognizing the number of objects at the position among the plurality of recognition means with an accuracy equal to or more than a predetermined lower limit.
Distribution estimation means for generating distribution information of the objects in the space from the number of objects acquired in the photographed image by the accuracy assurance recognition means stored in the storage means;
An object distribution estimation apparatus characterized by comprising: The lower limit value of the resolution with respect to the image of the object which can be recognized with an accuracy equal to or higher than the lower limit value is determined as the limit resolution of the recognition means.
The storage means associates, as the accuracy assurance recognition means, one of the plurality of recognition means in which the resolution of the image of the object at the position is equal to or higher than the limit resolution with respect to the position in the photographed image. Remember that,
The object distribution estimation apparatus according to claim 1, characterized in that The resolution is represented by the number of pixels of the image of the object, and the number of pixels corresponding to the limit resolution is the limit number of pixels.
A projection unit configured to project a model imitating the object present in the space onto a photographing surface of the photographing unit to generate a projected image;
The accuracy assurance recognition means for comparing the number of pixels of the projected image generated at the position with the limit number of pixels of each of the plurality of recognition means for the position in the photographed image and associating the number with the position Selection means to select,
The object distribution estimation apparatus according to claim 2, further comprising: The resolution is estimated by the distance from the imaging unit to the object in the space, and the distance corresponding to the limit resolution is set as a limit distance.
The storage means is, for the position within the photographed image, among the plurality of recognition means, one for which the distance for the object photographed at the position is equal to or less than the limit distance is the accuracy assurance recognition means Associate and memorize,
The object distribution estimation apparatus according to claim 2, characterized in that Projection means for back-projecting the position in the photographed image on a horizontal plane at a height according to the object in the space to obtain an object virtual position;
The accuracy in which the distance between the virtual position of the object corresponding to the position and the photographing unit and the limit distance of each of the plurality of recognition means are compared with the position in the photographed image, and the accuracy is associated with the position Selection means for selecting assurance recognition means;
The object distribution estimation apparatus according to claim 4, further comprising: An image acquisition unit that acquires a photographed image of a space in which a predetermined object photographed by the photographing unit may exist;
The density of the object photographed at an arbitrary position in the photographed image, using a density estimator which learns in advance features of respective density images obtained by photographing the space in which the object exists at the predetermined density for each predetermined density Density estimation means for recognizing
A single identification means for recognizing the presence or absence of the object photographed at an arbitrary position in the photographed image, using a single classifier which preliminarily learns features of a single image in which the single object is photographed;
The position in the photographed image is associated with the accuracy assurance recognition means capable of recognizing the density or presence / absence of the object at the position among the density estimation means and the single body identification means with an accuracy higher than a predetermined lower limit value. Storage means for storing
Distribution estimation means for generating distribution information of the object in the space from the density or presence of the object acquired in the photographed image by the accuracy assurance recognition means stored in the storage means;
An object distribution estimation apparatus characterized by comprising:

Images