JP2015052849A - Image processing apparatus, control method therefor, and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve the accuracy of passage detection.SOLUTION: A moving body area and a human body area being areas respectively showing a moving body and a human body are detected from frames constituting an image. At least one between the moving body area and the human body area detected from the respective frames is detected to pass through a predetermined line segment by comparing frames of the image at the different observation times. In the case of acquiring a moving body passage detection result showing that the moving body passes through the predetermined line segment, whether the moving body passage detection result is acquired when the moving body area corresponding to the human body area passes through the predetermined line segment is determined on the basis of a positional relation between the human body area and the moving body area.

Description

  The present invention relates to an image processing technique for detecting that a moving body or a human body, which is a subject in an image, passes a predetermined line segment.

  There are technologies for detecting a moving object region in an image, a technology for detecting a human body region, and a technology for tracking the movement of a detected moving object and a human body within a screen in a surveillance camera. As passage detection, a technique for determining that a detected moving body or human body has passed a predetermined line segment and a technique for determining entry / exit to a predetermined area are known.

  As a passage detection technique that combines moving body detection and human body detection, there is an intruder monitoring device that performs human body region detection only in a region where a moving body is detected and determines whether a person has entered a prohibited region ( Patent Document 1). In addition, there is a technique for tracking a moving object in an apparently overlapping or hidden scene by paying attention to a statistic based on the relevance of a local region in a spatiotemporal image, which has been implemented (Patent Documents 2 and 3).

Japanese Patent Laid-Open No. 9-50585 JP 2010-113732 A JP 2012-003780 A

  Conventional moving object detection methods have a problem that they are vulnerable to environmental fluctuation noise such as overlapping of objects on an image, hiding in other objects, and illumination conditions. On the other hand, there is a method for solving the above problem by paying attention to a statistic based on the relevance of the spatiotemporal image. However, when people are walking side by side, they may not be detected separately. For this problem, a method of tracking after detecting each person individually from the outline of the human body is effective. However, the detection accuracy of human body detection may deteriorate depending on the orientation and angle of a person, and cannot always be detected with high accuracy.

  SUMMARY An advantage of some aspects of the invention is that it provides an image processing apparatus that can improve the accuracy of passage detection, a control method therefor, and a program.

In order to achieve the above object, an image processing apparatus according to the present invention comprises the following arrangement. That is,
An image processing apparatus for detecting that a moving body or a human body as a subject in an image passes a predetermined line segment,
Detecting means for detecting a moving body region and a human body region, which are regions indicating the moving body and the human body, from the frame constituting the image;
Detection means for detecting that at least one of the moving body area and the human body area detected by the detection means from each frame passes through the predetermined line segment by comparing frames of images having different observation times. When,
When the detection means obtains a moving body passage detection result indicating that the moving body region has passed the predetermined line segment, the moving body region corresponding to the human body region has passed the predetermined line segment, so Determination means for determining whether or not a passage detection result is obtained based on a positional relationship between the human body region and the moving body region.

  According to the present invention, the accuracy of passage detection can be improved.

It is a block diagram of a passage detection determination device. It is principle explanatory drawing of a moving body human body integrated passage detection. It is principle explanatory drawing of a moving body human body integrated passage detection. It is a flowchart which shows the process of a passage detection determination apparatus. It is a flowchart which shows the detail of a human body tag provision process. It is a flowchart which shows the detail of the integrated process of a passage detection determination. It is a figure for demonstrating provision of the human body tag with respect to a moving body area | region. It is a flowchart which shows the passage detection result integration process including the case where a human body detection result fails.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<Embodiment 1>
The present embodiment relates to a passage detection process in which a subject (moving body / human body) in an image is detected and tracked and the subject detects passage of a predetermined line segment (detection line) in the image. In particular, the detection accuracy is improved by integrating the detection results of moving bodies and human bodies.

  A configuration example of a passage detection determination apparatus 100 that performs passage detection determination is shown in FIG.

  The passage detection determination device 100 is configured to assume a control device integrated with the imaging device 200 that is a monitoring camera that distributes an image on a network. The passage detection determination apparatus 100 includes an imaging system control unit 120 that controls zoom, focus, white balance, shooting posture, and the like of the imaging apparatus 200. In the passage detection determination device 100, the image acquisition unit 101 acquires an image from the imaging device 200. The acquired image is subjected to scaling processing by the scaling unit 102 in accordance with post-processing. The image distributed from the passage detection determination device 100 is distributed to the display device 150 and the recording device 160 connected to the network by obtaining the encoding unit 103, the image distribution control unit 104, and the communication unit 140.

  The moving object detection unit 105 detects a moving object region using the image after scaling as an input. For moving object detection, known image processing such as background difference using a mixed Gaussian distribution or region division by spatiotemporal image analysis is used. The moving object tracking unit 106 performs a tracking process on the detected moving object. This is to extract moving trajectories of moving objects by associating moving object regions between different frames. A method that pays attention to the similarity of the circumscribed rectangular shape (moving object region) of the moving object in the vicinity of the predicted moving position is generally used.

  The human body detection unit 107 detects a human body in the image using human body feature data (dictionary). For human body detection, there is a known method for detecting a silhouette of a head and a shoulder (so-called omega line). The human body tracking unit 108 tracks the detected human body. As the tracking algorithm, an algorithm common to the moving object tracking unit 106 can be used.

  After the moving body / human body tracking process, the passage detection determination unit 109 performs passage detection determination. The passage detection is determined based on whether or not a trajectory (movement vector) of a human body or a moving body passes (crosses) a predetermined detection line in the imaging screen of the imaging apparatus 200. In this determination, for example, by comparing image frames having different observation times, it is determined whether at least one of the moving body region and the human body region detected from each image frame has passed a predetermined line segment. Then, the passage detection determination result by the passage detection determination unit 109 is integrated by the passage detection result integration unit 110. Here, when a moving body passage detection result and a human body passage detection result overlap, a human body passage detection result is employ | adopted. The integrated detection result is transmitted to the display device 150 and the recording device 160 on the network via the communication unit 140.

  The setting control unit 130 performs settings for controlling various operations of the moving object detection unit 105, the human body detection unit 107, and the imaging system control unit 120. This setting may be based on a user input to the passage detection determination device 100, or may be based on a setting instruction received from the network via the communication unit 140.

  In the present embodiment, the passage detection determination device 100 is described assuming an image processing device with a built-in monitoring camera (imaging device 200). It can also be realized by a PC server type that realizes the function of the component. In this case, the image is acquired from an external network camera (surveillance camera) connected on the network and processed.

  Here, the information processing apparatus for realizing the passage detection determination apparatus 100 is a standard component (for example, CPU, RAM, ROM, hard disk, external storage device, network interface, display, keyboard, mouse) mounted on a general-purpose computer. Etc.). And the process which the passage detection determination apparatus 100 performs is realizable when CPU mounted in it performs and controls the program memorize | stored in memories, such as ROM, for example.

  The integration principle of the passage detection result will be described with reference to FIG. FIG. 2 shows a case where two persons pass through the detection line 211 from opposite directions. In FIG. 2, a) -1 to a) -3 are moving body detection processes, b) -1 to b) -3 are human body detection processes, and c) -1 to c) -3 are moving body / human body detection processes. It is a figure for demonstrating integration processing and each. Both show processing for three frames of images. The processing time (observation time) of each frame is represented as t1, t2, and t3.

  In addition, although the detection line 211 which is a predetermined line segment has shown the case where it is a straight line, the predetermined line segment used as this detection line is not limited to this. For example, the predetermined line segment serving as a passage detection mark may be configured by connecting a plurality of line segments, or may be a part of a side forming a polygon.

  First, examples of passage detection determination using moving body detection will be described with reference to a) -1 to a) -3.

  In a) -1, two persons 212 and 214 in the image 201 are detected and tracked as moving object regions 213 and 215, which are different moving objects, respectively.

  In a) -2, two persons 212 and 214 in the image 202 are tracked as one moving body region 216.

  In a) -3, one moving object region 217 in the image 203 passes the detection line 211. The reason why a plurality of persons are detected and tracked as one moving object is due to an apparent overlap or a plurality of moving object regions in the vicinity of each other being combined into one moving object region. In a) -3, one moving object detection is observed by the intersection of the detection line 211 and the movement of the moving object region (moving object region 216 → moving object region 217). As a result, the passage detection determination based only on the moving body detection cannot correctly detect the passage detection of two persons having different movement directions.

  Next, examples of passage detection determination using human body detection will be described with reference to b) -1 to b) -3.

  In b) −1, two persons are detected as two human body regions 221 and 222 in the image 204.

  In b) -2, two persons are detected and tracked as independent human body regions 223 and 224 in the image 205.

  In b) -3, two persons are detected and tracked as independent human body regions 225 and 226 in the image 206. Therefore, in b) -3, it is detected that the human body regions 225 and 226 have passed from different directions with respect to the detection line 211. As a result, human body passage detection of two persons having different moving directions is observed.

  As described above, in the case of FIG. 2, a correct passage detection result is obtained by employing human body passage detection. For this purpose, it is necessary to determine the overlap between human body passage detection and moving body passage detection.

  The overlap detection process for passage detection will be described. Look at the relationship between the human body area and the moving body area. Both the human body area and the moving body area are represented by rectangles. First, a human body region included in the moving body region is selected. Next, circumscribed rectangles of all human body region sets included are calculated. Finally, the positional relationship between at least three sides (upper side, left side, and right side) of the circumscribed rectangle and at least three sides (upper side, left side, and right side) of the original moving object region is compared to determine whether they match. To do. If they match, a human body tag is assigned to the moving object detection result (moving object tracking result).

  The reason why the lower side is not used when comparing the positional relationship is that the human body detection result may not match the feet of the person on the image when the lower body is hidden or the image is taken at an angle looking down at the person It is. However, in addition to the fact that the upper side, the left side, and the right side match, the lower side may match, so in such a case, a human body tag should be assigned to the moving object detection result (moving object tracking result). Can do.

  For example, in c) -1 of FIG. 2, two persons are detected as two moving body regions 213 and 215 in the image 207 and also as human body regions 221 and 222.

  c) -2, in the image 208, the moving body region 216 that is the moving body detection result includes the human body regions 223 and 224 that are the human body detection results, and all the sides are circumscribed rectangles of the included human body region set. Matches. In this case, a human body tag 227 is attached to the moving body region 216.

  c) -3, similarly, in the image 209, the moving body region 217 also includes the human body regions 225 and 226, and all sides coincide with the circumscribed rectangle of the included human body region set. In this case, a human body tag 228 is assigned to the moving object region 217. Thus, the moving object region 217 that is the moving object detection result coincides with the circumscribed rectangle of the human body region set that is a set of human body detection results.

  From the above, the passage detection result caused by the moving body detection result (moving body region 217) in c) -3 is not adopted (invalid), and the passage detection result caused by the human body detection result (human body regions 225 and 226) is adopted. To do. In this example, a case is described in which people pass each other, but it goes without saying that a human body tag can be used to detect correctly even when people pass side by side in the same direction.

  In the present embodiment, passage detection determination is performed on the moving object detection result, and a moving object passage detection flag is set. When the moving object passage detection result is adopted, the moving object passage detection flag is set to ON, and when not adopted, the moving object passage detection flag is set to OFF.

  Also in FIG. 2 c) -1, the moving body regions 213 and 215 include the human body regions 221 and 222, respectively, and all sides coincide with the circumscribed rectangle of the included human body region set. Therefore, human body tags 231 and 232 are assigned to the moving body regions 213 and 215, respectively.

  Next, an example in which a person and a vehicle intersect will be described with reference to FIG. FIG. 3 shows a case where a person and a vehicle pass through the detection line 311 from opposite directions. As in FIG. 2, a) -1 to a) -3 in FIG. 3 are moving body passage detection processes, b) -1 to b) -3 are human body passage detection processes, and c) -1 to c) -3 are It is a figure for demonstrating each integration processing of a moving body detection and a human body passage detection.

  First, examples of passage detection determination using moving body detection will be described with reference to a) -1 to a) -3.

  In a) -1, there is a detection line 311 in the center of the screen in the image 301, the person 312 moves from left to right, and the vehicle 313 moves from right to left. In this case, the correct value for correct detection is to detect the passage of a person and the passage of a vehicle as separate passage detections. In a) -1, the moving object detection results corresponding to the person 312 and the vehicle 313 are detected as separate moving object regions 134 and 315.

  In a) -2, a moving object region 318 is obtained as one moving object detection result for two objects (a person 316 and a vehicle 317) in the image 302.

  In a) -3, similarly, the moving object region 300 is obtained as one moving object detection result for two objects (person 319 and vehicle 320) in the image 303.

  From a) -2 to a) -3, the moving object region that is the moving object detection result passes through the detection line 311. Therefore, it can be determined that the moving object detection has occurred in a) -3.

  Next, examples of passage detection determination using human body detection will be described with reference to b) -1 to b) -3.

  In b) −1, one person is detected as the human body region 321 in the image 304.

  b) -2, one person is detected as the human body region 322 in the image 305.

  In b) -3, one person is detected as the human body region 323 in the image 306.

  In this case, since the detection line 311 is passed from b) -2 to b) -3, human body passage detection occurs in b) -3.

  Thus, in FIG. 3, the human body passage detection result and the moving body passage detection result do not overlap.

  In c) -1 in FIG. 3, the human body region 322 is included in the moving body region 318 in c) -2 from the state of the image 307. However, the right side and upper side of the moving body region 318 do not coincide with the right side and upper side of the human body region 322. Since there is no other human body region included in the moving body region, in c) -2, it is determined that the moving body region 318 and the human body region 322 are not the same object, and no human body tag is assigned to the moving body region 318. Similarly, in c) -3, no human body tag is assigned to the moving object region 300. Then, from c) -2 to c) -3, the moving body region 318 and the human body region 322 pass through the detection line 311 in different directions. Therefore, in c) -3, moving body detection and human body passage detection occur in the image 309. In this case, since the human body tag is not given to the moving body region 300, both passage detections are employed. As a result, the correct value for detecting the passage of the person and the passage of the vehicle as separate passage detection is obtained.

  Next, the processing procedure of the passage detection determination apparatus 100 of this embodiment will be described with reference to FIG.

  Note that the operation of the passage detection determination device 100 is realized by the CPU reading and executing a program stored in a memory such as a ROM.

  In step S401, the image acquisition unit 101 acquires an image. Next, in step S411, the moving object detection unit 105 performs a moving object detection process. In step S412, the moving object tracking unit 106 performs moving object tracking processing. Further, in step S413, the human body detection unit 107 performs human body detection processing. In step S414, the human body tracking unit 108 performs human body tracking processing. Thereafter, in step S415, the setting control unit 130 performs a human body tag attaching process. Details of this will be described later.

  In step S416, the passage detection determination unit 109 performs moving body passage detection determination. This process determines whether or not all moving objects have passed a predetermined detection line, and sets the moving object detection flag to ON for moving objects that have passed the detection line. In step S417, the passage detection determination unit 109 performs human body passage detection determination. In step S418, the passage detection result integration unit 110 performs an integration process for moving body / human body passage detection determination. Details of this process will be described later. Finally, in step S419, the communication unit 140 transmits the passage detection result. In step S420, the setting control unit 130 determines whether there is a setting end request. If there is no setting end request (NO in step S420), the process returns to step S410. On the other hand, if there is a setting end request (YES in step S420), the process ends.

  Next, details of the human body tag attaching process in step S415 will be described with reference to FIG.

  In step S501, the setting control unit 130 determines whether the processing has been completed for all moving object regions. If completed (YES in step S501), the process ends. On the other hand, if not completed (NO in step S501), in step S502, the setting control unit 130 extracts a human body region included in the moving body region. Next, in step S503, the setting control unit 130 calculates a circumscribed rectangle Rh that surrounds all the human body regions included in the moving body region.

  Next, in step S504, the setting control unit 130 compares the circumscribed rectangle Rm representing the moving body region with the circumscribed rectangle Rh of the human body region included therein. This compares whether the left side, the right side, and the top side of Rm and Rh match each other. The reason why the lower sides are not compared is that a case where a person's feet are not detected as a moving object is considered due to the hiding and the angle of the imaging device. Next, in step S505, the setting control unit 130 determines whether or not the circumscribed rectangle Rm representing the moving body region matches the circumscribed rectangle Rh of the human body region included therein. That is, whether the moving body area corresponds to the human body area is determined based on the positional relationship between the moving body area and the human body area. For example, it is determined whether at least three sides of the circumscribed rectangle of the moving body area and the circumscribed rectangle of the human body area included in the circumscribed rectangle of the moving body area match. If they match, it is determined that the moving body area corresponds to the human body area. If they do not match (NO in step S505), the process returns to step S501 to perform processing for the next moving object region. On the other hand, if they match (YES in step S505), the setting control unit 130 adds a human body tag to the moving object region.

  The process of integrating the passage detection results in step S418 will be described with reference to FIG.

  In step S601, the passage detection determination unit 109 determines whether or not processing for all moving object regions has been completed. If the process is complete (NO in step S601), the process ends. On the other hand, when the moving object region to be processed remains, that is, when the process is not completed (NO in step S601), in step S602, the passage detection determination unit 109 determines whether the moving object contributes to the passage detection. Determine. This determination is made based on whether or not the moving object passage detection flag is ON in the moving object region to be processed.

  The moving object passage detection flag is a flag that is given (ON) to the moving object region to be processed. When the moving object region is detected, the moving object passage detection flag is once set to ON. When the moving object region is adopted for passage detection (contributes to passage detection), the ON state is maintained as it is, and when it is not adopted, it is set to OFF.

  As a result of the determination, if the moving object passage detection flag is OFF, that is, if the moving object region does not contribute to the passage detection (NO in step S602), the process returns to step S601, and processing is performed for the next moving object region to be processed. On the other hand, if the moving object passage detection flag is ON, that is, if the moving object contributes to the passage detection (YES in step S602), in step S603, the passage detection determination unit 109 adds a human body tag to the moving object region to be processed. It is determined whether or not. When the human body tag is not assigned (NO in step S603), the process returns to step S601, and the process is executed for the next moving object region to be processed. On the other hand, when the human body tag is attached (YES in step S603), in step S604, the passage detection determination unit 109 turns off the moving body passage detection flag. Then, it returns to step S601 and continues the process. That is, when a moving object passage detection result indicating that the moving object region has passed the predetermined detection line is obtained (step S602), the moving object region corresponding to the human body region passes the predetermined detection line. It is determined whether or not it has been obtained (step S603). The association between the human body region and the moving body region is determined based on the positional relationship between the human body region and the moving body region, as described with reference to FIG.

  If it is determined that the moving body passage detection result has been obtained because the moving body region corresponding to the human body region has passed the predetermined detection line, the moving body passage detection result is invalidated in step S604. In this case, the human body passage detection result remains valid.

  On the other hand, if it is determined that a moving body passage detection result is obtained because a moving body region that does not correspond to the human body region has passed a predetermined detection line, the moving body passage detection result is not invalidated and remains valid. Also in this case, the human body passage detection result remains valid.

  For all moving object regions, the moving object passage detection flag is confirmed after the process of step S418 is completed. If there is at least one moving body region or human body region that contributes to passage detection with respect to the designated detection line, it is determined that passage detection has occurred in the detection line.

  As described above, according to the first embodiment, the moving object detection result and the human body detection result are combined, and at the time of passage detection, the combination or any one of them is adaptively selected and used for passage detection. By doing so, the accuracy of passage detection can be improved.

<Embodiment 2>
In the first embodiment, the human body passage detection is prioritized when the human body passage detection and the moving body passage detection are overlapped under the condition that the human body detection is stable and successful. However, in practice, human detection is not always successful. For example, when a person is close to the imaging apparatus 200 or when a person wears a hat or a mask, the human body detection accuracy decreases. Therefore, in the second embodiment, a description will be given of a configuration that is replaced by passage detection by moving body detection when human body detection and human body passage detection fail.

  FIG. 7 shows a case where a person moves in time series. In order to simplify the explanation, one person is assumed. Assume that the frame number specifying each time-series frame group constituting the captured image is indicated by “#”, and in FIG. 7, the frames of frame numbers # 001, # 003, # 006, and # 009 are shown.

  Moving object regions 701, 702, 703, and 704 are detected in the respective frames of frame numbers # 001, # 003, # 006, and # 009. On the other hand, human body regions 710, 711, and 713 are detected in the respective frames of frame numbers # 001, # 003, and # 009. In the frame of frame number # 006, a person exists on the screen but is not detected as a human body region. When a human body is detected in each frame, a human body tag is assigned to the moving body region according to the configuration of the first embodiment. As a result, human body tags 705, 706, and 707 are assigned to the moving object regions 701, 702, and 704 of the frames of frame numbers # 001, # 003, and # 009, respectively. In frame number # 006, no human body tag is added because no human body is detected.

  When a moving object passage detection event occurs in the moving object region 702 of the frame of frame # 003, since the human body tag 706 is assigned to the moving object region 702, has the corresponding human body region 711 contributed to the human body passage detection? Confirm. When the human body region 711 contributes to human body passage detection, it can be determined that it has overlapped with the moving body passage detection, so the moving body passage detection flag of the moving body region 703 is turned OFF and the human body passage detection is adopted.

  On the other hand, when a moving object passage detection event occurs in the moving object region 703 of the frame of frame # 006, since the moving object region 703 is not given a human body tag, moving object detection is effective.

  When a moving object passage detection event occurs in the moving object area 704 of the frame of frame # 009, the moving object area 704 is given a human body tag 707. However, the human body region 713 corresponding to the human body region 713 does not generate a human body passage detection event because the human body region is not detected in the immediately preceding (or previous) frame # 006. As a result, the moving object passage detection flag relating to the moving object region 704 remains ON, and the moving object passage detection event is valid.

  By the way, even if the human body region is not detected in a certain frame, it is possible to continue tracking for a short time by using the human body region detected in the previous frame. For example, even when the human body is not detected in a certain frame, it is possible to continue tracking while leaving the human body region in a predetermined number of frames before that frame.

  In the example of FIG. 7, this frame corresponds to the human body region 712. However, in this case, the human body passage detection determination is performed only for the frame in which the human body region is detected. In FIG. 7, the passage detection process is not applied to the human body region 712 in the frame of frame # 006. On the other hand, the passage detection process is applied to the human body region 713 in the frame of frame # 009. As a result of the tracking process, when it is determined that the human body region 712 and the human body region 713 are the same human body region, the human body region 713 is a detection target of the passage detection process.

  Next, details of the integration processing for moving body / human body passage detection determination in the present embodiment will be described with reference to FIG.

  In step S901, the passage detection determination unit 109 determines whether or not processing for all moving object regions has been completed. If the process is complete (NO in step S901), the process ends. On the other hand, when the moving object region to be processed remains, that is, when the process is not completed (NO in step S901), in step S902, the passage detection determination unit 109 determines whether the moving object contributes to the passage detection. Determine. This determination is made based on whether or not the moving object passage detection flag is ON in the moving object region to be processed.

  As a result of the determination, if the moving object passage detection flag is OFF, that is, if the moving object region does not contribute to the passage detection (NO in step S902), the process returns to step S901, and the process is executed for the next moving object region to be processed. On the other hand, when the moving object passage detection flag is ON and the moving object contributes to the passage detection (YES in step S902), in step S903, the passage detection determination unit 109 adds a human body tag to the moving object region to be processed. Determine whether or not. When the human body tag is not given (NO in step S903), the process returns to step S901, and the process is executed for the next moving object region to be processed.

  On the other hand, if a human body tag has been assigned (YES in step S903), whether or not the passage detection determination unit 109 has completed processing for all human body regions corresponding to the moving body region to which the human body tag has been assigned in step S910. Determine whether or not. If the process is complete (YES in step S910), the process returns to step S901, and the process is executed for the next moving object region to be processed. On the other hand, if the process has not been completed (NO in step S910), it is determined in step S911 whether or not the human body region has contributed to human body passage detection.

  If the human body region does not contribute to the passage detection (NO in step S911), the process returns to step S910, and the process is executed for the next human body region to be processed. On the other hand, if the human body region contributes to the passage detection (YES in step S911), the passage detection determination unit 109 turns off the moving body passage detection flag in step S912. As a result, the human body passage detection result is adopted. Then, it returns to step S910 and continues processing.

  As described above, according to the second embodiment, it is determined whether or not the human body region corresponding to the moving body region has been detected to pass in the processing target frame in which the moving body passage detection has occurred. Thereby, even if it is a case where a human body detection fails, passage detection can be performed using a moving body detection result. As a result, it is possible to prevent omission of detection while ensuring passage detection accuracy.

<Embodiment 3>
In the third embodiment, a case where the processing speeds of moving object detection and human body detection are different will be described.

  For example, consider a case where the human body detection process is slower than the moving body detection process, and the process is completed at a rate of one human body detection process for two or three times the moving body detection process. In this case, only the moving object detection processing result is obtained in frame # 002 (not shown) in FIG. The moving object passage detection event generated at this time is not issued immediately, but waits for the next human body detection in the frame of the subsequent frame # 003.

  Then, moving body detection and human body detection are executed in frame # 003. A human body tag 706 is attached to the moving body area 702 of the frame of frame # 003. At that time, moving body region information in a frame of frame # 002 (not shown) is held, and a human body tag is assigned to this moving body region on the assumption that the human body detection result 711 corresponds. Then, a human body passage detection determination is performed on the human body detection result 711 in the frame of frame # 003. If a human body passage is detected, the moving body passage detection flag in frames # 002 and # 003 is turned OFF.

  As described above, according to the third embodiment, the present invention can be applied even when the processing speeds of the moving object detection and the human body detection are different and the frequency of obtaining the passage detection output is different. As a result, the detection accuracy can be improved by combining the moving body passage detection and the human body passage detection.

  The present invention can also be realized by executing the following processing. That is, software (program) that realizes the functions of the above-described embodiments is supplied to a system or apparatus via a network or various storage media, and a computer (or CPU, MPU, or the like) of the system or apparatus reads the program. It is a process to be executed.

Claims (13)

An image processing apparatus for detecting that a moving body or a human body as a subject in an image passes a predetermined line segment,
Detecting means for detecting a moving body region and a human body region, which are regions indicating the moving body and the human body, from the frame constituting the image;
Detection means for detecting that at least one of the moving body area and the human body area detected by the detection means from each frame passes through the predetermined line segment by comparing frames of images having different observation times. When,
When the detection means obtains a moving body passage detection result indicating that the moving body region has passed the predetermined line segment, the moving body region corresponding to the human body region has passed the predetermined line segment, so An image processing apparatus comprising: a determination unit that determines whether or not a passage detection result is obtained based on a positional relationship between the human body region and the moving body region. The determination means invalidates the moving body passage detection result when it is determined that the moving body passage detection result is obtained because the moving body region corresponding to the human body region passes the predetermined line segment. The image processing apparatus according to claim 1. The determination means includes at least three sides of a circumscribed rectangle of the moving body region contributing to the moving body passage detection result and a circumscribed rectangle of the human body region contributing to the human body passage detection result included in the circumscribed rectangle of the moving body region. 2. The image processing apparatus according to claim 1, wherein in the case of coincidence, the moving object region corresponding to the human body region is determined to have obtained the moving object passage detection result by passing through the predetermined line segment. The determination means includes at least three sides of a circumscribed rectangle of the moving body region contributing to the moving body passage detection result and a circumscribed rectangle of the human body region contributing to the human body passage detection result included in the circumscribed rectangle of the moving body region. If they do not match, it is determined that the moving body passage detection result has been obtained by passing a predetermined moving segment of the moving body region that does not correspond to the human body region, and the moving body passage detection result and the detection result are detected as the detection result of the detection means. The image processing apparatus according to claim 1, wherein both the human body passage detection results are validated. The detecting means detects a first human body region and a corresponding first moving body region in a first frame, does not detect a human body region in a subsequent second frame, and further detects a human body region in a subsequent third frame. When the second human body region and the first moving body region corresponding to the second human body region are detected and it is determined that the first human body region and the second human body region are the same, the detection unit includes: The image processing apparatus according to any one of claims 1 to 4, wherein the second human body region is a detection target. When the processing speed of the human body detection process of the human body region by the detection unit is slower than the processing speed of the moving body detection process of the moving body region, it is detected by the moving body detection process executed until the human body detection process is executed. Holding means for holding moving body area information related to the moving body area,
When the human body region is detected by the human body detection process and the moving body region is detected by the moving body detection process, the determination unit contributes to the human body passage detection result by the detection unit. 6. The method according to claim 1, further comprising: determining whether the moving body region indicated by the moving body region information is a moving body region contributing to the human body passage detection result by the detection unit. The image processing apparatus according to any one of the above. The predetermined line segment is a line segment formed by connecting a plurality of line segments or a part of a side forming a polygon. 7. Image processing apparatus. A control method of an image processing apparatus for detecting that a moving object or a human body as a subject in an image passes a predetermined line segment,
A detection step of detecting a moving body region and a human body region, which are regions indicating the moving body and the human body, from the frame constituting the image;
A detection step of detecting that at least one of the moving body region and the human body region detected in the detection step from each frame passes through the predetermined line segment by comparing frames of images having different observation times. When,
When the moving step detection result indicating that the moving body region has passed the predetermined line segment is obtained by the detection step, the moving body region corresponding to the human body region has passed the predetermined line segment, so that the moving body is And a determination step of determining whether or not a passage detection result has been obtained based on a positional relationship between the human body region and the moving body region. The determination step includes at least three sides of a circumscribed rectangle of the moving body region contributing to the moving body region detection result and a circumscribed rectangle of the human body region included in the circumscribed rectangle of the moving body region and contributing to the human body passage detection result. The image processing apparatus according to claim 8, wherein, when they match, it is determined that the moving object passage detection result is obtained by moving the moving object region corresponding to the human body region through the predetermined line segment. Control method. The determination step includes at least three sides of a circumscribed rectangle of the moving body region contributing to the moving body region detection result and a circumscribed rectangle of the human body region included in the circumscribed rectangle of the moving body region and contributing to the human body passage detection result. If they do not match, it is determined that the moving body passage detection result has been obtained by passing a predetermined moving segment of the moving body region that does not correspond to the human body region, and as the detection result of the detection step, the moving body passage detection result and the The method for controlling an image processing apparatus according to claim 8, wherein both of the human body passage detection results are validated. The detecting step detects a first human body region and a corresponding first moving body region in a first frame, does not detect a human body region in a subsequent second frame, and further detects in a subsequent third frame. When the second human body region and the first moving body region corresponding to the second human body region are detected and it is determined that the first human body region and the second human body region are the same, the detection step includes: The method of controlling an image processing apparatus according to any one of claims 8 to 10, wherein the second human body region is a detection target. When the processing speed of the human body detection process of the human body region by the detection step is slower than the processing speed of the moving body detection process of the moving body region, it is detected by the moving body detection process executed until the human body detection process is executed. A holding process for holding the moving body area information on the moving body area in the storage medium,
When the human body region is detected by the human body detection process and the moving body region is detected by the moving body detection process, the determination step contributes to the human body passage detection result by the detection step. It is determined whether the moving body area | region which the said moving body area | region information shows according to whether it is a moving body area | region which contributes to the said human body passage detection result by the said detection process is characterized by the above-mentioned. The control method of the image processing apparatus of any one of Claims. A program for causing a computer to control the control of an image processing device that detects that a moving object or a human body that is a subject in an image passes a predetermined line segment,
The computer,
Detecting means for detecting a moving body region and a human body region, which are regions indicating the moving body and the human body, from the frame constituting the image;
Detection means for detecting that at least one of the moving body area and the human body area detected by the detection means from each frame passes through the predetermined line segment by comparing frames of images having different observation times. When,
When the detection means obtains a moving body passage detection result indicating that the moving body region has passed the predetermined line segment, the moving body region corresponding to the human body region has passed the predetermined line segment, so A program that functions as a determination unit that determines whether a passage detection result is obtained based on a positional relationship between the human body region and the moving body region.

Images