JP2018001115A - Potato determination device and potato selector - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a potato determination device and a potato selector capable of improving the accuracy of determining whether an object is a potato or not.SOLUTION: A potato selector comprises a texture feature amount extraction unit 14 for extracting a feature amount for a texture of an object based on image data for the object acquired by a line scan camera 8, a first comparison unit 17 for comparing the extracted feature amount with a basic feature amount recorded in a storage unit 16, a brightness information acquisition unit 15 for receiving light reflected at the object and acquiring brightness information for a wavelength based on the received light, a second comparison unit 18 for comparing the brightness information for the acquired object with basic brightness information recorded in the storage unit 16, and a determination unit 19 for determining whether or not the object is a potato based on results of comparison from the first comparison unit 17 and the second comparison unit 18.SELECTED DRAWING: Figure 4

Description

  The present invention relates to a wrinkle determination device and a wrinkle sorting device.

  When potatoes are harvested from the field using a harvester (harvester), foreign matter such as stones and soil blocks are also harvested in addition to potatoes. In order to store or transport the harvested potato, it is necessary to remove foreign matter from the harvested potato. In the following, potato is described, but the same applies to other potatoes such as sweet potatoes.

  Patent Document 1 below discloses an apparatus that can change a falling trajectory of an object by bringing a direction changing device into contact with the falling object. This apparatus selects an object according to quality or removes foreign matter by adjusting the position of the direction changing device based on a determination result obtained by optically scanning the object.

Special table 2014-533604 gazette

  As a technique for distinguishing potatoes from foreign substances, optical processing is generally performed. In this case, it is possible to reliably distinguish potatoes from foreign substances while the surface is covered with soil. It has been demanded.

  Since potatoes and stones harvested from the field are covered with soil, it is difficult to distinguish between them based on the overall shape and surface color when the stones approximate the shape of potatoes. is there. In addition, potatoes have various shapes such as a shape close to a sphere, a shape that is long in one direction, and a shape with a narrowed middle part, and potatoes can be distinguished from foreign objects such as artifacts and stems other than stones. It can be difficult.

  The present invention has been made in view of such circumstances, and provides a wrinkle determination device and a wrinkle sorting device capable of improving the determination accuracy in determining whether or not an object is a wrinkle. With the goal.

In order to solve the above-mentioned problems, the bag judging device and the bag sorting device of the present invention employ the following means.
That is, the wrinkle determination device according to the present invention includes an imaging unit that captures an image of an object, an irradiation unit that irradiates the object with light having a predetermined wavelength, and an object captured and acquired by the imaging unit. Based on image data, a feature extraction unit that extracts a feature amount related to the texture of the object, a first storage unit in which a basic feature amount related to the eyelid image data is recorded in advance, the extracted feature amount and the basic feature Based on a first comparison unit that compares feature amounts, a luminance information acquisition unit that receives light reflected by the object, acquires luminance information of the wavelength based on the received light, and light reflected by the eyelid A second storage unit in which basic luminance information of the wavelength is recorded in advance; a second comparing unit that compares the acquired luminance information of the object with the basic luminance information; the first comparing unit and the second comparing unit; Based on the comparison result in the comparison unit, Serial object and a determining section for determining whether the potatoes.

  In the above invention, the irradiation unit sequentially irradiates the object with a plurality of lights having different wavelengths, the luminance information acquisition unit acquires the luminance information for each wavelength, and the second storage unit Basic luminance information for each wavelength may be recorded in advance, and the second comparison unit may compare the acquired luminance information of the object and the basic luminance information for each wavelength.

  In the above invention, the feature extraction unit may extract a feature amount related to the texture of the object using a Local Binary Pattern.

  In the above invention, the luminance information acquisition unit may calculate an average value or a variance value of luminance in the image data of the object as the luminance information.

  In the above invention, a curvature information acquisition unit that extracts a contour shape of the target object and acquires curvature information based on the image data of the target object captured and acquired by the imaging unit; It may further include a third storage unit in which curvature information is recorded in advance, and a third comparison unit that compares the acquired curvature information with the basic curvature information, and the determination unit includes the first comparison unit, Based on the comparison results in the second comparison unit and the third comparison unit, it may be determined whether or not the object is a bag.

  In the above invention, the curvature information acquisition unit may calculate a curvature for each predetermined length on the contour shape, and calculate a frequency for each calculated curvature as the curvature information.

  A wrinkle sorting apparatus according to the present invention includes the above-described wrinkle determination apparatus and a sorting unit that sorts the object based on a result determined by the wrinkle determination apparatus.

  According to the present invention, it is possible to improve the determination accuracy in determining whether or not an object is a bag.

It is a lineblock diagram showing the potato sorting device concerning a 1st embodiment of the present invention. It is a timing chart regarding operation | movement of the rice scan camera and LED illumination part of the potato sorting apparatus which concerns on 1st Embodiment of this invention. It is the schematic which shows the image pick-up element of the rice scan camera of the potato sorting apparatus which concerns on 1st Embodiment of this invention. It is a block diagram showing a potato sorting device concerning a 1st embodiment of the present invention. It is a flowchart which shows operation | movement of the potato sorting apparatus which concerns on 1st Embodiment of this invention. It is a flowchart which shows operation | movement of the control part of the potato sorting apparatus which concerns on 1st Embodiment of this invention. It is a block diagram which shows the potato sorting apparatus which concerns on 2nd Embodiment of this invention. It is a flowchart which shows operation | movement of the control part of the potato sorting apparatus which concerns on 2nd Embodiment of this invention.

[First Embodiment]
Hereinafter, the potato sorting apparatus 1 according to the first embodiment of the present invention will be described.
The potato sorting apparatus 1 according to the first embodiment of the present invention is an apparatus that sorts foreign substances mixed in a large number of harvested and stored potatoes and potatoes. Foreign materials include stones, earth clots, artifacts, stems, wood fragments, shells, and the like. The potato sorting apparatus 1 can sort out foreign objects and potatoes with the dirt remaining on the surfaces of the foreign objects and potatoes.

  The potato sorting apparatus 1 freely drops an object such as a potato that has been transported, applies a sorting plate 6 to the falling object, and changes the dropping trajectory of the object. The potato sorting apparatus 1 optically determines whether or not the falling object is a potato, adjusts the position of the sorting plate 6 based on the determination result, and sets the fall trajectory between the case of potato and the case of foreign matter. Make it different. Thereby, the potato sorting apparatus 1 can distribute a potato and a foreign material to a different place.

  As shown in FIG. 1, the potato sorting apparatus 1 includes a transport unit 2, an optical unit 3, a control unit 4, a sorting unit 5, and the like. The optical unit 3 and the control unit 4 constitute a wrinkle determination device according to the present invention.

  The transport unit 2 is, for example, a conveyor having a certain width or more, transports an object in one direction from one end to the other end, and drops the object from the other end. When a plurality of objects are continuously or intermittently supplied to one end of the transport unit 2, the positions between the objects are leveled during transport by the transport unit 2. The transport unit 2 has a width of, for example, several to several tens of potatoes, so that a plurality of objects are dropped almost simultaneously from the other end of the transport unit 2.

  Further, in the transport unit 2, the distance between the objects in the transport direction is adjusted by a partition plate (not shown) provided in the transport unit 2 so that the objects are dropped at regular intervals. The partition plate is installed on a belt or the like of the transport unit 2 and is movable in the transport direction. By storing objects one by one in the transport direction between partition plates adjacent in the transport direction, the object is dropped from the other end portion at a constant interval.

  The sorting unit 5 has substantially the same width as the conveyance unit 2, and a plurality of sorting plates 6 are installed in the width direction. The selection position of the sorting plate 6 of the sorting unit 5 can be changed by an actuator 7 such that the sorting plate 6 is located on the falling track of the object or located outside the track. When the sorting plate 6 is positioned on the falling trajectory of the object, the object is bounced to change the falling trajectory of the object. On the other hand, when the sorting plate 6 is located outside the object dropping trajectory, the object is dropped as it is without changing the object dropping trajectory. Thereby, the selection unit 5 can distribute the object in two directions.

  For example, when it is determined that the object is a potato, the sorting plate 6 is located outside the dropping track, and the object can be dropped as it is. On the other hand, when it is determined that the foreign object is other than the potato, the sorting plate 6 is positioned on the dropping trajectory, the foreign object can be bounced, and the potato can be dropped to the location where it was bounced. At the destination, as shown in FIG. 1, the object may be placed on the conveyor 10 and conveyed to another place, or the object may be stored in a basket or the like.

  The sorting plate 6 of the sorting unit 5 is fixed so that the vicinity of one end thereof is rotatable, and the sorting plate 6 rotates around the axis, so that the plate surface of the sorting plate 6 is positioned on the dropping trajectory of the object or dropped. Or out of orbit.

  The width of each sorting plate 6 of the sorting unit 5 is equal to or less than one potato. The potato sorting apparatus 1 detects the position of an object dropped from the other end of the transport unit 2 and determines whether or not the sorting plate 6 placed at a location corresponding to the detected dropped position is a potato. Move based on. Even when a plurality of objects fall simultaneously in the width direction and the judgment results differ for each object, each sorting plate 6 operates individually, so that the fall trajectory can be made different for each object. The actuator 7 is driven based on the drive signal received from the sorting control unit 12 (see FIG. 4) of the control unit 4, and as a result, the position of each sorting plate 6 is changed.

  The sorting unit 5 may be provided in a plurality of stages (for example, two or more stages) in the vertical direction. When the sorting unit 5 is installed in two stages in the vertical direction, the falling object can be distributed in three directions. For example, in addition to distributing in two directions depending on whether it is potato or not, it can be further distributed in two directions depending on whether the quality is good or not, and distributed in a total of three directions.

  The optical unit 3 includes, for example, a line scan camera 8 and an LED illumination unit 9. The optical unit 3 is installed in a light shielding space where no external light enters. The falling object is irradiated with light by the LED illumination unit 9 while passing through the light shielding space, and the line scan camera 8 receives only the light reflected by the object.

  The line scan camera 8 images a falling object and transmits image data acquired by the imaging to the control unit 4. Two line scan cameras 8 are installed as a pair of two facing each other, and photograph an object from two opposite directions. Thereby, since the whole surface of one object can be image | photographed, the judgment whether it is a potato and the judgment of the quality of a potato can be performed with sufficient precision. Note that the line scan cameras 8 are not necessarily installed in pairs, and only one line scan camera 8 may be installed to capture an object from only one direction. In this case, the line scan camera 8 captures only one side of one object. When there is no difference in the sorting result when only one side is photographed and when both sides are photographed, the present invention can be realized with a simple configuration.

  The LED illumination unit 9 irradiates a falling object with light having a predetermined wavelength. The line scan camera 8 receives light of a predetermined wavelength that is irradiated by the LED illumination unit 9 and reflected by an object.

  The LED illumination units 9 are installed such that two units facing each other form one pair or four units form two pairs, and irradiate the object with light of the same wavelength from two opposite directions. Thereby, the light of the same wavelength can be irradiated to the whole surface of one object. The LED illumination units 9 do not necessarily have to be installed in pairs. When only one line scan camera 8 is installed, the LED illumination unit 9 emits light only from one direction to the object. It may be installed only on one side with respect to the object.

  The pair of LED illumination units 9 may each include a plurality of types of LEDs so as to irradiate light of different wavelengths. The LED illumination unit 9 irradiates light of a plurality of types of wavelengths, so that it is possible to accurately determine whether the potato is a potato or the quality of the potato.

  When a plurality of types of LEDs are provided, the LED illumination unit 9 sequentially irradiates light of different wavelengths in a time division manner as shown in the timing chart of FIG. The lighting of light of different wavelengths by the LED illumination unit 9 and the scanning for each line in the image sensor of the line scan camera 8 are synchronized. Based on the scanning timing for each line of the line scan camera 8 received from the control unit 4, the lighting control unit 23 turns on light of different wavelengths by the LED illumination unit 9. The image sensor of the line scan camera 8 receives light of a predetermined wavelength for each line in response to the lighting of the LED illumination unit 9. As a result, the line scan camera 8 can acquire image data of an object reflecting light of different wavelengths for each of a plurality of types of wavelengths.

  For example, when the LED illumination unit 9 includes four types of LEDs having different wavelengths to be irradiated, image data of an object reflecting light of four types of wavelengths can be acquired. That is, image data for each of the four types of wavelengths is obtained.

  As shown in FIG. 3, in the imaging device 20 having a plurality of lines, from above, a first line that receives light of the first wavelength, a second line that receives light of the second wavelength, and a third line. Are assigned in advance as a third line for receiving light of a wavelength of... Then, for example, luminance information based on the light received by the first line is acquired while the light of the first wavelength is irradiated, and then, while the light of the second wavelength is irradiated, Luminance information based on the light received by the second line is acquired. Similarly, luminance information is acquired for each light of each wavelength.

  When the first to fourth lines are one stage, in the example illustrated in FIG. 3, the imaging element 20 has a total of ten stages and 40 lines. Luminance information is simultaneously acquired for the first line of the first to tenth stages, and then the luminance information is acquired for the second line of the first to tenth stages simultaneously. Thereafter, the luminance information of the third line from the first stage to the tenth stage and the fourth line from the first stage to the tenth stage are acquired.

  In addition, since the object is moved by the shooting timing interval (line clock) between each line, the height of each line in the image sensor 20 is determined in consideration of the falling speed of the object and the line clock. Has been. When the object moves Amm in the in-plane direction of the image sensor 20 at intervals of the photographing timing between the lines, the height of one line is determined to be Amm. Thereby, although each line which receives the light of a different wavelength differs in the position of a height direction, it can image | photograph the same position of an object.

  The predetermined wavelength of light emitted from the LED illumination unit 9 is, for example, any wavelength in the range of 450 nm to 1000 nm. As the wavelength of the light emitted by the LED illumination unit 9, for example, a wavelength at which a difference in reflection characteristics between a potato and a foreign substance appears greatly is selected. In general, there is light having a wavelength at which a potato or an assumed foreign object is likely to be reflected or light having a wavelength that is easily absorbed, so that the wavelength of the light irradiated by the LED illumination unit 9 can be selected by specifying these in advance. .

  In addition, the LED illumination part 9 is good also as a structure which comprises only one type of LED, and irradiates the light of a single wavelength, and when irradiating the light of a multiple types of wavelength, 2 types, 3 types, or 5 types or more You may irradiate the light of the wavelength. Also in these cases, the image pickup device of the line scan camera 8 receives light of a predetermined wavelength for each line corresponding to the lighting of the LED illumination unit 9.

  As shown in FIG. 4, the control unit 4 includes an image processing unit 11 and a selection control unit 12.

  As shown in FIG. 4, the image processing unit 11 includes a position / size specifying unit 13, a texture feature amount extraction unit 14, a luminance information acquisition unit 15, a storage unit 16, a first comparison unit 17, 2 comparison part 18 and judgment part 19 etc. are provided.

  The position / size specifying unit 13 specifies the position and size of the falling object based on the image data acquired by the line scan camera 8. For example, the position / size specifying unit 13 specifies the position and size of the object by coordinates, and sends the specified result to the selection control unit 12.

  The texture feature quantity extraction unit 14 extracts a feature quantity related to the texture of the object based on the image data of the object captured and acquired by the line scan camera 8. Thereby, it is possible to determine whether or not the acquired image data is potato based on the texture of the object, that is, the pattern on the object surface. Since the potato epidermis has unique spots, bud depressions, and the like, and has characteristics not seen in other substances, potatoes and foreign substances are identified using the characteristics.

  The extraction of the feature quantity related to the texture is performed using, for example, a rotation invariant LBP having rotation invariance improved from the Local Binary Pattern (LBP). In the LBP, first, the central pixel value and the peripheral pixel value are compared, and the magnitude is recorded as 0 or 1. When 3 pixels × 3 pixels are used as a basic unit, eight comparison results are obtained by comparing the central pixel value and the peripheral pixel value, and the eight comparison results are arranged to obtain an 8-digit binary number. Then, when the 8-digit binary number is converted into a decimal number, a decimal number from 0 to 255 is obtained. When this is applied to all the pixels of the input image of the acquired object, the image acquired by LBP becomes an image of 256 gradations regardless of the number of gradations of the input image. As a result, the input image is normalized to a 256 gray scale image.

  By using the LBP, the same feature amount or the approximate feature amount can be acquired when the texture is the same or approximate regardless of the magnitude of the luminance value in the image acquired by the line scan camera 8. In LBP, the main calculation is the comparison of numerical values between pixels and conversion from binary numbers to decimal numbers using 3 pixels × 3 pixels as a basic unit, so that high-speed processing can be realized.

  Note that a feature generally used in image processing can be applied to the extraction of the feature amount related to the texture, and a method other than LBP may be used.

  When a feature amount is obtained based on image data from which texture-related features are extracted (in the case of LBP, image data of 256 gray scales), the feature amount is stored in advance in the storage unit 16 by the first comparison unit 17. It is compared with the recorded basic features. The basic feature amount is, for example, data acquired by image processing pattern learning. In the case of the present embodiment, basic feature amounts are acquired in advance based on image data relating to a plurality of potatoes. For example, in image processing pattern learning, a feature value is extracted from a large number of potato images acquired in advance using rotation invariant LBP, and potato data with similar features is extracted based on the extracted feature value. Classify and calculate the average vector of each class. Note that the potatoes in the image data of a plurality of potatoes acquired in advance are in the same state as the potatoes that are actually selected, for example, the state where soil is attached to the surface.

  The first comparison unit 17 includes a feature amount acquired by the texture feature amount extraction unit 14 based on the image data of the object acquired by the line scan camera 8 at the time of selection, and a basic feature amount recorded in the storage unit 16 in advance. Compare This comparison is, for example, calculation of a distance between a feature vector composed of feature amounts acquired at the time of selection and an average vector (256 dimensions) of each class composed of pre-recorded basic feature amounts. . The distance between the acquired feature vector and the average vector is obtained by a distance based on the Euclidean distance. It can be determined that the shorter distance is data that approximates the basic feature amount, that is, data that approximates potato. On the other hand, when the distance exceeds a predetermined threshold, it can be determined that the object is a foreign object.

  Further, the first comparison unit 17 compares a predetermined threshold value with the calculated distance. The first comparison unit 17 sends the result of whether or not the calculated distance is equal to or less than a predetermined threshold value to the determination unit 19.

  In the present embodiment, a plurality of classes are generated in advance based on potato image data acquired in advance, and an average vector of each class is obtained. And the feature vector comprised from the feature-value acquired at the time of a selection is compared with each of those average vectors. Unlike this embodiment, when only one average vector is obtained without generating a plurality of classes, since the potatoes themselves vary, the feature vector acquired at the time of selection is compared with this one average vector. In this case, it is difficult to identify a foreign object having a feature amount close to that of a potato. On the other hand, as in this embodiment, an average vector is obtained for each of a plurality of classes, and a feature vector acquired at the time of selection is compared with a plurality of average vectors, so that a foreign object having a feature amount close to potato can be obtained. In contrast, it is possible to identify with high accuracy.

  The luminance information acquisition unit 15 acquires luminance data of the image data of the object based on the image data of the object captured and acquired by the line scan camera 8. Then, the luminance information acquisition unit 15 calculates and acquires the average value and dispersion value (luminance information) of the luminance of the object image data based on the luminance data. Note that the luminance frequency of the image data can be represented by a histogram in which the vertical axis indicates the frequency and the horizontal axis indicates the luminance value.

  Since an image is obtained for each of a plurality of types of wavelengths irradiated by the LED illumination unit 9, the luminance information acquisition unit 15 can acquire luminance data (spectral reflection intensity) for each wavelength. Then, the average value and standard deviation of the brightness of the image data are calculated for each wavelength. For example, when the LED illumination unit 9 includes four types of LEDs having different wavelengths to be irradiated, images for each of the four types of wavelengths are obtained, and an average value and a standard deviation based on the average value and the dispersion value are calculated for each. And get. That is, in this case, 8-element data is obtained. The 8-element data may be an average value and a dispersion value instead of the average value and standard deviation of the luminance of the object surface in each wavelength image. Then, a feature vector consisting of 8 elements is constructed.

  Thereby, it is possible to determine whether or not the acquired image data is a potato, based on the luminance information of the image data captured by receiving light of a specific wavelength reflected by the object.

  When the average value and the standard deviation are calculated based on the image data for each wavelength, the average value and the standard deviation are calculated by the second comparison unit 18 as the basic average for each wavelength recorded in the storage unit 16 in advance. Compared to value and basic standard deviation. The basic average value or the basic standard deviation is, for example, data acquired by image processing pattern learning. In the case of this embodiment, a basic average value or a basic standard deviation is acquired in advance based on luminance data of image data relating to a plurality of potatoes. For example, in image processing pattern learning, average values and standard deviations are calculated for luminance data of a large number of potato images acquired in advance, and potatoes having similar characteristics are calculated based on the calculated average values and standard deviations. Are classified into classes, and the average vector of each class is calculated. Note that the potatoes in the image data of a plurality of potatoes acquired in advance are in the same state as the potatoes that are actually selected, for example, the state where soil is attached to the surface.

  The second comparison unit 18 compares the average value calculated based on the image data for each wavelength at the time of selection with the basic average value recorded in advance in the storage unit 16, and based on the image data for each wavelength. The calculated standard deviation is compared with the basic standard deviation recorded in advance in the storage unit 16. This comparison is based on the calculation of the distance between the feature vector composed of the average value and standard deviation acquired at the time of selection and the average vector of each class composed of the basic average value and basic standard deviation recorded in advance. is there. The distance between the acquired feature vector and the average vector is obtained by the distance based on the Mahalanobis distance. It can be determined that the shorter distance is the data that approximates the basic average value and the basic standard deviation, that is, the data that approximates potato. On the other hand, when the distance exceeds a predetermined threshold, it can be determined that the object is a foreign object.

  Further, the second comparison unit 18 compares the predetermined threshold value with the calculated distance. The second comparison unit 18 sends a result of whether or not the calculated distance is equal to or less than a predetermined threshold value to the determination unit 19.

  The determination unit 19 determines whether or not the photographed object is a potato based on the comparison results in the first comparison unit 17 and the second comparison unit 18.

  The determination unit 19 obtains a result that the calculated distance is smaller than a predetermined threshold value in the first comparison unit 17, and that the calculated distance is equal to or less than the predetermined threshold value in the second comparison unit 18. When the result is obtained, it is determined that the image data of the object acquired by the line scan camera 8 is image data related to potato. The determination unit 19 sends a determination result as to whether the image data is image data related to potatoes to the selection control unit 12.

  The sorting control unit 12 determines the position and number of the sorting plates 6 to be driven based on the position and size of the falling object specified by the position / size specifying unit 13. Further, the sorting control unit 12 moves the sorting plate 6 on the falling track of the object or moves it outside the dropping track based on the determination result of whether or not it is a potato by the determining unit 19. That is, the sorting control unit 12 determines the position and the number of the sorting plates 6 to be driven, and moves the sorting plates 6 on the object dropping trajectory or out of the dropping trajectory.

Next, the operation of the potato sorting apparatus 1 according to this embodiment will be described.
FIG. 5 is a flowchart showing the operation of the potato sorting apparatus 1 according to this embodiment. First, an object including a potato and a foreign substance other than a potato is conveyed by the conveying unit 2, and the object is dropped from the end (step S1). Note that the distance between the objects in the conveyance direction is adjusted by the partition plate provided in the conveyance unit 2 before the fall, and the fall of the objects from the end portion is performed at a constant interval.

  While falling, the optical unit 3 emits light having a predetermined wavelength, and an object is imaged when the light is emitted (step S2). Then, based on the imaging result, it is determined whether the object is a potato or a foreign object (step S3).

  Thereafter, the position of the sorting plate 6 of the sorting unit 5 is adjusted according to the determination result of whether or not the object is a potato and the fall timing, and the fall trajectory is made different between the case of potato and the case of foreign matter (step S5). ). At this time, the position and size of the falling object are specified (step S4), and the position and number of the sorting plates 6 to be driven are also determined. The falling position changes when the falling object hits the sorting plate 6. As a result, the potato sorting apparatus 1 can distribute potatoes and foreign substances to different places.

  In the present embodiment, the determination as to whether or not the object is a potato is made based on the feature information regarding the texture and the luminance information based on the wavelength of the reflected light. FIG. 6 is a flowchart showing the operation of the control unit 4 of the potato sorting apparatus 1 according to this embodiment. First, image data is acquired as a result of the line scan camera 8 imaging an object (step S11).

  Regarding the feature amount related to the texture, first, the feature amount related to the texture of the object is extracted based on the image data of the object captured and acquired by the line scan camera 8 (step S12). This extraction is performed using, for example, a rotation invariant Local Binary Pattern (LBP). When the feature amount is extracted, the feature amount is compared with the basic feature amount previously recorded in the storage unit 16 by the first comparison unit 17 (step S13). In this comparison, first, a distance (for example, Euclidean distance) between a feature vector composed of feature amounts acquired at the time of selection and an average vector (256 dimensions) of each class composed of pre-recorded basic feature amounts. ) Is then calculated, and then it is compared whether or not the calculated distance is less than or equal to a predetermined threshold.

  As for the luminance information based on the wavelength of the reflected light, first, the luminance data of the image data of the object is acquired based on the image data of the object captured and acquired by the line scan camera 8. Based on the luminance data, the average value and the variance value (luminance information) of the image data are calculated and acquired (step S14). When an image is obtained for each of a plurality of types of wavelengths, luminance data for each wavelength is acquired, and an average value and a standard deviation of the luminance of the image data are calculated for each wavelength.

  The calculated average value and standard deviation are compared with the basic average value and basic standard deviation recorded in advance in the storage unit 16 by the second comparison unit 18 (step S15). When an image is obtained for each of a plurality of types of wavelengths, it is compared with a basic average value and a basic standard deviation for each wavelength. In this comparison, first, a distance between a feature vector composed of an average value and a standard deviation acquired at the time of selection and an average vector of each class composed of a basic average value and a basic standard deviation recorded in advance ( For example, Mahalanobis distance) is calculated, and then it is compared whether or not the calculated distance is equal to or less than a predetermined threshold.

  And based on the comparison result in the 1st comparison part 17 and the 2nd comparison part 18, it is judged whether the image | photographed object is a potato. That is, it is determined whether or not the calculated feature amount, average value of brightness, and standard deviation regarding the texture satisfy the condition (step S16).

  With respect to the feature amount related to the texture, a result that the distance calculated by the first comparison unit 17 is equal to or less than a predetermined threshold is obtained, and luminance information based on the wavelength of the reflected light is calculated by the second comparison unit 18. When the result that the distance is equal to or smaller than the predetermined threshold is obtained, it is determined that the image data of the object acquired by the line scan camera 8 is image data related to potato (step S17).

  On the other hand, when the distance calculated by the first comparison unit 17 exceeds a predetermined threshold, or when the distance calculated by the second comparison unit 18 exceeds a predetermined threshold, the line scan camera 8 It is determined that the acquired image data of the object is not image data relating to potato but a foreign object (step S18).

  As described above, according to the present embodiment, whether or not an object is a potato is determined based on the texture-related feature amount and the luminance information based on the wavelength of the reflected light. When the determination is made based only on the feature amount related to the texture, there is a high possibility that an object whose pattern on the surface of the object approximates to potato (for example, a stone with soil attached to the surface) is a potato. In addition, when the determination is made based only on the luminance information based on the wavelength of the reflected light, there is a high possibility that the object that approximates the light of a specific wavelength reflected by the object is a potato. As described above, in some cases, potatoes and foreign substances may not be distinguished, but in the present embodiment, it is determined that a potato is a potato when both conditions are satisfied, thereby improving the sorting accuracy of potatoes and foreign substances. Can be made.

[Second Embodiment]
Hereinafter, a potato sorting apparatus 1 according to a second embodiment of the present invention will be described. In the following, detailed description of the same components and effects as those in the first embodiment will be omitted.
In the first embodiment described above, a case has been described in which whether or not an object is a potato is determined based on a texture-related feature amount and luminance information based on the wavelength of reflected light. However, in addition to these two determination conditions, Then, it may be determined whether or not the object is a potato in consideration of another determination condition.
For example, in this embodiment, it is determined whether the object is a potato by further using the curvature of the boundary of the object in the image data.

  As illustrated in FIG. 7, the image processing unit 11 of the control unit 4 according to the present embodiment further includes a curvature information acquisition unit 21 and a third comparison unit 22 in addition to the first embodiment described above.

  The curvature information acquisition unit 21 extracts the boundary line between the object and the background, that is, the contour shape of the object, based on the image data of the object captured and acquired by the line scan camera 8, and for each predetermined length on the contour shape. Calculate the curvature of. The contour of the object has a closed curve, the curved portion convex outward is a positive (plus) value, the concave curved portion recessed inward is a negative (minus) value, and the portion close to a straight line is The value is close to 0 (zero). The outline of a potato generally occupies a large proportion of a portion that is convex outward, and the curvature for each predetermined length often takes a positive value. In addition, the curvature of the potato has a gentle change in curvature along the contour. On the other hand, the contours of foreign substances such as stones and plant stems are often composed of straight lines, and the curvature largely varies positively and negatively along the contours. It should be noted that the foreign matter identification based on the contour shape is not easily affected by the soil adhering to the surface of the potato.

  The curvature information acquisition unit 21 normalizes the calculated curvature with reference to the maximum value of the calculated curvature, and accumulates (aggregates) the frequency for each normalized curvature. Then, the curvature information acquisition unit 21 sends frequency information (curvature information) regarding the curvature to the third comparison unit 22.

  When the frequency information regarding the curvature is obtained, the third comparison unit 22 calculates the ratio of the frequency of the curvature below a predetermined threshold Tc (for example, 0.07) to the whole. In addition, the third comparison unit compares the ratio of the curvature frequency equal to or lower than the predetermined threshold Tc with respect to the total to a predetermined threshold Tr% (for example, 20%). If the ratio of the curvature frequency equal to or lower than the predetermined threshold value Tc is equal to or lower than the predetermined threshold value Tr%, there is a high possibility of potato, and the ratio of the curvature frequency equal to or lower than the predetermined threshold value Tc is equal to the predetermined frequency. If the threshold value Tr% is exceeded, there is a high possibility of being a foreign object.

  Unlike potatoes, foreign substances such as stones and plant stems, potatoes have a low curvature ratio of around 0 (zero). Therefore, as described above, by setting the predetermined threshold Tc to 0.07 and setting the predetermined threshold Tr to 20%, not only a normal potato but also two potatoes are connected to form a saddle type Even if the center has a concave outline, it is possible to determine that it is a potato without being determined as a foreign object.

  The determination unit 19 determines whether the photographed object is a potato based on the comparison results in the first comparison unit 17, the second comparison unit 18, and the third comparison unit 22.

  The determination unit 19 obtains a result that the ratio of the curvature frequency equal to or less than the predetermined threshold Tc is less than or equal to the predetermined threshold Tr% in the third comparison unit 22, and is calculated in the first comparison unit 17. When the result that the calculated distance is smaller than the predetermined threshold is obtained and the second comparison unit 18 obtains the result that the calculated distance is equal to or smaller than the predetermined threshold, the line scan camera 8 acquires the distance. It is determined that the image data of the object is image data related to the potato. The determination unit 19 sends a determination result as to whether the image data is image data related to potatoes to the selection control unit 12.

Next, the operation of the potato sorting apparatus 1 according to this embodiment will be described. FIG. 8 is a flowchart showing the operation of the control unit 4 of the potato sorting apparatus 1 according to this embodiment.
The determination as to whether or not the potato is a potato is performed in consideration of the curvature of the object boundary in the image data in addition to the feature amount related to the texture and the luminance information based on the wavelength of the reflected light.

  Regarding the curvature of the boundary of the object, first, based on the image data of the object captured and acquired by the line scan camera 8, the boundary line between the object and the background, that is, the contour of the object is extracted, and a predetermined length on the contour is obtained. The curvature for each is calculated. Then, based on the calculated curvature, the frequency for each curvature is accumulated, and curvature frequency information is acquired (step S19).

  And if the frequency information regarding a curvature is obtained, the ratio which the frequency of the curvature below a predetermined threshold value Tc occupies with respect to the whole by the 3rd comparison part 22 will be calculated. Then, the third comparison unit 22 compares whether or not the ratio of the curvature frequency equal to or lower than the predetermined threshold Tc to the entire threshold is equal to or lower than the predetermined threshold Tr% (step S20).

  And based on the comparison result in the 1st comparison part 17, the 2nd comparison part 18, and the 3rd comparison part 22, it is judged whether the image | photographed object is a potato. That is, it is determined whether or not the calculated feature amount, average value and standard deviation of the brightness, and curvature frequency satisfy the condition (step S21).

  Regarding the curvature of the boundary of the object, if the third comparison unit 22 obtains a result that the ratio of the curvature frequency equal to or lower than the predetermined threshold Tc to the entire threshold is equal to or lower than the predetermined threshold Tr%, the texture-related feature With respect to the amount, a result that the distance calculated by the first comparison unit 17 is equal to or less than a predetermined threshold value is obtained, and the luminance information based on the wavelength of the reflected light is the predetermined distance calculated by the second comparison unit 18. If the result is less than or equal to the threshold value, it is determined that the image data of the object acquired by the line scan camera 8 is image data relating to potato (step S22).

  On the other hand, when the third comparison unit 22 obtains a result that the ratio of the curvature frequency equal to or lower than the predetermined threshold value Tc to the total exceeds the predetermined threshold value Tr%, the object acquired by the line scan camera 8 is obtained. This image data is not image data related to potatoes, but is determined to be a foreign object. As in the first embodiment, the distance calculated by the first comparison unit 17 exceeds a predetermined threshold value, or the distance calculated by the second comparison unit 18 exceeds a predetermined threshold value. Also in this case, it is determined that the object is a foreign object (step S23).

  As described above, according to the present embodiment, whether or not an object is a potato is determined based on three types of conditions: a texture-related feature amount, luminance information based on the wavelength of reflected light, and the curvature of an object boundary in image data. Is done. Compared to the first embodiment in which it is determined whether or not an object is a potato based on two types of conditions, that is, a feature amount related to texture and luminance information based on the wavelength of reflected light, in this embodiment, in the image data By considering the curvature of the boundary of the object, it is possible to further improve the accuracy of sorting potatoes and foreign objects.

  When the determination is made based only on the curvature of the object boundary in the image data, it is highly likely that an object (for example, a stone) whose curvature of the object boundary approximates to a potato is a potato. As described above, there is a case where the potato cannot be distinguished from the foreign matter only by the curvature of the boundary of the object in the image data. And the foreign matter sorting accuracy can be improved.

  In the first and second embodiments described above, the case where only one threshold is set in the first comparison unit 17 and the second comparison unit 18 has been described, but the present invention is not limited to this example. For example, the first comparison unit 17 and the second comparison unit 18 may set a plurality of threshold values, for example, two different values. At the time of selection, when the distance calculated by the first comparison unit 17 or the second comparison unit 18 is equal to or less than the lower threshold in either the first comparison unit 17 or the second comparison unit 18, the first comparison You may make it judge that it is a potato if it is below the other higher threshold of the part 17 and the 2nd comparison part 18. FIG. Thereby, when there is a correlation in the two determinations based on the texture and spectral reflection intensity regarding the potato, it becomes possible to identify with higher accuracy.

  Moreover, in embodiment mentioned above, although the potato sorting apparatus 1 was demonstrated as an example of the bag sorting apparatus which concerns on this invention, this invention is not limited to this example. The cocoon sorting apparatus according to the present invention can be applied to potatoes such as sweet potatoes other than potatoes.

  In the above-described embodiment, in the transport unit 2, the object is transported in one direction from one end to the other end, the object is dropped from the other end, and the sorting plate 6 of the sorting unit 5 is positioned on the dropping trajectory of the object. However, the present invention is not limited to this example. However, the present invention is not limited to this example. . For example, the technology used as a sorter such as fruits and vegetables can be applied to the apparatus constituting the transport unit and the sorting unit. For example, the sorter has a configuration in which an object is conveyed by a horizontal conveyor or the like, and the object is moved out of the conveyance route of the conveyor based on the determination result. For example, a device for moving an object out of the conveyor has a configuration in which a part of the conveyor moves downward to move the object downward, or a blade-like member picks up the object and conveys the object. There are some that have a configuration that eliminates to the side.

1: potato sorting device 2: transport unit 3: optical unit 4: control unit 5: sorting unit 6: sorting plate 7: actuator 8: line scan camera 9: LED illumination unit 10: conveyor 11: image processing unit 12: sorting control Unit 13: position / size specifying unit 14: texture feature amount extraction unit 15: luminance information acquisition unit 16: storage unit 17: first comparison unit 18: second comparison unit 19: determination unit 20: image sensor 21: curvature information Acquisition unit 22: third comparison unit 23: lighting control unit

Claims (7)

An imaging unit for imaging an object;
An irradiation unit for irradiating the object with light of a predetermined wavelength;
A feature extraction unit that extracts a feature amount related to the texture of the target object based on image data of the target object captured and acquired by the imaging unit;
A first storage unit in which basic feature values relating to the image data of the cocoon are recorded in advance;
A first comparison unit that compares the extracted feature quantity with the basic feature quantity;
A luminance information acquisition unit that receives light reflected by the object and acquires luminance information of the wavelength based on the received light;
A second storage unit in which basic luminance information of the wavelength based on the light reflected by the ridge is recorded in advance;
A second comparison unit for comparing the acquired luminance information and the basic luminance information of the object;
A determination unit that determines whether or not the object is a bag based on a comparison result in the first comparison unit and the second comparison unit;
A wrinkle determination device comprising: The irradiation unit sequentially irradiates the object with a plurality of lights having different wavelengths,
The luminance information acquisition unit acquires the luminance information for each wavelength,
The second storage unit records in advance basic luminance information for each wavelength,
The wrinkle determination device according to claim 1, wherein the second comparison unit compares the acquired luminance information and the basic luminance information of the object for each wavelength.   The wrinkle determination device according to claim 1, wherein the feature extraction unit extracts a feature amount related to a texture of the object using a Local Binary Pattern.   The wrinkle determination apparatus according to claim 1, wherein the luminance information acquisition unit calculates an average value or a variance value of luminance in the image data of the object as the luminance information. A curvature information acquisition unit that extracts the contour shape of the object based on the image data of the object acquired and acquired by the imaging unit, and acquires curvature information;
A third storage unit in which basic curvature information relating to the image data of the cocoon is recorded in advance;
A third comparison unit for comparing the acquired curvature information and the basic curvature information;
Further comprising
5. The device according to claim 1, wherein the determination unit determines whether or not the object is a bag based on a comparison result in the first comparison unit, the second comparison unit, and the third comparison unit. The wrinkle determination device according to item 1.   The wrinkle determination device according to claim 5, wherein the curvature information acquisition unit calculates a curvature for each predetermined length on the contour shape and calculates a frequency for each calculated curvature as the curvature information. A wrinkle determination device according to any one of claims 1 to 6,
Based on the result determined by the wrinkle determination device, a selection unit for selecting the object;
A basket sorting apparatus comprising:

Images