JPWO2016133175A1 - Fruit sorting system, fruit sorting device, and fruit sorting method - Google Patents

Abstract

When observing the seeds in the pod, it is possible to inspect the seed accommodation state (especially the number and size) of the seed in the pod by photographing the reflected light, eliminating the need for photographing with transmitted light. There are provided a fruit sorting system and a fruit sorting apparatus which inspects or sorts fruit according to the value of the vegetation index calculated from the above. This fruit sorting system examines or sorts the fruit by analyzing an image obtained by photographing a fruit with a bowl, and obtains an image obtained by the photographing means and an image obtaining means. Image analysis means for analyzing the image, and the image analysis means, based on the captured image, the length of each fruit, a size test for measuring the width of each fruit, and the length of the measured fruit Based on the above, a grain number inspection for calculating the number of seeds contained in the straw is executed, or alternatively, a fruit sorting system for inspecting or sorting the fruit according to the value of the vegetation index.

Description

The present invention relates to a fruit sorting system and a fruit sorting apparatus for sorting fruit and fruits such as edamame, and in particular, a fruit sorting system for sorting fruit based on appearance and / or vegetation index. And a fruit sorting apparatus.

  Fruits are fruits found in many leguminous plants, and include green beans, peas, nemkins, soybeans, red oaks, lupines, clovers, and crows. These fruits are often harvested in the state of strawberry, and in the soybeans harvested from immature soybeans, they are often provided on the table in the state of strawberry.

  Even when such a fruit is distributed in the market, it is necessary to select it in order to manage its quality and grade. However, the number of beans contained in each cocoon is different, and the size of the beans varies. Therefore, in the past, in order to automate the selection of fruit with a strawberry, several techniques have been proposed that use a camera to select fruit with a strawberry, such as green soybeans.

  For example, Patent Document 1 (Japanese Patent Application Laid-Open No. 2008-62116) includes transmitted light imaging means that uses transmitted light and picks up the fruits using the transmitted light that has passed through the fruits in order to select the fruits based on the grade of the fruits. A fruit sorting system for discriminating and sorting the grades of fruit based on the fruit image picked up by the transmitted light imaging means is proposed. Further, in Patent Document 2 (Japanese Patent Application Laid-Open No. 2005-279524), in order to achieve labor-saving and high efficiency of fine sorting of edamame koji, the two surfaces are joined in a V-shape in plan view at an angle of 120 °. We have proposed a method for fine sorting of edamame potatoes, which obtains an entire image of edamame candy by dropping the edamame candy in the vertical direction on the center plane of the connecting portion of the mirror.

  Further, when selecting fruits and vegetables, it is desirable to further select them according to the growth conditions such as the activity and water content of the fruits. In the past, a technique has also been proposed in which a spectrum analysis is performed on an image obtained by photographing a crop to determine the growth status and the like.

For example, Patent Document 3 (Japanese Patent Laid-Open No. 2006-250827) proposes a crop growth state analysis method for diagnosing the taste by estimating the protein content of rice from a spectrum image using remote sensing. Moreover, in patent document 4 (Unexamined-Japanese-Patent No. 2013-231645), vegetation is used using the optical data contained in the image information of a tea leaf in order to be able to judge the appropriateness of the picking of the shoot of a tea tree in a short time and with high precision easily. A tea leaf plucking aptitude evaluation method for calculating an index and evaluating tea leaf plucking suitability using this index has been proposed.

JP 2008-62116 A JP 2005-279524 A JP 2006-250827 A JP2013-231645A

  As described above, various techniques have been proposed in the past for using a camera to select a fruit with a wrinkle. However, in order to inspect the bean accommodation state in the basket, the transmitted light of the fruit must be photographed. In order to photograph the transmitted light in the fruit, a configuration for photographing the transmitted light, such as transporting with a mesh belt or photographing the falling fruit, is required.

  Therefore, the present invention eliminates the need for such transmitted light imaging, and captures the reflected light of the light irradiated to the fruit with a cocoon, thereby allowing the seed accommodation state (especially the number and size of grains) in the cocoon. It is a first object to provide a fruit sorting system and a fruit sorting apparatus that can inspect the fruit.

  In addition, the storage state of the quality control image can be obtained by using the image obtained by reflecting the reflected light even when inspecting the size and shape of the fruit or the tearing or chipping of the cocoon along with the state of the beans contained in the cocoon. It is a second object to provide a fruit sorting system and a fruit sorting apparatus that can reduce and improve the inspection processing speed.

  Conventionally, cameras used for sorting and inspecting fruits have taken visible light, and have only inspected the appearance of fruits only with a visible light image. Since inspection of such a surface image obtained by photographing visible light is determined solely by changes in the spectrum and color of the planar image, there is still room for improvement in accuracy. In particular, when inspecting flaws and the like in fruits, it is difficult to accurately inspect and sort because a difference between the color of the flaw part and the surrounding color is not clearly shown in a simple visible light image.

  Accordingly, the present invention provides a fruit sorting system and a fruit sorting system that can inspect and sort the appearance more accurately without sticking to a visible light image, and in particular, can inspect and sort scratches and the like in the same manner as visually. It is a third object to provide an apparatus and a fruit sorting method.

  In addition, the degree of growth of the fruits and vegetables, which are plants, varies depending on the growth conditions and harvest time. In addition, the dryness (water content) of the fruits varies depending on the passage of time and storage conditions after harvesting. However, when shipping fruit as a product, it is necessary to meet certain quality standards. Therefore, it is necessary to inspect and select the activity of the fruit to ensure thorough quality control.

Therefore, the present invention uses a vegetation index to further test or select at least one of fruit activity, moisture content, nitrogen content, fiber content, and protein content, and fruit sorting system. It is a fourth problem to provide an apparatus and a fruit sorting method.

  In order to solve the above problems, the present invention can determine the number of seeds in the pod by analyzing the acquired captured image, and can also determine the presence or absence of chipping or beard based on the captured image. Provided are a fruit sorting system and a fruit sorting apparatus.

  That is, in the present invention, by analyzing an image obtained by photographing a fruit with a jar, a fruit sorting system for inspecting or sorting the fruit, an image obtaining unit for obtaining an image photographed by the photographing unit, Image analysis means for analyzing the image acquired by the image acquisition means, the image analysis means is a size inspection for measuring the length of each fruit and the width of each fruit based on the photographed image, and measurement There is provided a fruit sorting system for performing a grain number inspection for calculating the number of seeds contained in a straw based on the length of the finished fruit.

In order to solve at least one of the above-mentioned problems, the present invention is a fruit sorting system for examining or sorting the fruits by analyzing an image obtained by photographing the fruits with a bowl. An image acquisition unit that acquires a reflected image of the captured fruit and an image analysis unit that analyzes the image acquired by the image acquisition unit, wherein the image acquisition unit includes a red image composed of reflected light in a red wavelength region; Extracting or obtaining a near-infrared image composed of reflected light in the near-infrared wavelength region, the image analysis means calculates a vegetation index according to the following formula 1, and examines or sorts the fruit by the value of the vegetation index A fruit sorting system is provided.
[Where:
R: reflectance of red wavelength region in red image IR: reflectance of near infrared wavelength region in near infrared image]

  Such a fruit sorting system can be formed using a computer. Specifically, it can be embodied as a computer in which processing contents and processing procedures are controlled by computer software. The image analysis means can be constituted by a CPU or a memory that performs processing by executing computer software.

<Grain count inspection>
In the fruit sorting system according to the present invention, the image analysis means executes “size inspection” for obtaining information on the length and width of each fruit from the acquired photographed image. The information on the length and width can be acquired as the actual length and width of the inspection object in addition to the number of dots (or the number of pixels) based on the pixels in the acquired captured image. When calculating the actual length and width of the inspection object from the photographed image, the length specified on the photographed image is multiplied by a certain enlargement or scale constant.

  Based on the measured length of the fruit, the number of seeds contained in the cocoon is calculated. The number of seeds can be set based on information that defines the number of seeds in association with the length of the fruit. For example, by preparing a length / number table in which the length of fruit and the number of seeds are associated in advance, the number of seeds corresponding to the length is retrieved by searching the length of the fruit read by image analysis as a search value. Can be calculated and extracted. The program can also be designed to output a certain value (number of seeds) according to the length of the obtained fruit. That is, in the present invention, the approximate number of seeds contained in the pod can be calculated without measuring the transmitted light by measuring the length of the pod and using this value.

  In the grain number inspection, the number of seeds in the pod can be calculated, and the width of the pod that changes depending on the presence of the seed can be calculated. That is, in the grain number inspection in the image analysis means, the individual fruit images are divided in the length direction by the number of the grain number. In addition, a position where the width is maximum in each of the divided areas is set as a convex portion, and a position where the width is narrower than the concave portion and minimum between the convex portions is set as a concave portion. And it can comprise so that the width | variety of the said convex part and a recessed part may be calculated. Thus, by obtaining the length of the ridge and the width of the convex portion and the concave portion, and comparing this with a preset allowable value, the size of the fruit to be inspected can be inspected in more detail. .

  Also, when setting the convex part and the concave part and calculating the width, it is possible to specify the convex part and the concave part and calculate the width without dividing the image of each result in the length direction. it can. That is, in the fruit sorting system, the image analyzing means is accommodated in the basket based on the size inspection for measuring the length of each fruit and the width of each fruit from the photographed image, and the measured fruit length. The number of seeds being set and the number of convex parts in the width direction are set based on the length of the fruit and / or the number of seeds, and the respective widths of the set convex parts and the convex parts It is possible to provide a fruit sorting system that executes a grain number inspection for obtaining the width of the concave portion that is the minimum width between the two.

  By performing the size inspection, the length of the fruit and / or the number of seeds is calculated, and then the number of convex portions in the width direction according to the length of the fruit and / or the number of seeds is set. . Here, the value set based on the length of the cocoon and / or the number of seeds is the number of convex portions, and does not specify the position of the convex portion. As for the position of the convex part, the width of the entire length of the fruit is scanned, the number of convex parts set in order from the largest value of the width is extracted, and the position of the extracted width is set as the convex part. Each width in the section can be calculated. For example, when the number of seeds in the pod is set to two, the number of convex portions in the width direction is set to two, and two places are set as convex portions in the order of width, and the width Is calculated. In addition, when the number of seeds in the pod is set to 3, the number of convex portions in the width direction is set to three, and three places are set as convex portions in the order of width, and each convex portion Get the width of the part.

  Next, a position where the width of the ridge becomes the minimum value is extracted between the convex portions, the position of the minimum width is set as the concave portion, and the width is acquired. That is, the location where the seeds are present in the cocoon can set a convex portion because the contour is swollen. And since a recessed part exists between seeds, it can confirm that a seed exists in a convex part by confirming presence of a recessed part.

  Moreover, in the said particle | grain number test | inspection, when it calculates as two or more convex parts, it is desirable to confirm the variation | change_quantity of the width between the said convex parts. If the number of grains is specified only by the length of the cocoon, even if it is the same cocoon length, depending on the size of the seeds in it, there may be one or two grains. is there. Therefore, it is desirable to confirm that the change in the width between the convex portions changes so that one trough portion (concave portion) appears.

  In the grain number inspection, it is also desirable to identify the fruit with only one seed contained in the pod. The fruit of one seed can be identified by the total length of the pod. That is, it is possible to assume that a fruit having a length shorter than the reference length has only one seed. By predicting the number of grains according to the length of the fruit, the processing speed in the grain number inspection can be improved. And about the fruit which has the length more than a reference value, it is judged that there are two or more seeds in the straw. However, even if it is determined that there are two or more grains, if the seed grains are large, only one grain may actually be accommodated. Therefore, the actual number of grains is calculated by reading the width from the photographed image, extracting and confirming the convex and concave portions. That is, when the length is equal to or greater than the reference value and the number of grains is set to two, the fruit is divided into two regions in the length direction, and the widest convex portion is extracted in each region. As a result, if there are two or more seeds contained in the pod, the convex portions should exist in two places, and there should be concave portions between the convex portions. However, when the seeds in the pod are one grain, even if the convex portions are set at two places, there is no narrower width between the convex portions than the two convex portions. There will be no recess. Therefore, it is possible to determine whether or not there is a single grain by checking whether or not there is a concave portion between the convex portions with respect to the length of fruit that can be one seed in the pod.

  Thus, the process in which the seeds in the pod identify one fruit is particularly effective when sorting green soybeans. This is because in the selection of green soybeans, the number of seeds in the basket also affects the commercial value, so that it is possible to increase the commercial value of the seed by identifying and selecting one fruit.

  Therefore, according to the fruit sorting system according to the present invention, it is not necessary to photograph the transmitted light, and the number of seeds contained in the basket is accurately grasped only by analyzing the photographed image acquired from the photographing means. I can do things.

  When analyzing a photographed image as described above, it is desirable that a plurality of fruits included in the photographed image be extracted for each fruit. Therefore, it is preferable that the image analysis unit performs a labeling process on the captured image acquired from the imaging unit and creates an image for each result. Further, it is desirable to perform angle correction on each image and measure the length in the major axis direction and the width in the minor axis direction for the corrected image subjected to the angle correction. This is because the contour shape of the fruit can be specified by performing the labeling process, and the length and width of the fruit can be measured more accurately by correcting the angle.

  In other words, in order to automate the selection of fruits, it is necessary to convey them by a conveying means such as a belt conveyor, and it is necessary to inspect the conveyed fruits by photographing them. However, it is practically difficult to accurately align the direction of the fruits conveyed by the conveying means. Therefore, by photographing the fruits that are transported at random, and correcting their orientation, a fruit sorting system that can accurately measure the length and width of the fruits while meeting the actual conditions of the inspection is realized. It is.

<Inspection of chipping of wrinkles>
In the fruit sorting system according to the present invention, it is preferable that the image analysis means inspects whether or not the fruit that is the inspection object is missing.
That is, the image analysis unit sets, as the inspection region, an area in which the amount of change in width is a predetermined range for the captured image acquired from the imaging unit, and uses the width of an arbitrary reference position in the inspection area as a reference width. It is desirable to obtain a width at a position moved in a certain direction from the reference position as a comparison width, and to perform a chip inspection in which a value obtained by calculating a difference between the reference width and the comparison width is compared with an allowable value.

  By setting an area where the amount of change in the width is within a predetermined range as an inspection area, it is possible to remove an area where noise is likely to occur (that is, an area that is easily recognized) such as an end in the length direction. Thus, the analysis speed can be improved. Then, within this inspection area, an arbitrarily defined reference position width is acquired as a reference width, a width at a position moved in a certain direction from the reference position is acquired as a comparison width, and the difference between the reference width and the comparison width The amount of change in width can be calculated by calculating. Then, the presence or absence of a chip can be inspected by comparing the change amount of the width with a preset allowable value. That is, since the width of the fruit changes abruptly at the portion where the chipping occurs, it can be determined that the chipping has occurred for the fruit having a large amount of change.

  In addition, the reference position for acquiring the reference width in the inspection region may be an arbitrary position, but for example, any one of the convex portion and the concave portion may be used as the reference value. When the convex portion is used as the reference position, the width at the position moved in a certain direction from the reference position should gradually become narrower until reaching the concave portion. On the other hand, when the concave portion is used as the reference position, the width at the position moved in a certain direction from the reference position should gradually increase until reaching the convex portion. Therefore, it is possible to inspect for the occurrence of chipping faster and more accurately by determining the increase / decrease in the amount of change in the comparison width using the characteristics of the shape of the fruits.

<Inspection of mustaches>
In the fruit sorting system according to the present invention, it is preferable that the image analysis unit further performs a mustache inspection for checking whether or not there is a mustache extending elongated from the wrinkle.
That is, the image analysis unit further performs binarization processing on the captured image acquired from the imaging unit, and determines the number of edge portions whose color changes in the width direction at an arbitrary plurality of points set in the length direction of the eyelid. It is desirable to detect and specify a point where the number of the edge portions is three or more as a beard portion existing location, and further, the length of the beard portion is calculated from the start point and the end point of the beard portion existing location It is desirable to do.

  By performing the binarization process on the photographed image, the outline of the fruit in the photographed image becomes clear, and the background part and the fruit part can be easily distinguished. As a result, it becomes easy to recognize the region of the fruit portion, so that the processing speed can be improved, the image can be recognized accurately, and the inspection accuracy can be increased.

  The inspection object of the mustache part inspection is a streak or the like that protrudes in a streak shape from the edge portion of the eyelid, which is read by image recognition and selected according to its length. Specifically, in the image created for each result, the determination is made based on the number of edges that are contour portions in the width direction. That is, the intersection of the line extending in the width direction and the contour portion over the length direction of the fruit is an edge. If the number of edges is two, it is obtained by the contour of the cocoon and exceeds 2. If it exists, it can be determined that there is a portion protruding from the contour of the ridge (that is, a mustache portion). And by executing this process over the entire length direction of the ridge, the length of the region from the portion where the number of edges exceeds two to the portion where the number of edges is two, It can be calculated as the length of the beard portion.

  As a result, it is possible to inspect the presence of the beard portion simply by reading and analyzing a two-dimensional image obtained by photographing the reflected light. In this regard, if the image is a photograph of transmitted light, the beard portion may not be photographed depending on the intensity of the light. In order to check the presence of the beard portion from the photographed image, reflected light is used. This was achieved because of the shooting.

<Inspection of fruit activity, water content, nitrogen content, fiber content, and protein content>
This test uses the vegetation index. That is, the vegetation index (particularly, the normalized difference vegetation index) is an index for judging the vegetation status and activity of the plant by the reflected light on the plant surface. That is, chlorophyll contained in many plants has a reflectance of less than 20% in the red visible light region of 0.5 to 0.7 μm, whereas the reflectance in the near infrared region of 0.7 to 1.3 μm is 60%. Over. Therefore, the light absorption and reflection in this specific wavelength range are used to calculate the normalized value between −1 and +1 in order to determine the vegetation status and activity of the plant.

  In addition, chlorophyll is unstable to heat and acidic environment, exhibits photodegradability by light, and has particularly remarkable resolution by ultraviolet rays. Therefore, the abundance of this chlorophyll can be estimated by calculating the vegetation index. As a result, it is possible to determine the activity, the elapsed time after harvesting (ie, freshness), and the like.

  Furthermore, there is a correlation between the vegetation index and the nitrogen content, fiber content, and protein content of the plant. For example, there is a correlation between the activity of the plant and the protein content in two wavelength ranges, a red wavelength and a near infrared wavelength. More specifically, there is a high positive correlation between the content of chlorophyll contained in plants and the protein content. Therefore, by calculating the vegetation index and estimating the chlorophyll content, the protein content can be estimated, and the taste of the target fruit can be estimated and selected.

  In addition, there is a correlation between the amount of nitrogen, the amount of fiber, and the vegetation index. That is, by calculating the vegetation index from the optical data of the red light region and the near-infrared light region, and estimating the nitrogen amount and the fiber amount based on the vegetation index, the growth status or nutritional state of the target fruit is determined. It can be estimated and can be sorted accordingly.

  Then, the reflectance in the red wavelength region in the red image and the reflectance in the near infrared wavelength region in the near-infrared image are acquired as the reflection luminance measured by the image sensor in the digital camera or the like, the intensity or digital value of the reflection spectrum. Vegetation parameters such as vegetation activity, vegetation coverage, and leaf area index (LAI) can be determined from the vegetation index calculated using them.

  For the red image composed of the reflected light in the red wavelength region and the near infrared image composed of the reflected light in the near infrared wavelength region, the spectrum of each wavelength region is extracted from the reflected image of the fruit taken by the photographing means. And can be obtained. In addition, a color filter can be installed in the photographing means to photograph and use a reflected image in the wavelength region. Therefore, for example, an ordinary commercial digital camera can be provided with two types of lens filters of visible light blocking and infrared light blocking to capture visible red and near infrared images.

  Further, the image analysis unit performs binarization processing on each of the red image and the near infrared image acquired by the image acquisition unit, and uses the binarized red image and near infrared image to generate the vegetation. It is desirable to calculate an index. By binarizing the image, a shaded image can be converted into two gradations of white and black according to a predetermined threshold value, and extraction of a detection target from the image can be facilitated. That is, by performing the binarization process, it is possible to easily extract the detection target, so that the process becomes easy even when calculating the vegetation index after that, and it is possible to process (or execute) at high speed. Become.

  Each of the red image and the near-infrared image is divided into a plurality of regions, and the image analysis unit calculates a vegetation index for each of the divided regions, and for each of the regions where the fruits are photographed, for each region. It is desirable to calculate the average value of the calculated vegetation index. The area to be divided is within the area where the fruits are reflected in the image captured by the photographing means. Therefore, it is desirable that the area where the fruits are reflected is extracted. When analyzing a captured image, it is desirable that a plurality of fruits included in the captured image are extracted for each result, so the acquired photographed image is labeled to identify the contour shape of the fruits, It is desirable to create an image for each fruit.

  By calculating the vegetation index for each of the divided areas, it is possible to inspect and determine defects such as black spots, discoloration, or cracks for each captured fruit. That is, since the value is calculated as a normalized value between −1 and +1 for each divided area, the degree of the defect can be recognized. For example, it is possible to recognize the size of black spots, the range of discoloration, the presence or absence of cracks, and the like. Therefore, by calculating the vegetation index for each sectioned area, it is not necessary to discard all the fruits recognized as defective.For example, the grown fruits can be used without waste by changing the shipping destination depending on the degree of defects. It becomes possible to do.

  Then, by calculating the average value of the vegetation index calculated for each region, it is possible to determine the activity of the fruit to be inspected / determined, the elapsed time after harvesting (ie, freshness), and the like. In other words, by taking the average value of the vegetation index for each divided area, one or more fruits are recognized as one group in the entire photographed area, and the vegetation index of that one group is calculated. The average value of the vegetation index calculated for each region is calculated. By calculating the average value of the vegetation index calculated for each region, it can be recognized as the vegetation index of the group, and the activity, water content, nitrogen content, fiber content, and protein content of the fruit in the group At least one of the evaluation items can be inspected and judged. Therefore, it becomes possible to select target fruits according to freshness, taste, nutritional state, and the like.

<Inspection of fruit cracking>
Further, it is desirable to use a visible light image (hereinafter also referred to as “RGB image”) composed of reflected light in the visible light region in order to make it easier to inspect and determine whether there is a crack in the fruit. That is, a visible light image composed of reflected light in the visible light region is also acquired by the image acquisition means. The image analysis means uses a value obtained by subtracting the reflectance of the red wavelength region in the red image from the reflectance of the visible light region in the visible light image and adding the value to the value of the vegetation index. For example, when inspecting and judging cracks in green soybeans, an image is composed of only two colors, green and white, and the part where the cracks are generated can be displayed in white, which makes it easy to visually judge. . In the image simply indicated by the vegetation index, the green portion of the green soybean has a small amount of red reflection, and the non-green portion has a large amount of red reflection. That is, since the part of the green soybean cracked and the part other than the green soybean are shown in red, it is difficult to determine the crack. Therefore, the amount of red reflection is reduced by subtracting the reflectance of the red wavelength region in the red image from the reflectance of the visible light region in the visible light image, and the value of the vegetation index (particularly binarized processing is added to that value). By adding (preferably), a crack portion which is a non-green portion can be shown in white. Thereby, recognition with the naked eye can be facilitated. In addition, regarding the determination of cracks, it is possible to inspect and determine by calculating the vegetation index for the entire photographed area without dividing it into areas as described above and then checking the color difference by the above method.

<Fruit sorting device>
In the present invention, the fruit sorting system configured as described above can be combined with fruit transporting means, etc., to provide a fruit sorting apparatus that can continuously sort the fruit. That is, in order to solve at least one of the above problems, a conveyance unit that conveys an inspection object, an imaging unit that images the inspection object conveyed by the conveyance unit, and an image captured by the imaging unit are acquired. In addition, a fruit sorting device for analyzing the image is provided, and a fruit sorting device using the fruit sorting system according to the present invention is provided as the fruit sorting means.

  According to such a fruit sorting device, the fruit to be inspected is continuously transported by the transport means, photographed by the photographing means, and analyzed by the fruit sorting system according to the present invention. Can be found and eliminated as needed.

  Moreover, in the fruit sorting apparatus according to the present invention, air is blown to the inspection object designated by the fruit sorting means among the inspection objects conveyed by the conveying means, and the inspection object is blown away by air. It is desirable to provide a discharging means. At this time, the image analyzing means in the fruit selecting means can specify the center of the length and the width from the result of the size inspection for measuring the length of each fruit and the width of each fruit from the photographed image. Therefore, it is desirable that the discharging means is configured to blow air toward the center of the length and width of the inspection object designated by the fruit sorting means.

By providing the discharging means, it is possible to discharge the fruit judged to be out of the standard by the fruit sorting system from the transport line, and it is possible to sort only the fruit that meets the standard. In addition, the image analysis means constituting this fruit sorting system can specify the length and width center from the result of the size inspection that measures the length of each fruit and the width of each fruit from the captured image, Therefore, the discharging means can blow air toward the center of the length and width of the inspection object designated by the fruit sorting means. As a result, air can be blown to the center of the non-standard fruit, and the non-standard fruit can be reliably extracted. Thus, even when a large amount of sorting is required, the harvested fruits can be inspected and sorted accurately and reliably.

  According to the edamame sorting system of the present invention, it is possible to inspect the accommodation state (especially the number and size) of seeds in the cocoon based on the image obtained by photographing the reflected light of the light irradiated on the fruit. The need for light photography can be eliminated. Further, even when an image taken for quality control is saved, it is only necessary to save an image taken of reflected light, so that the image saving area can be greatly reduced. In the inspection, it is only necessary to analyze one type of image obtained by photographing the reflected light. Therefore, the harvest time is limited to a certain period, and a large number of fruit must be inspected and selected within a short period. Even if it exists, inspection and sorting after harvesting can be performed quickly and in large quantities.

  Further, if the vegetation index is calculated based on the reflected image of the fruit photographed by the photographing means, that is, the red image composed of the reflected light in the red wavelength region and the near infrared image composed of the reflected light in the near infrared wavelength region. By using this index, it is possible to inspect and sort the fruit scratches and activity. That is, the appearance can be inspected and selected more accurately, and in particular, scratches and the like can be inspected and selected in the same manner as visually.

  In addition, by using the vegetation index, not only the appearance (particularly cracking) of the fruit but also the activity, water content, nitrogen content, fiber content, and protein content of the fruit are examined or selected. I can do it. Therefore, since the passage of time after harvesting and the dryness (moisture content) of the fruit can be recognized in detail, quality control can be more thoroughly implemented.

Furthermore, by utilizing the fruit sorting apparatus using the fruit sorting system according to the present invention, it becomes possible to inspect and sort the harvested fruits quickly and in large quantities. That is, in the present invention, by using the reflected light image and calculating the vegetation index, the accommodation state of seeds (particularly the number and size) in the pod, the presence or absence of pods or beards, Water content, nitrogen content, fiber content, and protein content can be examined. This eliminates the need to perform post-harvest inspection / sorting in multiple stages. Therefore, by using the fruit sorting apparatus according to the present invention, the harvesting time is limited to a certain period, and even if the fruit has to be inspected and sorted in a short period, Inspection and sorting can be performed quickly and in large quantities.

Whole block diagram which shows the green soybean sorter comprised using the green soybean selection system concerning this Embodiment Block diagram showing hardware configuration Flow chart showing the contents of basic processing Process content diagram showing basic process contents Flow chart showing the contents of processing for particle number inspection Contents of processing to check the number of grains Flow chart showing the contents of the single grain inspection process Flow chart showing processing contents in chip inspection Process details for chip inspection Flow chart showing processing contents in beard inspection Process content diagram for performing mustache inspection Flow chart showing the contents of vegetation index calculation processing Processing content diagram showing the content of vegetation index calculation processing Flow chart showing processing details for black spot, discoloration, and crack inspection Front view showing how to specify the sheath area / spot area in black spot, discoloration, and crack inspection Flow chart showing the contents of the activity inspection process The flowchart which shows the content of the process of the crack inspection concerning 2nd Embodiment The graph which shows the transmittance | permeability of IR filter used in Example 1 The perspective view which shows three types of test samples used in Example 1 IR image and RGB image of green soybean meal obtained by photographing in Example 1 NDVI (vegetation index) image obtained with the image processing software in Example 1 Binarized image based on NDVI (vegetation index) in Example 1 An image obtained by applying a binarized NDVI image to an RGB image obtained by reducing the red value in Example 1.

  Hereinafter, an embodiment of a fruit sorting system according to the present invention will be specifically described with reference to the drawings. In particular, in the present embodiment, a fruit sorting system for sorting green soybeans will be specifically described.

  FIG. 1 shows an overall configuration of a fruit sorting apparatus (hereinafter referred to as “green soybean sorting apparatus”) configured using a fruit sorting system according to the present embodiment (hereinafter referred to as “green soybean sorting system 40”). FIG. This green soybean sorting device accommodates green soybeans W to be selected, and continuously supplies the green beans to the sorting line, the first conveying means 20 for conveying the green beans W supplied from the hopper 10, and the The second conveying means 30 that inverts and conveys the green soybeans W conveyed by the first conveying means 20, the photographing means 21, 31 provided in each conveying means, and the images photographed by the photographing means 21, 31 are analyzed. The soy bean sorting system 40, the transport detecting means 22 and 32 provided in each transport means for detecting the transport speed or transport amount, and the discharge means 23 and 33 provided in the falling part of the soy beans W on the terminal side of each transport means. It consists of and.

  The hopper portion 10 includes a space portion for storing green soybeans W (fruits) to be selected. In addition, it is accompanied by a discharge structure for supplying green soybeans W accommodated in the hopper section 10 to the conveying means composed of the first conveying means 20 and the second conveying means 30. Such a discharge structure is a belt conveyor or the like provided on the bottom surface of the hopper portion 10 and moves the green soybeans W housed in the hopper portion 10 forward. However, the hopper portion 10 may have another configuration as long as the green soybean W accommodated in the hopper portion 10 can be sent out to the conveying means for performing the inspection.

  The green soybean W discharged from the hopper unit 10 is supplied to a conveying means composed of the first conveying means 20 and the second conveying means 30. At that time, a vibrating means (not shown) for conveying the green soybeans W discharged from the hopper while vibrating them may be provided in the dropped portion so that the green beans W dropped from the hopper 10 do not overlap each other. desirable.

  The first conveying means 20 and the second conveying means 30 can each be formed by a belt conveyor or the like. The green soybean W supplied from the hopper unit 10 is first transported by the first transport means 20. This first conveying means 20 can be preferably accompanied by an alignment portion for aligning the direction of the green soybeans W to be conveyed in a certain direction. Such an alignment portion can be formed, for example, by arranging alignment plates or wires (not shown) that are long in the conveying direction of the green beans W at a predetermined interval. The green soy beans W to be transported are corrected so as to be vertically oriented in the transport direction by hitting this alignment plate.

  Above the first conveying means 20, there is provided a photographing means 21 for photographing the green soybeans W conveyed by the first conveying means 20. As this photographing means 21, a camera or a video camera can be used. The photographing means 21 desirably captures the photographed image as electronic data so that the green soybean W being conveyed can be photographed as a still image or a moving image, and the photographed image can be directly taken into the green soybean sorting system 40 and analyzed. It is also desirable to provide lighting that stably illuminates the green soybean W existing in the photographing range of the photographing means 21. This is to stabilize the brightness of the photographed image by making the brightness of the light illuminating the green soybean W constant.

  The image photographed by the photographing means 21 is sent to a green soybean sorting system 40 constituted by a computer or the like. In the green soybean sorting system 40, based on the acquired photographed image, the green soybean W to be sorted is inspected, and at least the size and number of grains of the green soybean W, the presence or absence of chipping or whiskers, the presence or absence of black spots, discoloration, and the presence of crack Or at least one of the activities, and identify green soybean W that is out of specification. The configuration and processing of the green soybean sorting system 40 will be described later.

  The first transport means 20 is provided with a transport detection means 22 for acquiring the transport speed and transport distance. This is to identify where the green soybean W (non-standard green soybean W) selected by the green soybean selection system 40 is conveyed by associating the conveyance information of the green soybean W with the photographed image. By specifying the non-standard green soybeans W selected by the green soybean sorting system 40 in the conveyance process, the non-standard green soybeans W can be discharged out of the conveyance line. As the conveyance detection means 22, a rotary encoder, a linear encoder, or the like can be used.

  In the present embodiment, the non-standard green soybeans W are excluded when falling from the end of the first conveying means 20. That is, at the end of the first conveying means 20, there is provided a dropping portion for dropping the green soybean W being conveyed to the second conveying means 30, and when falling from the dropping portion, the green soybean sorting system 40 identifies The edamame W (non-standard edamame W) is discharged from the transport line by the discharge means 23. The discharging means 23 can be formed by blowing air to the identified green soybean W and blowing it out of the transport line, or by a configuration that allows the specified green soybean W to be taken out.

  As described above, the green soybeans W that have been transferred by the first transfer means 20 and determined to be out of specification by the green soybean sorting system 40 are discharged by the discharge means 23, and the remaining green beans W are transferred by the second transfer means 30. When transported by the second transport unit 30, the green soybean W is desirably transported in a state reversed from the transport state in the first transport unit 20. Therefore, it is desirable to provide a reversing means for reversing the green soybeans W to be conveyed after the first conveying means 20 or at the start end side (first conveying means 20 side) of the second conveying means 30. For example, the reversing means can be reversed by using a drop when dropping the green soybeans W to be transported, and can be reversed by applying a rotational force by a belt or a roller to any surface of the fruit to be transported.

  The green soybeans W transported by the second transport means 30 are also sorted in the same manner as the transport in the first transport means 20. That is, the green soybean W conveyed by the second conveying means 30 is photographed by the photographing means 31, and the photographed image is analyzed by the green soybean sorting system 40 to identify the non-standard green soybean W. At that time, the conveyance speed and distance of green soybeans are acquired from the conveyance detection means 32 comprising a rotary encoder. And the non-standard green soybeans W are discharged by blowing air by the discharge means 33 at the end of the transport means. Since the process in the second transport unit 30 is the same as the process in the first transport unit 20, in the drawing, the same configuration as the first transport line is indicated by adding a value of “10” to the reference numeral, and the details The detailed explanation is omitted.

  Next, with reference to FIGS. 2 to 23, the configuration of the green soybean sorting system for inspecting green beans conveyed in the green soybean sorting device and the processing contents thereof will be described in detail.

  FIG. 2 shows an example of a hardware configuration of a computer constituting the fruit sorting system according to the present embodiment. However, the computer 500 in FIG. 2 only exemplifies a typical configuration of the fruit sorting system, and the green soybean sorting system has a dedicated function as long as it executes an arithmetic device, a memory, and a program for image analysis. You may comprise as an apparatus.

  2 includes a CPU 501, a memory 502, an audio output device 503, a network interface 504, a display controller 505, a display 506, an input device interface 507, a keyboard 508, a mouse 509, an external storage device 510, and an external recording medium drive. The apparatus 511 includes a bus 512 that connects these components to each other.

  The CPU 501 controls the operation of each component of the computer 500, controls the execution of each process of the green soybean sorting system under the control of the OS, and controls its operation. The memory 502 is generally composed of a ROM (Read Only Memory) that is a non-volatile memory and a RAM (Random Access Memory) that is a volatile memory, and corresponds to a storage unit. The ROM stores a program executed by the green soybean selection system executed when the computer 500 is activated. The RAM is executed by the CPU 501 to analyze the image acquired from the photographing means, calculate the size and number of grains of green soybeans, and determine the presence or absence of chipping, whiskers, black spots, discoloration, cracks, activity, etc. Programs and data used by these programs during execution are temporarily stored.

  The audio output device 503 is a device that outputs sound, such as a speaker, and the network interface 504 is an interface for connecting to a network 520 for exchanging information with various devices. A display controller 505 is a dedicated controller for processing a drawing command issued by the CPU 501 and outputs drawing data to a display 506 serving as a display unit. The display 506 is a display device configured by an LCD or the like.

  The input device interface 507 receives a signal input from an input / output device such as a keyboard 508, a mouse 509, or a touch pad, and provides a predetermined command to the CPU 501 according to the signal pattern. A keyboard 508 and a mouse 509 are necessary for operations such as program execution and settings.

  The external storage device 510 is also included in the category of storage means in this specification. The external storage device 510 can be configured by a storage device such as a hard disk drive (HDD). The above-described program and data are recorded in this device, and are loaded from there into the RAM of the memory 502 as necessary at the time of execution.

  The external recording medium driving device 511 accesses a recording surface of a portable external recording medium 530 such as a CD (Compact Disc), an MO (Magnet-Optical Disc), a DVD (Digital Versatile Disc), etc., and is recorded there. It is a device that reads data.

  The green soybean sorting system according to the present embodiment is formed by using the computer configured as described above, and executes the processes shown in FIGS.

≪Basic processing≫
FIG. 3 is a flowchart showing the contents of the basic process executed on the captured image acquired from the photographing means, and FIG. 4 is a process contents diagram showing the contents of the basic process. When the basic process is executed, the green soybean selection system first acquires a captured image showing a plurality of green beans from the imaging means and records it in the memory (S11). Then, as shown in FIG. 4 (B), a labeling process is performed for each green soybean on an image showing a plurality of green beans (S12), and an image for each green soybean is extracted. Then, angle correction processing is performed on the extracted image for each green soybean (S13). In this angle correction processing, image processing is performed so that the long direction (length direction) faces sideways in each green soybean image. Then, as shown in FIG. 4C, binarization processing is performed on the angle-corrected image (S15) so that necessary numerical values and locations can be quickly identified and acquired in each subsequent inspection. I have to.

≪Grain number inspection≫
FIG. 5 is a flowchart showing the processing content for performing the grain number inspection, and FIG. 6 is a processing content diagram for performing the grain number inspection. In this grain number inspection, an image that has been subjected to the basic processing first (an image after binarization processing) is acquired as an inspection image. Since this inspection image is created for each green soybean transported by the transport means, this grain number inspection is executed for all green beans.

  When the inspection image is acquired, the CPU of the green soybean selection system constituted by a computer reads the number of pixels in the length direction in the region where the green soybean is projected in the inspection image, as shown in FIG. Based on this, the length of the green soybean is calculated (S52). This calculation can be calculated by multiplying a coefficient for making the captured image an actual measurement value. And if the length of green soybeans calculated by this calculation is acquired, the number of grains will be calculated based on this length (S54). As shown in FIG. 6, the calculation of the number of grains is performed by searching and extracting the file 41 in which the number of grains is defined in relation to the length, and without searching the file system. It is also possible to create a program that returns a value that specifies.

  It is determined whether the number of grains obtained as a result of this calculation is 1 (S56). This is because, in the case of green soybeans, one grain may have a low commercial value, so that the seed selects and removes one green soybean to increase the commercial value. In determining whether there is one seed contained in this pod, first, in the determination of the number of grains, one determined as one grain is excluded. Therefore, the green soybean for which the seed in the persimmon is determined to be one grain is determined to be a nonstandard product, and is specified as a discharge target by the discharge means.

  And in the said grain number judgment, the area | region is divided | segmented according to the grain number of the green soybeans by which the seed accommodated in the basket was judged to be 2 or more (S58). FIG. 6 shows an example in which it is determined that there are two grains. For this reason, the inspection image is divided into two regions (A1 and A2) in the length direction, as shown in FIG. 6B. Next, in each region, a point having the largest width value is set as a convex portion. In FIG. 6C, the widest point in the first region A1 is set as the first convex portion C1, and the widest point in the second region A2 is set as the second convex portion C2. ing. Then, a value (width) at each convex portion is calculated (S59).

  Next, as shown in FIG. 6 (D), the SA width between the respective convex portions (C1 and C2) is calculated, and as shown in FIG. 6 (E), a value smaller than the width of each convex portion. And the point where the width becomes the minimum value is set as the recess, and the width is calculated (S60). Then, the presence / absence of a recess is determined (S61). If the recess can be calculated, two or more seeds are contained in the pod, and it is determined to be within the standard in this grain number inspection as a green soybean that meets the standard ( S62). On the other hand, when the concave portion cannot be calculated, it is determined that the seed in the pod is one green soybean, it is determined that it is out of specification, and is specified as a discharge target by the discharge means. That is, the edamame sorting system executes the process of specifying the convex part and the concave part as described above as a result of judging that the number of seeds in the cocoon is two from the length of the cocoon (FIG. 7A). Specifically, the image of the green soybean is divided into two regions (A1 and A2) (FIG. 7B), and the widest region in each region is divided into the convex portions (C1 and C2). The setting is made (FIG. 7C). And the position of the narrowest width between these convex parts and narrower than each convex part is set to a recessed part (FIG.7 (D)). However, when the seed in the pod is one grain, the width of the concave portion is the same as any of the convex portions, and thus the concave portion cannot be set (FIG. 7E). Therefore, in this case, the seed in the pod is determined to be one grain, and is identified as a non-standard green soybean.

  By executing the particle number inspection by the above processing, the processing speed can be increased. That is, depending on the length of the pods, the seeds in the pods temporarily remove one edamame, so that the number of inspection objects can be reduced, thereby speeding up the processing. Furthermore, for the green soybeans that are within the standard in length, by further inspecting the convex part and the concave part, while the standard of the length is satisfied, the seeds in the pod can actually extract one green soybean, Sorting accuracy can be greatly improved. Therefore, by performing the grain number inspection by the above-described processing, it is possible to quickly and accurately select green soybeans (one soybean with one seed in the pod).

≪Chip inspection≫
Further, the green soybean sorting system can further perform the chip inspection by acquiring the inspection image created by performing the basic processing. In this chipping inspection, whether or not there is a chipping in the edamame is inspected. In this embodiment, the presence or absence of this chipping is determined from the amount of change in the width of the green soybeans in the inspection image.

  FIG. 8 is a flowchart showing the processing contents in this chip inspection, and FIG. 9 is a processing contents diagram for performing this chip inspection. As shown in FIG. 8, the inspection image created by the basic process is read by executing the chip inspection (S81). Then, an area having a large change in width in the inspection image, that is, a range excluding predetermined areas on both ends in the length direction is designated as the inspection area (S82). However, when specifying the inspection area, in addition to reading the width of the cocoon, it can also be specified as a percentage of the total length. For example, the range excluding the left and right 5% of the total length of the green soybean image is used as the inspection area. You can also specify.

  After designating the inspection area, an arbitrary position as a reference for inspection within the inspection range is specified (S83), the width of the reference position is calculated, and this is set as the reference width (S84). Then, a width at a point moved from the reference position by a predetermined amount in an arbitrarily determined direction is set as a comparison width, and values of the comparison width are sequentially acquired. In the present embodiment, the width at the position moved by a predetermined amount (for example, 2 pixels) from the reference position to the left is sequentially acquired as the comparison width (S85). Then, a difference between the acquired reference width and each comparison width is calculated (S86), and this is compared with a preset allowable value (S87). When calculating the difference between the reference width and each comparison width, the difference between the reference width and each comparison width can be simply calculated, and the change amount of the difference calculated by each comparison value can also be calculated. . In this case, an allowable value for the change amount of the difference is set in advance, and when the change amount exceeds the allowable value, it is possible to perform a non-standard determination.

  As a result of the comparison between the difference and the allowable value, if the difference exceeds the allowable value, it indicates that the width is abruptly narrower than the reference value. Judging that it exists. Therefore, if this difference exceeds an allowable value (preset allowable value), a non-standard determination is performed (S89). On the other hand, when the difference of the comparison width calculated based on an arbitrary reference position in the inspection area is equal to or smaller than the allowable value, the next reference position is set in the inspection area (S90). That is, a new reference position is set in a range not yet inspected. The reference position to be newly set is a position moved in the length direction by adding an arbitrary number to the previously set reference position, and the difference from the comparison width is set again with the position as a new reference position. And the value is compared with the allowable value to determine whether the value is within or outside the standard. The width of edamame, especially edamame, varies greatly depending on whether seeds are present in the pod. Therefore, if any one point is used as the reference position, the difference from the comparison value also changes significantly. As a result, it becomes difficult to determine whether the convex portion and the concave portion are normal shape changes or whether there is a chip. Therefore, in the present embodiment, when inspecting one green soybean, a plurality of reference positions are set in the inspection area set when inspecting for the presence or absence of a chip, and the difference between the width for each reference position and the comparison width Is calculated and the presence of a chip is inspected.

  Then, for each reference position set while shifting the position in the length direction as described above, the difference from the comparison width is compared with the allowable value. Then, when the comparison is completed for all of the inspection areas, if the difference from the comparison width is equal to or less than the allowable value at all the reference positions, it is determined that it is within the standard, and the lack inspection of the green soybean being inspected is completed. .

  By determining the presence or absence of chipping in green soybeans based on the above processing contents, chipping can be accurately detected even for green soybeans with a large amount of change in width. Moreover, since only the difference between the reference position and the comparison width is calculated, the processing speed can be increased, and the amount of green soybeans selected per unit time can be increased.

≪Beard inspection≫
In addition, the green soybean sorting system can acquire a test image created by performing a basic process and further perform a mustache test. In this mustache inspection, it is inspected whether or not there is a mustache portion that protrudes linearly from the pods of green soybeans. In the present embodiment, the presence or absence of this mustache portion is determined based on the number of edges in the width direction of green soybeans in the inspection image. This edge is a boundary point of the outline in the width direction, and can be specified as a point where the color value changes between the green soybean image and the background in the inspection image.

  FIG. 10 is a flowchart showing the processing content in this mustache test, and FIG. 11 is a processing content diagram for performing this mustache test. As shown in FIG. 10, the inspection image created by the basic process is read in response to the execution instruction of the mustache inspection (S101). For the acquired inspection image, the number of edges in the length direction is detected as shown in FIG. 11B (S102). Specifically, the number of points (hereinafter referred to as “edges”) existing in the contour in the width direction is detected in the green soybean image in the inspection image. Then, the number of detected edges is determined (S103). When determining the number of edges, as indicated by reference numeral E2 in FIG. 11C, if there are two edges, the outline of the normal cocoon in the green soybean is shown. Therefore, it is judged within the standard for green soybeans that do not have a mustache part (S105). On the other hand, as indicated by reference numeral E1, when the number of edges is three or more, specifically four, there is a protruding portion in addition to the outline of the eyelid. Since this protruding portion corresponds to a mustache portion in green soybeans, a point where the number of edges is four is specified as a point where the beard portion exists, and the presence or absence of the beard portion is determined (S104).

  For the green soybeans that have been determined to have a mustache portion, the length of the mustache portion is then calculated (S106). As shown in FIG. 11C, the length of the beard portion is calculated as a length from a point where the number of edges is calculated to a point where the number of edges converges to two or three. To do. Then, the calculated length of the beard portion is compared with a preset allowable value (S108), and if the length of the beard portion is equal to or smaller than the allowable value, it is determined that the length is within the standard. On the other hand, when the length of the beard portion exceeds the allowable value, it is determined that it is out of the standard and is recorded as a discharge target in the discharge means.

  By the above processing, it is possible to inspect not only whether there is a beard portion, but also the length of the beard portion, so that more accurate selection can be realized. It should be noted that the fruit sorting system and the fruit sorting apparatus according to the present invention can be variously modified without being limited to the mode shown in the above embodiment. For example, it can be configured to select and execute at least one of grain number inspection, chipping inspection, and whisker inspection, or can be manufactured by configuring a program for executing each process with different logic.

≪Vegetation index calculation process≫
FIG. 12 is a flowchart showing the contents of the vegetation index calculation process executed on the photographed image acquired from the photographing means, and FIG. 13 is a process contents diagram showing the contents of this vegetation index calculation process. When this vegetation index calculation process is executed, the green soybean selection system first displays a photographed image showing a plurality of green beans, specifically a red image consisting of reflected light in the red wavelength region (“R image” in FIG. 13). ) And a near-infrared image composed of reflected light in the near-infrared wavelength region (indicated as “IR image” in FIG. 13) is acquired from the imaging means and recorded in the memory (S111, S121). Then, as shown in FIG. 13 (B), a labeling process is performed for each edamame on the image showing the plurality of edamame (S112, S122), and an image for each edamame is extracted. An angle correction process is performed for each image (S113, S123). In this angle correction processing, image processing is performed so that the long direction (length direction) is vertically oriented in each green soybean image. Then, as shown in FIG. 13C, binarization processing is performed on the image that has been subjected to angle correction (S114, S124), and in subsequent inspections, the necessary numerical values and locations can be quickly identified and acquired. I can do it. That is, using the binarized red image and near-infrared image, a vegetation index is calculated as shown in FIG. 13D (S115) and output as an inspection image (S116). In the present embodiment, the labeling process, the angle correction process, and the binarization process are performed. However, the vegetation index can be calculated without necessarily performing the process.

≪Inspection of black spots, discoloration, cracks≫
FIG. 14 is a flowchart showing the contents of processing for performing black spot / discoloration / cracking inspection, and FIG. 15 is a front view showing a mode of designating a sheath area / black spot area in the black spot / discoloration / crack inspection. In such black spot / discoloration / cracking inspection, first, an inspection image of the target green soybean is acquired by a photographing means such as a camera (S151), and a wrinkle region and a defective region in the acquired image are designated (S152). FIG. 15 shows a state in which the edamame cocoon area 61 and the black spot area 62 are designated, and shows the processing contents when the black spot inspection is performed.

  As described above, after designating the cocoon region and the defective region, the vegetation index for each region is calculated (S153). In addition, it is possible to detect sunspots, discoloration, and cracks without necessarily calculating the vegetation index when carrying out black spot, discoloration, and cracking inspections, but using the vegetation index makes it easier to recognize with the naked eye, Subsequent processing can be facilitated.

  Then, it is possible to determine whether or not the target green soybean is within the standard by calculating the ratio of the defective area occupying the entire basket (S154) and determining the ratio of the defective area (S155). That is, by calculating the ratio of the defective areas as described above, it is determined whether or not the ratio of the defective areas is within the set allowable value range (S156). If the ratio of the defective area of the target green soybean is within the allowable value, it is transported to the subsequent process as an in-standard product (S158). On the other hand, if the ratio of the defective area of the target green soybean exceeds the allowable value, it is determined as a non-standard product and specified as a discharge target by the discharge means (S157).

≪Activity test≫
Next, the case where the activity test of the target green soybean is performed will be described based on the flowchart shown in FIG. In such an activity test, first, an inspection image of the target green soybean is acquired by a photographing means such as a camera (S171), and an inspection region in the acquired image is designated (S172). Then, a vegetation index for each designated area is calculated (S173), and an average value of the vegetation index for each area is calculated (S174). Here, by taking the average value of the vegetation index for each region, one or a plurality of green soybeans are recognized as one group in the entire photographed region. By calculating the average value of the vegetation index for each region, the vegetation index of the group can be recognized, and the activity of green soybeans in the group can be recognized. Therefore, since the activity level of the target green soybeans can be recognized, selection based on the activity level is possible.

  Then, it is determined whether or not the degree of activity calculated as described above, that is, the average value of the vegetation index for each region is within a preset allowable value range (S175). As a result, when the activity of the target green soybean is within the allowable value, it is conveyed to the subsequent process as an in-standard product (S177). On the other hand, if the activity of the target green soybeans exceeds the allowable value, it is determined as a non-standard product, and is specified as a discharge target by the discharge means (S176).

≪Inspection of cracks using visible light images≫
Next, the processing content of the crack inspection according to the second embodiment will be described based on the flowchart shown in FIG. In the crack inspection according to the second embodiment, a crack inspection is performed using a visible light image made of reflected light in the visible light region. First, a photographic image showing a plurality of green soybeans from the photographing means, specifically a visible light image obtained by photographing reflected light in the visible light wavelength region (hereinafter also referred to as `` RGB image ''), and a near infrared wavelength region A near-infrared image obtained by photographing the reflected light (hereinafter also referred to as “IR image”) is acquired (S181, S191). In addition, when acquiring the above image, in addition to extracting and acquiring the spectrum of each wavelength region from the reflected image of green soybean imaged by the imaging unit, installing a color filter in the imaging unit, You can shoot and use reflection images

  In each image, an inspection area to be a target is designated (S182, S192). After designating the inspection area, it is desirable to perform image composition processing (or pixel matching) on the acquired RGB image and IR image in order to correct pixel displacement due to camera parallax (S183).

  In the synthesized image, a vegetation index is calculated (S184) and binarization processing is performed (S185). On the other hand, the step of subtracting the reflectance of the red region from the reflectance of the visible light region in the synthesized image is also performed simultaneously (S193). Then, by adding the binary value of the vegetation index to the value obtained by subtracting the reflectance of the red region from the reflectance of the visible light region (S186), it becomes possible to extract a crack that can be recognized with the naked eye.

  Then, it is determined whether or not the green soybeans extracted as described above are cracked (S187). As a result, when there is no crack in the target green soybean, it is conveyed to the subsequent process as an in-standard product (S189). On the other hand, when the target green soybean is cracked, it is determined as a non-standard product and specified as a discharge target by the discharge means (S188).

In Example 1, it was verified whether the green soybean crack inspection is an effective method that can be recognized with the naked eye.
<Use machine>
In this example, in order to obtain a four-band image of the pods of green soybeans, an infrared cut filter (hereinafter also referred to as “IR cut filter”) is removed from the digital camera (CANON PowerShot G10) of about 14.7 million pixels, and the infrared ray is transmitted. An infrared image (hereinafter also simply referred to as “IR” image) was obtained by attaching a filter (hereinafter also simply referred to as “IR filter”). The IR filter used was IR-76, 78, 80, 82, 84, 86, 88, 90, 92, 94 manufactured by FUJIFILM, and the transmittance of each is shown in the graph of FIG. This IR filter is a film that absorbs visible light on the short wavelength side of 700 nm or less and transmits infrared light.

  Next, an RGB three-band image (hereinafter also simply referred to as “RGB” image) was obtained with a non-modified digital camera (CANON PowerShot G10). Both images were taken under the condition that they were fixed at a location 30 cm high from the sample so that the optical axis was not shifted.

  In the 4-band image processing of captured IR images and RGB images, it is difficult for the lens optical axes of both cameras to be strictly parallel in a three-dimensional space, so images are formed in different forms on the screen. Pixel displacement due to camera parallax also occurs. Therefore, image deformation correction and RGB image and IR image synthesis processing (pixel matching) are indispensable for processing. Therefore, pixel matching of the acquired image was performed, and NDVI (vegetation index) was calculated and visualized. This NDVI (vegetation index) calculation and visualization process were performed as follows. That is, first, 1) read an RGB / IR image to be processed / synthesized, and 2) specify a measurement section from the RGB / IR image. Next, 3) the images are combined, and 4) the distortion of the combined image is corrected. 5) The vegetation index was calculated, visualized, and graphed.

<Verification method>
As test samples, edamame (variety “secret”) was used, and three types were prepared: non-defective rice cakes, rice cakes with cracks that can be recognized with the naked eye, and rice cakes with cracks that are difficult to recognize with the naked eye. . In this experiment, 1) Place the above three types of specimens so that they fit within the frame of a white paper with a 10cm square drawn (see Fig. 10), and then 2) install a light and provide sufficient light. We secured a digital camera with an IR filter and secured it with a tripod. The height from the edamame bowl to the digital camera, shutter speed, and aperture (F value) were the same. Then, 3) the above photographing was performed with IR filters 76 to 94, and 4) NDVI (vegetation index) was calculated and visualized for the obtained images, and whether or not cracking of green soybeans could be determined was verified.

<Results and discussion>
FIG. 20 is an IR image and RGB image of green soybean meal obtained by photographing in this experiment, and FIG. 21 is an NDVI (vegetation index) image obtained by image processing software. FIG. 22 is a binarized image based on NDVI (vegetation index), and FIG. 23 shows an image obtained by covering the binarized NDVI image with the red value reduced from the RGB image. ing.

  As shown in FIG. 21, in the NDVI image obtained by applying the captured image to the image processing software, a crack that can be recognized with the naked eye can be recognized in all the images.

  In addition, the binarization process was applied to the NDVI image, and the possibility of recognizing cracks that could not be recognized with the naked eye was examined. The result is shown in FIG. The binarization process makes it easier to recognize cracks that can be recognized with the naked eye than simple NDVI images. In order to reduce the effects of light intensity, it is desirable to use a filter of IR-86 or lower, and it has been confirmed that the use of a filter of IR-86 or lower makes it easier to recognize cracks. did it.

  In addition, the NDVI image has the property that the green portion has a small amount of red reflection and the non-green portion has a large amount of red reflection. FIG. 23 shows the result of examining whether or not a non-green crack can be recognized by applying a binarized NDVI image to an RGB image obtained by greatly reducing the red value using this property. ing. As a result, cracks that could be recognized with the naked eye were free from black spots and could be confirmed more easily than those obtained by binarizing an NDVI image.

  From the above, in the above verification, the process of applying the binarized NDVI image to the image obtained by greatly reducing the red value from the RGB image is more unaided than the simple NDVI image and the binarized image. It was confirmed that it was easy to confirm the cracks that could be recognized by. That is, by selecting green soybeans by the above process, it becomes possible to inspect and select accurately and quickly.

10 Hopper section
20 First transfer means
21, 31 Photography means
22, 32 Transport detection means
23, 33 Discharge means
30 Second transport means
31 Shooting means
32 Transport detection means
33 Discharge means
40 Edamame sorting system
W Edamame

Claims (12)

A fruit sorting system for inspecting or sorting the fruits by analyzing an image obtained by photographing the fruits with a bowl,
An image acquisition means for acquiring an image taken by the imaging means; and an image analysis means for analyzing the image acquired by the image acquisition means,
The image analysis means includes
A size test that measures the length of each fruit and the width of each fruit based on the captured image;
A fruit sorting system that performs a grain number test for calculating the number of seeds contained in the straw based on the measured fruit length.
Grain number inspection in the image analysis means,
The image of each fruit is divided in the length direction by the number of the number of grains, and the position where the width is maximum in each of the divided areas is set as a convex portion, and the width between the convex portions is smaller than the concave portion. The fruit sorting system according to claim 1, wherein a narrow and minimum position is set as a concave portion, and the width of the convex portion and the concave portion is calculated.
The image analysis unit performs a labeling process on the captured image acquired from the imaging unit, creates an image for each result, performs angle correction on each image, and performs a long axis direction on the corrected image on which the angle correction has been performed. The fruit sorting system according to claim 1 or 2 which measures the length which becomes and the width which becomes the direction of a minor axis.
Furthermore, the image analysis means includes
For the captured image acquired from the imaging unit, an area whose width change amount is within a predetermined range is set as an inspection area, and the width of an arbitrary reference position in the inspection area is acquired as a reference width, and from the reference position 3. The fruit sorting system according to claim 1, wherein a defect inspection is performed in which a width at a position moved in a certain direction is acquired as a comparison width, and a missing inspection is performed in which a value obtained by calculating a difference between the reference width and the comparison width is compared with an allowable value.
Furthermore, the image analysis means includes
Perform binarization processing on the captured image acquired from the imaging means,
Detecting the number of edge portions whose color changes in the width direction at any plurality of points set in the length direction of the eyelid, specifying a point where the number of edge portions is 3 or more as a beard portion existing location, and The fruit sorting system according to claim 1 or 2, wherein the length of the beard portion is calculated from the start point and the end point of the beard portion existing location.
The image acquisition means includes
Extract or acquire a red image consisting of reflected light in the red wavelength region and a near infrared image consisting of reflected light in the near infrared wavelength region,
The image analysis means calculates the vegetation index according to the following formula 1 together with the size inspection and the grain number inspection or instead of the size inspection and the grain number inspection, and inspects or sorts the fruit according to the value of the vegetation index. The fruit sorting system according to claim 1 or 2, characterized by things.
[Where:
R: reflectance of red wavelength region in red image IR: reflectance of near infrared wavelength region in near infrared image]
Further, the image analysis unit performs binarization processing on each of the red image and the near infrared image acquired by the image acquisition unit, and uses the binarized red image and near infrared image to generate the vegetation. The fruit sorting system according to claim 6 which calculates an index.
Each of the red image and the near-infrared image is divided into a plurality of regions,
The image analysis means calculates a vegetation index for each divided area, calculates an average value of the vegetation index calculated for each area for the entire area where the fruit is photographed, and examines or sorts the fruit. 6. The fruit sorting system according to 6.
The image acquisition means further acquires a visible light image composed of reflected light in a visible light region,
The image analysis means subtracts the reflectance of the red wavelength region in the red image from the reflectance of the visible light region in the visible light image, and adds the value to the value of the vegetation index to determine the crack in the fruit. The fruit sorting system according to claim 6, wherein the presence or absence of the fruit is inspected and determined.
A transport means for transporting the inspection object;
Photographing means for photographing the inspection object conveyed by the conveying means;
The image capturing means acquires an image captured by the image capturing means, and has a fruit sorting means for analyzing the image,
A fruit sorting apparatus in which the fruit sorting system according to claim 1, 2, or 6 is used as the fruit sorting means.
Furthermore, it comprises a discharge means for blowing air to the inspection object designated by the fruit sorting means among the inspection objects conveyed by the conveying means, and blowing the inspection object with air,
The image analysis means in the fruit sorting means specifies the length and width center from the result of size inspection that measures the length of each fruit and the width of each fruit from the captured image,
The fruit sorting apparatus according to claim 6, wherein the discharging means blows air toward the center of the length and width of the inspection object designated by the fruit sorting means.
A fruit sorting method for inspecting or sorting the fruit by analyzing an image obtained by photographing the fruit with a bowl,
A reflection image of the fruit taken by the photographing means is acquired, and a process of selecting by analyzing the acquired image is executed,
Extracting or obtaining a red image composed of reflected light in the red wavelength region and a near infrared image composed of reflected light in the near infrared wavelength region as the reflected image of the fruit to be obtained,
A fruit selection method, wherein a vegetation index is calculated by the following mathematical formula 1, and the fruit is examined or selected based on the value of the vegetation index.
[Where:
R: reflectance of red wavelength region in red image IR: reflectance of near infrared wavelength region in near infrared image]