Bread inspection system

Abstract

This invention provides a food inspection system that can verify whether bread, which does not have a consistent shape, meets predetermined quality standards through visual inspection. [Solution] A bread inspection system comprising: a transport means for transporting bread on top; an upper imaging means for imaging the bread from above while it is being transported; an analysis means for acquiring and analyzing the image captured by the imaging means; and a determination means for determining whether the bread is a good product or not. The upper imaging means is a 3D camera, and when the bread is transported to a predetermined position, it images the entire bread from above to acquire an image of the entire bread and thickness information of at least a part of the bread. The analysis means analyzes the size of the bread from the image of the entire bread, acquires the thickness of the bread from the thickness information, and calculates the abnormality rate by comparing the image of the entire bread with a pre-prepared sample of bread shape. The determination means determines that the bread is a good product if the size, thickness, and abnormality rate all meet the pre-set good product criteria.

Description

[Technical Field]

[0001] The present invention relates to a bread inspection system for verifying whether bread meets predetermined quality standards, and more particularly to a bread inspection system that primarily performs visual inspection based on images of bread captured by an imaging means. [Background technology]

[0002] Traditionally, the inspection of bread produced in bakeries and other facilities typically involved humans visually checking whether it met specified quality standards. This inspection included not only checking for foreign objects but also visual inspections to ensure that the size and color of the bread met the required specifications. However, in mass-production bakeries, for example, tens of thousands of loaves of bread are produced daily, and visually checking each one individually was an extremely laborious task. The inspection work required a high level of concentration, placing a heavy burden on workers and increasing the risk of inspection errors due to fatigue. In addition, physical and mental state could potentially affect the inspection results, and since factors such as color were judged based on individual perception, variability in individual inspection results was also a problem.

[0003] Patent Document 1 discloses a food inspection device for checking whether any ingredients are missing from food products that have multiple ingredients in their packaging, such as bento boxes. The food inspection apparatus of Patent Document 1 includes a standard setting means for pre-setting criteria for each ingredient from image data of a food that serves as a standard for determining whether or not ingredients are missing based on image data of a food having multiple ingredients in a package captured by a color camera, and a determination means for comparing the criteria set by the standard setting means with image data of the food to be inspected to make a determination. The standard setting means includes a color conversion means for converting RGB color information within an inspection area defined for each ingredient from the image data of the standard food into HSL information of hue, saturation, and brightness, a color extraction means for extracting color information from the HSL information, an area measurement means for measuring the area of ​​the ingredient based on the color information extracted from the color extraction means, and an ingredient information storage means for registering the inspection area, color information, and area as ingredient standard information. The determination means includes a color conversion means, a color extraction means, and an area measurement means for obtaining color information and area within an inspection area in the image data of the food to be inspected, and an ingredient omission determination means for comparing the obtained area with the area recorded in the ingredient information storage means to determine whether it is within a pre-set tolerance range. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2014-145639 [Overview of the project] [Problems that the invention aims to solve]

[0005] The food inspection device described in Patent Document 1 can detect the absence of predetermined ingredients placed in fixed positions within a container of a consistent shape, such as a bento box, but it is difficult to use for inspecting bread. Bread does not have a consistent shape after baking; each loaf is different. The degree of fermentation and rising of the bread also varies depending on the temperature and humidity of the day, resulting in variations in size and shape. There are also prepared breads in which various ingredients are mixed into the dough before baking, but the position of the ingredients is not fixed, and even if the ingredients are exposed on the surface, their position and number vary from loaf to loaf, making inspection by machine difficult. Furthermore, the color of the baked bread is similar in color to the undercooked and burnt parts of the dough, making it difficult to distinguish between them.

[0006] Therefore, the present invention aims to provide a food inspection system that can confirm whether predetermined quality standards are met in visual inspection of bread whose shape is not uniform. Furthermore, the present invention aims to provide a bread inspection system that can be easily incorporated into existing production lines. [Means for solving the problem]

[0007] To solve the above problems, the present invention provides a bread inspection system comprising: a conveying means for transporting bread on top; an upper imaging means for imaging the bread from above while it is being transported; an analysis means for acquiring and analyzing the image captured by the imaging means; and a determination means for determining whether the bread is a good product or not. The upper imaging means is a 3D camera, and when the bread is transported to a predetermined position, it images the entire bread from above to acquire an image of the entire bread and thickness information of at least a part of the bread. The analysis means analyzes the size of the bread from the image of the entire bread, acquires the thickness of the bread from the thickness information, and calculates the abnormality rate by comparing the image of the entire bread with a pre-prepared sample of bread shape. The determination means determines that the bread is a good product if the size, thickness, and abnormality rate of the bread all meet the pre-set good product criteria.

[0008] Furthermore, in the bread inspection system described above, the determination means may determine that a product is substandard if at least one of the size, thickness, and abnormality rate of the bread meets a predetermined substandard criterion, and may also determine that a product requires reinspection if at least one of the size, thickness, and abnormality rate of the bread does not meet the good product criterion, and all of the size, thickness, and abnormality rate of the bread do not meet the substandard criterion.

[0009] Furthermore, the bread inspection system described above may include sorting means for sorting the bread being transported by the transport means according to the determination result of the determination means.

[0010] Furthermore, in the bread inspection system described above, the shape sample may be an average shape image created from multiple images of bread that have been determined to be good products.

[0011] Furthermore, in the bread inspection system described above, the analysis means may calculate the major and minor axes of the bread from the overall image of the bread to analyze its size, and may obtain thickness information at one or more positions on the major and minor axes. Alternatively, it may obtain thickness information at the intersection of the major and minor axes and thickness information at the midpoint between the intersection and the ends of the major and minor axes, and use the average of these as the thickness of the bread.

[0012] Furthermore, another bread inspection system of the present invention includes a transport means for transporting bread, an imaging means for imaging the bread while it is being transported, an illumination means for illuminating the bread within the imaging range of the imaging means, an analysis means for acquiring and analyzing the image captured by the imaging means, and a determination means for determining whether the bread is a good product or not, wherein the imaging means captures the entire bread when the bread is transported to a predetermined position to acquire a full-color image of the bread, and the analysis means divides the full-color image of the bread into multiple regions and divides The method generates a hue histogram with the hue of the pixels within the region as the class, a saturation histogram with the saturation as the class, and a lightness histogram with the lightness as the class for each of the divided regions. The method then selects the multiple regions into regions to be judged and regions to be excluded based on the hue histogram, the saturation histogram, and / or the lightness histogram. The method calculates the degree of baking of the bread based on the distribution of the lightness histogram in the regions to be judged. The method determines that the bread is a good product if the degree of baking meets a predetermined standard for good products.

[0013] Furthermore, in the bread inspection system described above, the analysis means may calculate the degree of baking of the bread based on the frequency of the burnt class and / or the frequency of the underbaked class in the brightness histogram of the area to be judged.

[0014] Furthermore, in the bread inspection system described above, the determination means may determine that the bread is substandard if its baking level meets a preset substandard criterion, and may determine that the bread is a product that requires re-inspection if its baking level does not meet the good product criterion and does not meet the substandard criterion.

[0015] Furthermore, the bread inspection system described above may include sorting means for sorting the bread being transported by the transport means according to the determination result of the determination means.

[0016] In addition, another bread inspection system of the present invention includes a conveying means for conveying bread with ingredients exposed on the surface, an imaging means for imaging a color image of the bread during conveyance, an analysis means for acquiring and analyzing the image captured by the imaging means, and a determination means for determining whether the bread is a good product. The imaging means captures an entire image of the bread from above to obtain an entire bread color image when the bread is conveyed to a predetermined position. The analysis means includes a normal ingredient detection model learned using an image of a normal ingredient of the bread and an abnormal ingredient detection model learned using an image of an abnormal ingredient of the bread. The entire bread color image is input into the normal ingredient detection model to count the number of normal ingredients, and the entire bread color image is input into the abnormal ingredient detection model to count the number of abnormal ingredients. The determination means preferably determines that the bread is a good product when the counted number of normal ingredients and the counted number of abnormal ingredients meet a preset good product standard.

[0017] Furthermore, in the above bread inspection system, the image of the abnormal ingredient may be an image of the dough after the ingredient has peeled off from the surface of the bread.

[0018] Furthermore, in the above bread inspection system, reflection means for reflecting the side surface of the bread toward the imaging means may be provided on both sides of the conveying means.

[0019] Furthermore, in the above bread inspection system, lighting means for illuminating the bread within the imaging range of the imaging means may be provided.

[0020] Furthermore, in the above bread inspection system, the determination means may determine that the product is out of specification when at least one of the counted number of normal ingredients and the counted number of abnormal ingredients meets a preset out-of-specification standard, and may determine that the product is a re-inspection product when one of the counted number of normal ingredients and the counted number of abnormal ingredients does not meet the good product standard and the other does not meet the out-of-specification standard.

[0021] Furthermore, in the above bread inspection system, a sorting means for sorting the conveyance destination of the bread according to the determination result of the determination means for the bread being conveyed by the conveyance means may be included.

[0022] Also, another bread inspection system of the present invention includes a conveyance means for conveying bread, an imaging means for imaging a color image of the back surface of the bread being conveyed by the conveyance means, an analysis means for acquiring and analyzing the image captured by the imaging means, and a determination means for determining whether the bread is a good product. The imaging means captures the back surface of the bread to obtain a bread back surface color image. The analysis means includes a back surface detection model learned using the image of the back surface of the bread, inputs the bread back surface color image into the back surface detection model to detect abnormalities on the back surface, and the determination means determines that it is a good product when the detected abnormality meets a preset good product standard.

Effect of the Invention

[0023] In the bread inspection system of the present invention, for the appearance inspection of bread with an irregular shape, specifically, one or more of a molding inspection for inspecting the size and shape, a baking color inspection for inspecting the baking color of the surface, a topping inspection for inspecting the toppings exposed on the surface of the bread, and a back surface inspection for inspecting burnt scraps and foreign substances on the back surface can be automatically inspected by a machine based on the image acquired by the image acquisition unit. Therefore, not only can it replace the human inspection work, but it can also achieve the sophistication of inspection and the improvement of quality based on data. Other effects will be described in the embodiments.

Brief Description of the Drawings

[0024] [Figure 1] Schematic diagram of the basic configuration common to the bread inspection system of the present invention [Figure 2] Block diagram of the bread inspection system used in the molding inspection [Figure 3] (a) is an example of an image of the imaging range of the upper imaging means, (b) is an image of the cut-out bread, (c) is a diagram showing an example of the analysis result by the image analysis unit, (d) is a black-and-white image with the bread as white, and (e) is an average shape image of the bread [Figure 4] An example flowchart for a process that detects the timing of image acquisition using a 3D camera and obtains an image of bread. [Figure 5] An example flowchart for analyzing the size of bread [Figure 6] An example flowchart for analyzing the shape of bread. [Figure 7] Example of judgment criteria [Figure 8] Block diagram of the bread inspection system used for browning inspection. [Figure 9] (a) is a diagram showing the entire color image of the bread divided into multiple regions, and (b) to (e) are the hue histogram (left), saturation histogram (center), and lightness histogram (right) for each region shown in (a) b to e. [Figure 10] An example flowchart for the process of checking browning. [Figure 11] Block diagram of the bread inspection system used for ingredient inspection. [Figure 12] (a) is an example of an image of the imaging range of the imaging means, (b) is an image of a sliced ​​piece of bread. [Figure 13] An example flowchart for the process of analyzing ingredients [Figure 14] Block diagram of the bread inspection system used for back-side inspection. [Figure 15] (a) is a diagram showing the underside of the bread, (b) to (d) are images of the imaging range of the imaging device. [Figure 16] Diagram showing an example of a bread inspection system. [Modes for carrying out the invention]

[0025] Figure 1 is a schematic diagram of the basic configuration common to the bread inspection system 10 of the present invention. The bread inspection system 10 includes a transport means 12 for transporting bread 11, an image acquisition unit 13 including an imaging means for capturing images of the bread's appearance, and an information processing device 14 that functions as an analysis means and a determination means for the images captured by the imaging means of the image acquisition unit 13. Furthermore, the bread inspection system 10 may also include sorting means 15a, 15b for sorting bread based on the determination result, and other sorting destinations 16a, 16b. In the bread inspection system 10 of the present invention, the appearance of the bread is mainly inspected based on the images acquired by the image acquisition unit 13. Specifically, there are four types of inspections: a molding inspection to check the size and shape, a browning inspection to check the browning of the surface, a filling inspection to check the fillings exposed on the surface of the bread, and a backside inspection to check for burnt residue and foreign matter on the backside. The bread inspection system 10 may perform all four inspections, perform one or more of them, or perform them in combination with other inspections. These four inspections may be performed by one image acquisition unit 13 or by multiple image acquisition units 13. Furthermore, multiple inspections may be performed from images captured by a single imaging means, or one or more imaging means may be provided for each inspection. The configuration and function of each inspection will be described later, but they can be used as independent inspection systems or as an inspection system capable of performing several inspections simultaneously.

[0026] The bread 11 is the object of inspection, and the required inspection is selected depending on the type of bread. Preferably, the surface of the bread 11 to be inspected is exposed, and to improve the accuracy of the inspection, it is preferable that the bread is not covered, even with a transparent sheet. The surface of the bread 11 is topped with ingredients 11a. The conveying means 12 is a means for conveying the bread, and depending on the inspection, the bread may be conveyed with the same orientation (up and down or left and right), or it may be conveyed in a line. As the conveying means 12, for example, a belt conveyor or roller conveyor may be used to convey a large quantity of bread freely or in an orderly manner, or each loaf of bread may be placed in a container such as a plate and conveyed. The image acquisition unit 13 is the part that captures images of the bread during transport using imaging means, and one or more imaging means are arranged therein. The timing of imaging with the imaging means can be determined by installing a sensor that detects when the bread reaches the imaging position and imaging the bread based on the sensor's detection result. However, if the imaging means is a 3D camera, by measuring the distance information of the inspection line that is set in advance within the imaging range with the 3D camera, the distance information changes when the bread reaches the inspection line (it becomes shorter than the distance to the transport surface by the thickness of the bread), so it is possible to detect when the bread has reached the inspection line and image the bread that has reached the inspection line.

[0027] The information processing device 14 analyzes and judges images captured by the imaging means of the image acquisition unit 13. It includes a CPU that performs calculations to realize various control functions, volatile RAM (Random Access Memory: e.g., SRAM, DRAM, etc.) which serves as a work area for calculations, non-volatile storage (e.g., hard disk, flash memory, etc.) for storing various data, and a display device. The information processing device 14 reads programs pre-stored in the storage into the RAM, and the CPU performs calculations according to the program to execute various processes (e.g., image analysis, image judgment), and the acquired images, analysis results, and judgment results can be displayed on the display device. Furthermore, it is preferable that the information processing device 14 controls sorting means 15a and 15b based on the judgment results to sort the bread to be transported. For example, the judgment means may determine that a product is good if all judgment elements meet the good product criteria, determine that a product is substandard if one or more judgment elements meet the substandard criteria, and determine that a re-inspection is required in all other cases.

[0028] The sorting means 15a and 15b are means for sorting the destination of the bread by changing the direction of travel of the conveyed bread. They are not particularly limited as long as they can sort the conveyed bread, and for example, a diverter system, a slide shoe system, a cross belt system, a pusher system, a tilt tray system, etc., can be used. In Figure 1, two guide plates (diverters) installed on the conveying means allow the user to choose whether to move the guide plates away from the bread's path to allow the bread to travel straight, slide the guide plate of sorting means 15a to position it diagonally on the bread's path and transport it to sorting destination 16a, or slide the guide plate of sorting means 15b to position it diagonally on the bread's path and transport it to sorting destination 16b. Thus, in Figure 1, the bread is sorted into three destinations: straight, sorting destination 16a, and sorting destination 16b. However, it is preferable to have as many sorting destinations as there are judgment results from the judgment means. It is preferable to have at least two sorting destinations, one for good products and one for substandard products, in which case one sorting means is sufficient. When the determination means classifies products into three stages, such as good products, substandard products, and products requiring re-inspection, it is preferable to have three sorting destinations: straight ahead, sorting destination 16a, and sorting destination 16b, as shown in Figure 1. At the destination, good bread is packaged, substandard bread is discarded, and bread requiring re-inspection is re-inspected by a person or machine to determine whether it is good or substandard.

[0029] [Molding inspection] Figure 2 is a block diagram of the bread inspection system 20 used for molding inspection, showing a plan view and a cross-sectional view above and below. The bread inspection system 20 includes a transport means 22 for transporting the bread 21 on top, an upper imaging means 23 for imaging the bread 21 from above while it is being transported, an analysis means 24a for acquiring and analyzing the image captured by the upper imaging means 23 within an information processing device 14, and a determination means 24b for determining whether the bread 21 is a good product or not. Furthermore, the bread inspection system 20 may also include sorting means 15a, 15b for sorting the bread based on the determination result, and other sorting destinations 16a, 16b. The bread inspection system 20 can inspect the size and shape of the bread 21, but it may also inspect only the size or only the shape. There are no particular restrictions on the bread 21 that is inspected in the molding inspection of the bread inspection system 20, and any bread that needs to be inspected for size and / or shape can be inspected.

[0030] The conveying means 22 carries the bread 21 on top of it, and there are no particular restrictions as long as the entire bread 21 can be transported in a way that allows imaging from above. At the imaging position, it is sufficient that the space between the top of the bread 21 and the imaging means 23 is open for imaging. The conveying means 22 may be, for example, a belt conveyor or a roller conveyor that transports a large quantity of bread freely or in an orderly manner, or each loaf of bread may be transported on a plate or other container. In order to determine the exact size and shape, it is preferable to transport the bread so that it passes directly below the upper imaging means 23 in a single line without overlapping.

[0031] The upper imaging means 23 is a 3D camera capable of acquiring video information and distance information (depth) to each pixel. When bread is transported to a predetermined position, it images the entire bread from above to acquire an image of the entire bread and thickness information of at least a portion of the bread. To accurately determine the size and shape, it is preferable to position the upper imaging means 23 directly above the bread so that its optical axis is perpendicular to the transport surface of the transport means. The image of the entire bread may be a color image or a black and white image. The timing of imaging the bread can be determined by installing a sensor to detect when the bread reaches the imaging position and imaging the bread based on the sensor's detection result. However, since the upper imaging means 23 is a 3D camera, the distance information of an inspection line set in advance within the imaging range can be measured with the 3D camera. By detecting the change in distance information when the bread reaches the inspection line, it is possible to detect when the bread has reached the inspection line and image the bread that has reached the inspection line. The upper imaging means 23 can acquire distance information to each pixel, but since measuring and recording distance information for all pixels would increase the processing load and data volume, it is preferable to acquire distance information (bread thickness information) only partially.

[0032] Figure 3(a) is an example of an image of the imaging range of the upper imaging means 23, with the inspection line 23a set within the imaging range. The dotted rectangle represents the cropping range 23b for cropping the image of the bread to be inspected. Figure 3(b) is an image of the cropped bread. Figures 3(a) and (b) are whole-bread images in which the entire bread is captured. Figure 3(c) is an example of the results of analysis by the image analysis unit, showing the coordinates of the major axis 21a and minor axis 21b of the cropped bread, the coordinates of its ends, and the coordinates of positions C1 to C5 for acquiring thickness information. Figure 3(d) is a black and white image with the cropped bread in white, and Figure 3(e) is an average shape image 27 of the bread.

[0033] Figure 4 is an example of a flowchart for the process of detecting the imaging timing by a 3D camera and acquiring an image of bread. As shown in Figures 3(a), 3(b), and 4, the information processing device 14 acquires video of the imaging range of the upper imaging means 23 (S41), acquires distance information of the inspection line 23a measured by the upper imaging means 23 (S42), and determines whether there is a change in the distance information (S43). If there is no change (No in S43), it returns to acquiring distance information again (S42). If there is a change (Yes in S43), it acquires an image for inspection (S44). After that, if necessary, a trimming range 23b (see Figure 3(a)) may be set to cut out the image of the bread to be inspected, and the image of the bread to be inspected may be cut out (S45) (see Figure 3(b)). For example, if other bread is included in the image, the process of cutting out the image of the bread to be inspected may be performed to remove the bread other than the bread to be inspected. Here, the bread to be inspected may be, for example, a bread that has reached inspection line 23a, or a bread located in the center of the image. Subsequently, if necessary, the cropped image may be processed, for example, by converting it to a black and white image, or by separating the bread from the background to obtain an image of only the bread (S46). For the molding inspection, the image may remain in color without any particular image processing, but in order to clarify the outline of the bread and make it easier to compare with the shape sample, the color of the bread and the conveying surface may be changed and converted to a black and white image (see Figure 3(d)).

[0034] The analysis means 24a analyzes the size and / or shape of the bread 21 from the image of the bread captured by the upper imaging means 23 by having the CPU of the information processing device 14 execute a program stored in storage. The method for calculating the size of the bread 21 varies depending on the shape of the bread, but for example, in the case of a roughly elliptical loaf of bread, the size may be calculated using the length of the major axis and minor axis and the thickness of the bread 21, or it may be calculated using the area and thickness of the bread, or it may be done by any other method. It is preferable to set the position of the diameter to be measured and the position from which to acquire thickness information based on the shape of a good loaf of bread to be inspected. Regarding thickness information, it can be calculated from the distance information of the 3D camera, but as mentioned above, if distance information is measured and recorded at all pixels, the amount of processing and data will increase, so it is preferable to limit the locations where thickness is measured or recorded. The thickness information obtained from one location may be used as the thickness of the bread, the thickness distribution may be calculated from the thickness information from multiple locations, or the average of the thickness information from multiple locations may be used as the thickness of the bread. Alternatively, the thickness of the bread may be calculated from thickness information obtained from one or more positions (e.g., intersections) on the major and minor axes, or distance information may be measured at all pixels and thickness information at the position where the thickness is maximum may be obtained (recorded). In Figure 3(c), thickness information at the intersection point C1 of the major axis 21a and the minor axis 21b, and thickness information at the midpoints C2 to C5 between the intersection point C1 and the ends of the major and minor axes are obtained, and the average of these is calculated as the thickness of the bread. In the case of a special shape, such as a cornet bread with a long, thin piece of dough wrapped around a cone, the bottom side of the cone is naturally thicker, so the length of the base and height of the triangle when the cone is viewed from above may be measured, and the thickness at the midpoint of the base may be measured. Since the shapes of the individual breads 21 are different, the difference in shape when comparing the overall bread image with a prepared shape sample can be quantified as an abnormality rate, and the quality of the shape of the bread 21 can be determined based on that value. The abnormality rate is the degree of dissimilarity (degree of difference) in shape. If the abnormality rate is small, the shapes are similar, and if the abnormality rate increases, the shape will be different from the shape sample. For example, taking the case of exact match as 0, the abnormality rate may be increased as the difference becomes larger, or only a certain range may be extracted and normalized to 0 to 1 (0% to 100%) for output. As the shape sample, for example, an image of the average shape (average shape image 27) may be created from a large number of bread images determined to be good products and used (see Fig. 3(e)). Also, an ideal shape (ideal image) may be used as the shape sample. In the analysis means 24a, the analysis method for comparing shapes is not particularly limited. For example, the overall bread image may be rotated and compared using chamfer distance or Hausdorff distance that compares the coordinate differences of the bread contours, or the overall bread image may be rotated and / or enlarged or reduced by Procrustes analysis to compare the shape with the shape sample (average shape image or ideal shape).

[0035] FIG. 5 is an example of a flowchart of a process for analyzing the size of bread. As shown in FIGS. 3(c) and 5, the information processing device 14 acquires an overall bread image (S51), and from the overall bread image, the coordinates A1(x<> A1 , y<> A1 ), A2(x<> A2 , y<> A2 ) of both ends of the major axis 21a of the bread, the coordinates B1(x<> B1 , y<> B1 ) and B2(x<> B2 , y<> B2 ) of both ends of the minor axis 21b, and the coordinates C1(x<> C1 , y<> C1 ) of the intersection of the major axis and the minor axis, the midpoints C2(x<> C2 , y<> C2 ), C3(x<> C3 , y<> C3 ), C4(x<> C4 , y<> C4 ), C5(x<> C5 , y<> C5The size is calculated (S52). The actual length corresponding to one pixel on the image is calculated in advance (S53), the distance on the image between the coordinates of the major axis and minor axis is calculated, and the actual lengths of the major axis and minor axis are calculated from the actual length corresponding to one pixel (S54). Furthermore, thickness information (distance information to the transport surface - distance information at the time of imaging) at the intersection coordinate C1 and the midpoint coordinates C2~5 is obtained from the upper imaging means 23 (3D camera), and the average thickness is calculated (S55). This size information (actual lengths of the major axis and minor axis and thickness information at coordinate C) is input to the determination means 24b. As described above, when measuring the size of the bread using other standards, the size and thickness are obtained according to the standards.

[0036] Figure 6 is an example of a flowchart for the process of analyzing the shape of bread. As shown in Figure 6, the information processing device 14 acquires an image of the entire bread (S61). Here, a black and white image as shown in Figure 3(d) was used to eliminate noise from ingredients and clarify the outline of the bread, making it easier to compare with a shape sample. However, it may be processed into an edge image with further emphasis on the outline, or a color image is also acceptable as long as the shape of the bread is clear. Next, a pre-prepared shape sample is acquired, for example, by reading it from storage (S62), and the entire bread image and the shape sample are compared using, for example, chamfer distance or Procrustes analysis (S63), and the anomaly rate is calculated (S64). The calculated anomaly rate is input to the determination means 24b.

[0037] The determination means 24b determines that a loaf of bread is good if its size, thickness, and abnormality rate all meet the pre-set good product standards. If the inspection is only for the size of the bread, it determines that a loaf is good if both the size and thickness meet the pre-set good product standards. If the inspection is only for the shape of the bread, it determines that a loaf is good if the abnormality rate meets the pre-set good product standards. The good product standards are set appropriately depending on the type of bread, but it is preferable that they have a certain range. The determination means 24b may also determine that a loaf is substandard if at least one of the size, thickness, and abnormality rate of the bread falls outside the good product standards. Alternatively, it may set a substandard standard different from the good product standard, and determine that a loaf is substandard if it meets the substandard standard. In this case, the loaf may fall outside the good product standards and also not meet the substandard standard. In this case, for example, the loaf may be re-inspected and judged again by a human or machine. The determination means may determine that a loaf is good if all inspection results meet the good product standards, determine that a loaf is substandard if one or more inspection results meet the substandard standard, and determine that a re-inspection is necessary in all other cases.

[0038] Figure 7 shows an example of the judgment criteria. A product is judged as good if it falls within the upper and lower limits of the good product standard. The non-standard standard is set with a range from the upper and lower limits of the good product standard. For the long diameter, short diameter, and thickness of the bread, a predetermined range of lengths is set, and the percentage of shape abnormality is set within a predetermined range. For example, for the long diameter of the bread, 150mm to 170mm is considered good, and if the long diameter is less than 145mm or more than 200mm, it is considered non-standard. If the long diameter is between 145mm and 150mm or between 170mm and 200mm, it is judged to require re-inspection.

[0039] The information processing device 14 may control the sorting means 15a and 15b to change the destination of the bread based on the judgment result. For example, good products may be sent in a straight line, products requiring re-inspection may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15a and transported to sorting destination 16a, and off-spec products may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15b and transported to sorting destination 16b.

[0040] [Grilling inspection] Figure 8 is a block diagram of the bread inspection system 30 used for browning inspection, with a plan view and a cross-sectional view shown above and below. The bread inspection system 30 includes a transport means 32 for transporting bread 31, an imaging means 33 for imaging the bread 31 while it is being transported, an illumination means 37 for illuminating the bread within the imaging range of the imaging means 33, and an analysis means 34a within the information processing device 14 for acquiring and analyzing the image captured by the imaging means, and a determination means 34b for determining whether the bread is a good product or not. Furthermore, the bread inspection system 30 may also include sorting means 15a, 15b for sorting the bread based on the determination result, and other sorting destinations 16a, 16b. The bread inspection system 30 can identify the browning of the bread 31 and inspect the degree of baking (normal, undercooked, burnt). There are no particular restrictions on the bread 31 that can be inspected in the browning inspection of the bread inspection system 30; any bread that needs to be inspected for browning can be inspected.

[0041] The conveying means 32 is for conveying the bread 31, and there are no particular restrictions as long as the entire bread 31 can be imaged while being conveyed. It is sufficient that the space between the bread 31 and the imaging means 33 is open enough to allow image capture at the imaging position. As the conveying means 32, for example, a belt conveyor or roller conveyor may be used to convey a large quantity of bread freely or in an orderly manner, or each loaf of bread may be placed in a container such as a plate and conveyed. Brightness is important for determining the browning of the bread, and it is preferable to convey the bread so that it is not shaded by other loaves of bread.

[0042] The imaging means 33 is a camera capable of acquiring color images, and when the bread 31 is transported to a predetermined position, it images the entire bread to acquire a color image of the entire bread. The imaging means 33 may be a 2D camera or a 3D camera. In order to determine the browning of the bread, it is preferable for the imaging means 33 to image the entire top surface from above, but it may also be possible to image from the side and determine the browning from the browning of one side, or both sides may be imaged by two imaging means. The timing of imaging the bread may be determined by installing a sensor that detects when the bread reaches the imaging position and imaging the bread based on the sensor's detection result, or if the imaging means 33 is a 3D camera, imaging may be started at the timing when the distance changes, as shown in S42 to S44 in Figure 4.

[0043] The illumination means 37 illuminates the bread 31 within the imaging range of the imaging means 33, and may consist of one light source or multiple lights. For example, it may be uniformly illuminated from all sides by an annular ring light or multiple spotlights, or multiple rod-shaped lights may be arranged in a line according to the size of the bread. In Figure 8, an annular ring light is used to surround the imaging means 33, and it is configured so that the bread to be inspected is illuminated uniformly from all directions. The browning of the bread is difficult to distinguish from the dough or burnt parts, so the brightness at the time of shooting is important. Although the illumination means 37 is provided to maintain a constant brightness, it is preferable to perform color correction when the device is started up or before inspection in order to compensate for the influence of ambient light and color deviations due to individual differences in camera sensors. Color correction is performed by starting the system with a dedicated color sample set up, which creates a setting file that records the correction information, and then the color correction is performed according to the recorded correction information.

[0044] The analysis means 34a analyzes the degree of baking of the bread in a full-color image of the bread by having the CPU of the information processing device 14 execute a program stored in storage. The analysis means 34a divides the full-color image of the bread into multiple regions, and generates a hue histogram with the hue of the pixels in the region as the class, a saturation histogram with the saturation as the class, and a brightness histogram with the brightness as the class for each divided region. Based on the generated histograms, the multiple regions are selected as regions to be judged and regions to be excluded, and the degree of baking of the bread in each region is analyzed based on the distribution of the brightness histogram of the regions to be judged. The size and shape of the regions to be divided are set appropriately according to the type of bread. If the bread has fillings, it is preferable that the size of the region be such that the fillings can be contained within it. The shape of the region is not particularly limited and may be a triangle, square, pentagon, or hexagon. It is preferable that each region be the same size. In addition, before dividing, image processing may be performed to separate the bread from the background to obtain an image of only the bread. Hue, saturation, and lightness are the three attributes of color. Hue indicates the difference in color (wavelength of monochromatic light in the spectrum), saturation indicates the degree of vividness of the color (the degree to which white, gray, and black are mixed), and lightness indicates the brightness of the color. A hue histogram allows us to understand the distribution of colors present within that region, and a saturation histogram allows us to understand the distribution of vividness within that region. These can be used to distinguish between bread and other materials, and the bread region may be selected as the region to be judged. In addition to the hue histogram and saturation histogram, a lightness histogram may also be used to select the region to be judged and the region to be excluded. Then, the lightness histogram of the region to be judged is used to determine whether the region has a normal browning, or whether it contains burnt or undercooked areas. A lightness histogram shows the distribution of brightness of pixels present within that region. Browning will be brighter in the case of undercooked items compared to normal items, and darker in the case of burnt items compared to normal items. Therefore, by defining the range of the lighter classes in the brightness histogram as the undercooked class and the range of the darker classes as the burnt class, the frequencies included in the undercooked class and / or burnt class can be calculated to analyze whether the food is undercooked or burnt.

[0045] Figure 9(a) shows a color image of the entire bread divided into multiple regions, and Figures 9(b) to (e) show the hue histogram (left), saturation histogram (center), and lightness histogram (right) for each region shown in Figures 9(a) b to e. Figure 9(b) is the histogram of the region of the conveyor, which is the transport means, Figure 9(c) is the histogram of the region containing the ingredients, Figure 9(d) is the histogram of the region containing the burnt parts, and Figure 9(e) is the histogram of the region with normal browning. First, comparing it with the region with normal browning in Figure 9(e), the hue histogram (left) of the conveyor region in Figure 9(b) is clearly different, so the conveyor region can be excluded. In the region containing the ingredients in Figure 9(c), it is difficult to distinguish based on the hue histogram alone, but the saturation histogram is distributed throughout, so it can be distinguished from the region with normal browning in Figure 9(e), and can be excluded. The hue histogram (left) and saturation histogram (center) in Figures 9(d) and 9(e) are similar and both are characteristic histograms of bread dough, and are considered as areas for determination. The lightness histogram (right) in Figure 9(d) has a darker class compared to the lightness histogram (right) in Figure 9(e). The area in Figure 9(d) is determined to be an area containing burnt parts, while the area in Figure 9(e) is determined to be an area with normal browning.

[0046] Figure 10 is an example of a flowchart for the process of inspecting browning. As shown in Figures 9 and 10, the information processing device 14 acquires a full-color image of the bread (S101). The full-color image of the bread may be the image as captured by the imaging means 42, or it may be an image of only the bread to be inspected, as shown in S45 of the flowchart in Figure 4. Next, the full-color image of the bread is divided into multiple regions (S102) (see Figure 9(a)), and a hue histogram, saturation histogram, and brightness histogram are generated for each divided region (S103) (see Figures 9(b) to (e)). Then, based on the generated histograms, the multiple regions are sorted into regions to be judged and regions not to be judged (S104), and it is determined whether the frequency of the browning class in the brightness histogram of the regions to be judged is higher or lower than a predetermined set value (S105). If the frequency of the burnt category is higher than a predetermined setting value (Yes in S105), the area is determined to contain burnt areas (S106), and the process is terminated. If the frequency of the burnt category is lower than a predetermined setting value (No in S105), it is determined whether the frequency of the undercooked category in the brightness histogram of the area to be judged is higher or lower than a predetermined setting value (S107). If the frequency of the undercooked category is higher than a predetermined setting value (Yes in S107), the area is determined to contain undercooked areas (S108), and the process is terminated. If the frequency of the undercooked category is lower than a predetermined setting value (No in S107), the area to be judged is determined to be a normally browned area (S109), and the process is terminated.

[0047] The determination means 34b determines that the bread is a good product if its degree of baking meets a predetermined standard for good products. The standard for good products is set appropriately depending on the type of bread, but it is preferable that it has a certain range. The determination means may determine that the bread is a substandard product if its degree of baking falls outside the standard for good products, or it may set a substandard standard that is different from the numerical value of the standard for good products, and determine that the bread is a substandard product if it meets the substandard standard. In this case, the bread may fall outside the standard for good products and also not meet the substandard standard, in which case it may be re-inspected and judged again by a human or machine, for example. The determination means 34b may determine the product by the burnt area ratio (= number of burnt areas / number of areas to be judged × 100) and the undercooked area ratio (= number of undercooked areas / number of areas to be judged × 100), or by the number of burnt areas and undercooked areas. In the criteria shown in Figure 7, the standard for a good product is when the burnt area percentage is 0-0.5% and the undercooked area percentage is 0-0.5%, while the standard for a non-standard product is 0.5% or higher for both. Therefore, there is no determination of a product requiring re-inspection. However, if the standard for a non-standard product is set to 5% or higher, then products with a burnt area percentage and / or undercooked area percentage of 0.5-5% will require re-inspection.

[0048] The information processing device 14 may control the sorting means 15a and 15b to change the destination of the bread based on the judgment result. For example, good products may be sent in a straight line, products requiring re-inspection may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15a and transported to sorting destination 16a, and off-spec products may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15b and transported to sorting destination 16b.

[0049] [Ingredient inspection] Figure 11 is a block diagram of a bread inspection system 40 used for ingredient inspection, with a plan view and a cross-sectional view shown above and below. The bread inspection system 40 includes a transport means 42 for transporting bread 41 with ingredients 41a exposed on its surface, an imaging means 43 for capturing a color image of the bread 41 during transport, an analysis means 44a within an information processing device 14 for acquiring and analyzing the image captured by the imaging means 43, and a determination means 44b for determining whether the bread 41 is a good product or not. The bread inspection system 40 may also include an illumination means 47 for illuminating the bread within the imaging range of the imaging means 43, and a reflection means 48 on both sides of the transport means 42 for reflecting the sides of the bread 41 toward the imaging means 43. Furthermore, the bread inspection system 40 may include sorting means 15a, 15b for sorting the bread based on the determination result, and other sorting destinations 16a, 16b. The bread inspection system 40 can inspect the ingredients 41a exposed on the surface of the bread 41. The bread 41 inspected by the ingredients inspection of the bread inspection system 40 is bread in which the ingredients 41a are exposed on the surface. There are no particular restrictions on the ingredients 41a, but they can be, for example, beans, fruits, vegetables, meat, seafood, etc. The ingredients may be mixed with the dough and baked, or the ingredients may be placed on the surface after the bread has been baked.

[0050] The conveying means 42 is for conveying the bread 41, and there are no particular restrictions as long as it is possible to image the surface of the bread 41 with its fillings exposed during conveyance. It is sufficient that the space between the bread 41 and the imaging means 43 is open at the imaging position so that imaging can be performed. As the conveying means 42, for example, a belt conveyor or roller conveyor may be used to convey a large quantity of bread freely or in an orderly manner, or each loaf of bread may be placed in a container such as a plate and conveyed. Brightness is important for determining the fillings of the bread, and it is preferable to convey the bread so that it is not obscured by other loaves of bread.

[0051] The imaging means 43 is a camera capable of acquiring color images, and when the bread 41 is transported to a predetermined position, it images the entire bread to acquire a full-color image of the bread. The imaging means 43 may be a 2D camera or a 3D camera. The imaging means 43 preferably images the entire top surface from above in order to determine the ingredients of the bread, but both sides may be imaged by multiple imaging means. In addition, by installing reflective means 48 on both sides of the transport means 42, the top surface and both sides can be imaged at the same time. The timing of imaging the bread can be determined by installing a sensor to detect when the bread reaches the imaging position and imaging the bread based on the sensor's detection result, or, if the imaging means 43 is a 3D camera, imaging may be started at the timing when the distance changes, as shown in S42 to S44 in Figure 4.

[0052] The illumination means 47 illuminates the bread 41 within the imaging range of the imaging means 43, and may consist of one light source or multiple lights. For example, it may be uniformly illuminated from all sides by an annular ring light or multiple spotlights, or multiple rod-shaped lights may be arranged in a line according to the size of the bread. In Figure 11, an annular ring light is used to surround the imaging means 43, and it is configured so that the bread to be inspected is illuminated uniformly from all directions. The ingredients of the bread may be embedded in the dough, so the brightness at the time of shooting is important. Although the illumination means 47 is provided to maintain a constant brightness, it is preferable to perform color correction when the device is started or before inspection in order to compensate for the influence of ambient light and color deviations due to individual differences in camera sensors. Color correction is performed by starting the system with a dedicated color sample set up, which creates a setting file that records the correction information, and then the color correction is performed according to the recorded correction information.

[0053] The reflective means 48 is positioned diagonally on both sides of the transport means 42 within the imaging range of the imaging means 43, and reflects the sides of the bread 41 toward the imaging means 43. With the reflective means 48, the imaging means 43 can image the top surface and both sides at once, allowing for more accurate identification of the ingredients. However, if only the top surface is to be inspected, or if imaging is performed using multiple imaging means, the reflective means may not be necessary.

[0054] The analysis means 44a analyzes the ingredients of a whole-bread color image by having the CPU of the information processing device 14 execute a program stored in storage. The analysis means 44a includes a normal ingredient detection model 441 trained using images of normal ingredients of bread, and an abnormal ingredient detection model 442 trained using images of abnormal ingredients of bread. The whole-bread color image is input to the normal ingredient detection model 441 to count the number of normal ingredients, and the whole-bread color image is input to the abnormal ingredient detection model 442 to count the number of abnormal ingredients. Images of abnormal ingredients include, for example, images of dough after the ingredients have peeled off the surface of the bread, or images of damaged ingredients. Although the normal ingredient detection model 441 and the abnormal ingredient detection model 442 are described separately due to their functional differences, the ingredient detection model may also be trained using both images of normal and abnormal ingredients. If the whole-bread color image is input to the ingredient detection model, both normal and abnormal ingredients can be detected and counted.

[0055] Figure 12(a) is an example of an image of the imaging range of the imaging means 43, showing the top surface of the bread 41 in the center, and the sides of the bread 41 reflected by the reflection means 48 above and below it. The dotted rectangles in Figure 12(a) are the cropping ranges 43a, 43b, and 43c for cropping the image of the bread to be inspected, cropping not only the image of the top surface in the center but also the sides of the bread 41 reflected by the reflection means 48. Figure 12(b) is an image of the cropped bread, where ingredients detected by the normal ingredient detection model 441 are enclosed in rectangular frames 441a to 441i, and ingredients detected by the abnormal ingredient detection model 442 are enclosed in a rectangular frame 442a.

[0056] Figure 13 is an example of a flowchart for the process of analyzing the ingredients. As shown in Figures 12(b) and 13, the information processing device 14 acquires a color image of the entire bread (S131), inputs the color image of the entire bread to the normal ingredient detection model, and counts the number of normal ingredients (S132). As shown in Figure 12(b), the normal ingredient detection model detects four normal ingredients in frames 441a to 441d on the top surface and five normal ingredients in frames 441e to 441i on the sides. Since the ingredients in frames 441a on the top surface and 441e on the side, the ingredients in frames 441b and 441f, the ingredients in frames 441c and 441h, and the ingredients in frames 441d and 441i are the same, five normal ingredients were detected in the bread 41. Note that if the presence or absence of ingredients is important and the number is not, it may be permitted to count them twice. Furthermore, a color image of the entire bread is input to the abnormal ingredient detection model to count the number of abnormal ingredients (S133). As shown in Figure 12(b), the abnormal ingredient detection model detected one abnormal ingredient in the side frame 442a.

[0057] The determination means 44b determines that the product is good if the number of normal ingredients and the number of abnormal ingredients meet the pre-set good product criteria. The good product criteria are set appropriately depending on the type of bread, but it is preferable that they have a certain range. The determination means 44b may also determine that the product is substandard if the counted number of normal ingredients and the number of abnormal ingredients fall outside the good product criteria, or it may set a different substandard criterion than the good product criterion, and determine that the product is substandard if it meets the substandard criterion. In this case, the product may fall outside the good product criteria and also not meet the substandard criterion, in which case it may be re-inspected and judged again by a human or machine. In the determination criteria of Figure 7, the good product criterion for normal ingredients is 3 or more, the substandard criterion is 0, and if there are 1 or 2, a re-inspection is required. In the determination criteria of Figure 7, the good product criterion for abnormal ingredients is 0, the substandard criterion is 2 or more, and if there is 1, a re-inspection is required.

[0058] The information processing device 14 may control the sorting means 15a and 15b to change the destination of the bread based on the judgment result. For example, good products may be sent in a straight line, products requiring re-inspection may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15a and transported to sorting destination 16a, and off-spec products may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15b and transported to sorting destination 16b.

[0059] [Inspection of the reverse side] Figure 14 is an example of a block diagram of a bread inspection system 50 used for backside inspection, with a plan view and a cross-sectional view shown above and below. The bread inspection system 50 includes transport means 52a and 52b for transporting bread 51, an imaging means 53 for capturing a color image of the backside of the bread 51 while it is being transported, an analysis means 54a for acquiring and analyzing the image captured by the imaging means 53 within the information processing device 14, and a determination means 54b for determining whether the bread 51 is a good product or not. The bread inspection system 50 may also include an illumination means 57 for illuminating the bread within the imaging range of the imaging means 53. Furthermore, the bread inspection system 50 may include sorting means 15a and 15b for sorting the bread based on the determination result, and other sorting destinations 16a and 16b. The bread inspection system 50 is capable of inspecting the backside of the bread 51. The bread 51 inspected by the back-side inspection of the bread inspection system 50 has no particular restrictions and can be inspected for any bread that requires inspection of the back side.

[0060] The transport means 52a and 52b transport the bread 51, and there are no particular restrictions as long as the transport means can transport the bread 41 in a way that allows imaging of the underside of the bread during transport. In Figure 14, a gap is provided between the upstream transport means 52a and the downstream transport means 52b, and the imaging means 53 is positioned upward below this gap. By positioning it in this way, the underside of the bread can be inspected without changing its orientation. If the bread is transported with the underside facing upwards, the imaging means can also be placed on top, as in other inspection systems, but in the case of bread with toppings on top, there is a possibility that the toppings may fall, so it is preferable to image from below as in Figure 14. A transparent platform may also be placed in the gap.

[0061] The imaging means 53 is a camera capable of acquiring color images, and when the bread 51 is transported to a predetermined position, it continuously images the underside of the bread to acquire a color image of the underside of the bread. The imaging means 53 may be a 2D camera or a 3D camera. The timing of imaging the bread can be determined by installing a sensor to detect when the bread reaches the imaging position and imaging the bread based on the sensor's detection result, or, if the imaging means 53 is a 3D camera, imaging may be started at the timing when the distance changes, as shown in S42 to S44 in Figure 4.

[0062] The illumination means 57 illuminates the pan 51 within the imaging range of the imaging means 53, and may consist of one light source or multiple lights. For example, it may be uniformly illuminated from all sides by an annular ring light or multiple spotlights, or multiple rod-shaped lights may be arranged in a line according to the size of the pan. In Figure 14, an annular ring light is used to surround the imaging means 53, and is configured to illuminate the back surface of the pan to be inspected. Since the back surface of the pan is prone to shadows, it is preferable to provide the illumination means 57 when the imaging means is positioned below it. In addition, it is preferable to perform color correction when the device is started or before inspection in order to compensate for the influence of ambient light and color shifts due to individual differences in camera sensors. Color correction is performed by starting the system with a dedicated color sample set up, which creates a setting file containing correction information, and then performs color correction according to the recorded correction information.

[0063] The analysis means 54a analyzes the underside of the bread in a color image of the underside of the bread by having the CPU of the information processing device 14 execute a program stored in storage. The analysis means 54a includes an underside detection model 541 that has been trained using images of the underside of the bread, and inputs the color image of the underside of the bread to the underside detection model 541 to detect abnormalities on the underside. Burnt residue may adhere to the underside of the bread, and for example, the underside detection model 541 may be trained using images of burnt residue attached to the underside of the bread. In this case, the underside detection model 541 can detect burnt residue attached to the underside of the bread.

[0064] Figure 15(a) shows the underside of the bread 51, with burnt residue 51a attached. Figure 15(b) is an example of an image of the imaging range of the imaging means 53, with the inspection line 53a set within the imaging range. Here, the imaging means 53 is a 3D camera, and by measuring the distance information of the inspection line 53a with the 3D camera, the change in distance information caused by the bread 51 reaching the inspection line is detected, and the system starts capturing a color video of the bread that has reached the inspection line 53a. As shown in Figure 15(b), when the bread 51 reaches the inspection line, a color video of the underside of the bread is captured and input to the underside detection model 541. Figure 15(c) is another frame image, showing the burnt residue 51a on the underside of the bread 51. When this image is input to the underside detection model 541, the underside detection model 541 displays the detected burnt residue 51a surrounded by a rectangular frame 541a. Subsequently, as shown in Figure 15(d), when the edge of the pan 51 passes the inspection line 53a, the change in distance information ceases, and the inspection is terminated.

[0065] The determination means 54b determines that a product is good if the underside meets a predetermined standard for good products. The standard for good products is set appropriately depending on the type of bread, but it is preferable that it has a certain range. The determination means 54b may determine that a product is substandard if the underside does not meet the standard for good products, but it may also set a substandard standard that is different from the numerical value of the standard for good products, and determine that a product is substandard if it meets the substandard standard. In this case, there may be cases where the product does not meet the standard for good products and also does not meet the substandard standard, in which case it may be re-inspected and judged again by a human or machine, for example. In the determination criteria of Figure 7, the standard for good products is when the number of burnt parts on the underside is 0 to 1, and the standard for substandard products is 2 or more, but the determination may also be made by the burnt area on the underside (= area of ​​burnt parts / area of ​​the underside × 100). In the determination criteria of Figure 7, there is no determination of products that require re-inspection, but for example, if the standard for substandard products is set to 3 or more, a product with 2 burnt parts will require re-inspection.

[0066] The information processing device 14 may control the sorting means 15a and 15b to change the destination of the bread based on the judgment result. For example, good products may be sent in a straight line, products requiring re-inspection may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15a and transported to sorting destination 16a, and off-spec products may be positioned diagonally on the bread's path by sliding the guide plate of the sorting means 15b and transported to sorting destination 16b. [Examples]

[0067] Figure 16 shows an embodiment of the bread inspection system 60, which is a system that can continuously inspect the underside of bread 61 with ingredients 61a exposed on the surface, as well as perform a shaping inspection, a browning inspection, and an ingredients inspection. The bread inspection system 60 includes a first conveying means 62a and a second conveying means 62b, a 3D camera 63a and lighting means 67a positioned below the gap between them to capture a color image of the underside of the bread 61 being conveyed, an upper imaging means 63b and lighting means 67b positioned above the second conveying means 62b, and reflective means 68 positioned on both sides to reflect the sides of the bread 61 toward the imaging means 63b. The information processing device 14 includes an analysis means 64a that acquires and analyzes images captured by the imaging means 63a and the upper 3D camera 63b, a determination means 64b that determines whether the bread 61 is a good product or not, sorting means 15a, 15b that sort the bread based on the determination result, and other sorting destinations 16a, 16b. The analysis means 64a includes a back surface detection model 641, a normal ingredient detection model 642, and an abnormal ingredient detection model 643.

[0068] First, the analysis means 64a acquires a color image of the back surface of the transported bread 61 using a 3D camera 63a and an illumination means 67a in the gap between the first transport means 62a and the second transport means 62b, and inputs it into the back surface detection model 641 to perform a back surface inspection. Next, the analysis means 64a acquires a color image of the entire bread and thickness information of at least a part of the bread using the upper 3D camera 63b and illumination means 67b at a predetermined position on the second transport means 62b, and performs image processing as necessary to perform a molding inspection (Figures 5 and 6) and a browning inspection (Figure 10). Furthermore, the color image of the entire bread is input into the normal ingredient detection model 642 and the abnormal ingredient detection model 643 to perform an ingredient inspection (Figure 13). The determination means 64b determines whether the results of the back surface inspection, molding inspection, browning inspection, and ingredient inspection meet the standards for good products. The determination means 64b determines that a product is good if all inspection results meet the good product standard, and that it is a substandard product if even one inspection result does not meet the substandard standard. In all other cases, that is, if at least one inspection result does not meet the good product standard and all inspection results do not meet the substandard standard, it may determine that the product needs to be re-inspected. Based on the determination result, the information processing device 14 controls the sorting means 15a and 15b to allow good products to proceed straight, to control the sorting means 15a to transport re-inspected products to sorting destination 16a, and to control the sorting means 15b to transport substandard products to sorting destination 16b. [Explanation of Symbols]

[0069] 10. Bread Inspection System 11 Bread 12 Conveying means 13 Image acquisition unit 14 Information Processing Devices 15 Sorting means 16 Sorting destinations

Claims

1. A transporting device that carries bread on top, An upper imaging means for imaging the bread from above while it is being transported, An analysis means for acquiring and analyzing images captured by the aforementioned imaging means, A determination means for determining whether the bread is of good quality or not, A bread inspection system that includes, The upper imaging means is a 3D camera, and when the bread is transported to a predetermined position, it images the entire bread from above to obtain an image of the entire bread and thickness information of at least a part of the bread. The analysis means analyzes the size of the bread from the overall image of the bread, obtains the thickness of the bread from the thickness information, and calculates the abnormality rate by comparing the overall image of the bread with a pre-prepared sample of bread shape. The aforementioned determination means is a bread inspection system that determines a bread to be good if all of its size, thickness, and abnormality rate meet the pre-set standards for good products.

2. The bread inspection system according to claim 1, wherein the determination means determines that the bread is substandard if at least one of the size, thickness, and abnormality rate of the bread meets a preset substandard criterion.

3. The bread inspection system according to claim 2, wherein the determination means determines that the bread is a product that needs to be re-inspected if at least one of the size, thickness, and abnormality rate of the bread does not meet the good product standard, and all of the size, thickness, and abnormality rate of the bread do not meet the out-of-spec standard.

4. The bread inspection system according to claim 1, further comprising a sorting means for sorting the bread being transported by the transporting means according to the determination result of the determination means, to determine the destination of the bread.

5. The bread inspection system according to claim 1, wherein the shape sample is an average shape image created from multiple images of bread determined to be good products.

6. The bread inspection system according to claim 1, wherein the analysis means calculates the major and minor diameters of the bread from the overall image of the bread to analyze the size of the bread, and obtains thickness information at one or more positions on the major and minor diameters.

7. The bread inspection system according to claim 6, wherein the analysis means acquires thickness information at the intersection of the major axis and the minor axis, and thickness information at the midpoint between the intersection and the ends of the major axis and the minor axis, and the average of these is used as the thickness of the bread.

8. A means of transporting bread, An imaging means for imaging the bread while it is being transported, An illumination means for illuminating the pan within the imaging range of the imaging means, An analysis means for acquiring and analyzing images captured by the aforementioned imaging means, A determination means for determining whether the bread is of good quality or not, A bread inspection system that includes, When the bread is transported to a predetermined position, the imaging means captures an image of the entire bread to obtain a full-color image of the bread. The analysis means divides the overall color image of the bread into multiple regions, generates a hue histogram with the hue of the pixels within the region as the class, a saturation histogram with the saturation as the class, and a brightness histogram with the brightness as the class for each divided region, selects the multiple regions into regions to be judged and regions to be excluded based on the hue histogram, the saturation histogram, and / or the brightness histogram, and calculates the degree of baking of the bread based on the distribution of the brightness histogram of the regions to be judged. The aforementioned determination means is a bread inspection system that determines if the degree of baking of the bread meets a predetermined standard for good quality.

9. The bread inspection system according to claim 8, wherein the analysis means calculates the degree of baking of the bread based on the degree of burning and / or undercooked in the brightness histogram of the area to be judged.

10. The bread inspection system according to claim 9, wherein the determination means determines that the bread is an off-spec product if the degree of baking of the bread meets a predetermined off-spec standard.

11. The bread inspection system according to claim 10, wherein the determination means determines that the bread is a product that needs to be re-inspected if the degree of baking does not meet the standard for good products and does not meet the standard for non-standard products.

12. The bread inspection system according to claim 8, further comprising a sorting means for sorting the bread being transported by the transporting means according to the determination result of the determination means, to determine the destination of the bread.

13. A conveying means for transporting bread with ingredients exposed on the surface, An imaging means for capturing a color image of the bread while it is being transported, An analysis means for acquiring and analyzing images captured by the aforementioned imaging means, A determination means for determining whether the bread is of good quality or not, A bread inspection system that includes, When the bread is transported to a predetermined position, the imaging means captures an image of the entire bread from above to obtain a full-color image of the bread. The analysis means includes a normal ingredient detection model trained using images of normal ingredients in the bread, and an abnormal ingredient detection model trained using images of abnormal ingredients in the bread. The whole color image of the bread is input to the normal ingredient detection model to count the number of normal ingredients, and the whole color image of the bread is input to the abnormal ingredient detection model to count the number of abnormal ingredients. The aforementioned determination means determines that a bread is good if the number of normal ingredients and the number of abnormal ingredients counted meet a preset standard for good products.

14. The bread inspection system according to claim 13, wherein the image of the abnormal ingredients is an image of the dough after the ingredients have peeled off the surface of the bread.

15. The bread inspection system according to claim 13, further comprising reflective means on both sides of the conveying means for reflecting the sides of the bread toward the imaging means.

16. The bread inspection system according to claim 13, further comprising an illumination means for illuminating the bread within the imaging range of the imaging means.

17. The bread inspection system according to claim 13, wherein the determination means determines that the product is substandard if at least one of the counted number of normal ingredients and the number of abnormal ingredients meets a preset substandard criterion.

18. The bread inspection system according to claim 17, wherein the determination means determines that a product is to be re-inspected if either the number of normal ingredients or the number of abnormal ingredients counted does not meet the standard for good products, and the other does not meet the standard for non-standard products.

19. The bread inspection system according to claim 13, further comprising a sorting means for sorting the bread being transported by the transporting means according to the determination result of the determination means, to determine the destination of the bread.

20. A means of transporting bread, The conveying means includes an imaging means for capturing a color image of the underside of the bread while it is being conveyed, An analysis means for acquiring and analyzing images captured by the aforementioned imaging means, A determination means for determining whether the bread is of good quality or not, A bread inspection system that includes, The imaging means captures the back surface of the bread to obtain a color image of the back surface of the bread. The analysis means includes a back surface detection model trained using an image of the back surface of the bread, and inputs the color image of the back surface of the bread into the back surface detection model to detect abnormalities on the back surface. The aforementioned determination means determines that the product is good if the detected abnormality meets a preset standard for good products, and this is a bread inspection system.

Images