Soybean peeling machine peeling quality detection method and system

Abstract

The present invention relates to the field of image data processing technology, and in particular to a soybean peeling machine peeling quality detection method and system, comprising: obtaining a plurality of super-pixel blocks in each frame of soybean peeling image in a soybean peeling image sequence, determining a damage index value of each super-pixel block based on the difference in the distance from the pixel point on the boundary of each super-pixel block to the center point, obtaining a damaged soybean region composed of a plurality of super-pixel blocks based on the difference in the damage index values ​​of the super-pixel blocks in each frame of soybean peeling image and the distance between the super-pixel blocks, determining the amount of damage in the damaged soybean region based on the correlation between the distance from the center point of the super-pixel block to the center point of the damaged soybean region and the damage index value of the super-pixel block, and judging whether the soybean peeling quality is qualified based on the size of the damage in the damaged soybean region. The present invention has a soybean peeling quality recognition feature with high applicability and can obtain accurate evaluation indicators.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a method and system for detecting the peeling quality of a soybean peeling machine. Background Art

[0002] A soybean peeling machine is a mechanical device used to remove the hulls (skins) of soybeans. This equipment is widely used in industries such as food processing and oil extraction to improve soybean processing efficiency. After the hulls are removed, the soybeans are generally more suitable for subsequent processing, such as milling and oil extraction. The machine typically works by mechanically separating the soybean hulls from the beans, primarily through mechanical impact: high-frequency impacts are used to break the hulls and separate them from the beans. The quality of soybean peeling is primarily assessed by factors such as peeling rate, residual hull percentage, breakage rate, membrane detachment, and uniformity. To inspect the quality of soybean peeling, a machine vision system uses a camera and image processing technology to capture and analyze the soybeans. This involves real-time image acquisition, image processing, and real-time defect detection. Defect detection involves quantitative analysis of the quality of soybean peeling.

[0003] Existing Problem: During the actual peeling process, the quality of the peeling process not only considers the residual soybean hull percentage, i.e., the peeling effect, but also the soybean integrity rate after peeling, i.e., the degree of soybean damage. Due to the physical limitations of peeling, soybeans require high-speed movement and vibration, which means they are subjected to mechanical shock. This can lead to some soybean skin damage under high external mechanical vibration conditions. Furthermore, due to these physical conditions (high-speed movement and vibration), soybeans are prone to overlapping or accumulation during peeling, resulting in a loss of surface information and an inability for machine vision to accurately detect the presence of damaged beans. Summary of the Invention

[0004] The invention provides a soybean peeling machine peeling quality detection method and system to solve the existing problems.

[0005] The soybean peeling machine peeling quality detection method and system of the present invention adopts the following technical solutions:

[0006] One embodiment of the present invention provides a method for detecting the peeling quality of a soybean peeling machine, the method comprising the following steps:

[0007] Acquire soybean peeling image sequence;

[0008] Acquire multiple superpixel blocks within each frame of a soybean peeling image sequence; determine the damage index value of each superpixel block based on the difference in distances from pixels on the boundary of each superpixel block to the center point; and obtain a damaged soybean region consisting of multiple superpixel blocks based on the difference in damage index values ​​of the superpixel blocks within each frame of the soybean peeling image and the distances between the superpixel blocks;

[0009] Determine the damage amount of the damaged soybean area based on the correlation between the distance from the center point of the superpixel block in the damaged soybean area to the center point of the damaged soybean area and the damage index value of the superpixel block;

[0010] The quality of soybean peeling is judged based on the amount of damage in the damaged soybean area.

[0011] Furthermore, the determination of the damage index value of each superpixel block includes the following specific steps:

[0012] Record all the straight line segments connecting the pixel points on the boundary of the nth superpixel block to the center point of the nth superpixel block as target straight line segments, obtain the average of the lengths of all target straight line segments, and record the sum of the absolute values ​​of the differences between the lengths of all target straight line segments and the average as the irregularity of the nth superpixel block;

[0013] The sum of the absolute values ​​of the differences between the lengths of the target straight line segments corresponding to all two adjacent pixels on the boundary of the n-th superpixel block is recorded as the tortuosity of the n-th superpixel block;

[0014] A damage index value of the nth superpixel block is determined according to the irregularity and tortuosity of the nth superpixel block.

[0015] Furthermore, the step of determining the damage index value of the nth superpixel block according to the irregularity and tortuosity of the nth superpixel block includes the following specific steps:

[0016] The normalized value of the sum of the irregularity and the tortuosity of the n-th superpixel block is recorded as the damage index value of the n-th superpixel block.

[0017] Furthermore, the specific steps of obtaining the damaged soybean area composed of a plurality of superpixel blocks are as follows:

[0018] In each frame of soybean peeling image, all superpixel blocks are divided into several clusters according to the difference in damage index values ​​of superpixel blocks and the distance between superpixel blocks;

[0019] The average of the damage index values ​​of all superpixel blocks in each cluster is recorded as the damage index value of each cluster;

[0020] The area consisting of all superpixel blocks in the cluster corresponding to the maximum damage index value is recorded as the damaged soybean area.

[0021] Furthermore, the method of dividing all superpixel blocks into a number of clusters includes the following specific steps:

[0022] In any frame of soybean peeling image in the soybean peeling image sequence, the normalized value of the shortest distance between any two super-pixel blocks is obtained, and then the absolute value of the difference between the damage index values ​​of the any two super-pixel blocks is obtained. The average of the normalized value of the shortest distance and the absolute value of the difference between the damage index values ​​is recorded as the clustering distance of the any two super-pixel blocks. The DBSCAN clustering method is used to perform clustering operations on all super-pixel blocks to obtain several cluster clusters.

[0023] Furthermore, the specific steps of determining the damaged amount of damaged soybeans include the following:

[0024] In any frame of soybean peeling image in the soybean peeling image sequence, the distances from the center points of all superpixel blocks in the damaged soybean area to the center point of the damaged soybean area are obtained, and all are recorded as target distances. All target distances are arranged from small to large to obtain a target distance sequence;

[0025] In the target distance sequence, the damage index value of the super pixel block corresponding to each target distance is obtained in turn to form a damage index value sequence;

[0026] According to the correlation between the target distance sequence and the damage index value sequence, combined with the damage index value of the superpixel block in the damaged soybean area, the damage amount of the damaged soybean area is determined.

[0027] Furthermore, the damage amount of the damaged soybean area is determined based on the correlation between the target distance sequence and the damage index value sequence, in combination with the damage index values ​​of the superpixel blocks in the damaged soybean area, and the specific steps include the following:

[0028] Obtain the inversely proportional normalized value of the Pearson correlation coefficient between the target distance sequence and the damage index value sequence, then obtain the mean of the damage index values ​​of all superpixel blocks in the damaged soybean area, and record the normalized value of the product of the inversely proportional normalized value of the Pearson correlation coefficient and the mean of the damage index value as the damage amount of the damaged soybean area.

[0029] Furthermore, judging whether the soybean peeling quality is qualified based on the amount of damage in the damaged soybean area includes the following specific steps:

[0030] In the soybean peeling image sequence, the damaged soybean area in the first frame of the soybean peeling image is recorded as the target area, and the areas overlapping with the target area in all other soybean peeling image frames except the first frame are obtained and recorded as reference areas;

[0031] Obtain the damage amount of each reference area according to the method for obtaining the damage amount of the target area;

[0032] The quality of soybean peeling is judged based on the damage amount of the target area and all reference areas.

[0033] Furthermore, judging whether the soybean peeling quality is qualified based on the damage amount of the target area and all reference areas includes the following specific steps:

[0034] When the mean of the damage amount of the target area and all reference areas is greater than the preset first threshold, the ARIMA prediction model is used to predict the damage amount of the target area and all reference areas to obtain several predicted damage amounts. If any predicted damage amount is greater than the preset second threshold, the soybean peeling quality at the current moment is judged to be unqualified.

[0035] The present invention also proposes a soybean peeling machine peeling quality detection system, including a memory, a processor and a computer program stored in the memory and runnable on the processor. The processor executes the computer program stored in the memory to implement the steps of the aforementioned soybean peeling machine peeling quality detection method.

[0036] The beneficial effects of the technical solution of the present invention are:

[0037] In an embodiment of the present invention, several superpixel blocks are obtained within each frame of a soybean peeling image sequence, and the damage index value of each superpixel block is determined based on the difference in distance from the pixel point on the boundary of each superpixel block to the center point. Thus, by analyzing the inherent universal characteristics of soybeans, namely the irregularity and tortuosity manifested in the image, a soybean peeling quality identification characteristic with high applicability is obtained, and a specific evaluation index is obtained to ensure the accuracy of peeling quality detection. Based on the difference in damage index values ​​of superpixel blocks within each frame of the soybean peeling image and the distance between superpixel blocks, a damaged soybean region consisting of several superpixel blocks is obtained. Based on the correlation between the distance from the center point of the superpixel block to the center point of the damaged soybean region within the damaged soybean region and the damage index value of the superpixel block, the damage amount of the damaged soybean region is determined. Thus, by analyzing the physical peeling process, a local overall machine vision relationship between local damage and actual physical damage is obtained, and the cumulative degree of damage amount is calculated to determine the data premise of fuzzy prediction. The soybean peeling quality is judged based on the amount of damage in the damaged soybean area. This improves the problem of inaccurate judgment of overlapping parts by analyzing local quality, combining physical laws, and using fuzzy prediction. Therefore, the present invention has highly applicable soybean peeling quality identification characteristics and can obtain accurate evaluation indicators. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 This is a flowchart of the steps of a soybean peeling machine peeling quality detection method of the present invention;

[0040] Figure 2 Schematic diagram of soybeans after peeling. DETAILED DESCRIPTION

[0041] To further illustrate the technical means and effectiveness of the present invention in achieving its intended objectives, the following, in conjunction with the accompanying drawings and preferred embodiments, describes in detail a soybean peeling quality detection method and system for a soybean peeling machine, including its specific implementation, structure, features, and effectiveness. In the following description, different references to "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics of one or more embodiments may be combined in any suitable manner.

[0042] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0043] The following describes in detail a soybean peeling machine peeling quality detection method and system provided by the present invention with reference to the accompanying drawings.

[0044] See also Figure 1 , which shows a flowchart of a soybean peeling machine peeling quality detection method provided by one embodiment of the present invention, the method comprising the following steps:

[0045] Step S001: Acquire a soybean peeling image sequence.

[0046] In this embodiment, a real-time machine vision detection environment is configured to collect real-time soybean peeling images, and then a real-time soybean peeling machine visual image is obtained. The soybean visual unit damage index is determined based on the irregular tortuosity of the pixel block, and then the regional damage amount at the location of the damage index is determined. Combined with the index results of multiple frames of images, the cumulative degree of damage is obtained, and the cumulative degree is fuzzy predicted to obtain the fuzzy predicted damage degree of the overlapping area, to determine the real-time soybean peeling quality, and to improve the comprehensiveness and accuracy of the detection.

[0047] It should be noted that the real-time monitoring camera used in this embodiment is an industrial camera, typically a CMOS camera, capable of providing high resolution and high frame rates. Its resolution is 1920 x 1080, ensuring clear image detail. Based on the speed of the peeling process, an acquisition frequency of 10 frames per second was selected to ensure that information about rapidly moving soybeans is captured and to prevent loss of information in the peeled soybean images due to insufficient frame counts. The light source used in the detection environment was selected by using a uniform lighting system (LED) to ensure uniform brightness on the soybean surface in the image. The light intensity was adjusted to avoid reflections or shadows (to prevent misidentification of peeling defects due to shadows). The camera was installed based on the location of the soybean peeling machine, with a position reference indicator selected (sufficient for capturing real-time images of the peeled soybeans). The camera was then activated and image acquisition software (OpenCV) was used to acquire each frame of the soybean peeling process in real time. The captured image data was then transferred to a computer via a high-speed USB 3.0 interface for subsequent processing. Among them, after the soybean peeling image is grayed, the mean filter is used to remove noise, and then the histogram equalization algorithm is used to enhance contrast to improve image quality. Image graying, mean filtering and histogram equalization are all well-known technologies, and the specific methods will not be introduced here. Schematic diagram of soybean after peeling, as shown Figure 2 shown.

[0048] Thus, a soybean peeling image sequence is obtained, and the last soybean peeling image in the soybean peeling image sequence corresponds to the current moment. Setting the time period to 5 seconds, taking this as an example, the soybean peeling image sequence consists of soybean peeling images from the current moment to the last 5 seconds.

[0049] Step S002: Acquire several super-pixel blocks in each frame of soybean peeling image in the soybean peeling image sequence; determine the damage index value of each super-pixel block based on the difference in the distance from the pixel point on the boundary of each super-pixel block to the center point; acquire the damaged soybean area composed of several super-pixel blocks based on the difference in the damage index value of the super-pixel blocks in each frame of soybean peeling image and the distance between the super-pixel blocks.

[0050] It should be noted that in the actual production process, after a high-kinetic energy physical process such as vibration, beans are more likely to produce an overlapping effect, which affects the judgment of the soybean peeling quality. In order to achieve the purpose of comprehensive analysis, this embodiment analyzes the visual effect of bean damage and combines the physical correlation properties of the local area of ​​beans during the peeling process to determine the quality of beans in the overlapping area. Therefore, it is necessary to first analyze the damage index of beans after peeling. The index value reflects the quality of the beans themselves after peeling, such as whether there is a certain degree of incompleteness, irregularity and cracks. For soybeans, the visual characteristics of their damage can be summarized as follows: after the physical changes (such as impact, friction or vibration) during the peeling process, the damage effects produced on their surface are mainly: (1) Incompleteness: Part of the soybean skin falls off or breaks, resulting in an irregular shape. (2) Irregular shape: Due to the impact or vibration, the soybean produces a certain deformation, making its surface shape no longer symmetrical or smooth. (3) Cracks and depressions: Mechanical impact, vibration and other factors will form cracks or depressions on the soybean surface, affecting the overall appearance. After analysis and collation, the aforementioned soybean quality characteristics are primarily manifested in images as irregularities in localized pixel targets (i.e., irregular shapes caused by skin shedding or cracking) and the degree of tortuosity (tortuousness) of some target edges. This tortuosity manifests itself as depressions, deformations, and asymmetry on the soybean surface. Therefore, the damage index can be derived from analyzing these characteristics in the image.

[0051] Preferably, in one embodiment of the present invention, the method for obtaining the damaged soybean area includes:

[0052] In the soybean peeling image sequence, the superpixel segmentation algorithm is used to segment each frame of soybean peeling image to obtain several superpixel blocks.

[0053] It should be noted that the superpixel segmentation algorithm is a well-known technique, and the specific method will not be described here. In the image, individual soybeans have distinct edges that separate them from their surroundings. Therefore, after segmentation, each superpixel represents the soybean's outline. To determine the damage index for each soybean, the corresponding superpixel block is analyzed.

[0054] Get the center point of the nth superpixel block, record the straight line segments connecting all pixel points on the boundary of the nth superpixel block to the center point as target straight line segments, obtain the mean of the lengths of all target straight line segments, calculate the absolute value of the difference between the length of each target straight line segment and the mean, and record the sum of the absolute values ​​of the differences between the lengths of all target straight line segments and the mean as the irregularity of the nth superpixel block.

[0055] Obtain the absolute value of the difference between the target straight line segment lengths of two adjacent pixel points on the boundary of the nth superpixel block, and record the sum of the absolute values ​​of the difference between the target straight line segment lengths of all two adjacent pixel points on the boundary of the nth superpixel block as the tortuosity of the nth superpixel block.

[0056] It should be noted that the greater the difference between the target straight line segment length and the mean length, the more random the variation in the edge distance from the geometric center within the block. Therefore, in the soybean peeling image, the edge of a soybean unit will appear irregular, indicating a higher irregularity in the superpixel block, a higher likelihood of soybean damage, and a higher damage index. Conversely, a smaller difference indicates a smoother and more regular edge, and a higher overall smoothness of the soybean edge. Larger length differences between adjacent line segments indicate less smooth edge variation within the block, reflecting the smoothness stability of local regions within the block. This suggests that in some local areas, the soybean edge may have cracks due to the physical peeling process. These cracks manifest as a sudden decrease in the distance between the edge and the geometric center within a short interval, reflecting a higher tortuosity.

[0057] The normalized value of the sum of the irregularity and the tortuosity of the n-th superpixel block is recorded as the damage index value of the n-th superpixel block.

[0058] It should be noted that: In this embodiment, the The linear normalization function normalizes the sum of irregularity and tortuosity to a value between 0 and 1. That is, the closer the value is to 1, the higher the damage index value is, and the higher the possibility of damage to soybeans after peeling.

[0059] In any frame of soybean peeling image in the soybean peeling image sequence, the normalized value of the shortest distance between any two superpixel blocks is obtained, and then the absolute value of the difference between the damage index values ​​of the arbitrary two superpixel blocks is obtained. The average of the normalized value of the shortest distance and the absolute value of the difference between the damage index values ​​is recorded as the clustering distance of the arbitrary two superpixel blocks. The DBSCAN clustering method is used to cluster all superpixel blocks to obtain several cluster clusters.

[0060] It should be noted that the DBSCAN clustering method (Density-Based Spatial Clustering of Applications with Noise) is a well-known technology and the specific method will not be introduced here.

[0061] The average of the damage index values ​​of all superpixel blocks in each cluster is recorded as the damage index value of each cluster.

[0062] In any soybean peeling image in the soybean peeling image sequence, the area consisting of all superpixel blocks in the cluster corresponding to the maximum damage index value is recorded as the damaged soybean area.

[0063] It should be noted that the larger the damage index value of a cluster, the more likely it is that damaged soybeans after peeling are concentrated within the area formed by all superpixel blocks in the cluster. This means that the damage status of soybeans in the damaged soybean area will be fuzzy predicted.

[0064] Step S003: Determine the damage amount of the damaged soybean area based on the correlation between the distance from the center point of the super pixel block in the damaged soybean area to the center point of the damaged soybean area and the damage index value of the super pixel block.

[0065] It should be noted that during the soybean peeling process, the physical relationship between the degree of damage in a particular area and the degree of damage per soybean unit is as follows: Because soybeans are subjected to high-speed motion and mechanical impact during the peeling process, their physical properties and the characteristics of the external forces acting on them directly influence the probability and extent of soybean damage. First, high-speed motion and impact. When soybeans experience high-speed motion or intense vibration during the peeling process, the impact forces cause cracks or damage to their internal structure and surface. This physical force can exacerbate damage in certain areas, often with more damage occurring in localized areas. Therefore, when concentrated damage occurs in a particular area, the degree of damage in that area is generally higher. Second, the amount of regional damage. During the peeling process, uneven stress distribution creates a phenomenon known as "stress concentration" in material mechanics. Specifically, after high-speed impact, soybean surfaces are prone to concentrated damage in certain areas, especially where soybeans collide with each other. Higher damage in these areas indicates a greater probability of soybean damage within those areas. Consequently, when concentrated damage occurs in a particular area, the degree of damage in that area is generally higher. After the above analysis, when a high amount of damage occurs in an area, the soybean quality of the overlapping part of the area that does not appear in the visual field will also approach the same damage index value.

[0066] It should be further explained that, under ideal stress distribution, regional damage can be simply calculated by taking the average damage index value of the superpixels within the region. However, in actual production, when soybeans undergo physical collisions, stress distribution is not uniform in local areas. Therefore, the damage index value within a cluster is related to the damage index values ​​of the superpixels within the region and their spatial distribution. The specific spatial distribution pattern is as follows: During the soybean peeling process, as analyzed above, a large number of grains in a certain area accumulate, overlap, or are squeezed, forming a phenomenon similar to "grain agglomeration" or "localized blockage." In this case, due to the high contact and restricted movement of soybeans, external mechanical shock and vibration are transmitted to this area, with the impact force concentrated along the center of the accumulation. Grains in the center of the region are subjected to pressure from all directions, forming a so-called "stress concentration center." This prevents the soybeans from moving freely and instead becomes a point of force convergence. Therefore, soybeans in the center of the region are more susceptible to damage and suffer more severe damage. Therefore, the damage characteristics at the center of the region are dense, irregular, deep, and with pronounced cracks. As the damage radiates outward, the damage becomes sparser, and the crack length decreases, until the edges of the region, where the soybean surface is intact or slightly scratched and shallowly sunken. In other words, the damage gradually decreases from the center outward. Therefore, the regional damage value is not simply an average. When the regional damage distribution closely matches the above characteristics, the actual regional damage value should be calculated.

[0067] Preferably, in one embodiment of the present invention, the method for obtaining the damaged amount of damaged soybean area includes:

[0068] In any frame of soybean peeling image in the soybean peeling image sequence, the center point of the damaged soybean area is obtained, and then the distances from the center points of all superpixel blocks in the damaged soybean area to the center point of the damaged soybean area are obtained, all of which are recorded as target distances. All target distances are arranged from small to large to obtain a target distance sequence.

[0069] In the target distance sequence, the damage index value of the super pixel block corresponding to each target distance is obtained in turn to form a damage index value sequence.

[0070] Get the Pearson correlation coefficient between the target distance series and the damage index value series The inverse proportional normalized value of the damaged soybean area is obtained, and then the mean of the damage index values ​​of all superpixel blocks in the damaged soybean area is obtained. The Pearson correlation coefficient The normalized value of the product of the inverse proportional normalized value and the mean of the damage index value is recorded as the damage amount of the damaged soybean area.

[0071] It should be noted that the Pearson correlation coefficient is a well-known technology and the specific method will not be introduced here. The value range of the Pearson correlation coefficient is between -1 and 1. The closer it is to -1, the more negatively correlated the two sequences are. That is, as the distance to the target increases, the damage index value gradually decreases, which is consistent with the characteristic that the damage in the area gradually decreases from the center to the surrounding areas. As the Pearson correlation coefficient The inverse normalized value of It is a linear normalization function used to normalize the data value to between 0 and 1. The inverse proportional normalized value of is used to adjust the mean of the damage index value to obtain the damage amount of the damaged soybean area.

[0072] Step S004: judging whether the soybean peeling quality is qualified according to the amount of damage in the damaged soybean area.

[0073] It should be noted that if the same region still has a high damage amount in multiple frames, it is reasonable to infer that the damage index of the overlapping soybeans is higher. Conversely, if the damage amount of a region is low, the damage index of the soybeans in the overlapping part is also low.

[0074] In the soybean peeling image sequence, the damaged soybean area in the first frame of the soybean peeling image is recorded as the target area. In all other soybean peeling image frames except the first frame, the areas overlapping with the target area (the overlapping areas are the areas at the same position in different images) are obtained and recorded as reference areas.

[0075] The damage amount of each reference area is obtained according to the method of obtaining the damage amount of the target area.

[0076] It should be noted that when an incomplete superpixel block exists within the reference region, that is, a superpixel block is partially within the reference region, the complete superpixel block is assigned to the reference region. The damage level of the reference region is determined based on the damage index value of the superpixel block within the reference region and the distance from the center point of the damage index value to the center point of the reference region.

[0077] The first threshold is preset to 0.75, and the second threshold is preset to 0.8, which is used as an example for description.

[0078] When the mean of the damage amount of the target area and all reference areas is greater than the preset first threshold, the ARIMA prediction model is used to predict the damage amount of the target area and all reference areas to obtain several predicted damage amounts. If any predicted damage amount is greater than the preset second threshold, the soybean peeling quality at the current moment is judged to be unqualified and an early warning is issued.

[0079] It should be noted that the ARIMA prediction model (Autoregressive Integrated Moving Average Model) is a well-known technology, and the specific method will not be introduced here. The prediction step size is set to 2 seconds, and this is used as an example to predict the amount of damage in the reference area of ​​the soybean peeling image within 2 seconds after the current moment. The soybean information in each frame is only partial information about the soybeans. The remaining information will be displayed in other frames as the soybean peeling machine vibrates, causing the soybeans to turn over. Therefore, the amount of soybean damage in the same area of ​​different frames is obtained. When the amount of damage is large, the damage information displayed by the soybeans is more complete through prediction, which is used to determine whether the soybean peeling quality is qualified. In this way, when there is overlapping soybean damage, a more specific amount of damage can be obtained, solving the problem of severe overlap in actual production processes.

[0080] The present invention also provides a soybean peeling machine peeling quality detection system, including a memory, a processor and a computer program stored in the memory and runnable on the processor, wherein the processor executes the computer program stored in the memory to implement the steps of the aforementioned soybean peeling machine peeling quality detection method.

[0081] In summary, in an embodiment of the present invention, several super-pixel blocks are obtained in each frame of soybean peeling image in a soybean peeling image sequence, and the damage index value of each super-pixel block is determined based on the difference in the distance from the pixel point on the boundary of each super-pixel block to the center point. According to the difference in the damage index value of the super-pixel blocks in each frame of soybean peeling image and the distance between the super-pixel blocks, a damaged soybean area composed of several super-pixel blocks is obtained. According to the correlation between the distance from the center point of the super-pixel block to the center point of the damaged soybean area in the damaged soybean area and the damage index value of the super-pixel block, the damage amount of the damaged soybean area is determined. According to the size of the damage amount of the damaged soybean area, whether the soybean peeling quality is qualified is judged. The present invention has a soybean peeling quality recognition feature with high applicability and can obtain accurate evaluation indicators.

[0082] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A soybean peeling machine peeling quality detection method, characterized in that, The method comprises the following steps: Acquire soybean peeling image sequence; Acquire multiple superpixel blocks within each frame of a soybean peeling image sequence; determine the damage index value of each superpixel block based on the difference in distances from pixels on the boundary of each superpixel block to the center point; and obtain a damaged soybean region consisting of multiple superpixel blocks based on the difference in damage index values ​​of the superpixel blocks within each frame of the soybean peeling image and the distances between the superpixel blocks; Determine the damage amount of the damaged soybean area based on the correlation between the distance from the center point of the superpixel block in the damaged soybean area to the center point of the damaged soybean area and the damage index value of the superpixel block; According to the size of the damaged soybean area, the soybean peeling quality is judged to be qualified; The specific steps of determining the damage index value of each superpixel block include the following: Record all the straight line segments connecting the pixel points on the boundary of the nth superpixel block to the center point of the nth superpixel block as target straight line segments, obtain the average of the lengths of all target straight line segments, and record the sum of the absolute values ​​of the differences between the lengths of all target straight line segments and the average as the irregularity of the nth superpixel block; The sum of the absolute values ​​of the differences between the lengths of the target straight line segments corresponding to all two adjacent pixels on the boundary of the n-th superpixel block is recorded as the tortuosity of the n-th superpixel block; The normalized value of the sum of the irregularity and the tortuosity of the n-th superpixel block is recorded as the damage index value of the n-th superpixel block.

2. A soybean peeling machine peeling quality detection method according to claim 1, characterized in that, The specific steps of obtaining the damaged soybean area composed of a plurality of superpixel blocks are as follows: In each frame of soybean peeling image, all superpixel blocks are divided into several clusters according to the difference in damage index values ​​of superpixel blocks and the distance between superpixel blocks; The average of the damage index values ​​of all superpixel blocks in each cluster is recorded as the damage index value of each cluster; The area consisting of all superpixel blocks in the cluster corresponding to the maximum damage index value is recorded as the damaged soybean area.

3. A soybean peeling machine peeling quality detection method according to claim 2, characterized in that, The specific steps of dividing all superpixel blocks into several clusters are as follows: In any frame of soybean peeling image in the soybean peeling image sequence, the normalized value of the shortest distance between any two super-pixel blocks is obtained, and then the absolute value of the difference between the damage index values ​​of the any two super-pixel blocks is obtained. The average of the normalized value of the shortest distance and the absolute value of the difference between the damage index values ​​is recorded as the clustering distance of the any two super-pixel blocks. The DBSCAN clustering method is used to perform clustering operations on all super-pixel blocks to obtain several cluster clusters.

4. A soybean peeling machine peeling quality detection method according to claim 1, characterized in that, The specific steps of determining the damaged amount of damaged soybeans include the following: In any frame of soybean peeling image in the soybean peeling image sequence, the distances from the center points of all superpixel blocks in the damaged soybean area to the center point of the damaged soybean area are obtained, and all are recorded as target distances. All target distances are arranged from small to large to obtain a target distance sequence; In the target distance sequence, the damage index value of the super pixel block corresponding to each target distance is obtained in turn to form a damage index value sequence; According to the correlation between the target distance sequence and the damage index value sequence, combined with the damage index value of the superpixel block in the damaged soybean area, the damage amount of the damaged soybean area is determined.

5. A soybean peeling machine peeling quality detection method according to claim 4, characterized in that, The method of determining the damage amount of the damaged soybean area based on the correlation between the target distance sequence and the damage index value sequence and the damage index value of the super pixel block in the damaged soybean area includes the following specific steps: Obtain the inversely proportional normalized value of the Pearson correlation coefficient between the target distance sequence and the damage index value sequence, then obtain the mean of the damage index values ​​of all superpixel blocks in the damaged soybean area, and record the normalized value of the product of the inversely proportional normalized value of the Pearson correlation coefficient and the mean of the damage index value as the damage amount of the damaged soybean area.

6. A soybean peeling machine peeling quality detection method according to claim 1, characterized in that, The specific steps of judging whether the soybean peeling quality is qualified according to the amount of damage in the damaged soybean area are as follows: In the soybean peeling image sequence, the damaged soybean area in the first frame of the soybean peeling image is recorded as the target area, and the areas overlapping with the target area in all other soybean peeling image frames except the first frame are obtained and recorded as reference areas; Obtain the damage amount of each reference area according to the method for obtaining the damage amount of the target area; The quality of soybean peeling is judged based on the damage amount of the target area and all reference areas.

7. A soybean peeling machine peeling quality detection method according to claim 6, characterized in that, The specific steps of judging whether the soybean peeling quality is qualified based on the damage amount of the target area and all reference areas are as follows: When the mean of the damage amount of the target area and all reference areas is greater than the preset first threshold, the ARIMA prediction model is used to predict the damage amount of the target area and all reference areas to obtain several predicted damage amounts. If any predicted damage amount is greater than the preset second threshold, the soybean peeling quality at the current moment is judged to be unqualified.

8. A soybean peeling machine peeling quality detection system, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is executed by a processor, the steps of the soybean peeling machine peeling quality detection method according to any one of claims 1 to 7 are implemented.

Images