Method for detecting early recessive damage of Korla pear based on artificial intelligence

By subjecting Korla fragrant pears to compression treatments of varying degrees and conducting image comparison analysis, the shortcomings of existing damage detection technologies have been addressed. This enables proactive analysis of damage causes and optimized transport, ensuring low-damage or damage-free transportation of Korla fragrant pears.

CN118941503APending Publication Date: 2024-11-12XINJIANG UNIV OF SCI & TECH
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202410924079.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-11
Publication Date
2024-11-12

Smart Images

  • Figure CN118941503A_ABST
    Figure CN118941503A_ABST
Patent Text Reader

Abstract

The invention discloses a Korla pear early recessive damage detection method based on artificial intelligence, and relates to the technical field of agricultural product detection. According to the Korla pear early-stage recessive damage detection method based on artificial intelligence, the Korla pear is subjected to extrusion treatment with different strengths in advance to obtain extrusion damage shape information and corresponding active extrusion force numerical value information, and then the X image of the to-be-detected Korla pear is obtained to obtain the early-stage recessive damage of the Korla pear. After the image of the extrusion damage part is acquired, the pressure bearing degree of the korla pear with the extrusion damage can be conveniently analyzed, and the subjectivity of the extrusion damage can be judged by comparing the similarity of the damage image, so that a reliable reference condition is provided for judging whether the korla pear is manually and actively damaged or not; the method can provide an optimization direction for the transfer defects of the Korla pears, and also can provide a reliable guarantee for low-loss and even lossless transportation of the Korla pears.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of agricultural product detection, and in particular to an artificial intelligence-based method for detecting early hidden damage of Korla fragrant pear. Background Art

[0002] The Korla Pear is a hybrid of the white pear and the Western pear. It is known as the "treasure of pears" and the "prince of fruits". It is a small fruit with a single fruit weight of about 90g to 150g. It has the characteristics of thin skin, fine flesh, crispy and refreshing taste, high juice and little residue, long-term storage, rich nutrition and strong aroma. The Korla Pear is favored by consumers for its delicious taste and high nutritional value.

[0003] However, mechanical damage to pears can reduce their quality and seriously affect consumers' purchasing decisions. Mechanical damage is particularly prone to occur during the picking and storage of Korla pears, of which hidden damage is one of the most common and difficult to identify types. Due to the thin-skinned and juicy characteristics of Korla pears, they are prone to bumps during the picking, transportation and sorting process, resulting in slight changes in the color of the flesh and softening of the tissue. This early hidden damage usually does not damage the peel or show any difference in appearance, making it difficult to identify with the naked eye. Over time, the damaged parts are also susceptible to microbial invasion, causing the fruit to rot and mildew, which in turn affects the quality of the surrounding undamaged fruit. In addition, traditional manual sorting methods are time-consuming and labor-intensive, and cannot meet the high-efficiency requirements of modern fruit sorting.

[0004] As described in the application number 202311472395.0, a method for detecting early hidden damage to Korla fragrant pears that integrates hyperspectral image features and spectral variables introduces the idea of ​​spectral fusion, integrates the texture feature information of the hyperspectral image with the spectral feature variable information, and applies it to the detection of early hidden damage to Korla fragrant pears. Compared with the traditional method that only uses spectral information, it not only makes full use of the image and spectral data of hyperspectral imaging technology, but also can detect smaller areas of hidden damaged fragrant pears, which makes the present invention have obvious advantages.

[0005] Based on the retrieval of the above information, it can be seen that when using hyperspectral images to detect early invisible damage of Korla fragrant pear, it is often only possible to detect whether the Korla fragrant pear has extrusion damage, and there is a lack of active analysis of the causes of extrusion damage. This cannot provide reliable guarantees for the transportation optimization of Korla fragrant pear and the low-loss or even lossless transportation of Korla fragrant pear. For this reason, a detection method for early hidden damage of Korla fragrant pear based on artificial intelligence is proposed. By pre-extruding the Korla fragrant pear with different intensities to obtain the shape of the extrusion damage, it is not only convenient to analyze the pressure intensity of the Korla fragrant pear with extrusion damage, but also to determine whether the occurrence of extrusion damage is caused by human initiative through the similarity comparison of the damage images. While providing optimization direction for the transportation defects of Korla fragrant pear, it can also provide reliable guarantees for the low-loss or even lossless transportation of Korla fragrant pear. Summary of the invention

[0006] 1. Technical issues to be resolved

[0007] In view of the shortcomings of the prior art, the present invention provides a method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, which solves the problem that when conventionally using hyperspectral images to detect early hidden damage of Korla fragrant pear, it is often only possible to detect whether the Korla fragrant pear has compression damage, but lacks active analysis of the causes of compression damage, which cannot provide reliable guarantee for the transportation optimization of Korla fragrant pear and the low-loss or even loss-free transportation of Korla fragrant pear.

[0008] (II) Technical solution

[0009] To achieve the above objectives, the present invention is implemented by the following technical scheme: a method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, specifically comprising the following steps:

[0010] Step 1: Collect several groups of Korla fragrant pears with intact appearance, mark them with numbers, and obtain reference X-ray images of Korla fragrant pears with corresponding numbers;

[0011] Step 2: applying artificial simulated squeezing damage treatment with different forces to the Korla fragrant pear to obtain the processed X-ray images of the Korla fragrant pear with corresponding numbers;

[0012] Step 3: Compare the processed X-ray image with the reference X-ray image to obtain a compression damage image;

[0013] Step 4: Perform early hidden damage detection on the Korla fragrant pear to be inspected based on the squeeze damage image;

[0014] Step 5: Analyze the correlation between the squeeze damage image and force, establish a damage library, and compare and analyze it with the early hidden damage detection results.

[0015] The present invention is further configured as follows: the processed X-ray images in step 2 are X-ray images of Korla fragrant pears with intact appearance, and the different forces in step 2 include 0N-15N.

[0016] The present invention is further configured as follows: the placement angle and X-ray image acquisition angle of the Korla fragrant pear in the processed X-ray image in step 2 are consistent with those in the reference X-ray image in step 1, and the X-ray equipment parameters used in the two are consistent.

[0017] The present invention is further configured as follows: the acquisition of the compression damage image in step 3 includes:

[0018] D(x, y)=∣I1(x, y)-I2(x, y)∣

[0019] Where I1(x, y) is the reference X image, I2(x, y) is the processed X image, D(x, y) is the damaged image, and x and y are pixel coordinates.

[0020] The present invention is further configured as follows: the method of performing early hidden damage detection on the Korla fragrant pear to be detected according to the squeeze damage image in step 4 includes:

[0021] A1. Obtain an X-ray image of the Korla fragrant pear to be tested;

[0022] A2. Mark abnormal areas of X-ray images.

[0023] The present invention is further configured as follows: the abnormal area in A2 includes a dark area and a fracture line area.

[0024] The present invention is further configured as follows: the establishment of the damage library in step 5 includes:

[0025] B1. Calculate the damage area based on the damage image and determine the relationship between the damage area and the applied force;

[0026] B2. Collect the shape of the damage image and build a damage library corresponding to the damage area and applied force.

[0027] The present invention is further configured as follows: the comparison and analysis method with the early hidden damage detection result in step 5 includes:

[0028] C1. Calculate the area of ​​the marked abnormal region as the analysis area, and select the damage image with the area closest to the analysis area from the damage image library as the comparison image;

[0029] C2. Perform a similarity comparison between the marked abnormal area and the comparison image, and preset a correlation coefficient value as a standard threshold. When the similarity comparison result is greater than or equal to the standard threshold, the abnormal area is judged to be man-made damage. When the similarity comparison result is less than the standard threshold, the abnormal area is judged to be non-man-made damage.

[0030] The present invention is further configured as follows: the similarity comparison in C2 includes:

[0031]

[0032] Where A and B are the feature vectors of the marked abnormal area and the comparison image, respectively. Each feature vector contains n elements. Ai and Bi are the i-th elements of A and B, respectively. and are the average values ​​of A and B respectively, and r(A, B) is the correlation coefficient between A and B, and its value range is [-1, 1];

[0033] The closer the absolute value of r(A, B) is to 1, the more similar the shapes of A and B are; the closer the absolute value is to 0, the greater the difference in the shapes of A and B is. Specifically:

[0034] When r(A, B) equals 1, it means that A and B are completely positively correlated, that is, their shapes are very similar;

[0035] When r(A, B) is equal to -1, it means that A and B are completely negatively correlated, which is unlikely to occur in actual shape comparisons because we usually do not expect two shapes to be completely opposite in characteristics;

[0036] When r(A, B) is equal to 0, it means that A and B are unrelated, that is, their shapes have no obvious similarities in features.

[0037] The present invention is further configured as follows: in the step 4, if the result of the early invisible damage detection on the Korla fragrant pear shows a squeeze damage, the number of Korla fragrant pears with squeeze damage in the same batch is determined as the abnormal number, and when the abnormal number reaches a preset abnormal value, an abnormal step analysis is performed; otherwise, when the abnormal number does not reach the preset abnormal value, the abnormal analysis is terminated;

[0038] The abnormal step analysis includes:

[0039] D1. Obtain the damage pressure value of Korla fragrant pear according to the applied force value corresponding to the comparison image;

[0040] D2. According to the pressure value, manually determine the link that may be subjected to this kind of squeezing force during the transportation of Korla fragrant pears and regard it as the problematic link;

[0041] D3. Optimize the problematic links, including manual optimization and optimization of the transportation strength of Korla fragrant pear.

[0042] (III) Beneficial effects

[0043] The present invention provides a method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, which has the following beneficial effects:

[0044] (1) The present invention pre-squeezes Korla fragrant pears with different intensities to obtain squeeze damage shape information and corresponding active squeeze force numerical information. After obtaining X-ray images of the Korla fragrant pear to be tested and images of the squeeze damage site, it is not only convenient to analyze the pressure intensity of the Korla fragrant pear with squeeze damage, but also to judge the subjectivity of the squeeze damage by similarity comparison of the damage images. In this way, by combining the area comparison and similarity analysis of the squeeze damage image, a reliable reference condition is provided for judging whether the Korla fragrant pear is artificially actively damaged. In addition, combined with the actual transportation process of the Korla fragrant pear, it can provide an optimization direction for the transportation defects of the Korla fragrant pear and provide reliable guarantee for the low-loss or even lossless transportation of the Korla fragrant pear.

[0045] (2) The present invention uses the number of Korla fragrant pears with extrusion damage in the same batch as the abnormal number. When the abnormal number reaches a preset abnormal value, an abnormal step analysis is performed to perform a safety analysis on the transportation process of the Korla fragrant pear. When the Korla fragrant pear reaches the preset extrusion damage loss standard during the transportation process, the transportation process of the Korla fragrant pear can be optimized to ensure low-loss or even lossless transportation of the Korla fragrant pear. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] See also Figure 1 , the embodiment of the present invention provides the following technical solutions:

[0049] Embodiment 1

[0050] A method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, specifically comprising the following steps:

[0051] Step 1, collecting a number of groups of Korla fragrant pears with intact appearance, wherein the number of the groups is preferably fifteen, and the number of Korla fragrant pears in a group is set to at least three, and after marking with numbers, obtaining a reference X-ray image of the Korla fragrant pear with the corresponding number;

[0052] Step 2, applying artificial simulated squeezing damage treatment with different forces to the Korla fragrant pear, the different forces include 0N-15N, preferably 1N, 2N, 3N, 4N, 5N, 6N, 7N, 8N, 9N, 10N, 11N, 12N, 13N, 14N and 15N, for applying squeezing force to a group of Korla fragrant pears respectively, and obtaining processed X-ray images of Korla fragrant pears with corresponding numbers, wherein the processed X-ray images are X-ray images of Korla fragrant pears with intact appearance;

[0053] It should be noted that the placement angle of the Korla fragrant pear and the X-ray image acquisition angle in the processed X-ray image in step 2 and the reference X-ray image in step 1 are consistent, and the X-ray equipment parameters used in the two are consistent;

[0054] Step 3: Compare the processed X-ray image with the reference X-ray image to obtain the compression damage image. Specifically:

[0055] D(x, y)=∣I1(x, y)-I2(x, y)∣

[0056] Where I1(x, y) is the reference X image, I2(x, y) is the processed X image, D(x, y) is the damaged image, and x and y are pixel coordinates;

[0057] Step 4: Perform early invisible damage detection on the Korla pear to be tested based on the squeeze damage image. The specific methods include:

[0058] A1. Obtain an X-ray image of the Korla fragrant pear to be tested;

[0059] A2. Marking abnormal areas of the X-ray image, where the abnormal areas include dark areas and fracture line areas;

[0060] Step 5: Analyze the correlation between the squeeze damage image and force, establish a damage library, and compare and analyze it with the early hidden damage detection results.

[0061] As a preferred solution, the establishment of the damage library in step 5 includes:

[0062] B1. Calculate the damage area based on the damage image and determine the relationship between the damage area and the applied force;

[0063] B2. Collect the shape of the damage image and build a damage library corresponding to the damage area and applied force.

[0064] As a preferred solution, the comparison and analysis method with the early hidden damage detection results in step 5 includes:

[0065] C1. Calculate the area of ​​the marked abnormal region as the analysis area, and select the damage image with the area closest to the analysis area from the damage image library as the comparison image;

[0066] C2. Perform a similarity comparison between the marked abnormal area and the comparison image, and preset a correlation coefficient value as a standard threshold. When the similarity comparison result is greater than or equal to the standard threshold, the abnormal area is judged to be man-made damage. When the similarity comparison result is less than the standard threshold, the abnormal area is judged not to be man-made damage. The specific process of similarity comparison includes:

[0067]

[0068] Where A and B are the feature vectors of the marked abnormal area and the comparison image, respectively. Each feature vector contains n elements. Ai and Bi are the i-th elements of A and B, respectively. and are the average values ​​of A and B respectively, and r(A, B) is the correlation coefficient between A and B, and its value range is [-1, 1];

[0069] The closer the absolute value of r(A, B) is to 1, the more similar the shapes of A and B are; the closer the absolute value is to 0, the greater the difference between the shapes of A and B. Specifically:

[0070] When r(A, B) equals 1, it means that A and B are completely positively correlated, that is, their shapes are very similar;

[0071] When r(A, B) is equal to -1, it means that A and B are completely negatively correlated, which is unlikely to occur in actual shape comparisons because we usually do not expect two shapes to be completely opposite in characteristics;

[0072] When r(A, B) is equal to 0, it means that A and B are unrelated, that is, their shapes have no obvious similarities in features.

[0073] In this embodiment, the Korla fragrant pear is squeezed with different intensities in advance to obtain squeeze damage shape information and corresponding active squeeze force numerical information. After obtaining the X-ray image of the Korla fragrant pear to be detected and the squeeze damage site image, it is not only convenient to analyze the pressure intensity of the Korla fragrant pear with squeeze damage, but also possible to judge the subjectivity of the squeeze damage by similarity comparison of the damage images. In this way, by combining the area comparison and similarity analysis of the squeeze damage image, a reliable reference condition is provided for judging whether the Korla fragrant pear is artificially actively damaged.

[0074] Embodiment 2

[0075] A method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, specifically comprising the following steps:

[0076] Step 1, collecting a number of groups of Korla fragrant pears with intact appearance, wherein the number of the groups is preferably fifteen, and the number of Korla fragrant pears in a group is set to at least three, and after marking with numbers, obtaining a reference X-ray image of the Korla fragrant pear with the corresponding number;

[0077] Step 2, applying artificial simulated squeezing damage treatment with different forces to the Korla fragrant pear, the different forces include 0N-15N, preferably 1N, 2N, 3N, 4N, 5N, 6N, 7N, 8N, 9N, 10N, 11N, 12N, 13N, 14N and 15N, for applying squeezing force to a group of Korla fragrant pears respectively, and obtaining processed X-ray images of Korla fragrant pears with corresponding numbers, wherein the processed X-ray images are X-ray images of Korla fragrant pears with intact appearance;

[0078] It should be noted that the placement angle of the Korla fragrant pear and the X-ray image acquisition angle in the processed X-ray image in step 2 and the reference X-ray image in step 1 are consistent, and the X-ray equipment parameters used in the two are consistent;

[0079] Step 3: Compare the processed X-ray image with the reference X-ray image to obtain the compression damage image. Specifically:

[0080] D(x, y)=∣I1(x, y)-I2(x, y)∣

[0081] Where I1(x, y) is the reference X image, I2(x, y) is the processed X image, D(x, y) is the damaged image, and x and y are pixel coordinates;

[0082] Step 4: Perform early invisible damage detection on the Korla pear to be tested based on the squeeze damage image. The specific methods include:

[0083] A1. Obtain an X-ray image of the Korla fragrant pear to be tested;

[0084] A2. Marking abnormal areas of the X-ray image, where the abnormal areas include dark areas and fracture line areas;

[0085] Step 5: Analyze the correlation between the squeeze damage image and force, establish a damage library, and compare and analyze it with the early hidden damage detection results.

[0086] As a preferred solution, the establishment of the damage library in step 5 includes:

[0087] B1. Calculate the damage area based on the damage image and determine the relationship between the damage area and the applied force;

[0088] B2. Collect the shape of the damage image and build a damage library corresponding to the damage area and applied force.

[0089] As a preferred solution, the comparison and analysis method with the early hidden damage detection results in step 5 includes:

[0090] C1. Calculate the area of ​​the marked abnormal region as the analysis area, and select the damage image with the area closest to the analysis area from the damage image library as the comparison image;

[0091] C2. Perform a similarity comparison between the marked abnormal area and the comparison image, and preset a correlation coefficient value as a standard threshold. When the similarity comparison result is greater than or equal to the standard threshold, the abnormal area is judged to be man-made damage. When the similarity comparison result is less than the standard threshold, the abnormal area is judged not to be man-made damage. The specific process of similarity comparison includes:

[0092]

[0093]

[0094] Where A and B are the feature vectors of the marked abnormal area and the comparison image, respectively. Each feature vector contains n elements. Ai and Bi are the i-th elements of A and B, respectively. and are the average values ​​of A and B respectively, and r(A, B) is the correlation coefficient between A and B, and its value range is [-1, 1];

[0095] The closer the absolute value of r(A, B) is to 1, the more similar the shapes of A and B are; the closer the absolute value is to 0, the greater the difference between the shapes of A and B. Specifically:

[0096] When r(A, B) equals 1, it means that A and B are completely positively correlated, that is, their shapes are very similar;

[0097] When r(A, B) is equal to -1, it means that A and B are completely negatively correlated, which is unlikely to occur in actual shape comparisons because we usually do not expect two shapes to be completely opposite in characteristics;

[0098] When r(A, B) is equal to 0, it means that A and B are unrelated, that is, their shapes have no obvious similarities in features.

[0099] As a preferred solution, in step 4, the early invisible damage detection result of the Korla fragrant pear shows squeeze damage, and the number of Korla fragrant pears with squeeze damage in the same batch is determined as the abnormal number. When the abnormal number reaches a preset abnormal value, an abnormal step analysis is performed. On the contrary, when the abnormal number does not reach the preset abnormal value, the abnormal analysis is terminated.

[0100] The abnormal step analysis includes:

[0101] D1. Obtain the damage pressure value of Korla fragrant pear according to the applied force value corresponding to the comparison image;

[0102] D2. According to the pressure value, manually determine the link that may be subjected to this kind of squeezing force during the transportation of Korla fragrant pears and regard it as the problematic link;

[0103] D3. Optimize the problematic links, including manual optimization and optimization of the transportation strength of Korla fragrant pear.

[0104] The advantage of the second embodiment over the first embodiment is that the number of Korla fragrant pears with extrusion damage in the same batch is judged as the abnormal number. When the abnormal number reaches a preset abnormal value, an abnormal step analysis is performed to perform a safety analysis on the transportation process of the Korla fragrant pear. When the Korla fragrant pear reaches the preset extrusion damage loss standard during the transportation process, the transportation process of the Korla fragrant pear can be optimized to ensure low-loss or even lossless transportation of the Korla fragrant pear.

[0105] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0106] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting early hidden damage of Korla pear based on artificial intelligence. Features: The specific steps include: Step 1: Collect several groups of Korla fragrant pears with intact appearance, mark them with numbers, and obtain reference X-ray images of Korla fragrant pears with corresponding numbers; Step 2: applying artificial simulated squeezing damage treatment with different forces to the Korla fragrant pear to obtain the processed X-ray images of the Korla fragrant pear with corresponding numbers; Step 3: Compare the processed X-ray image with the reference X-ray image to obtain a compression damage image; Step 4: Perform early hidden damage detection on the Korla fragrant pear to be inspected based on the squeeze damage image; Step 5: Analyze the correlation between the squeeze damage image and force, establish a damage library, and compare and analyze it with the early hidden damage detection results.

2. According to claim 1, a method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, Features: The processed X-ray images in step 2 are X-ray images of Korla fragrant pears with intact appearance, and the different forces in step 2 include 0N-15N.

3. According to claim 1, a method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence, Features: The placement angle and X-ray image acquisition angle of the Korla fragrant pear in the processed X-ray image in step 2 are consistent with those in the reference X-ray image in step 1, and the X-ray equipment parameters used in the two are consistent.

4. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 1, Features: The acquisition of the compression damage image in step 3 includes: D(x,y)=∣I 1 (x,y)-I 2 (x,y)∣ In the formula, I 1 (x, y) is the reference X image, I 2 (x, y) is the processed X image, D(x, y) is the damaged image, and x and y are both pixel coordinates.

5. The method for detecting early hidden damage of Korla pear based on artificial intelligence according to claim 1, Features: The method of performing early hidden damage detection on the Korla fragrant pear to be detected according to the squeeze damage image in step 4 includes: A1. Obtain an X-ray image of the Korla pear to be tested; A2. Mark abnormal areas of X-ray images.

6. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 5, Features: The abnormal area in A2 includes a dark area and a fracture line area.

7. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 6, Features: The establishment of the damage library in step 5 includes: B1. Calculate the damage area based on the damage image and determine the relationship between the damage area and the applied force; B2. Collect the shape of the damage image and build a damage library corresponding to the damage area and applied force.

8. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 7, Features: The comparison and analysis method with the early hidden damage detection results in step 5 includes: C1. Calculate the area of ​​the marked abnormal region as the analysis area, and select the damage image with the area closest to the analysis area from the damage image library as the comparison image; C2. Perform a similarity comparison between the marked abnormal area and the comparison image, and preset a correlation coefficient value as a standard threshold. When the similarity comparison result is greater than or equal to the standard threshold, the abnormal area is judged to be man-made damage. When the similarity comparison result is less than the standard threshold, the abnormal area is judged to be non-man-made damage.

9. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 8, Features: The similarity comparison in C2 includes: Where A and B are the feature vectors of the marked abnormal area and the comparison image, respectively. Each feature vector contains n elements. Ai and Bi are the i-th elements of A and B, respectively. and are the average values ​​of A and B respectively, and r(A, B) is the correlation coefficient between A and B, and its value range is [-1, 1]; The closer the absolute value of r(A, B) is to 1, the more similar the shapes of A and B are; the closer the absolute value is to 0, the greater the difference between the shapes of A and B is.

10. The method for detecting early hidden damage of Korla fragrant pear based on artificial intelligence according to claim 9, Features: In the step 4, the early hidden damage detection result of the Korla fragrant pear shows the squeeze damage, and the number of Korla fragrant pears with squeeze damage in the same batch is determined as the abnormal number. When the abnormal number reaches a preset abnormal value, the abnormal step analysis is performed. On the contrary, when the abnormal number does not reach the preset abnormal value, the abnormal analysis is terminated; The abnormal step analysis includes: D1. Obtain the damage pressure value of Korla fragrant pear according to the applied force value corresponding to the comparison image; D2. According to the pressure value, manually determine the link that may be subjected to this kind of squeezing force during the transportation of Korla fragrant pears and regard it as the problematic link; D3. Optimize the problematic links, including manual optimization and optimization of the transportation strength of Korla fragrant pear.

Citation Information

Patent Citations

  • Method for detecting early recessive damage of Korla pear by fusing hyperspectral image features and spectral variables

    CN117491382A