Intelligent detection method and device applied to product sorting

Abstract

The application provides an intelligent detection method applied to product sorting, which comprises the following steps: obtaining a real-time image of a product to be sorted on a conveying belt, and pre-processing the real-time image to obtain a pre-processed real-time image; identifying the pre-processed real-time image based on a Canny edge detection algorithm to obtain an edge detection result; and matching the edge detection result based on a template matching algorithm to obtain a matching result and sorting based on the matching result. The real-time image is processed by the Canny edge detection algorithm to obtain the edge detection result; finally, the edge detection result is matched based on the template matching algorithm to obtain the matching result, and the sorting is performed based on the matching result, thereby solving the technical problems of poor real-time performance and low sorting accuracy in the prior art.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of machine vision, in particular to an intelligent detection method applied to product sorting. BACKGROUND

[0002] Product sorting is an important production link in industrial automation production, and traditional manual sorting or teaching industrial robot sorting cannot meet the needs of modern industry for labor productivity and production process automation level, which will directly affect the economic benefits and product quality of food processing enterprises, therefore, an automatic and intelligent detection system needs to be developed, the system has practical significance for improving the automation degree of enterprises, improving the production efficiency of enterprises and improving the product quality, and combination of machine vision technology and industrial robots can make the sorting operation have higher flexibility and reliability.

[0003] Lu Zhenli et al. start from the feature detection of cigarette boxes, recognize the true and false cigarettes by using the Matlab tool, and sort the cigarettes by controlling the robot hand. Zhou Jun et al. research a method for sorting Xinjiang walnut kernels according to the color and integrity of the walnut kernels based on machine vision technology. Tuong Phuoc Tho et al. sort and locate tomatoes according to the color, shape and other characteristics, and realize automatic sorting of tomatoes. However, there are some limitations in the application of these intelligent recognition technologies, such as poor flexibility, high cost, high power consumption, low accuracy and the like.

[0004] The application provides an intelligent detection method applied to product sorting, realizes a full-automatic assembly line, and realizes appearance detection of products by processing images on an assembly line transmitted by an industrial camera through edge detection and model matching, and accurately classifies the products, so that the technical problems of too large calculation amount, long time consumption, very slow speed, poor real-time performance, too long matching time and low sorting accuracy in the prior art are solved. SUMMARY

[0005] Therefore, it is necessary to provide an intelligent detection method applied to product sorting, so as to solve the technical problems of too large calculation amount, long time consumption, very slow speed, poor real-time performance, too long matching time and low sorting accuracy in the prior art.

[0006] The application provides an intelligent detection method applied to product sorting, which comprises the following steps:

[0007] obtaining real-time images of products to be sorted on a conveying belt, and pre-processing the real-time images to obtain pre-processed real-time images;

[0008] identifying the pre-processed real-time images based on a Canny edge detection algorithm to obtain an edge detection result;

[0009] The edge detection result is matched based on a template matching algorithm to obtain a matching result, and the matching result is used for sorting.

[0010] In some possible implementation manners, the real-time image is preprocessed to obtain a preprocessed real-time image, including:

[0011] The image is subjected to noise reduction and smoothing processing by a Gaussian filter with a Gaussian convolution kernel of 5x5 to obtain the preprocessed real-time image.

[0012] In some possible implementation manners, the preprocessed real-time image is identified based on a Canny edge detection algorithm to obtain an edge detection result, including:

[0013] The gradient amplitude and the gradient direction of each pixel point in the preprocessed real-time image are calculated based on the Canny edge detection algorithm to determine the gradient value and the gradient direction of each pixel point in the preprocessed real-time image.

[0014] Each pixel point gradient value in the preprocessed real-time image is subjected to non-maximum suppression based on the Canny edge detection algorithm to output a non-maximum suppression image.

[0015] The non-maximum suppression image is subjected to double-threshold detection based on the Canny edge detection algorithm to obtain image edge detection data.

[0016] Boundary tracing processing is performed based on the image edge detection data to obtain the edge detection result.

[0017] In some possible implementation manners, the gradient amplitude and the gradient direction of each pixel point in the preprocessed real-time image are calculated based on the Canny edge detection algorithm to determine the gradient value and the gradient direction of each pixel point in the preprocessed real-time image, including:

[0018] The gradient value and the gradient direction of each pixel point in the preprocessed real-time image are calculated based on a gradient value calculation formula and a gradient direction angle calculation formula.

[0019] In some possible implementation manners, each pixel point gradient value in the preprocessed real-time image is subjected to non-maximum suppression based on the Canny edge detection algorithm to output a non-maximum suppression image, including:

[0020] The gradient value of each pixel point in the real-time image is compared with the gradient value of an adjacent pixel point.

[0021] When the gradient value of the pixel point is the largest compared with the gradient value of the adjacent pixel point, the pixel point is reserved as an edge point, and otherwise, the pixel point is suppressed.

[0022] Output the non-maximum suppression image based on the edge points retained after the non-maximum suppression processing.

[0023] In some possible implementation manners, the non-maximum suppression image is subjected to double-threshold detection based on a Canny edge detection algorithm to obtain image edge data, including:

[0024] The high threshold value is obtained through a maximum entropy algorithm, and the low threshold value is obtained based on the high threshold value;

[0025] The non-maximum suppression image is subjected to double-threshold detection based on the high threshold value and the low threshold value, to determine strong edge points and weak edge points of the image data;

[0026] The image edge data is output based on the strong edge points and the weak edge points of the image data;

[0027] The edge points with the gradient value greater than the high threshold value are divided into strong edge points, the edge points with the gradient value less than the high threshold value and greater than the low threshold value are divided into weak edge points, and the edge points with the gradient value less than the low threshold value are suppressed.

[0028] In some possible implementation manners, the boundary tracing processing is performed based on the image edge detection data to obtain an edge detection result, including:

[0029] The strong edge points are connected as a fine line as a boundary based on the image edge data, the effective low gray scale surrounding the fine line is detected and connected, and a low gray scale mixed connection image edge is formed;

[0030] The low gray scale mixed connection image edge is subjected to binarization processing, the high and low gray scale areas or the perimeters of the low gray scale mixed connection image edge after the binarization processing are obtained, and the edge detection result is output.

[0031] In some possible implementation manners, the matching processing is performed on the edge detection result based on a template matching algorithm to obtain a matching result, including:

[0032] A standard information template of a product to be sorted is obtained;

[0033] The edge detection result is traversed based on the standard information template until the entire standard information module area is traversed to obtain the matching result;

[0034] The point matching step between the pre-extracted single product standard information template and the specific area with the maximum similarity in the selected template in the edge detection result image is as follows: any one pre-extracted single product standard information template is selected;

[0035] All feature points located in the pre-extracted single product standard information template are extracted;

[0036] extract all feature points in the image block in the result of edge detection, and use a brute-force matching algorithm to match the two sets of feature points;

[0037] select those feature point pairs with smaller Hamming distance of descriptors as the final matching result, and repeat the whole process until all templates are traversed.

[0038] In some possible implementations, the obtaining the matching result until the entire standard information module region is traversed comprises:

[0039] selecting a certain feature point in the quasi-information template as a selected point;

[0040] calculating the Hamming distance of descriptors between each selected point and a corresponding point in the result of edge detection;

[0041] when the Hamming distance of descriptors between the selected point and the corresponding point in the result of edge detection is within a preset first threshold range, the selected point and the corresponding point in the result of edge detection are high matching degree points;

[0042] when the Hamming distance of descriptors between the selected point and the corresponding point in the result of edge detection is not within the preset first threshold range, the selected point and the corresponding point in the result of edge detection are low matching degree points;

[0043] obtaining a matching result based on the proportion of high matching degree points in the corresponding points.

[0044] In some possible implementations, the sorting based on the matching result comprises:

[0045] when the matching result is greater than a preset second threshold, the product is a qualified product;

[0046] when the matching result is less than the preset second threshold, the product is an unqualified product.

[0047] Compared with the prior art, the beneficial effects of the above embodiment are that: the application obtains a real-time image of a product to be sorted on a conveying belt, and obtains a preprocessed real-time image by preprocessing the real-time image; an edge detection result is obtained by identifying the preprocessed real-time image based on a Canny edge detection algorithm; a matching result is obtained by matching the edge detection result based on a template matching algorithm, and the product is sorted based on the matching result. The real-time image is processed by the Canny edge detection algorithm to obtain an edge detection result; finally, the edge detection result is matched based on the template matching algorithm to obtain a matching result, and the product is sorted based on the matching result, thereby solving the technical problems of poor real-time performance and low sorting accuracy in the prior art. BRIEF DESCRIPTION OF DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description only show some embodiments of the present application, and all other drawings obtained by those skilled in the art without creative effort based on these drawings are within the scope of the present application.

[0049] Figure 1 An embodiment flow diagram of an intelligent detection method applied to product sorting provided by the present application is shown in the figure. DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of the present application.

[0051] It should be understood that the schematic drawings are not drawn according to the actual proportions. The flowchart used in the present application shows the operations implemented according to some embodiments of the present application. It should be understood that the operations of the flowchart can not be implemented in sequence, and the steps without logical context relationship can be reversed in sequence or implemented simultaneously. In addition, one or more other operations can be added to the flowchart or removed from the flowchart by those skilled in the art under the guidance of the content of the present application.

[0052] Some block diagrams shown in the drawings are functional entities, which do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0053] In this paper, the phrase "embodiment" means that the specific features, structures or characteristics described in conjunction with the embodiment can be included in at least one embodiment of the present application. The phrase appears in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment to other embodiments. Those skilled in the art explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0054] The embodiments of the present application provide an intelligent detection method applied to product sorting, which will be described below respectively.

[0055] Please refer to Figure 1 , Figure 1An embodiment flow diagram of an intelligent detection method applied to product sorting is provided in the application, and the intelligent detection method applied to product sorting comprises the following steps:

[0056] In the application, the real-time image of the product to be sorted on the conveying belt is acquired, and the real-time image is preprocessed to obtain a preprocessed real-time image.

[0057] In the application, the preprocessed real-time image is recognized based on a Canny edge detection algorithm to obtain an edge detection result.

[0058] In the application, the edge detection result is matched based on a template matching algorithm to obtain a matching result, and the product is sorted based on the matching result.

[0059] Compared with the prior art, the application acquires the real-time image of the product to be sorted on the conveying belt, pre-processes the real-time image to obtain a preprocessed real-time image, recognizes the preprocessed real-time image based on a Canny edge detection algorithm to obtain an edge detection result, matches the edge detection result based on a template matching algorithm to obtain a matching result, and sorts the product based on the matching result. The real-time image is processed based on the Canny edge detection algorithm to obtain the edge detection result, and finally the edge detection result is matched based on the template matching algorithm to obtain the matching result, and the product is sorted based on the matching result, thereby solving the technical problems of poor real-time performance and low sorting accuracy in the prior art.

[0060] In specific embodiments, an industrial camera of a DASLS series with a parameter of 800,000 pixels is used to scan the product on the conveying line at a speed of 64 frames per second, after the set information is collected, a photograph is triggered and a high-definition image of the product is acquired, and the image acquisition card transmits the image into the computer.

[0061] In some embodiments of the application, the preprocessed real-time image is obtained by pre-processing the real-time image, which comprises the following steps:

[0062] The real-time image is denoising and smoothing processed by a Gaussian filter with a Gaussian convolution kernel parameter of 5x5 to obtain the preprocessed real-time image.

[0063] In specific embodiments, the read-in image is filtered by a Gaussian first derivative filter, that is, the original image f(x, y) is filtered by a Gaussian function to obtain a smooth image data matrix, and (x, y) are image pixel coordinates.

[0064] The original image is convolved by the Gaussian filter function to obtain a smooth image after filtering, and the specific formula is as follows:

[0065] s(x, y) = f(x, y) * G(x, y, sigma)

[0066] wherein, in the formula, s(x, y) is a smooth image, and G(x, y, sigma) is a Gaussian filter function.

[0067] The Gaussian filter function is:

[0068] G(x, y, sigma) is a two-dimensional Gaussian function, and in a direction n, it is a first-order directional inverse of G(x, y), so:

[0069] wherein

[0070] wherein, in the formula, n is a direction vector, is a gradient vector, and the original image f(x, y) is convolved with G n , and the direction of n is changed, then G n *f(x, y) reaches the maximum value, and the n is the direction of the detected edge, and the value range of sigma is 0.3-3.

[0071] In some embodiments of the present application, the preprocessed real-time image is identified based on the Canny edge detection algorithm to obtain an edge detection result, including:

[0072] The gradient amplitude and gradient direction of each pixel point in the preprocessed real-time image are calculated based on the Canny edge detection algorithm to determine the gradient value and gradient direction of each pixel point in the preprocessed real-time image;

[0073] The gradient value of each pixel point in the preprocessed real-time image is non-maximum suppression based on the Canny edge detection algorithm to output a non-maximum suppression image;

[0074] The non-maximum suppression image is double-threshold detection based on the Canny edge detection algorithm to obtain image edge data;

[0075] The boundary tracking processing is performed based on the image edge detection data to obtain an edge detection result.

[0076] In some embodiments of the present application, the gradient amplitude and gradient direction of each pixel point in the preprocessed real-time image are calculated based on the Canny edge detection algorithm to determine the gradient value and gradient direction of each pixel point in the preprocessed real-time image, including:

[0077] The gradient value and gradient direction of each pixel point in the preprocessed real-time image are calculated based on the gradient value calculation formula and the gradient direction angle calculation formula.

[0078] In a specific embodiment, the gradient amplitude M and the direction angle A of each pixel point are calculated by the following formula:

[0079] M = ‖f(x,y)*G(x,y,σ)‖

[0080]

[0081] where ‖x‖≥0, and ‖x‖=0 if and only if x=0;

[0082] When ‖ax‖=|a|*‖x‖, where a is a number in the number field corresponding to the linear space;

[0083] When ‖x+y‖≤‖x‖+‖y‖.

[0084] In some embodiments of the application, the gradient value of each pixel point in the pre-processed real-time image is subjected to non-maximum suppression based on the Canny edge detection algorithm to output a non-maximum suppression image, comprising:

[0085] Comparing the gradient value of each pixel point in the real-time image with the gradient value of adjacent pixel points;

[0086] When the gradient value of the pixel point is the largest compared with the gradient value of adjacent pixel points, the pixel point is reserved as an edge point, otherwise the pixel point is suppressed;

[0087] Outputting a non-maximum suppression image based on the edge points reserved after non-maximum suppression processing.

[0088] In specific embodiments, the elimination of the scattered response accompanying edge detection can be achieved by comparing the gradient intensity currently possessed by the pixel with the gradient intensity of two pixels possessed along the positive and negative gradient directions. When the gradient intensity of the current pixel is found to be the largest compared with the gradient intensity of the other two pixels at the same time, the current pixel point can be saved and can serve as an edge point. If the result is not so, the current pixel point is suppressed.

[0089] In some embodiments of the application, the non-maximum suppression image is subjected to double-threshold detection based on the Canny edge detection algorithm to obtain image edge data, comprising:

[0090] Obtaining a high threshold value by means of the maximum entropy algorithm and calculating a low threshold value therefrom;

[0091] Detecting the non-maximum suppression image by means of double thresholds according to the high threshold value and the low threshold value to determine strong edge points and weak edge points of the image data;

[0092] Outputting image edge data based on the strong edge points and the weak edge points of the image data;

[0093] The edge points with the gradient value greater than the high threshold value are divided into strong edge points, the edge points with the gradient value less than the high threshold value and greater than the low threshold value are divided into weak edge points, and the edge points with the gradient value less than the low threshold value are suppressed.

[0094] In specific embodiments, the high threshold value is obtained by a maximum entropy algorithm, and the low threshold value is obtained based on the high threshold value, and the method for reducing the number of false edges is to use a threshold value for the non-maximum value suppressed image N(i,j), and assign zero values to all values less than the threshold value, and the thresholding result of the non-maximum value suppressed image amplitude is an edge array of an image I(i,j), the edge array obtained after thresholding still has false edges, a high threshold value Q1 is obtained by using a cumulative statistical histogram, and then a low threshold value Q2 is obtained.

[0095] If the response of the image signal is greater than the high threshold value, it must be an edge; if it is less than the low threshold value, it must not be an edge; if it is between the low threshold value and the high threshold value, it is seen whether there is an edge greater than the high threshold value in the eight adjacent pixels, if there is, it is an edge, otherwise it is not an edge.

[0096] In some embodiments of the present application, the boundary tracing processing is performed based on the image edge detection data to obtain an edge detection result, including:

[0097] Based on the image edge data, a strong edge point is made into a thin line as a boundary, and the effective low gray scale surrounding the thin line is connected to form a low gray scale mixed connection image edge.

[0098] The low gray scale mixed connection image edge is binarized to obtain the high and low gray scale areas or the perimeters of the binarized low gray scale mixed connection image edge, and the edge detection result is output.

[0099] In specific embodiments, based on the image edge data, a strong edge point is made into a thin line as a boundary, and the effective low gray scale surrounding the thin line is connected to form a low gray scale mixed connection image edge.

[0100] Specifically, the effective low gray scale surrounding a width of 3-11Pix of the thin line is connected to form the boundary tracing and low gray scale mixed connection image edge.

[0101] The image itself is composed of pixels with different gray scales, and the gray scale values exist, and the maximum and minimum functions are used to register the maximum gray scale value H and the minimum gray scale value L of the periphery, the gray scale threshold is an adjustable part on the software interface, that is, a gray scale value is set, and the gray scale value is greater than or equal to the value, and the gray scale value is reset to 255, and the binary image is obtained, after the following edge extraction is completed, the binary gray scale processing mode is used for processing, and the edge detection result is output.

[0102] Among them, a binary gray scale processing mode is specifically to determine the high and low gray scale areas or perimeters of the mixed connection image edge after binary gray scale processing.

[0103] First, the highest gray scale H and the lowest gray scale L of the current picture are obtained, wherein,

[0104]

[0105]

[0106] In the formula, k represents the number of each mechanism calculated.

[0107] According to the difference between the two gray scales, a suitable gray scale threshold is set, the binary pixel areas SH and SL of the detection region after binary processing are calculated, and the edge detection structure is obtained, that is, the high and low gray scale areas or perimeters of the mixed connection image edge.

[0108] It should be further pointed out that the effective low gray scale refers to being greater than 25Pix and being in the range of 20% to 60% of the effective highest gray scale.

[0109] In some embodiments of the present application, the matching processing based on the template matching algorithm on the edge detection result obtains a matching result, comprising:

[0110] Obtain the standard information template of the product to be sorted;

[0111] Based on the standard information template, the edge detection result is traversed until the entire standard information module area is traversed to obtain the matching result;

[0112] The point matching step between the pre-extracted single product standard information template and the specific area with the maximum similarity of the selected template in the result image of the edge detection is: selecting any one piece of pre-extracted single product standard information template;

[0113] Extract all feature points located in the pre-extracted single product standard information template;

[0114] Extract all feature points in the image block in the result of the edge detection, and use the brute force matching algorithm to match the two groups of feature points.

[0115] Select those feature point pairs with smaller Hamming distance of descriptors as the final matching result, and loop the whole process until all templates are traversed.

[0116] In some embodiments of the present application, the matching result is obtained by traversing the entire standard information module region, including:

[0117] Select a certain feature point in the quasi-information template as a selected point;

[0118] Calculate the Hamming distance of descriptors between each selected point and the corresponding point in the edge detection result;

[0119] When the Hamming distance of descriptors between the selected point and the corresponding point in the edge detection result is within a preset first threshold range, the selected point and the corresponding point in the edge detection result are high matching degree points;

[0120] When the Hamming distance of descriptors between the selected point and the corresponding point in the edge detection result is not within a preset first threshold range, the selected point and the corresponding point in the edge detection result are low matching degree points;

[0121] Based on the proportion of high matching degree points in the corresponding points, a matching result is obtained.

[0122] In some embodiments of the present application, the sorting based on the matching result includes:

[0123] When the matching result is greater than a preset second threshold, the product is a qualified product;

[0124] When the matching result is less than a preset second threshold, the product is an unqualified product.

[0125] Compared with the prior art, the present application obtains a real-time image of a product to be sorted on a conveying belt, and pre-processes the real-time image to obtain a pre-processed real-time image; based on a Canny edge detection algorithm, the pre-processed real-time image is identified to obtain an edge detection result; based on a template matching algorithm, the edge detection result is matched to obtain a matching result and the product is sorted based on the matching result. The real-time image is processed by the Canny edge detection algorithm to obtain an edge detection result; finally, the edge detection result is matched based on the template matching algorithm to obtain a matching result, and the product is sorted based on the matching result, solving the technical problems of poor real-time performance and low sorting accuracy in the prior art.

[0126] The recognition precision is improved by pre-processing the real-time image first, then the product is intelligently recognized based on the Canny edge detection algorithm and the template matching and machine learning algorithm, the recognition speed of the product is further improved, the detection speed is improved by one step based on the two algorithms for the detection of the substandard product and classification, the real-time performance of the detection is improved, the technical problems of poor real-time performance and low sorting accuracy in the prior art are solved, and multiple sorting devices can be linked, and automatic reminding can be performed when the distance between the products is abnormal and the running frequency of the sorting device is higher than the normal level.

[0127] The above describes in detail the intelligent detection method applied to product sorting provided by the application, the principle and implementation mode of the application are described by applying specific examples in this paper, the above embodiment description is only used to help understand the method and core idea of the application; meanwhile, for those skilled in the art, the specific implementation mode and application range will be changed according to the idea of the application, and the above description should not be understood as a limitation on the application.

Claims

1. An intelligent detection method applied to product sorting, characterized in that, The method comprises the following steps: acquiring a real-time image of a product to be sorted on a conveying belt, and pre-processing the real-time image to obtain a pre-processed real-time image; based on a Canny edge detection algorithm, identifying the pre-processed real-time image to obtain an edge detection result, including: based on the Canny edge detection algorithm, calculating the gradient amplitude and gradient direction of each pixel point in the pre-processed real-time image to determine the gradient value and gradient direction of each pixel point in the pre-processed real-time image; based on the Canny edge detection algorithm, performing non-maximum suppression on the gradient value of each pixel point in the pre-processed real-time image to output a non-maximum suppression image; based on the Canny edge detection algorithm, performing double-threshold detection on the non-maximum suppression image to obtain image edge detection data; based on the image edge detection data, performing boundary tracing processing to obtain the edge detection result; based on a template matching algorithm, performing matching processing on the edge detection result to obtain a matching result, and based on the matching result, sorting; the matching processing on the edge detection result based on the template matching algorithm to obtain the matching result, including: acquiring a standard information template of the product to be sorted; based on the standard information template, traversing the edge detection result until the entire standard information module area is traversed to obtain the matching result; the point matching step between the pre-extracted single product standard information template and the specific area with the maximum similarity of the selected template in the result image of the edge detection is: selecting an arbitrary piece of pre-extracted single product standard information template; extracting all feature points located in the pre-extracted single product standard information template; extracting all feature points in the image block in the result of the edge detection, and using a brute-force matching algorithm to match the two groups of feature points; selecting those feature point pairs with smaller Hamming distance of descriptors as the final matching result, and repeating the whole process until all templates are traversed; the step of obtaining the matching result by traversing the entire standard information module area includes: selecting a certain feature point in the standard information template as a selected point; calculating the descriptor Hamming distance between each selected point and the corresponding point in the result of the edge detection; when the descriptor Hamming distance between the selected point and the corresponding point in the result of the edge detection is within a preset first threshold range, the selected point and the corresponding point in the result of the edge detection are high-matching-degree points; when the descriptor Hamming distance between the selected point and the corresponding point in the result of the edge detection is not within the preset first threshold range, the selected point and the corresponding point in the result of the edge detection are low-matching-degree points; based on the proportion of the high-matching-degree points in the corresponding points, obtaining the matching result.

2. The intelligent detection method applied to product sorting according to claim 1, characterized in that, The pre-processing of the real-time image to obtain the pre-processed real-time image includes: performing noise reduction and smoothing processing on the image by a Gaussian filter with a 5x5 Gaussian convolution kernel to obtain the pre-processed real-time image.

3. The intelligent detection method applied to product sorting according to claim 1, characterized in that, the step of calculating the gradient amplitude and gradient direction of each pixel point in the pre-processed real-time image to determine the gradient value and gradient direction of each pixel point in the pre-processed real-time image includes: The gradient value and the gradient direction of each pixel point in the preprocessed real-time image are calculated based on the gradient value calculation formula and the gradient direction angle calculation formula.

4. The intelligent detection method applied to product sorting according to claim 1, characterized in that, The gradient value of each pixel point in the preprocessed real-time image is subjected to non-maximum suppression based on the Canny edge detection algorithm to output a non-maximum suppression image, including: Comparing the gradient value of each pixel point in the real-time image with the gradient value of adjacent pixel points; When the gradient value of the pixel point is the largest compared with the gradient value of adjacent pixel points, the pixel point is reserved as an edge point, otherwise the pixel point is suppressed; The non-maximum suppression image is output based on the edge points reserved after the non-maximum suppression processing.

5. The intelligent detection method for product sorting according to claim 1, characterized in that, The image edge data is obtained by performing double-threshold detection on the non-maximum suppression image based on the Canny edge detection algorithm, including: The high threshold value is obtained by the maximum entropy algorithm, and the low threshold value is calculated based on the high threshold value; The double-threshold detection is performed on the non-maximum suppression image based on the high threshold value and the low threshold value to determine the strong edge points and the weak edge points of the image data; The image edge data is output based on the strong edge points and the weak edge points of the image data; The edge points with the gradient value greater than the high threshold value are divided into strong edge points, the edge points with the gradient value less than the high threshold value and greater than the low threshold value are divided into weak edge points, and the edge points with the gradient value less than the low threshold value are suppressed.

6. The intelligent detection method for product sorting according to claim 1, wherein, The boundary tracing processing is performed based on the image edge detection data to obtain the edge detection result, including: The strong edge points are connected as a fine line as a boundary based on the image edge data, and the effective low gray surrounding the fine line is connected to form a low gray mixed connection image edge; The low gray mixed connection image edge is subjected to binarization processing to obtain the high and low gray area or the perimeter of the low gray mixed connection image edge after binarization processing, and the edge detection result is output.

7. The intelligent detection method for product sorting according to claim 1, wherein, The sorting based on the matching result includes: When the matching result is greater than a preset second threshold value, the product is a qualified product; When the matching result is less than the preset second threshold value, the product is an unqualified product.

Images