Plastic particle quality detection method and system based on optic nerves

Abstract

The invention relates to the technical field of optic nerve detection, in particular to a plastic particle quality detection method and system based on optic nerves, and the method comprises the following steps: capturing a plastic particle image through an industrial camera, carrying out the balance processing of the image contrast, extracting the brightness and color features of the image, and removing the noise interference; and adjusting the image brightness and optimizing the particle edge to obtain a particle optimization image. According to the invention, the accuracy and automation level of plastic particle quality detection are greatly enhanced by using the visual neural network and the image processing technology, and the edge and shape features of the particles can be accurately extracted from a complex background by automatically adjusting the image contrast and brightness and applying an advanced edge detection algorithm. According to the method, the visual quality of the image is optimized, tiny flaws such as cracks and bubbles of the particles can be effectively recognized and analyzed, and more detailed data support can be provided compared with a traditional method by accurately calculating the surface roughness and texture uniformity of the particles.

Description

Technical Field

[0001] The present invention relates to the technical field of visual nerve detection, and in particular to a method and system for detecting plastic particle quality based on visual nerve. Background Art

[0002] The field of visual inspection technology encompasses the use of image processing tools and methods to identify, classify, and measure objects. This field utilizes images captured by cameras or other imaging devices and analyzes image content through software algorithms to automatically detect and evaluate target objects. Core areas include image capture, image preprocessing, feature extraction, object detection, and classification. The overall technical field encompasses everything from basic image acquisition equipment to complex image processing and analysis techniques, including but not limited to image enhancement, noise removal, edge detection, and object recognition, among other image processing methods. These methods improve image quality and recognition accuracy to meet the needs of various industrial and commercial applications.

[0003] Among them, the visual neural-based plastic particle quality detection method refers to the use of vision-based technology to detect and evaluate the quality of plastic particles. The technical matters targeted include capturing images of plastic particles through imaging equipment, using visual neural networks to analyze images, and identifying defects or non-standard features in the particles. Specific methods include using specific image processing technology to improve the visibility of particles in the image, and applying trained neural network models to perform quality classification. This process involves image segmentation, feature standardization, and quality judgment.

[0004] Although existing technologies have slow processing speeds and limited accuracy in high-speed and automated industrial environments, especially in the detection of tiny defects in particles, conventional methods have difficulty accurately distinguishing subtle differences in the surface of particles due to the limitations of image processing algorithms, affecting product quality control standards. In addition, traditional visual inspection systems require a lot of manual adjustment and verification when processing large-scale data, which not only reduces production efficiency but also increases operational complexity. The lack of sufficiently intelligent analysis systems leads to inaccuracies in particle classification and quality assessment, which will cause waste of resources and increased costs in practical applications. Summary of the Invention

[0005] In order to solve the existing problems in the detection of minor defects in particles, conventional methods are difficult to accurately distinguish subtle differences in the surface of particles due to the limitations of image processing algorithms, which affects the quality control standards of products. In addition, traditional visual inspection systems require a lot of manual adjustment and verification when processing large-scale data, which not only reduces production efficiency but also increases operational complexity. The lack of a sufficiently intelligent analysis system leads to inaccurate particle classification and quality assessment, which will cause waste of resources and increased costs in practical applications. The embodiment of the present invention provides a plastic particle quality detection method and system based on visual nerves. The technical solution is as follows:

[0006] In one aspect, a method for detecting the quality of plastic particles based on visual nerves is provided, comprising the following steps:

[0007] S1: Capture plastic particle images using an industrial camera, balance the image contrast, extract image brightness and color features, adjust image brightness, and optimize particle edges to obtain optimized particle images.

[0008] S2: Based on the particle optimization image, edge detection is performed on the image through a visual neural network, pixel grayscale gradient values are calculated and change trends are analyzed, segmentation boundary connected areas are detected, particle contours are identified, and a particle shape feature set is obtained;

[0009] S3: Based on the particle shape feature set, extract the pixel points on the particle surface, calculate the pixel grayscale change rate, analyze the light reflectivity change, measure the brightness fluctuation range, identify the particle surface roughness, analyze the surface texture uniformity, and obtain the surface feature analysis results;

[0010] S4: using the surface feature analysis results, analyzing particle color difference, detecting particle cracks, bubbles and impurity distribution, analyzing particle morphology deviation, screening particle surface deformation areas, and obtaining particle defect indicators;

[0011] S5: Classify the defects according to the particle defect indicators, compare with the standard quality, analyze the defect differences, calculate the overall defect ratio, evaluate the quality grade of the particles, and obtain the particle evaluation results.

[0012] On the other hand, the particle optimization image includes the color saturation of the particles, the clarity of the particle edges and the noise suppression effect; the particle shape feature set includes the boundary length, closedness and shape complexity; the surface feature analysis results include the light reflectivity, brightness fluctuation range and surface uniformity index of the particles; the particle defect index includes the color difference, brightness offset, roughness deviation and contour deformation of the particles; the particle evaluation results include the defect type classification, defect ratio and particle quality grade of the particles.

[0013] On the other hand, the step of obtaining the particle optimization image is specifically as follows:

[0014] S101: Captures images of plastic particles using an industrial camera, extracts the initial brightness and color characteristics of the image, performs image equalization, adjusts light intensity to highlight image details, optimizes image visual quality, and generates contrast-optimized images.

[0015] S102: Based on the contrast-optimized image, extract key brightness and color features of the image, perform noise suppression to remove high-frequency interference signals in the image, optimize the edge contours of the particles, and obtain a particle contour image;

[0016] S103: Based on the particle contour image, sharpen the particle edges, adjust the local contrast to optimize the recognizability of the particle shape, eliminate edge blur, identify the shape characteristics of the particles, and obtain a particle optimized image.

[0017] On the other hand, the steps of obtaining the particle shape feature set are specifically as follows:

[0018] S201: Based on the particle optimized image, edge detection is performed on the image using a visual neural network, the grayscale gradient value of each pixel is calculated, pixels whose grayscale changes exceed a standard are screened, the particle boundary area is determined, and the pixel gradient change trend is analyzed to detect boundary connectivity and obtain a boundary detection result;

[0019] S202: Based on the boundary detection results, connected regions are extracted for segmentation, boundary regions that meet the particle shape standards are marked, non-continuous boundary regions are eliminated, and morphological parameters of each particle are calculated, including boundary length, closedness, and shape complexity, to obtain particle morphological analysis data;

[0020] S203: Calculate the mean value of the curvature change of the particle boundary based on the particle morphology analysis data, measure the smoothness of the contour, identify the particle morphology characteristics, screen the irregular contour area based on the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set.

[0021] On the other hand, the mean value of the curvature change of the particle boundary is calculated using the formula:

[0022]

[0023] Measure the smoothness of the contour, identify the particle morphology, screen the irregular contour area according to the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set;

[0024] Among them, K ε represents the mean value of curvature change, θ α Represents the angle change at the boundary pixel point α, (x α,y α ) represents the coordinate value of the particle boundary point α, and N represents the total number of boundary points on the particle contour.

[0025] On the other hand, the steps of obtaining the surface feature analysis results are specifically as follows:

[0026] S301: Based on the particle shape feature set, extracting pixel data on the particle surface, calculating the average grayscale change rate between the pixels, analyzing the change trend of light reflectivity on the particle surface, determining the particle surface roughness, and obtaining a particle roughness index;

[0027] S302: Calling the particle roughness index, extracting the surface light reflectivity, identifying the particle surface brightness fluctuation range, determining the brightness abnormality area, extracting the distribution of brightness abnormality pixels, and obtaining the particle brightness fluctuation characteristics;

[0028] S303: Based on the particle brightness fluctuation characteristics, calculate the particle surface brightness uniformity, set the particle texture reference value, compare the particle surface roughness with the particle texture reference value, analyze the particle surface texture uniformity, screen the texture abnormal area, and obtain the surface feature analysis result.

[0029] On the other hand, the grayscale change rate average between pixels is calculated using the formula:

[0030]

[0031] Analyze the change trend of light reflectivity on the particle surface, determine the particle surface roughness, and obtain the particle roughness index;

[0032] Among them, R η Represents the average grayscale change rate of the particle surface, I μ Represents the grayscale value of the pixel μ on the particle surface, N η Represents the total number of measurement points, I μ+1 -I μ Represents the grayscale difference between adjacent pixels, M η Represents the number of pixels in the window area used to calculate the local standard deviation, I μ,ν Represents the grayscale value of pixel ν in the window area, I μ,avg Represents the grayscale mean within the window area.

[0033] On the other hand, the steps for obtaining the particle defect index are specifically as follows:

[0034] S401: Using the surface feature analysis results, extracting color information of the particle surface, analyzing color deviations of adjacent areas, identifying areas with color difference abnormalities on the particle surface, and obtaining color difference features of the particle surface;

[0035] S402: Invoking the particle surface color difference feature, calculating the brightness offset of the particle surface, analyzing the change in light reflection intensity, determining the area where the brightness of the particle surface changes suddenly, identifying the distribution of bubbles or impurities, screening the brightness offset abnormal points, and obtaining the brightness offset degree of the particle;

[0036] S403: Based on the particle brightness deviation, the particle surface roughness deviation value is calculated, the crack morphology characteristics are extracted, the particle contour morphology deviation is analyzed, the abnormal area of the particle contour is determined, and compared with the particle standard quality, the bubbles, impurities and contour deformation areas on the particle surface are screened, and the particle defect index is obtained.

[0037] On the other hand, the steps of obtaining the particle evaluation results are specifically as follows:

[0038] S501: Extracting particle surface cracks, deformations, and impurity areas based on the particle defect indicators, analyzing the distribution characteristics of each defect, calculating the particle defect area ratio, classifying the particles according to defect type, and obtaining particle defect classification data;

[0039] S502: Based on the particle defect classification data, analyze key defect features on the particle surface, compare with quality standard particle data, retrieve color, morphology, and structural deviations of the defect area, count the number and distribution of particles of each defect type, and obtain particle quality comparison results;

[0040] S503: Based on the particle quality comparison results, analyze the defect area ratio of the particles, measure the overall quality uniformity of the particles, count the proportion of particles that meet the quality standards, calculate the overall defect ratio of the particles, evaluate the quality grade of the particles, and obtain the particle evaluation results.

[0041] On the other hand, a plastic particle quality detection system based on visual nerves is provided, which is applied to a plastic particle quality detection method based on visual nerves, including:

[0042] The image acquisition and optimization module uses an industrial camera to capture images of plastic particles, balance the image contrast, extract image brightness and color features, remove noise interference, optimize particle edges, and obtain particle optimized images;

[0043] The edge detection and shape recognition module uses a visual neural network to perform edge detection based on the particle optimization image, calculates pixel grayscale gradient values and analyzes change trends, detects and segments boundary connected areas, extracts particle contours, and obtains a particle shape feature set;

[0044] The surface feature analysis module extracts pixel points on the particle surface based on the particle shape feature set, calculates the grayscale change rate, analyzes the reflectivity change, identifies the particle roughness, measures the brightness fluctuation, analyzes the surface texture uniformity, and obtains the surface feature analysis results;

[0045] The particle defect detection module uses the surface feature analysis results to analyze particle color difference, detect particle cracks, bubbles and impurities, analyze particle morphology deviation, and obtain particle defect indicators;

[0046] The particle quality assessment module classifies defective particles according to the particle defect index, compares the defective particles with the standard quality, calculates the defect ratio, assesses the quality grade of the particles, and obtains the particle assessment result.

[0047] The beneficial effects brought about by the technical solution provided by the embodiment of the present invention include at least:

[0048] The use of visual neural networks and image processing technology has greatly enhanced the accuracy and automation level of plastic particle quality inspection. By automatically adjusting image contrast and brightness and applying advanced edge detection algorithms, it can accurately extract the edge and shape features of particles from complex backgrounds, which not only optimizes the visual quality of the image, but also effectively identifies and analyzes tiny defects in particles, such as cracks and bubbles. By accurately calculating the surface roughness and texture uniformity of the particles, it can provide more detailed data support than traditional methods, providing a more scientific and comprehensive evaluation basis for particle quality control, thereby reducing the misjudgment rate and improving the processing efficiency and automation level on the production line. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0050] Figure 1 It is a main step flow chart of the present invention;

[0051] Figure 2 is a flow chart of the steps of S1 of the present invention;

[0052] Figure 3 This is a flow chart of the steps of S2 of the present invention;

[0053] Figure 4 This is a flow chart of the steps of S3 of the present invention;

[0054] Figure 5 This is a flow chart of the steps of S4 of the present invention;

[0055] Figure 6 This is a flow chart of the steps of S5 of the present invention;

[0056] Figure 7 It is a system block diagram of the present invention. DETAILED DESCRIPTION

[0057] The technical solution of the present invention is described below in conjunction with the accompanying drawings.

[0058] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to indicate examples, illustrations, or explanations. Any embodiment or design described as an "exemplary" in the present invention should not be interpreted as being preferred or advantageous over other embodiments or designs. Rather, the use of the word "exemplary" is intended to present concepts in a concrete manner. Furthermore, in the embodiments of the present invention, "and / or" can mean both or either of the two.

[0059] In the embodiments of the present invention, the terms "image" and "picture" may be used interchangeably. It should be noted that, when the distinction between them is not emphasized, their intended meanings are the same. The terms "of," "corresponding," and "corresponding" may be used interchangeably. It should be noted that, when the distinction between them is not emphasized, their intended meanings are the same.

[0060] In the embodiments of the present invention, sometimes a subscript such as W1 may be written as a non-subscript such as W1. When the difference is not emphasized, the meanings to be expressed are the same.

[0061] In order to make the technical problems, technical solutions and advantages to be solved by the present invention clearer, a detailed description will be given below with reference to the accompanying drawings and specific embodiments.

[0062] The embodiment of the present invention provides a method for detecting the quality of plastic particles based on visual nerves, such as Figure 1 As shown, the following steps are included:

[0063] S1: Capture plastic particle images using an industrial camera, balance the image contrast, extract key brightness and color features, perform noise suppression to remove high-frequency interference signals, adjust the image brightness balance, and optimize the particle edge contour to obtain an optimized particle image.

[0064] S2: Based on the particle optimization image, the image is edge detected through a visual neural network, the grayscale gradient values of adjacent pixels are calculated, the pixel gradient change trend is analyzed to detect boundary connectivity, the connected areas are segmented, the boundary curvature is calculated, the boundary area that conforms to the particle outline is identified, and the particle shape feature set is obtained;

[0065] S3: Based on the particle shape feature set, extract the pixel points in the particle surface area, calculate the average grayscale change rate between the pixels, analyze the change in light reflectivity of the particle surface, measure the brightness fluctuation range and screen out the brightness abnormality area, identify the particle surface roughness, and then analyze the surface texture uniformity based on the particle texture reference value to obtain the surface feature analysis results;

[0066] S4: Using the surface feature analysis results, analyze the surface color difference of the differential particles, calculate the particle surface brightness offset, determine the distribution of bubbles or impurities, detect particle cracks based on the particle surface roughness, analyze the morphological deviation of the particle contour, screen the bubble, impurity and contour deformation areas on the particle surface, and obtain the particle defect index;

[0067] S5: Classify particle defects according to particle defect indicators, record information on cracks, deformations, or impurity particles, and compare them with particles of standard quality. Analyze the difference between each defect and the standard, calculate the overall defect ratio of the particles, evaluate the quality grade of the particles, and obtain the particle evaluation results.

[0068] The particle optimization image includes the color saturation of the particles, the clarity of the particle edges and the noise suppression effect. The particle shape feature set includes the boundary length, closedness and shape complexity. The surface feature analysis results include the light reflectivity, brightness fluctuation range and surface uniformity index of the particles. The particle defect indicators include the color difference, brightness offset, roughness deviation and contour deformation of the particles. The particle evaluation results include the defect type classification, defect ratio and particle quality grade of the particles.

[0069] like Figure 2 As shown in FIG, the steps for obtaining particle optimization images are as follows:

[0070] S101: Captures images of plastic particles using an industrial camera, extracts the initial brightness and color characteristics of the image, performs image equalization, adjusts light intensity to highlight image details, optimizes image visual quality, and generates contrast-optimized images.

[0071] Capture the plastic particle image through the industrial camera, obtain the original RGB image data, and extract the initial brightness value B of the pixel point i,j and color feature C i,j , use histogram equalization to calculate the equalized brightness distribution B' i,j , the formula is as follows: Among them B low and B high are the minimum and maximum brightness values respectively, ensuring brightness normalization. If the brightness standard deviation σ B <500, then adjust the contrast parameter to increase the brightness difference, and calculate the brightness deviation ΔB=|B' i,j -B mean|, if ΔB>30, perform gamma correction to adjust the brightness distribution: Where β is dynamically adjusted according to the average brightness of the image. If B mean <100, set β = 1.2, otherwise set β = 0.8. Next, calculate the local contrast gain for the uneven illumination area: Among them B local_mean and B local_std are the mean and standard deviation in the local window respectively, if G p,q >2, perform local smoothing to reduce over-enhancement, and finally use bilateral filtering to remove high-frequency noise, optimize the visual quality of the image, and generate a contrast-adjusted image.

[0072] S102: Based on the contrast optimization image, key brightness and color features of the image are extracted, noise suppression is performed to remove high-frequency interference signals in the image, and the edge contours of the particles are optimized to obtain a particle contour image;

[0073] Extract the key brightness feature B″ of the pixel i,j and color distribution C″ i,j , calculate the brightness variance like >50, indicating that there is a large deviation in the image brightness distribution, and further equalization is required. Adaptive histogram equalization is used to divide the image into N×M sub-blocks, and the brightness histogram of each sub-block is calculated and equalized. The equalization results of all sub-blocks are fused to optimize the image contrast, perform Gaussian noise suppression, set the filter window size k=5, and calculate the noise variance of local pixels. If the local noise level Perform low-pass filtering to suppress high-frequency noise, apply edge detection technology to extract particle contour information, and calculate the gradient amplitude M p,q and direction Θ p,q : in and Represents the brightness gradient in the horizontal and vertical directions respectively. If M p,q >20 and the gradient direction Θ p,q If the pixel is within the edge detection threshold range, the pixel is marked as an edge point, all edge points are extracted to form an edge contour, the particle boundary information is optimized, and the particle contour image is obtained.

[0074] S103: Based on the particle contour image, the particle edges are sharpened, the local contrast is adjusted to optimize the recognizability of the particle shape, the edge blur is eliminated, the shape characteristics of the particles are identified, and the particle optimized image is obtained.

[0075] Calculate the local contrast improvement factor C f (p,q): Among them B local_meanand B local_std are the mean and standard deviation of the local area respectively, if C f (p,q)>2, enhance the edge contrast of the area, and use adaptive filtering to remove the noise caused by over-sharpening. Then calculate the contour details after sharpening through edge gradient analysis, and calculate the contour clarity score S λ : Where W p,q is the edge area weight matrix, if S λ >80, indicating a clear outline. If S λ <50, perform secondary enhancement and calculate the sharpening gain factor K of the local area p,q And make dynamic adjustments: Where ζ is the sharpening adjustment coefficient and is set to ζ = 0.5. Morphological analysis is performed on the enhanced image to match the particle contours and calculate the shape characteristic parameters of the particles, including boundary length, morphological complexity, and closure, to obtain the particle optimized image.

[0076] like Figure 3 As shown in Figure 2, the steps for obtaining the particle shape feature set are as follows:

[0077] S201: Based on the particle optimization image, perform edge detection on the image through a visual neural network, calculate the grayscale gradient value of each pixel, filter out pixels with grayscale changes exceeding the standard, determine the particle boundary area, analyze the pixel gradient change trend, detect boundary connectivity, and obtain boundary detection results;

[0078] Use the visual neural network to extract the multi-layer features of the image, perform edge detection on the image, input the image to the first layer of the convolutional neural network (CNN), and use the convolution kernel W m,n Extract local pixel gradient information: B' i,j =∑ m,n B i+m,j+n W m,n , where B i,j is the pixel value of the input image, B' i,j The feature map value after convolution calculation is used to enhance the edge features using the ReLU activation function of the neural network to ensure nonlinear transformation, making the edge detection robust to particle images of different contrasts: B″ i,j =max(0,B' i,j ), calculate the gradient component B″ of the image in the horizontal and vertical directions x and B″ y , and calculate its gradient magnitude: Among them, G i,j is the gradient amplitude of the pixel point, if G i,jIf the pixel is >20, the pixel is determined to be a candidate boundary point. In order to enhance the boundary connectivity of the particles, the long short-term memory (LSTM) unit of the visual neural network is used to analyze the pixel gradient change trend, define the boundary pixel change within the time step t, and calculate its state update: H t =σ(W h ·H t-1 +W x ·G t +b), where H t is the hidden state at time t, G t is the gradient information of the pixel at the current step, W h and W x is the weight matrix of the neural network, and σ is the activation function. This process can effectively track the changes in the particle boundaries and avoid the appearance of broken or blurred boundaries. Based on the marked boundary pixels, connectivity detection is performed and the total number of pixels N in the connected area is calculated using the depth-first search (DFS) method. c : If N c If is <10, isolated noise points are removed, all valid boundary points are integrated to form the particle boundary region, and the boundary detection result is obtained.

[0079] S202: Based on the boundary detection results, connected regions are extracted for segmentation, boundary regions that meet the particle shape standards are marked, non-continuous boundary regions are eliminated, and morphological parameters of each particle are calculated, including boundary length, closedness, and shape complexity, to obtain particle morphological analysis data;

[0080] Based on the boundary detection results, all boundary connected areas are extracted and the pixel group of the connected area is defined as R k , calculate its area: If A k <50, the region is removed as an invalid boundary. Then, the boundary length P of the particle is calculated. k : Among them, δ(i,j) indicates whether the pixel point belongs to the boundary. If δ(i,j)=1, the pixel point is included in the perimeter calculation to evaluate the closedness of the particle shape C k : If C k <0.8, then mark the area as irregular shaped particles and calculate the shape complexity S k , defined as the fractal dimension of the boundary: If S k If the value is >1.2, the particle boundary is considered to be complex, and the morphological parameters of all particles are sorted out to obtain particle morphological analysis data.

[0081] S203: Based on the particle morphology analysis data, calculate the mean curvature change of the particle boundary, measure the smoothness of the contour, identify the particle morphology characteristics, screen the irregular contour area based on the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set.

[0082] Calculate the mean curvature change of the particle boundary using the formula:

[0083]

[0084] Measure the smoothness of the contour, identify the particle morphology, screen the irregular contour area according to the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set;

[0085] Among them, K ε represents the mean value of curvature change, θ α Represents the angle change at the boundary pixel point α, (x α ,y α ) represents the coordinate value of the particle boundary point α, and N represents the total number of boundary points on the particle contour;

[0086] Calculate the angle change θ α , the angle change of the particle boundary point is calculated by the gradient direction of the boundary pixel, and the gradient direction is defined as:

[0087]

[0088] Among them, x α ,y α Represents the coordinates of the particle boundary point α. The parameters are measured by visual detection equipment to obtain the position of the particle edge point and calculate its gradient change;

[0089] Calculate the cumulative absolute value of the angle change at all boundary points:

[0090]

[0091] The actual measured particle boundary point coordinates are as follows:

[0092] (x1,y1)=(10,15), (x2,y2)=(14,18), (x3,y3)=(18,22);

[0093] Calculate the gradient direction:

[0094]

[0095]

[0096] |θ2-θ1|=|45 ° -36.87 ° |=8.13° ;

[0097] Calculation of cumulative angle change:

[0098]

[0099] Compute the Euclidean distance between adjacent boundary points:

[0100]

[0101] For the above point set:

[0102]

[0103]

[0104] Total distance accumulation:

[0105]

[0106] Calculate the mean curvature change

[0107] Bring in data:

[0108]

[0109] Curvature change mean K ε =0.763 indicates the degree of change in the average curvature of the particle boundary. This value is used to identify whether the boundary has sharp angles, bends or irregular shapes. If the value is greater than 1, it indicates that the particle contour is significantly deformed, and the irregular boundary area is analyzed.

[0110] like Figure 4 As shown in FIG, the steps for obtaining the surface feature analysis results are as follows:

[0111] S301: Based on the particle shape feature set, extract the pixel data of the particle surface, calculate the average grayscale change rate between the pixels, analyze the change trend of the light reflectivity of the particle surface, determine the particle surface roughness, and obtain the particle roughness index;

[0112] Calculate the mean grayscale change rate between pixels using the formula:

[0113]

[0114] Analyze the change trend of light reflectivity on the particle surface, determine the particle surface roughness, and obtain the particle roughness index;

[0115] Among them, R η Represents the average grayscale change rate of the particle surface, I μ Represents the grayscale value of the pixel μ on the particle surface, N ηRepresents the total number of measurement points, I μ+1 -I μ Represents the grayscale difference between adjacent pixels, M η Represents the number of pixels in the window area used to calculate the local standard deviation, I μ,ν Represents the grayscale value of pixel ν in the window area, I μ,avg Represents the grayscale mean within the window area;

[0116] Calculate the pixel grayscale difference I μ+1 -I μ ;

[0117] Gray value of pixel point on particle surface I μ The grayscale value is obtained from an industrial camera sensor and ranges from 0 to 255. The sampling pixels of the particle image are as follows:

[0118] I1=120, I2=125, I3=135, I4=140;

[0119] Calculate the grayscale difference between pixels:

[0120] |I2-I1|=|125-120|=5;

[0121] |I3-I2|=|135-125|=10;

[0122] |I4-I3|=|140-135|=5;

[0123] The accumulation of all grayscale differences:

[0124]

[0125] Calculate the grayscale standard deviation within the local window;

[0126] The local area pixel window is selected M η = 4 pixels, the pixel values are as follows: I 1,1 =120,I 1,2 =122,I 1,3 =123,I 1,4 =124;

[0127] Calculation of local area grayscale mean:

[0128]

[0129]

[0130] Calculation of grayscale standard deviation within a local window:

[0131]

[0132]

[0133] Calculate the mean grayscale change rate;

[0134]

[0135]

[0136]

[0137] The mean value of the grayscale change rate of the particle surface R η =9.87 represents the grayscale change trend of the particle surface. The higher R η The value indicates that the gray distribution on the particle surface is uneven, there are rough areas, and the lower R η A value of 0 indicates that the particle surface is smooth. This parameter can be used to further analyze the optical reflection characteristics and quality assessment of the particle surface.

[0138] S302: Calling the particle roughness index, extracting the surface light reflectivity, identifying the particle surface brightness fluctuation range, determining the brightness abnormality area, extracting the distribution of brightness abnormality pixels, and obtaining the particle brightness fluctuation characteristics;

[0139] Extract particle surface light reflectance data R f , analyze the light reflection characteristics and calculate the brightness change curve L(x): If the fluctuation of L(x) in a certain area exceeds the set standard value T L =20, it is determined that the brightness fluctuation in this area is large and there is a brightness anomaly. Calculate the local light intensity difference ΔL on the particle surface. i,j :ΔL i,j =|L i+1,j -L i,j |, if ΔL i,j >25, the point is identified as a brightness abnormal area and its pixel coordinates are recorded. Then, the distribution of brightness abnormal areas is analyzed and the proportion of brightness abnormal areas P is calculated. L : Among them, N abnormal is the total number of abnormal brightness pixels, N total is the total number of pixels on the particle surface, if P L If the value is >10%, it indicates that there is a large range of brightness fluctuation on the particle surface, and the particle brightness fluctuation characteristics are obtained.

[0140] S303: Based on the particle brightness fluctuation characteristics, calculate the particle surface brightness uniformity, set the particle texture baseline value, compare the particle surface roughness with the particle texture baseline value, analyze the particle surface texture uniformity, screen the texture abnormal areas, and obtain the surface feature analysis results.

[0141] Calculate the particle surface brightness uniformity U L : Among them, σ L is the standard deviation of surface brightness, L mean is the average surface brightness, if U L <0.8, it indicates that the brightness distribution of the particle surface is uneven. Then, the particle surface texture benchmark value T is established. T , defined as the mean value of the grayscale fluctuation on the surface of normal particles: If the grayscale fluctuation of a certain area exceeds T T If the particle surface texture uniformity is 1.5 times of that, it is marked as a texture abnormal area, and the particle surface texture uniformity C is calculated. T : Among them, N smooth is the total number of pixels that meet the texture benchmark, N total is the total number of pixels on the particle surface, if C T If the value is less than 0.85, the surface texture of the particle is judged to be abnormal, and the coordinates of the abnormal area are recorded. All the texture abnormal areas are screened to obtain the surface feature analysis results.

[0142] like Figure 5 As shown in the figure, the steps for obtaining particle defect indicators are as follows:

[0143] S401: Using the surface feature analysis results, extracting color information of the particle surface, analyzing the color deviation of adjacent areas, identifying areas with color difference abnormalities on the particle surface, and obtaining color difference characteristics of the particle surface;

[0144] Extract the RGB color channel data R of the particle surface i,j ,G i,j ,B i,j , calculate the color deviation value ΔC of adjacent areas i,j :

[0145]

[0146] If ΔC i,j >T C (where T C is the color difference threshold set, such as 15), then mark the area as the color difference abnormal area, and calculate the color difference mean C of the particle surface mean and color difference standard deviation σ C :

[0147]

[0148] If σ CIf the value is >10, it indicates that the color difference distribution on the particle surface is uneven. Further analysis of the color difference change area is required, and all color difference abnormal points are screened and marked to obtain the color difference characteristics of the particle surface.

[0149] S402: Invoke the color difference characteristics of the particle surface, calculate the brightness offset of the particle surface, analyze the change in light reflection intensity, determine the area where the brightness of the particle surface changes suddenly, identify the distribution of bubbles or impurities, screen out the brightness offset abnormal points, and obtain the brightness offset degree of the particle;

[0150] Calculate the brightness distribution L on the particle surface i,j :L i,j =0.299R i,j +0.587G i,j +0.114B i,j Calculate the brightness offset ΔL i,j :ΔL i,j =|L i+1,j -L i,j |, if ΔL i,j >T L (Set the brightness offset threshold, such as 20), then determine that the brightness of the pixel has a sudden change. Analyze the change in light reflection intensity and calculate the reflection ratio R f : Among them, L specular is the mirror reflected light, L diffuse is diffuse reflected light, L ambient is the ambient light. If R f If the value is >0.5, there are bubbles or impurities on the particle surface. All abnormal brightness deviation points are screened and marked to obtain the particle brightness deviation degree.

[0151] S403: Based on the particle brightness deviation, the particle surface roughness deviation value is measured, the crack morphology characteristics are extracted, the particle contour morphology deviation is analyzed, the abnormal area of the particle contour is determined, and compared with the particle standard quality, the bubble, impurity and contour deformation area of the particle surface are screened, and the particle defect index is obtained.

[0152] Calculate the roughness deviation R of the particle surface q : If R q >10, there is a significant change in the roughness of the particle surface, extract the crack morphology characteristics, and calculate the boundary discontinuity D k : Among them, P k is the crack boundary length, A k is the crack area, if D k If the value is >5, the crack morphology is obvious. According to the particle shape standard, the abnormal contour areas are screened, including bubbles, impurities and morphological deformation areas on the particle surface, to obtain the particle defect index.

[0153] like Figure 6 As shown in the figure, the steps for obtaining the particle evaluation results are as follows:

[0154] S501: Based on the particle defect indicators, the particle surface cracks, deformations, and impurity areas are extracted, the distribution characteristics of each defect are analyzed, the particle defect area ratio is calculated, the particles are classified according to the defect type, and the particle defect classification data is obtained;

[0155] According to the particle defect index, the particle surface cracks, deformation and impurity areas are extracted, and the binary marking matrix D of the defect area is defined. α,β , where D α,β =1 represents a defective pixel, D α,β =0 represents a normal pixel. Calculate the distribution characteristics of each defect type, including area A δ , density ρ δ And shape parameter: A δ =∑ α,β D α,β , Among them, A total is the total area of the particles. δ >0.05, the particle is marked as a serious defect particle. In addition, the shape complexity S of the defect is calculated. δ : Among them, P δ is the perimeter of the defect boundary, if S δ If the value is >10, it indicates that the defect is a crack or foreign matter. The particles are classified according to the defect type to obtain particle defect classification data.

[0156] S502: Based on the particle defect classification data, analyze the key defect features on the particle surface, compare with the quality standard particle data, call the color, morphology and structural deviation of the defect area, count the number and distribution of particles of each defect type, and obtain the particle quality comparison results;

[0157] Based on the particle defect classification data, key defect characteristic values of the particle surface are extracted, including color deviation ΔC δ 、Morphological characteristics F δ and structural deviation S δ , calculate the average color difference of the defect area: ∑ (α,β)∈D |C α,β -C ref |, where C α,β is the color value of the defect area, C ref is the color mean of the standard particles. If ΔC δ If the value is >20, there is a serious color deviation. Compare the data of the quality standard particles and calculate the difference in morphological characteristics: Among them, Anorm is the morphological parameter of the standard particle, if F δ <0.9 or F δ If the value is >1.1, it indicates that the particles have obvious morphological deviations. The number of particles with each defect type is counted, and the proportion of defects in particles from different batches is calculated to obtain the particle quality comparison results.

[0158] S503: Based on the particle quality comparison results, analyze the defect area ratio of the particles, measure the overall quality uniformity of the particles, count the proportion of particles that meet the quality standards, calculate the overall defect ratio of the particles, evaluate the quality grade of the particles, and obtain the particle evaluation results.

[0159] Based on the particle mass comparison results, calculate the defect area ratio P of the entire particle Γ : If P Γ >0.1, it indicates that there is a problem with the quality of the particles in this batch. Measure the uniformity of the overall quality of the particles U Γ : Among them, σ Γ is the standard deviation of the particle quality score, Q mean is the average particle mass, if U Γ If the value is <0.85, it indicates that the particle quality distribution is uneven. The proportion of particles that meet the quality standards is counted, the overall defect ratio is calculated, and the particles are graded according to the quality grade standards to obtain the particle evaluation results.

[0160] like Figure 7 As shown, a plastic particle quality detection system based on visual nerves includes:

[0161] The image acquisition and optimization module uses an industrial camera to capture images of plastic particles, balance the image contrast, extract image brightness and color features, remove noise interference, optimize particle edges, and obtain particle optimized images;

[0162] The edge detection and shape recognition module uses a particle optimization image and a visual neural network to perform edge detection, calculate pixel grayscale gradient values and analyze change trends, detect and segment boundary connected areas, extract particle contours, and obtain a particle shape feature set.

[0163] The surface feature analysis module extracts pixel points on the particle surface based on the particle shape feature set, calculates the grayscale change rate, analyzes reflectivity changes, identifies particle roughness, measures brightness fluctuations, analyzes surface texture uniformity, and obtains surface feature analysis results;

[0164] The particle defect detection module uses surface feature analysis results to analyze particle color difference, detect particle cracks, bubbles and impurities, analyze particle morphology deviations, and obtain particle defect indicators;

[0165] The particle quality assessment module classifies defective particles according to particle defect indicators, compares them with standard quality, calculates the defect ratio, evaluates the quality level of the particles, and obtains the particle assessment results.

[0166] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0167] In this disclosure, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of a, b, or c" can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0168] It should be understood that in various embodiments of the present invention, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0169] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present invention.

[0170] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described equipment, devices and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0171] In the several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interface, indirect coupling or communication connection of the device or unit, which can be electrical, mechanical or other forms.

[0172] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0173] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0174] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0175] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A method for detecting the quality of plastic particles based on visual nerves, characterized in that: The method comprises: S1: Capture plastic particle images using an industrial camera, balance the image contrast, extract image brightness and color features, adjust image brightness, and optimize particle edges to obtain optimized particle images. S2: Based on the particle optimization image, edge detection is performed on the image through a visual neural network, pixel grayscale gradient values are calculated and change trends are analyzed, segmentation boundary connected areas are detected, particle contours are identified, and a particle shape feature set is obtained; S3: Based on the particle shape feature set, extract the pixel points on the particle surface, calculate the pixel grayscale change rate, analyze the light reflectivity change, measure the brightness fluctuation range, identify the particle surface roughness, analyze the surface texture uniformity, and obtain the surface feature analysis results; S4: using the surface feature analysis results, analyzing the color difference of the particles, detecting the distribution of cracks, bubbles and impurities in the particles, analyzing the morphological deviation of the particles, screening the deformation area on the particle surface, and obtaining the particle defect index; S5: Classify the defects according to the particle defect indicators, compare with the standard quality, analyze the defect differences, calculate the overall defect ratio, evaluate the quality grade of the particles, and obtain the particle evaluation results.

2. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The particle optimization image includes the color saturation, particle edge clarity and noise suppression effect of the particles; the particle shape feature set includes boundary length, closedness and shape complexity; the surface feature analysis results include the light reflectivity, brightness fluctuation range and surface uniformity index of the particles; the particle defect indicators include the color difference, brightness offset, roughness deviation and contour deformation of the particles; and the particle evaluation results include the defect type classification, defect ratio and particle quality grade of the particles.

3. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The steps of obtaining the particle optimization image are specifically as follows: S101: Captures images of plastic particles using an industrial camera, extracts the initial brightness and color characteristics of the image, performs image equalization, adjusts light intensity to highlight image details, optimizes image visual quality, and generates contrast-optimized images. S102: Based on the contrast-optimized image, extract key brightness and color features of the image, perform noise suppression to remove high-frequency interference signals in the image, optimize the edge contours of the particles, and obtain a particle contour image; S103: Based on the particle contour image, sharpen the particle edges, adjust the local contrast to optimize the recognizability of the particle shape, eliminate edge blur, identify the shape characteristics of the particles, and obtain a particle optimized image.

4. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The steps of obtaining the particle shape feature set are specifically as follows: S201: Based on the particle optimized image, edge detection is performed on the image using a visual neural network, the grayscale gradient value of each pixel is calculated, pixels whose grayscale changes exceed a standard are screened, the particle boundary area is determined, and the pixel gradient change trend is analyzed to detect boundary connectivity and obtain a boundary detection result; S202: Based on the boundary detection results, connected regions are extracted for segmentation, boundary regions that meet the particle shape standards are marked, non-continuous boundary regions are eliminated, and morphological parameters of each particle are calculated, including boundary length, closedness, and shape complexity, to obtain particle morphological analysis data; S203: Based on the particle morphology analysis data, calculate the mean curvature change of the particle boundary, measure the smoothness of the contour, identify the particle morphology characteristics, screen the irregular contour area according to the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set.

5. The method for detecting plastic particle quality based on visual nerve according to claim 4, characterized in that: The calculation of the mean curvature change of the particle boundary adopts the formula: Measure the smoothness of the contour, identify the particle morphology, screen the irregular contour area according to the particle shape parameters, mark the characteristic points of the particle contour, and obtain the particle shape feature set; Among them, K ε represents the mean value of curvature change, θ α Represents the angle change at the boundary pixel point α, (x α ,y α ) represents the coordinate value of the particle boundary point α, and N represents the total number of boundary points on the particle contour.

6. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The steps of obtaining the surface feature analysis results are specifically as follows: S301: Based on the particle shape feature set, extracting pixel data on the particle surface, calculating the average grayscale change rate between the pixels, analyzing the change trend of light reflectivity on the particle surface, determining the particle surface roughness, and obtaining a particle roughness index; S302: Calling the particle roughness index, extracting the surface light reflectivity, identifying the particle surface brightness fluctuation range, determining the brightness abnormality area, extracting the distribution of brightness abnormality pixels, and obtaining the particle brightness fluctuation characteristics; S303: Based on the particle brightness fluctuation characteristics, calculate the particle surface brightness uniformity, set the particle texture reference value, compare the particle surface roughness with the particle texture reference value, analyze the particle surface texture uniformity, screen the texture abnormal area, and obtain the surface feature analysis result.

7. The method for detecting plastic particle quality based on visual nerve according to claim 6, characterized in that: The calculation of the average grayscale change rate between pixels adopts the formula: Analyze the change trend of light reflectivity on the particle surface, determine the particle surface roughness, and obtain the particle roughness index; Among them, R η Represents the average grayscale change rate of the particle surface, I μ Represents the grayscale value of the pixel μ on the particle surface, N η Represents the total number of measurement points, I μ+1 -I μ Represents the grayscale difference between adjacent pixels, M η Represents the number of pixels in the window area used to calculate the local standard deviation, I μ,ν Represents the grayscale value of pixel ν in the window area, I μ,avg Represents the grayscale mean within the window area.

8. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The steps for obtaining the particle defect index are specifically as follows: S401: Using the surface feature analysis results, extracting color information of the particle surface, analyzing color deviations of adjacent areas, identifying areas with color difference abnormalities on the particle surface, and obtaining color difference features of the particle surface; S402: Invoking the particle surface color difference feature, calculating the brightness offset of the particle surface, analyzing the change in light reflection intensity, determining the area where the brightness of the particle surface changes suddenly, identifying the distribution of bubbles or impurities, screening the brightness offset abnormal points, and obtaining the brightness offset degree of the particle; S403: Based on the particle brightness deviation, the particle surface roughness deviation value is calculated, the crack morphology characteristics are extracted, the particle contour morphology deviation is analyzed, the abnormal area of the particle contour is determined, and compared with the particle standard quality, the bubbles, impurities and contour deformation areas on the particle surface are screened, and the particle defect index is obtained.

9. The method for detecting plastic particle quality based on visual nerve according to claim 1, characterized in that: The steps for obtaining the particle evaluation results are specifically as follows: S501: Extracting particle surface cracks, deformations, and impurity areas based on the particle defect indicators, analyzing the distribution characteristics of each defect, calculating the particle defect area ratio, classifying the particles according to defect type, and obtaining particle defect classification data; S502: Based on the particle defect classification data, analyze key defect features on the particle surface, compare with quality standard particle data, retrieve color, morphology, and structural deviations of the defect area, count the number and distribution of particles of each defect type, and obtain particle quality comparison results; S503: Based on the particle quality comparison results, analyze the defect area ratio of the particles, measure the overall quality uniformity of the particles, count the proportion of particles that meet the quality standards, calculate the overall defect ratio of the particles, evaluate the quality grade of the particles, and obtain the particle evaluation results.

10. A plastic particle quality detection system based on visual nerves, wherein the plastic particle quality detection system based on visual nerves is used to implement the plastic particle quality detection method based on visual nerves according to any one of claims 1 to 9, characterized in that: The system comprises: The image acquisition and optimization module uses an industrial camera to capture images of plastic particles, balance the image contrast, extract image brightness and color features, remove noise interference, optimize particle edges, and obtain particle optimized images; The edge detection and shape recognition module uses a visual neural network to perform edge detection based on the particle optimization image, calculates pixel grayscale gradient values and analyzes change trends, detects and segments boundary connected areas, extracts particle contours, and obtains a particle shape feature set; The surface feature analysis module extracts pixel points on the particle surface based on the particle shape feature set, calculates the grayscale change rate, analyzes the reflectivity change, identifies the particle roughness, measures the brightness fluctuation, analyzes the surface texture uniformity, and obtains the surface feature analysis results; The particle defect detection module uses the surface feature analysis results to analyze particle color difference, detect particle cracks, bubbles and impurities, analyze particle morphology deviation, and obtain particle defect indicators; The particle quality assessment module classifies defective particles according to the particle defect index, compares the defective particles with the standard quality, calculates the defect ratio, assesses the quality grade of the particles, and obtains the particle assessment result.

Images