Industrial camera production line data analysis control method and system

Abstract

The invention belongs to the technical field of industrial automation, and discloses an industrial camera production line data analysis control method and system, and the method comprises the steps: obtaining industrial camera production line data which comprise industrial camera production parameter data, production equipment operation data and industrial camera production image data; identifying and removing abnormal values in the industrial camera production parameter data, the production equipment operation data and the industrial camera production image data; a wavelet transform denoising technology and a multi-dimensional feature extraction technology are fused, an image denoising method based on two-dimensional convolution and weight dynamic adjustment, Sobe l edge detection, gray level co-occurrence matrix texture analysis and color histogram extraction are combined, and a production parameter feature data set, an equipment operation feature data set and a production image feature data set are obtained; and the production efficiency and the productivity of the industrial camera are improved.

Description

Technical Field

[0001] The present invention relates to the field of industrial automation technology. More specifically, the present invention relates to a method and system for data analysis and control of an industrial camera production line. Background Art

[0002] A patent with the patent publication number CN115618269A discloses a big data analysis method and system based on industrial sensor production. In this application, a production status analysis result of an industrial sensor production line is obtained by processing a production detection flow data set through a production detection data analysis model. Through artificial intelligence technology, the accuracy of production status analysis in online monitoring of industrial sensor production and the efficiency of data analysis can be improved. During the optimization process of the production detection data analysis model, the accuracy of determining key sensor production data and industrial sensor production status in the production detection flow data set of the industrial sensor production line is also improved.

[0003] The existing methods and systems for data analysis and control of industrial camera production lines have the following main problems:

[0004] During the production process, many defects or details are usually more obvious in the edge part of the image. If the extraction of edge information is ignored, it may be difficult for industrial cameras to accurately detect subtle production defects or abnormalities; the lack of extraction of texture features will reduce the accuracy of defect detection and cannot effectively distinguish different types of surface defects; fixed denoising strategies may not be able to adapt to these changes, resulting in ineffective noise removal, affecting the quality of images and signals, and further affecting defect detection and data analysis during the production process; human judgment is easily affected by experience, subjective factors, and errors, resulting in inaccurate and unreliable evaluation results, and further affecting the quality control effect during the production process; the lack of quantification and systematic control of performance indicators may lead to production optimization relying on intuition or simple adjustments and cannot be precisely optimized through data-driven methods;

[0005] Without the combination of global optimization and local fine-tuning, it may lead to waste of resources and inefficient operation of each link in the production process; without optimizing the production scheduling plan, the production tasks may be irrationally allocated, resulting in waste of resources and poor production process. Without optimized production scheduling, some equipment may be idle for a long time, while other equipment is overloaded, affecting production efficiency and product quality; without optimizing the equipment operating parameters, the key parameters such as equipment accuracy, temperature, exposure time, etc. may not reach the optimal state, resulting in unstable equipment performance, affecting production quality, and even causing failures or quality defects in the production process, which in turn affects the quality standards of the final product; the optimization process may lack flexibility and adaptability, and without a dynamic adjustment mechanism, it may cause the optimization process to fall into a local optimal solution too early, or the optimization effect may be poor; without synchronous optimization of production scheduling and equipment parameters, the quality control of industrial cameras may not be comprehensive enough, resulting in unstable product quality.

[0006] In view of this, the present invention proposes an industrial camera production line data analysis control method and system to solve the above problems. Summary of the invention

[0007] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a data analysis and control method for an industrial camera production line, comprising:

[0008] S1. Obtain industrial camera production line data; the industrial camera production line data includes industrial camera production parameter data, production equipment operation data and industrial camera production image data;

[0009] S2. Identify and remove outliers in industrial camera production parameter data, production equipment operation data, and industrial camera production image data; integrate wavelet transform denoising and multidimensional feature extraction technology, and combine image denoising method based on two-dimensional convolution and dynamic weight adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain production parameter feature data set, equipment operation feature data set, and production image feature data set;

[0010] S3. Build a production quality assessment model, input the production line comprehensive feature data set into the trained production quality assessment model, and predict the production quality assessment label;

[0011] S4. Determine whether the industrial cameras produced are qualified according to the production quality assessment label; if the industrial cameras produced are qualified, analyze and obtain the quality grade of the industrial cameras; issue sorting instructions through the production intelligent control terminal, and automatically sort the industrial cameras produced according to the quality grade of the industrial cameras;

[0012] S5. If the industrial cameras produced are unqualified, optimize the collaborative work among production equipment through the hybrid intelligent optimization algorithm, collaboratively execute production scheduling optimization and equipment operation status optimization, perform production scheduling optimization through the genetic algorithm, and use the comprehensive quality score of the industrial camera as the fitness function to guide the optimization; use the particle swarm optimization algorithm to adjust the equipment operation parameters, control the constraint conditions through the dynamic penalty coefficient, and adjust and control the production of industrial cameras until the produced industrial cameras are qualified.

[0013] Preferably, the production parameter data of the industrial camera includes imaging parameter data, optical component parameter data, electronic component parameter data, and interface communication parameter data; the production equipment operation data includes equipment status data, production control data, production environment data, and production log data; the industrial camera production image data includes production process images, imaging quality assessment images, defect detection images, and classification annotation images.

[0014] Preferably, the method for identifying and removing outliers from the production parameter data of the industrial camera, the production equipment operation data, and the industrial camera production image data includes:

[0015] Identify and remove the outliers existing in the production parameter data of the industrial camera, the production equipment operation data, and the industrial camera production image data through the K-Means clustering algorithm, perform denoising and multi-dimensional feature extraction on the production parameter data of the industrial camera, the production equipment operation data, and the industrial camera production image data after removing the outliers, and obtain the production parameter feature dataset, the equipment operation feature dataset, and the production image feature dataset;

[0016] Perform standard deviation normalization processing on the production parameter feature dataset, the equipment operation feature dataset, and the production image feature dataset after feature extraction, convert them into a standard normal distribution with a mean of 0 and a standard deviation of 1, and obtain the normalized production parameter feature dataset, the equipment operation feature dataset, and the production image feature dataset.

[0017] Preferably, the acquisition methods for the production parameter feature dataset, the equipment operation feature dataset, and the production image feature dataset include:

[0018] Apply wavelet transform to the production parameter data of the industrial camera and the production equipment operation data, decompose the signal into different frequency sub-bands, remove noise through thresholding operations, perform signal reconstruction on the denoised signal, and extract statistical features and frequency domain features, and finally obtain the denoised production parameter feature dataset and the equipment operation feature dataset;

[0019] For the industrial camera production image data, perform denoising on it through the two-dimensional convolution formula, and dynamically design and adjust the weight value through the weight value adjustment formula; use the Sobel operator to detect the edges in the industrial camera production images;

[0020] Calculate the gradients of the industrial camera production images in the horizontal and vertical directions respectively through the horizontal gradient calculation formula and the vertical gradient calculation formula; extract the texture features of the industrial camera production images through the gray-level co-occurrence matrix, and calculate the texture energy of the industrial camera production images using the energy feature formula; extract the color features of the industrial camera production images by calculating the color histograms of the industrial camera production images, and finally obtain the production parameter feature dataset, equipment operation feature dataset, and production image feature dataset after feature extraction.

[0021] Preferably, the method for obtaining the production line comprehensive feature dataset includes:

[0022] Fuse the normalized production parameter feature dataset, equipment operation feature dataset, and production image feature dataset through a weighted model to obtain the production line comprehensive feature dataset.

[0023] Preferably, the method for constructing the production quality assessment model includes:

[0024] Construct and train a production quality assessment model. The input data of the model is the historical production line comprehensive feature dataset, and the output label of the model is the production quality assessment label. The production quality assessment label is divided into two categories: qualified label and unqualified label; use the sigmoid function as the activation function, and the production quality assessment model is a gradient boosting decision tree model; use binary cross-entropy as the loss function to measure the error of the model prediction; use the trained production quality assessment model to predict the current production line comprehensive feature dataset to obtain the production quality assessment label.

[0025] Preferably, the method for obtaining the industrial camera quality grade includes:

[0026] Further score the industrially produced cameras that pass the inspection through the industrial camera quality assessment formula to obtain the comprehensive industrial camera quality score G, and obtain the industrial camera quality grade based on the comprehensive industrial camera quality score G.

[0027] Preferably, the method for automatically sorting the produced industrial cameras includes:

[0028] Preset the comprehensive industrial camera quality score threshold, and further divide the comprehensive industrial camera quality score threshold into the first comprehensive industrial camera quality score threshold G1 and the second comprehensive industrial camera quality score threshold G2; compare the comprehensive industrial camera quality score G with the first comprehensive industrial camera quality score threshold G1 and the second comprehensive industrial camera quality score threshold G2 respectively;

[0029] If G < G1, it is determined that the quality grade of the industrial camera is A. For A-grade industrial cameras, they are sorted into the high-quality product area; if G1 < G ≤ G2, it is determined that the quality grade of the industrial camera is B. For B-grade cameras, they are sorted into the medium-quality product area; if G > G2, it is determined that the quality grade of the industrial camera is C. For C-grade cameras, they are sorted into the qualified product area.

[0030] Preferably, the method for adjusting and controlling the production of industrial cameras includes:

[0031] Dividing the optimization of the collaborative work among production equipment into the optimization of production scheduling problems and the optimization of equipment operation status problems, and realizing the comprehensive adjustment of production scheduling and equipment parameters through a hybrid intelligent optimization algorithm; the hybrid intelligent optimization algorithm is a two-layer optimization structure, including a genetic algorithm and a particle swarm optimization algorithm; globally optimizing the task allocation of equipment through the genetic algorithm to find the optimal production scheduling plan, and based on the optimal production scheduling plan, optimizing the equipment operation parameters through the particle swarm optimization algorithm;

[0032] Define the objective function as maxG. By optimizing the production scheduling plan and equipment parameters, maximize the comprehensive quality score of industrial cameras while satisfying the constraints of production and equipment operation; the constraints include production scheduling constraints and equipment operation status constraints; the production scheduling constraints include the constraints of equipment utilization rate and scheduling time. Divide the working time T p of the p-th equipment by the total production cycle T al to obtain the equipment utilization rate U p . Preset the maximum threshold of the equipment utilization rate as U max , and limit the equipment utilization rate U p to be less than or equal to the preset maximum threshold U of the equipment utilization rate max ;

[0033] Automatically record the start time of each task executed by the equipment and the end time of each task through a timer. Subtract the start time of each task from the end time of each task to obtain the allocation time T rw of each task. Accumulate the allocation times of all tasks to obtain the total scheduling time T wa of all tasks; preset the maximum threshold of the total scheduling time of all tasks as T max , and limit the total scheduling time of all tasks to be less than or equal to the preset maximum threshold T max ;

[0034] The equipment operation status constraints include defining the range of equipment operation parameters, optimizing production scheduling through the genetic algorithm, defining each chromosome to represent a production scheduling plan, and forming a production scheduling plan set d; define the fitness function of the genetic algorithm as the comprehensive quality score of industrial cameras, and calculate the fitness value of each production scheduling plan;

[0035] The optimal chromosome is selected according to the fitness value by the roulette wheel selection method; two chromosomes are randomly selected, and single-point crossover is performed on the production scheduling part of the chromosomes to form new individuals; the task order in the chromosomes is randomly adjusted for mutation to generate the next-generation scheduling plan until the fitness value of the production scheduling plan no longer changes or reaches the maximum number of iterations preset by the genetic algorithm, and then stop, and output the optimal production scheduling plan;

[0036] The operation parameters of the equipment are optimized by the particle swarm optimization algorithm. Each particle is defined to represent a set of equipment operation parameter configurations, which form the equipment operation parameter configuration set h; the fitness function of the particle swarm optimization algorithm is defined, and the penalty coefficient μ in the fitness function is dynamically adjusted and constrained through the penalty coefficient adjustment formula; the operation parameter h of the z-th equipment is adjusted z to update the velocity and position of the particle, and limit the position update of the particle to satisfy: greater than or equal to the minimum value h of the operation parameter of the z-th equipment set z,min and less than or equal to the maximum value h of the operation parameter of the z-th equipment z,man ;

[0037] Continuously update the velocity and position of the particle until the fitness value of the operation parameter of the equipment no longer changes or reaches the maximum number of iterations preset by the particle swarm optimization algorithm, and then stop, and output the optimal equipment operation parameter configuration.

[0038] An industrial camera production line data analysis and control system, comprising:

[0039] A data acquisition module, which acquires industrial camera production line data, and the industrial camera production line data includes industrial camera production parameter data, production equipment operation data, and industrial camera production image data;

[0040] A data processing module, which identifies and eliminates outliers in the industrial camera production parameter data, production equipment operation data, and industrial camera production image data; fuses wavelet transform denoising and multi-dimensional feature extraction technologies, and combines an image denoising method based on two-dimensional convolution and weight dynamic adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain a production parameter feature data set, an equipment operation feature data set, and a production image feature data set;

[0041] A production quality evaluation module, which constructs a production quality evaluation model, and inputs the production line comprehensive feature data set into the trained production quality evaluation model to predict and obtain production quality evaluation labels;

[0042] An automatic sorting module determines whether the industrial cameras produced are qualified according to the production quality assessment tags. If the produced industrial cameras are qualified, it analyzes and obtains the quality grades of the industrial cameras, issues sorting instructions through the production intelligent control terminal, and automatically sorts the produced industrial cameras according to the quality grades of the industrial cameras.

[0043] A production line collaborative control module, if the produced industrial cameras are unqualified, optimizes the collaborative work among production equipment through a hybrid intelligent optimization algorithm, collaboratively executes production scheduling optimization and equipment operation state optimization, performs production scheduling optimization through a genetic algorithm, and guides the optimization with the comprehensive quality score of the industrial camera as the fitness function. It uses a particle swarm optimization algorithm to adjust the equipment operation parameters, controls the constraint conditions through a dynamic penalty coefficient, and adjusts and controls the production of industrial cameras until the produced industrial cameras are qualified.

[0044] The technical effects and advantages of a data analysis and control method and system for an industrial camera production line according to the present invention:

[0045] By using the Sobel operator to extract the edge information of the image, the gray-level co-occurrence matrix to extract texture features, and the color histogram to extract color features, various types of data in the production process can be comprehensively described, providing a rich information source for subsequent data analysis. Dynamically adjusting the weight values helps to flexibly optimize the denoising effect according to different production situations or environmental changes, ensuring effective noise removal under different conditions and improving the quality of images and signals. The extraction of texture energy and color histogram provides more detailed features for the image, helping to improve the accuracy of industrial camera image analysis, especially suitable for detecting complex production defects and changes.

[0046] By quantifying and evaluating multiple performance indicators of the industrial camera (such as resolution, signal-to-noise ratio, contrast, lens distortion coefficient, frame rate, etc.) through formulas, the quality of the camera can be scientifically and objectively evaluated, avoiding the deviation of human judgment. By setting weight factors for controlling the comprehensive clarity, contrast, lens distortion coefficient, and frame rate of the image, the quality levels of various important factors can be controlled and optimized during the production process, thereby improving the overall quality and consistency of the product.

[0047] Combining genetic algorithms and particle swarm optimization algorithms, through a two-layer optimization structure, simultaneously optimize production scheduling and equipment operating status, which can comprehensively improve production efficiency and product quality. The genetic algorithm provides global optimization, while the particle swarm optimization algorithm makes local fine adjustments to equipment parameters to ensure all-round optimization of the production process; by optimizing the production scheduling plan through the genetic algorithm, production tasks can be scientifically allocated, resource waste can be avoided, and the smoothness and efficiency of the production process can be ensured; it can dynamically adjust production tasks according to the actual operating conditions of the equipment, improve equipment utilization rate and ensure task completion on time; by optimizing the equipment operating parameters through the particle swarm optimization algorithm, key parameters such as the accuracy, temperature, and exposure time of the equipment can be adjusted, thereby improving production quality. The optimized equipment operating parameters help ensure the quality standards of industrial cameras and prevent production problems caused by improper parameter settings; dynamically adjusting the penalty coefficient can make the optimization process more flexible and adaptive. During the optimization process, according to the current iteration number, the penalty coefficient will be gradually adjusted, so that the optimization process can gradually stabilize over time, avoiding the situation where the penalty coefficient is too large or too small at the beginning, resulting in an unsatisfactory optimization effect; the comprehensively optimized production scheduling and equipment parameters can maximize resource utilization rate, shorten the production cycle, reduce equipment idle time, and improve overall production efficiency and production capacity. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic flow chart of a data analysis and control method for an industrial camera production line according to the present invention;

[0049] Figure 2 It is a schematic structural diagram of a data analysis and control system for an industrial camera production line according to the present invention;

[0050] Figure 3 It is a schematic flow chart of a method for constructing a production quality evaluation model provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0051] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0052] Embodiment 1

[0053] Please refer to Figure 1 and Figure 3 As shown, this embodiment further illustrates a data analysis and control method for an industrial camera production line proposed by the present invention, including:

[0054] With the continuous improvement of industrial automation levels, industrial cameras, as key visual inspection devices in the production process, are widely used in multiple links such as product quality monitoring, defect identification, and precision measurement. To improve the production efficiency and imaging quality of industrial cameras, it is usually necessary to combine data analysis and control methods to optimize image information, equipment parameters, and production processes in multiple dimensions. However, there are still many technical problems to be solved urgently in the existing data analysis and control methods and systems for industrial camera production lines.

[0055] In the process of image feature extraction, traditional methods often ignore the information in the image edge region, which is often the main area where minute defects or abnormal changes occur, resulting in insufficient ability to identify detailed defects. Secondly, the existing methods lack in-depth extraction of image texture features, limiting the ability to distinguish different types of surface defects, thereby reducing the accuracy and robustness of defect detection.

[0056] Most current image denoising strategies use fixed parameters or static models and are difficult to adapt dynamically to noise changes in the actual production scenario, easily causing a decline in image and signal quality, further affecting subsequent data analysis and quality judgment. At the same time, some quality assessments still rely on manual experience judgment, are easily interfered by subjective factors, lack objectivity and consistency, and thus affect the accuracy and stability of quality control.

[0057] In terms of production optimization, the existing systems generally lack quantitative control and systematic analysis of production performance indicators, resulting in the optimization process relying on experience or simple rules and being difficult to achieve precise tuning through data-driven means. In addition, production scheduling and equipment operation optimization are usually independent of each other, lacking a linkage mechanism, failing to achieve reasonable allocation of scheduling resources and fine adjustment of equipment operation parameters, and prone to problems such as uneven equipment load, resource waste, and process bottlenecks, thus affecting the overall production efficiency and product quality.

[0058] The existing optimization methods generally have problems of insufficient flexibility and adaptability, lack a dynamic adjustment mechanism, are easily trapped in local optimal solutions, and cannot meet the complex and ever-changing working conditions requirements in actual production, resulting in poor optimization effects, incomplete quality control of industrial cameras, and affecting the consistency and reliability of the final product.

[0059] To effectively solve the above problems, the present invention proposes a data analysis and control method for an industrial camera production line, including:

[0060] S1. Obtain industrial camera production line data, where the industrial camera production line data includes industrial camera production parameter data, production equipment operation data, and industrial camera production image data;

[0061] S2. Identify and remove outliers from the industrial camera production parameter data, production equipment operation data, and industrial camera production image data; fuse wavelet transform denoising and multi-dimensional feature extraction techniques, and combine an image denoising method based on two-dimensional convolution and weight dynamic adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain a production parameter feature dataset, a device operation feature dataset, and a production image feature dataset;

[0062] S3. Build a production quality assessment model, input the comprehensive production line feature dataset into the trained production quality assessment model, and predict to obtain a production quality assessment label;

[0063] S4. According to the production quality assessment label, determine whether the produced industrial camera is qualified; if the produced industrial camera is qualified, analyze and obtain the quality grade of the industrial camera; issue a sorting instruction through the production intelligent control terminal, and automatically sort the produced industrial cameras according to the quality grade of the industrial camera;

[0064] S5. If the produced industrial camera is unqualified, optimize the collaborative work between production equipment through a hybrid intelligent optimization algorithm, collaboratively execute production scheduling optimization and equipment operation state optimization, perform production scheduling optimization through a genetic algorithm, and use the comprehensive quality score of the industrial camera as the fitness function to guide the optimization; use the particle swarm optimization algorithm to adjust the equipment operation parameters, and control the constraint conditions through a dynamic penalty coefficient to adjust and control the production of the industrial camera until the produced industrial camera is qualified.

[0065] The industrial camera production parameter data includes imaging parameter data, optical component parameter data, electronic component parameter data, and interface communication parameter data; the imaging parameter data includes the resolution of the industrial camera, the contrast of the image, the exposure time, the aperture size, the signal-to-noise ratio of the image, and the image frame rate collected by the industrial camera per second; the optical component parameter data includes the lens focal length, the lens distortion coefficient, and the filter parameters; the electronic component parameter data includes the image sensor type, the image sensor size, and the pixel size; the interface communication parameter data includes the communication interface type, the data transmission bandwidth, and the communication protocol standard;

[0066] The production equipment operation data includes equipment status data, production control data, production environment data, and production log data; the industrial camera production image data includes production process images, imaging quality assessment images, defect detection images, and classification annotation images;

[0067] The device status data includes the device startup time, device running time, device failure rate, device maintenance time, and device maintenance frequency; the production control data includes the cooperation parameters between devices, the production volume completed per unit time, and the speed, voltage, and current of the drive motor; the production environment data includes the production environment temperature, production environment humidity, and drive motor vibration frequency; the production log data includes the industrial camera production start time, industrial camera production end time, and production batch number; the production process images include the sample images of industrial camera imaging tests, black and white board test images, and black screen images of sensor dark current tests; the imaging quality evaluation images include resolution target images, distortion test charts, and contrast test charts of high and low brightness regions; the defect detection images include bright spots, dark spots, and pixel unevenness images in the images; the classification and annotation images include the annotation of abnormal regions in the images, such as stains, scratches, and contamination.

[0068] The method for identifying and removing outliers from the industrial camera production parameter data, production equipment operation data, and industrial camera production image data includes:

[0069] Identifying and removing the outliers existing in the industrial camera production parameter data, production equipment operation data, and industrial camera production image data through the K-Means clustering algorithm, denoising and extracting multi-dimensional features from the industrial camera production parameter data, production equipment operation data, and industrial camera production image data after removing the outliers, to obtain a production parameter feature data set, a device operation feature data set, and a production image feature data set;

[0070] Performing standard deviation normalization processing on the production parameter feature data set, device operation feature data set, and production image feature data set after feature extraction, and converting them into a standard normal distribution with a mean of 0 and a standard deviation of 1, to obtain the normalized production parameter feature data set, device operation feature data set, and production image feature data set.

[0071] The acquisition methods for the production parameter feature data set, device operation feature data set, and production image feature data set include:

[0072] S41. Applying wavelet transform to the industrial camera production parameter data and production equipment operation data, decomposing the signal into different frequency sub-bands, removing noise through thresholding operations, reconstructing the signal after denoising, and extracting statistical features and frequency domain features, finally obtaining the denoised production parameter feature data set and device operation feature data set;

[0073] S42. For the industrial camera production image data, performing denoising on it through a two-dimensional convolution formula, and the two-dimensional convolution formula is: where, I(x+i,y+j) is the pixel value of the industrial camera production image at the coordinate (x+i,y+i); I pre(x, y) is the pixel value at the coordinate (x, y) of the denoised industrial camera production image; θ(i, j) is the weight value at the position (i, j) in the Gaussian filter window; k is the window size of the Gaussian filter; i and j are the offsets of the current pixel from the center of the Gaussian filter; θ is the weight value symbol;

[0074] S43. Dynamically design and adjust the weight value θ(i, j) through the weight value adjustment formula. The weight value adjustment formula is: where σ is the standard deviation of the Gaussian distribution, used to control the width of the weight distribution; q is a constant for adjusting the ratio between the window size k and the standard deviation σ of the Gaussian distribution; exp is the exponential function with base e;

[0075] S44. Use the Sobel operator to detect the edges in the industrial camera production image; calculate the gradients of the industrial camera production image in the horizontal and vertical directions respectively through the horizontal gradient calculation formula and the vertical gradient calculation formula. The horizontal gradient calculation formula is: where G′ x is the gradient of the industrial camera production image in the horizontal direction; S x (i, j) is the Sobel horizontal operator, which is a weight matrix (convolution kernel) used to detect edges in the horizontal direction;

[0076] The vertical gradient calculation formula is: where G′ y is the gradient of the industrial camera production image in the vertical direction; is the Sobel vertical operator, which is also a weight matrix (convolution kernel) used to calculate the gradient in the vertical direction;

[0077] By adding the squares of G′ x and G′ y and taking the square root, the total edge intensity of each pixel in the industrial camera production image is obtained

[0078] S45. Extract the texture features of the industrial camera production image through the gray-level co-occurrence matrix, and calculate the texture energy of the industrial camera production image using the energy feature formula. The gray-level co-occurrence matrix is a matrix describing the image texture, which statistically counts the occurrence frequencies of the gray values of pixel pairs in the industrial camera production image. The dimension of the matrix is n×n, where n is the number of gray levels. For an 8-bit gray image, the number of gray levels is 256, so the co-occurrence matrix is a 256×256 matrix. The energy feature formula is: Among them, Ey is the texture energy of the image produced by the industrial camera, and the texture energy describes the uniformity of the image texture. In the gray-level co-occurrence matrix, the larger the sum of the squares of all elements, the more uniform the image texture. An image with high energy means that the combination of its gray values is more consistent and the texture is more regular; P(a, b) is an element in the gray-level co-occurrence matrix, representing the probability that pixel pairs with gray values a and b appear in the image produced by the industrial camera; n is the number of gray levels of the image produced by the industrial camera, that is, the number of different gray values in the image produced by the industrial camera. In a grayscale image, the gray value usually ranges from 0 (black) to 255 (white). Therefore, for an 8-bit grayscale image, n = 256; a is the row index in the gray-level co-occurrence matrix; b is the column index in the gray-level co-occurrence matrix;

[0079] It means adding the squared values of all elements P(a, b) of the gray-level co-occurrence matrix to obtain the energy feature of the image;

[0080] For example, the gray-level co-occurrence matrix of the image produced by the industrial camera is as follows (simplified to a 3×3 matrix, and the number of gray levels n = 3): Since the number of gray levels n = 3, the upper limit of the summation is n - 1 = 2, that is, a, b ∈ {0, 1, 2}; for this matrix, the calculation of the texture energy Ey is as follows:

[0081] Ey = (0.1) 2 +(0.2) 2 +(0.3) 2 +(0.4) 2 +(0.5) 2 +(0.6) 2 +(0.7) 2 +(0.8) 2 +(0.9) 2 = 2.85.

[0082] S46. Extract the color features of the image produced by the industrial camera by calculating its color histogram, and finally obtain the production parameter feature dataset, equipment operation feature dataset, and production image feature dataset after feature extraction.

[0083] The method for obtaining the comprehensive feature dataset of the production line includes:

[0084] Fuse the normalized production parameter feature dataset, equipment operation feature dataset, and production image feature dataset through a weighted model to obtain the comprehensive feature dataset of the production line;

[0085] Denote the production parameter feature dataset as F1, the equipment operation feature dataset as F2, and the production image feature dataset as F3; the weighted model is: YE = F1·δ1 + F2·δ2 + F3·δ3; where YE is the production line comprehensive feature dataset; δ1 is the weight coefficient of the production parameter feature dataset; δ2 is the weight coefficient of the equipment operation feature dataset; δ3 is the weight coefficient of the production image feature dataset.

[0086] According to the importance of the production parameter feature dataset in the production comprehensive feature dataset, set the initial value range of the weight factor δ1 to 0.3 - 0.5, and the initial default value is 0.4. Use the historical production feature dataset to analyze the contribution degree of the production parameter feature dataset to the production quality evaluation result. If the production parameter feature data has a greater impact on production quality (such as the proportion affecting the evaluation result exceeds 50%), then increase the value of δ1; otherwise, decrease its value. Gradually optimize the weight value according to the feedback of the real-time feature dataset to ensure that the model can effectively adapt to the fluctuations of parameters in the production process.

[0087] According to the influence degree of the equipment operation feature dataset on the production comprehensive feature dataset, set the initial value range of the weight factor δ2 to 0.4 - 0.6, and the initial default value is 0.5. Use the historical equipment operation feature dataset and through the regression analysis method, evaluate the contribution degree of the equipment operation feature dataset to production quality. If the equipment operation feature dataset makes a high contribution to production quality (such as the change of equipment parameters directly affects production efficiency), then appropriately increase the value of δ2; otherwise, decrease its value. Dynamically adjust the weight in combination with the real-time equipment operation feature dataset to ensure that the weight of the equipment operation feature dataset is consistent with the actual production situation.

[0088] According to the importance of the production image feature dataset in the production comprehensive feature dataset, set the initial value range of the weight factor δ3 to 0.2 - 0.4, and the initial default value is 0.3. Determine its contribution degree by analyzing the correlation between the production image data and production quality (such as the role of image data in equipment status recognition or production process optimization). If the image feature data plays a key role in the actual production process (such as accurately reflecting equipment abnormalities), then increase the value of δ3; if its influence is weak, then appropriately decrease it.

[0089] Analyze the feedback effect of the production image feature dataset on production evaluation in real time, and dynamically optimize the value range of δ3 in combination with the model performance.

[0090] The construction method of the production quality evaluation model includes:

[0091] Divide the dataset into a training set, a validation set, and a test set, and construct a production quality assessment model; the sample set is a subset of the dataset, and each sample set includes the historical production line comprehensive feature dataset and the corresponding production quality assessment label;

[0092] The input data of the model is the historical production line comprehensive feature dataset, and the output label of the model is the production quality assessment label. The production quality assessment label is divided into two categories: a qualified label and an unqualified label; the sigmoid function is used as the activation function; the production quality assessment model is a gradient boosting decision tree model;

[0093] Initialize the production quality assessment model, and set the hyperparameters of the number of trees, the learning rate, and the depth of the tree; through k-fold cross-validation, use different combinations of hyperparameters to train the model, and use the results of cross-validation to tune the initially set hyperparameters, and select the best-performing combination of hyperparameters; use binary cross-entropy as the loss function to measure the error of the model prediction;

[0094] Train the production quality assessment model on the training set by training a new decision tree based on the residuals of the current model in each iteration; adjust the hyperparameters of the model according to the model performance feedback to optimize the model, retrain the model with the tuned hyperparameters, and select the optimal hyperparameters through cross-validation after the training process ends;

[0095] Stop when the training reaches the preset stopping condition or the model performance no longer improves, and obtain the finally trained production quality assessment model; use the trained production quality assessment model to predict the current production line comprehensive feature dataset to obtain the production quality assessment label.

[0096] If the produced industrial camera is qualified, the method for analyzing and obtaining the quality grade of the industrial camera includes: further scoring the qualified industrial camera through the industrial camera quality assessment formula to obtain the industrial camera quality grade; the industrial camera quality assessment formula is: where G is the comprehensive quality score of the industrial camera; R is the resolution of the industrial camera; SNR is the signal-to-noise ratio of the image; K is the contrast of the image; E is the lens distortion coefficient; H(t) is the image frame rate collected by the industrial camera per second; ω1 is the weight factor for controlling the overall image sharpness; ω2 is the weight factor for controlling the contrast of the image; ω3 is the weight factor for controlling the lens distortion coefficient; ω4 is the weight factor for controlling the image frame rate collected by the industrial camera per second; t0 is the start time of image frame rate collection; t1 is the end time of image frame rate collection; t is the time of image frame rate collection;

[0097] According to the importance of the comprehensive clarity of industrial cameras in the quality evaluation of industrial cameras, the initial value range of the weight factor ω is set to 0.3 - 0.5, and the initial default value is 0.4. By analyzing the influence degree of clarity indicators (such as resolution R and signal-to-noise ratio SNR) on the comprehensive quality score of industrial cameras, the weight factor is optimized. If the clarity has a greater impact on the quality score (such as significant changes in resolution and signal-to-noise ratio significantly affect the evaluation results), then increase the value of ω1; otherwise, decrease its value. In the camera usage scenarios (such as complex lighting conditions or high-speed acquisition), analyze the role of clarity in real-time and dynamically optimize the value range of ω1.

[0098] According to the importance of the image contrast of industrial cameras in the quality evaluation, the initial value range of the weight factor ω2 is set to 0.2 - 0.4, and the initial default value is 0.3. Using the historical image contrast K data, through regression analysis, evaluate the contribution degree of contrast to the comprehensive score of industrial cameras. If the contrast has a greater impact on the evaluation results (such as the image quality drops sharply in low-contrast scenarios), appropriately increase ω2; otherwise, decrease its value. Combine specific application scenarios (such as high-contrast environments or low-illumination acquisition) to dynamically adjust the value of ω2 to ensure reasonable distribution of the weight of the contrast factor.

[0099] According to the influence degree of lens distortion on the image quality of industrial cameras, the initial value range of the weight factor ω3 is set to 0.1 - 0.3, and the initial default value is 0.2. Analyze the degree of damage of lens distortion to the image clarity and geometric features of industrial cameras, and quantify its influence on the comprehensive score. If the lens distortion significantly affects the image usability (such as severe barrel or pincushion distortion), then appropriately increase ω3; otherwise, decrease its value. In the camera configuration (such as wide-angle lenses or special optical conditions), dynamically optimize the weight of ω3 to make it match the actual application.

[0100] According to the influence of the image frame rate H(t) collected by industrial cameras per second on the image quality in dynamic scenarios, the initial value range of the weight factor ω4 is set to 0.1 - 0.3, and the initial default value is 0.2. Using the historical frame rate data, analyze the influence of the frame rate on the image clarity and real-time performance of industrial cameras in dynamic scenarios. If the frame rate makes a greater contribution to the quality score (such as in the scenario of collecting high-speed moving targets), then appropriately increase the value of ω4; otherwise, decrease its value. According to the actual acquisition requirements (such as real-time or high-precision requirements), combine the frame rate characteristics to dynamically optimize the weight value of ω4.

[0101] Resolution directly affects the clarity of the image and the ability to capture details. The higher the resolution, the better the image quality. Therefore, resolution and quality score are positively correlated. Frame rate reflects how many images the camera can capture per second. A higher frame rate usually means the camera can still maintain clear images during high-speed movement. Therefore, frame rate and quality are positively correlated. Contrast is one of the key factors of image quality, which affects the clarity and visibility of the image. High contrast can enhance the detail performance of the image. Signal-to-noise ratio is an important criterion for measuring image quality. A higher signal-to-noise ratio indicates less noise and higher clarity in the image. Lens distortion can cause the image to stretch or distort, which will seriously affect the image quality. The smaller the lens distortion coefficient, the less image distortion and the higher the quality. If the frame rate varies within the time range, then the use of integration can comprehensively consider the time-varying characteristics of the frame rate and obtain an overall quality score. Each factor directly or indirectly affects the clarity, accuracy, stability, and performance of the image in reality. For example, resolution affects the detail clarity of the image, frame rate and contrast affect the fluency and visual effect of the image, while the lens distortion coefficient determines whether the image is distorted. Therefore, considering multiple factors helps to comprehensively evaluate the camera quality and avoid the situation where a single indicator cannot fully reflect the camera quality;

[0102] For example, the resolution R of an industrial camera is 4000; the signal-to-noise ratio SNR of the image is 50; the contrast K of the image is 0.9; the lens distortion coefficient E is 0.02; the frame rate H(t) of the images captured by the industrial camera per second is 10 frames per second; the weight factor ω1 for controlling the comprehensive clarity of the image is 0.4; the weight factor ω for controlling the contrast of the image is 0.3; the weight factor ω3 for controlling the lens distortion coefficient is 0.2; the weight factor ω4 for controlling the frame rate of the images captured by the industrial camera per second is 0.1; the starting time t0 for frame rate acquisition is 0 seconds; the ending time t1 for frame rate acquisition is 10 seconds; the time t for frame rate acquisition is 10 - 0 = 10 seconds; then the comprehensive quality score of the industrial camera

[0103] The method for automatically sorting the produced industrial cameras includes:

[0104] Preset the comprehensive quality score threshold of the industrial camera, and further divide the comprehensive quality score threshold of the industrial camera into the first threshold G1 and the second threshold G2 of the comprehensive quality score of the industrial camera; compare the comprehensive quality score G of the industrial camera with the first threshold G1 and the second threshold G2 of the comprehensive quality score of the industrial camera respectively;

[0105] If G < G1, it is determined that the quality grade of the industrial camera is Grade A. For Grade A industrial cameras, they are sorted into the high-quality product area; if G1 < G ≤ G2, it is determined that the quality grade of the industrial camera is Grade B. For Grade B cameras, they are sorted into the medium-quality product area; if G > G2, it is determined that the quality grade of the industrial camera is Grade C. For Grade C cameras, they are sorted into the qualified product area.

[0106] The method for adjusting and controlling the production of industrial cameras includes:

[0107] Dividing the optimization of the collaborative work among production equipment into the optimization of production scheduling problems and the optimization of equipment operation status problems, and realizing the comprehensive adjustment of production scheduling and equipment parameters through a hybrid intelligent optimization algorithm; the hybrid intelligent optimization algorithm is a two-layer optimization structure, including a genetic algorithm and a particle swarm optimization algorithm; globally optimizing the task allocation of equipment through the genetic algorithm to find the optimal production scheduling plan, and based on the optimal production scheduling plan, optimizing the equipment operation parameters through the particle swarm optimization algorithm;

[0108] Defining the objective function as maxG, and maximizing the comprehensive quality score of industrial cameras by optimizing the production scheduling plan and equipment parameters, while satisfying the constraints of production and equipment operation; the constraints include production scheduling constraints and equipment operation status constraints;

[0109] The production scheduling constraints include the constraint of equipment utilization rate and the constraint of scheduling time. Divide the working time T p of the p-th equipment by the total production cycle T al to obtain the equipment utilization rate U p , and preset the maximum threshold of the equipment utilization rate as U max , restricting the equipment utilization rate U p to be less than or equal to the preset maximum threshold U of the equipment utilization rate max ;

[0110] Automatically record the start time of each task executed by the equipment and the end time of each task through a timer, and subtract the start time of each task from the end time of each task to obtain the allocation time T of each task rw , and accumulate the allocation time of all tasks to obtain the total scheduling time of all tasks where L is the total number of tasks; preset the maximum threshold of the total scheduling time of all tasks as T max , restricting the total scheduling time of all tasks to be less than or equal to the preset maximum threshold T of the total scheduling time of all tasks max ;

[0111] The equipment operation status constraints include defining the range of equipment operation parameters as: c p ∈[c p,min , c p,max ; where c pare the operating parameters of the p-th device; p is the index of the device, p = 1, 2,..., M; M is the total number of devices; find the rate of change of the operating parameters of the p-th device with respect to the device operating time t * rate of change The maximum rate of change threshold of the operating parameters of the preset device with respect to the device operating time t * is Δc max , and limit the rate of change of each device with respect to the device operating time t * to be less than or equal to the maximum rate of change threshold Δc of the operating parameters of the preset device with respect to the device operating time t * max ;

[0112] Optimize the production scheduling through the genetic algorithm. Define each chromosome to represent a production scheduling plan, and form a set of production scheduling plans: d = {d1, d2,..., d g ..., d N}; where d is the set of production scheduling plans; d g is the g-th production scheduling plan; g is the index of the production scheduling plan; N is the total number of production scheduling plans;

[0113] Define the fitness function of the genetic algorithm as the comprehensive quality score of the industrial camera, and calculate the fitness value of each production scheduling plan; the fitness function of the genetic algorithm is: where f A′ (d) is the fitness value based on the current production scheduling plan; G(d g ) is the comprehensive quality score of the industrial camera based on the current production scheduling plan; λ1 is the weight coefficient for balancing the comprehensive quality score of the industrial camera; λ2 is the weight coefficient for balancing the total task scheduling time;

[0114] According to the importance of the comprehensive quality score of the industrial camera in the fitness function, set the initial value range of λ1 to be 0.4 - 0.6, and the initial default value is λ1 = 0.5. In production scheduling, if the comprehensive quality score G(d g ) of the industrial camera is more important for scheme optimization (such as high-quality images being a key requirement), then increase the value of λ1; otherwise, decrease its value. If the change range of G(d g ) is small, and the optimization mainly focuses on the task scheduling time, then decrease λ1 and increase λ2. The adjustment strategy should optimize the value range of λ1 in real time according to the actual production requirements to ensure that the influence of the quality score matches the weight.

[0115] According to the importance of the total task scheduling time for scheme optimization, set the initial value range of λ2 to be 0.4 - 0.6. The initial default value is λ2 = 0.5. If the production task time requirement is urgent (such as minimizing the total task scheduling time T wa ​), then increase the value of λ2; conversely, decrease its value. If T wa The change range is small, and the quality score of the industrial camera has a significant impact on the optimization, then decrease λ2 and increase λ1. In different production environments and task requirements, dynamically adjust the value of λ2 so that it can truly reflect the contribution of the task scheduling time to the optimization result.

[0116] For example, based on the comprehensive quality score G(d g ) of the industrial camera in the current production scheduling scheme is 85, the total task scheduling time T wa is 10, λ1 = 0.5, λ2 = 0.5, then the fitness value based on the current production scheduling scheme If the actual demand has a higher requirement for the quality of the industrial camera, then adjust the weights λ1 = 0.6, λ2 = 0.4, then the new fitness value

[0117] Use the roulette wheel selection method to select the optimal chromosome according to the fitness value f A (d); randomly select two chromosomes, perform single-point crossover on the production scheduling part of the chromosomes to form new individuals; on the basis of the current high-quality chromosomes, generate new chromosomes (solutions) through gene exchange to explore new solution spaces, promote gene mixing within the population, thereby discovering potential high-quality solutions, improving the diversity of the population, and avoiding premature convergence of the population on some sub-optimal solutions; randomly adjust the task order in the chromosomes for mutation to generate the next-generation scheduling scheme until the fitness value of the production scheduling scheme no longer changes or reaches the preset maximum number of iterations of the genetic algorithm and then stop, output the optimal production scheduling scheme; break the similarity of the chromosomes within the population, maintain the diversity of the population, make up for the limitations of the crossover operation, and help the population escape from sub-optimal solutions;

[0118] Optimize the equipment operation parameters through the particle swarm optimization algorithm. Define that each particle represents a set of equipment operation parameter configurations, and form a set of equipment operation parameter configurations h = {h1, h2,..., h z ,..., h Q}; where, h z is the operation parameter of the z-th equipment; z is the index of the operation parameter of the equipment; Q is the total number of operation parameters of the equipment; the configuration of the equipment operation parameters includes the focusing accuracy of the lens assembly equipment, the calibration value of the optical axis deviation, the temperature of the welding equipment, the welding time, the exposure time of imaging, and the light source intensity; randomly generate the initial positions (equipment operation parameter configurations) and speeds of the particles.

[0119] Define the fitness function of the particle swarm optimization algorithm as:

[0120] where, f B′(h) is the fitness value based on the operating parameters of the current device; G(h) is the comprehensive quality score of the industrial camera based on the operating parameters of the current device; μ is the penalty coefficient, which is used to ensure that the position of the particle (i.e., the device operating parameters) is within the specified range. If the hyperparameters exceed the bounds, the penalty term will increase, thus affecting the value of the fitness function; h z,max is the maximum value of the operating parameters of the z-th device; h z,min is the minimum value of the operating parameters of the z-th device; max(0, h z - h z,max ) represents the penalty when h z exceeds the maximum value, that is, the deviation of h z from the maximum value; max(0, h z,min - h z ) represents the penalty when h z is less than the minimum value, that is, the deviation of h z from the minimum value;

[0121] The penalty coefficient μ is dynamically adjusted and constrained through the penalty coefficient adjustment formula. The penalty coefficient adjustment formula is: where μ′ is the dynamically adjusted penalty coefficient; β is the adjustment factor of the penalty coefficient; t′ is the current iteration number of the particle swarm optimization algorithm; t′ max is the maximum iteration number of the particle swarm optimization algorithm;

[0122] In the formula, the penalty coefficient μ decreases as the iteration number t′ increases. That is, as the optimization process progresses, the penalty coefficient gradually decreases; during the iteration process of the particle swarm optimization algorithm, the particles continuously explore the optimal solution and gradually converge. When the algorithm approaches the optimal solution, the operating parameters of the device basically meet the constraint conditions. Therefore, reducing the penalty coefficient can make the optimization process more focused on improving the performance of the device rather than constraint violations; by linearly reducing the penalty coefficient, it can ensure stronger penalties for out-of-range parameters in the initial stage, preventing the algorithm from falling into an inappropriate solution space early. In the later stage of iteration, the reduced penalty coefficient helps to reduce unnecessary forced intervention on the parameter range limit; when the number of particles increases, the search space becomes more extensive, meaning that more iterations are required to find the optimal solution, and the adjustment of the penalty coefficient should be more gentle. By affecting the penalty coefficient through the total number of operating parameters of the device, that is, the number of particles Q, it can make the constraint penalty coefficient of a larger particle swarm higher in the initial stage to ensure that the solutions in the extensive search stage can meet the constraint conditions;

[0123] For example, the total number of operating parameters Q of the device is 5, and the adjustment factor β of the penalty coefficient is 0.5; the current iteration number t′ of the particle swarm optimization algorithm is 50, and the maximum iteration number t′ of the particle swarm optimization algorithm maxIf it is 100 and the penalty coefficient μ is 0.1, then the dynamically adjusted penalty coefficient is

[0124] Adjust the operating parameter h of the z-th device z , update the velocity and position of the particle, and limit the position update of the particle to satisfy: greater than or equal to the minimum value h of the operating parameter of the z-th device set z,min , less than or equal to the maximum value h of the operating parameter of the z-th device z,max ; Continuously update the velocity and position of the particle until the fitness value of the operating parameter of the device no longer changes or reaches the maximum number of iterations preset by the particle swarm optimization algorithm, and then stop, and output the optimal device operating parameter configuration.

[0125] The preset comprehensive quality score threshold of the industrial camera is set by the staff. The comprehensive quality scores of different industrial cameras are collected through the production intelligent control terminal, and the average value of the comprehensive quality scores of multiple industrial cameras is taken as the preset comprehensive quality score threshold of the industrial camera; similarly, the first threshold G1 of the industrial camera quality comprehensive score, the second threshold G2 of the industrial camera quality comprehensive score, the maximum utilization rate threshold U max , the maximum threshold T of the total task scheduling time max and the maximum change rate threshold Δc of the operating parameter of the device with respect to time t max ;

[0126] In this embodiment, the edge information of the image is extracted by the Sobel operator, the texture features are extracted by the gray-level co-occurrence matrix, and the color features are extracted by the color histogram, which can comprehensively describe various types of data in the production process and provide a rich information source for subsequent data analysis; dynamically adjusting the weight value helps to flexibly optimize the denoising effect according to different production situations or environmental changes, ensuring that noise can be effectively removed under different conditions and improving the quality of images and signals; the texture energy and color histogram extraction provide more detailed features for the image, which helps to improve the accuracy of industrial camera image analysis and is particularly suitable for detecting complex production defects and changes;

[0127] Quantitatively evaluating multiple performance indicators (such as resolution, signal-to-noise ratio, contrast, lens distortion coefficient, frame rate, etc.) of the industrial camera through formulas can scientifically and objectively evaluate the camera quality and avoid the deviation of human judgment; by setting the weight factors for controlling the comprehensive clarity, contrast, lens distortion coefficient and frame rate of the image, the quality levels of various important factors can be controlled and optimized during the production process, thereby improving the overall quality and consistency of the product;

[0128] Combining the genetic algorithm and the particle swarm optimization algorithm, the production scheduling and the equipment operation status are simultaneously optimized through a two-layer optimization structure, which can comprehensively improve production efficiency and product quality. The genetic algorithm provides global optimization, while the particle swarm optimization algorithm makes local fine-tuning for equipment parameters to ensure all-round optimization of the production process; by optimizing the production scheduling plan with the genetic algorithm, production tasks can be scientifically allocated, resource waste can be avoided, and the smoothness and efficiency of the production process can be ensured; it can dynamically adjust production tasks according to the actual operation of the equipment, improve equipment utilization rate and ensure that tasks are completed on time; by optimizing the equipment operation parameters with the particle swarm optimization algorithm, key parameters such as the accuracy, temperature, and exposure time of the equipment can be adjusted, thereby improving production quality. The optimized equipment operation parameters help to ensure the quality standards of industrial cameras and prevent production problems caused by improper parameter settings of the equipment; dynamically adjusting the penalty coefficient can make the optimization process more flexible and adaptive. During the optimization process, the penalty coefficient will be gradually adjusted according to the current number of iterations, so that the optimization process can gradually tend to be stable over time, avoiding the situation where the penalty coefficient is too large or too small at the beginning, resulting in an unsatisfactory optimization effect; the comprehensively optimized production scheduling and equipment parameters can maximize the utilization rate of resources, shorten the production cycle, reduce equipment idle time, and improve overall production efficiency and production capacity.

[0129] Embodiment 2

[0130] Please refer to Figure 2 shown. A data analysis and control system for an industrial camera production line in this embodiment includes:

[0131] A data acquisition module that acquires data of the industrial camera production line. The data of the industrial camera production line includes industrial camera production parameter data, production equipment operation data, and industrial camera production image data;

[0132] A data processing module that identifies and eliminates outliers in the industrial camera production parameter data, production equipment operation data, and industrial camera production image data; integrates wavelet transform denoising and multi-dimensional feature extraction technologies, and combines an image denoising method based on two-dimensional convolution and weight dynamic adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain a production parameter feature data set, an equipment operation feature data set, and a production image feature data set;

[0133] A production quality evaluation module that constructs a production quality evaluation model and inputs the production line comprehensive feature data set into the trained production quality evaluation model to predict and obtain a production quality evaluation label;

[0134] The automatic sorting module determines whether the industrial cameras produced are qualified according to the production quality evaluation tags. If the produced industrial cameras are qualified, it analyzes and obtains the quality grades of the industrial cameras, issues sorting instructions through the production intelligent control terminal, and automatically sorts the produced industrial cameras according to the quality grades of the industrial cameras.

[0135] The production line collaborative control module, if the produced industrial cameras are unqualified, optimizes the collaborative work among production equipment through a hybrid intelligent optimization algorithm, collaboratively executes production scheduling optimization and equipment operation state optimization, conducts production scheduling optimization through a genetic algorithm, and guides the optimization with the comprehensive quality score of the industrial camera as the fitness function. It uses the particle swarm optimization algorithm to adjust the equipment operation parameters, controls the constraint conditions through a dynamic penalty coefficient, and adjusts and controls the production of industrial cameras until the produced industrial cameras are qualified. Each module is connected by wired and / or wireless means.

[0136] Since the electronic device introduced in this embodiment is the electronic device used to implement the method and system for analyzing and controlling the data of an industrial camera production line in the embodiments of the present application, based on the method and system for analyzing and controlling the data of an industrial camera production line introduced in the embodiments of the present application, those skilled in the art can understand the specific implementation manners and various variations of the electronic device in this embodiment. Therefore, the specific implementation of how this electronic device implements the method in the embodiments of the present application will not be described in detail here. As long as those skilled in the art implement the electronic device used in the method and system for analyzing and controlling the data of an industrial camera production line in the embodiments of the present application, it falls within the protection scope of the present application.

[0137] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain a formula that is closest to the actual situation. The preset parameters and threshold selection in the formulas are set by those skilled in the art according to the actual situation.

[0138] The above is only the preferred implementation manner of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions falling within the idea of the present invention belong to the protection scope of the present invention. It should be noted that for ordinary technical users in the technical field, several improvements and refinements made without departing from the principle of the present invention should also be regarded as within the protection scope of the present invention.

Claims

1. A data analysis and control method for an industrial camera production line, characterized in that: include: S1. Acquire industrial camera production line data, which includes industrial camera production parameter data, production equipment operation data, and industrial camera production image data; S2. Identify and remove outliers in industrial camera production parameter data, production equipment operation data, and industrial camera production image data; integrate wavelet transform denoising and multidimensional feature extraction technology, and combine image denoising method based on two-dimensional convolution and dynamic weight adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain production parameter feature data set, equipment operation feature data set, and production image feature data set; S3. Build a production quality assessment model, input the production line comprehensive feature data set into the trained production quality assessment model, and predict the production quality assessment label; S4. Determine whether the industrial cameras produced are qualified according to the production quality assessment label; if the industrial cameras produced are qualified, analyze and obtain the quality grade of the industrial cameras; issue sorting instructions through the production intelligent control terminal, and automatically sort the industrial cameras produced according to the quality grade of the industrial cameras; S5. If the industrial cameras produced are unqualified, the collaborative work between production equipment is optimized through a hybrid intelligent optimization algorithm, and production scheduling optimization and equipment operation status optimization are collaboratively performed. Production scheduling optimization is performed through a genetic algorithm, and the comprehensive quality score of the industrial camera is used as the fitness function to guide the optimization; the particle swarm optimization algorithm is used to adjust the equipment operating parameters, and the constraints are controlled through a dynamic penalty coefficient to adjust and control the production of industrial cameras until the produced industrial cameras are qualified.

2. The data analysis and control method for an industrial camera production line according to claim 1, characterized in that: The industrial camera production parameter data includes imaging parameter data, optical component parameter data, electronic component parameter data and interface communication parameter data; the production equipment operation data includes equipment status data, production control data, production environment data and production log data; the industrial camera production image data includes production process images, imaging quality assessment images, defect detection images and classification and annotation images.

3. The data analysis and control method for an industrial camera production line according to claim 2, characterized in that: The method for identifying and eliminating abnormal values ​​in industrial camera production parameter data, production equipment operation data, and industrial camera production image data includes: The K-Means clustering algorithm is used to identify and eliminate outliers in the industrial camera production parameter data, production equipment operation data, and industrial camera production image data. The industrial camera production parameter data, production equipment operation data, and industrial camera production image data after the outliers are eliminated are denoised and multi-dimensional feature extracted to obtain production parameter feature data sets, equipment operation feature data sets, and production image feature data sets. The production parameter feature data set, equipment operation feature data set and production image feature data set after feature extraction are normalized by standard deviation and converted into a standard normal distribution with a mean of 0 and a standard deviation of 1 to obtain the normalized production parameter feature data set, equipment operation feature data set and production image feature data set.

4. The data analysis and control method for an industrial camera production line according to claim 3 is characterized in that: The method for acquiring the production parameter feature data set, the equipment operation feature data set and the production image feature data set comprises: Apply wavelet transform to the production parameter data of industrial cameras and the operation data of production equipment, decompose the signal into different frequency sub-bands, remove noise through thresholding operations, reconstruct the denoised signal, and extract statistical features and frequency domain features. Finally, obtain the denoised production parameter feature dataset and equipment operation feature dataset; For the production image data of industrial cameras, denoise it through the two-dimensional convolution formula and dynamically design and adjust the weight value through the weight value adjustment formula; Use the Sobel operator to detect the edges in the production images of industrial cameras; Calculate the gradients of the production images of industrial cameras in the horizontal and vertical directions respectively through the horizontal gradient calculation formula and the vertical gradient calculation formula; Extract the texture features of the production images of industrial cameras through the gray-level co-occurrence matrix, and calculate the texture energy of the production images of industrial cameras using the energy feature formula; Extract the color features of the production images of industrial cameras by calculating the color histograms of the production images of industrial cameras. Finally, obtain the production parameter feature dataset, equipment operation feature dataset, and production image feature dataset after feature extraction.

5. The data analysis and control method for an industrial camera production line according to claim 4, characterized in that: The method for obtaining the comprehensive feature dataset of the production line includes: Fuse the normalized production parameter feature dataset, equipment operation feature dataset, and production image feature dataset through a weighted model to obtain the comprehensive feature dataset of the production line.

6. The data analysis and control method for an industrial camera production line according to claim 5, characterized in that: The method for constructing the production quality assessment model includes: Construct and train a production quality assessment model. The input data of the model is the historical comprehensive feature dataset of the production line, and the output label of the model is the production quality assessment label. The production quality assessment label is divided into two categories: qualified label and unqualified label; Use the sigmoid function as the activation function, and the production quality assessment model is a gradient boosting decision tree model; Use binary cross-entropy as the loss function to measure the error of model prediction; Use the trained production quality assessment model to predict the current comprehensive feature dataset of the production line to obtain the production quality assessment label.

7. The data analysis and control method for an industrial camera production line according to claim 6, characterized in that: The method for obtaining the quality grade of industrial cameras includes: Further score the industrial cameras that pass the production through the industrial camera quality assessment formula to obtain the comprehensive quality score G of the industrial cameras, and obtain the quality grade of the industrial cameras according to the comprehensive quality score G of the industrial cameras.

8. The data analysis and control method for an industrial camera production line according to claim 7, characterized in that: The method for automatically sorting the produced industrial cameras includes: Preset the comprehensive quality score threshold of industrial cameras, and further divide the comprehensive quality score threshold of industrial cameras into the first threshold G1 and the second threshold G2 of the comprehensive quality score of industrial cameras; Compare the comprehensive quality score G of industrial cameras with the first threshold G1 and the second threshold G2 of the comprehensive quality score of industrial cameras respectively; If G < G1, it is determined that the quality grade of the industrial camera is A level. For A-level industrial cameras, sort them to the high-quality product area; If G1 < G ≤ G2, it is determined that the quality grade of the industrial camera is B level. For B-level cameras, sort them to the medium-quality product area; If G > G2, it is determined that the quality grade of the industrial camera is C level. For C-level cameras, sort them to the qualified product area.

9. The data analysis and control method for an industrial camera production line according to claim 8, characterized in that: The method for adjusting and controlling the production of industrial cameras includes: The optimization of the collaborative work between production equipment is divided into the optimization of production scheduling problems and the optimization of equipment operating status problems. The hybrid intelligent optimization algorithm is used to achieve comprehensive adjustment of production scheduling and equipment parameters. The hybrid intelligent optimization algorithm is a two-layer optimization structure, including genetic algorithm and particle swarm optimization algorithm. The genetic algorithm is used to globally optimize the task allocation of equipment to find the optimal production scheduling solution. Based on the optimal production scheduling solution, the particle swarm optimization algorithm is used to optimize the equipment operating parameters. Define the objective function as maxG, and maximize the comprehensive quality score of industrial cameras by optimizing the production scheduling plan and equipment parameters, while satisfying the constraints of production and equipment operation; the constraints include production scheduling constraints and equipment operation status constraints; production scheduling constraints include equipment utilization constraints and scheduling time constraints, and the working time T of the pth equipment is set as p Divide by the total production cycle T al , and get the equipment utilization rate U p , the preset maximum threshold of equipment utilization is U max , limit the equipment utilization U p Less than or equal to the preset maximum threshold value of equipment utilization U max ; The timer automatically records the start time and end time of each task executed by the device, and subtracts the start time of each task from the end time of each task to get the allocated time T of each task. rw , add up the allocation time of all tasks to get the total scheduling time of all tasks T wa ; The maximum threshold of the total time for all task scheduling is preset to T max , limit the total time of all task scheduling to be less than or equal to the preset maximum threshold T of the total time of all task scheduling max ; The equipment operation status constraints include defining the range of equipment operation parameters, optimizing production scheduling through genetic algorithms, defining each chromosome to represent a production scheduling plan, and forming a production scheduling plan set d; defining the fitness function of the genetic algorithm as the comprehensive quality score of the industrial camera, and calculating the fitness value of each production scheduling plan; The roulette wheel selection method is used to select the best chromosome according to the fitness value; two chromosomes are randomly selected, and the production scheduling part of the chromosome is crossovered to form a new individual; the task order in the chromosome is randomly adjusted to mutate and generate the next generation scheduling plan until the fitness value of the production scheduling plan no longer changes or reaches the maximum number of iterations preset by the genetic algorithm, and the optimal production scheduling plan is output; Optimize the equipment operating parameters through the particle swarm optimization algorithm, define each particle to represent a set of equipment operating parameter configurations, and form the equipment operating parameter configuration set h; define the fitness function of the particle swarm optimization algorithm, and dynamically adjust the penalty coefficient μ in the fitness function through the penalty coefficient adjustment formula; adjust the operating parameter h of the zth equipment z , update the speed and position of the particle, and limit the particle position update to satisfy: greater than or equal to the minimum value h of the operating parameters of the zth device z,min , less than or equal to the maximum value h of the operating parameters of the zth device z,max ; The speed and position of the particles are continuously updated until the fitness value of the equipment's operating parameters no longer changes or the maximum number of iterations preset by the particle swarm optimization algorithm is reached, and the optimal equipment operating parameter configuration is output.

10. An industrial camera production line data analysis control system, used to implement an industrial camera production line data analysis control method according to any one of claims 1 to 9, characterized in that: include: A data acquisition module acquires industrial camera production line data, which includes industrial camera production parameter data, production equipment operation data, and industrial camera production image data; The data processing module identifies and removes outliers in the industrial camera production parameter data, production equipment operation data, and industrial camera production image data; integrates wavelet transform denoising and multidimensional feature extraction technology, and combines image denoising methods based on two-dimensional convolution and dynamic weight adjustment, Sobel edge detection, gray-level co-occurrence matrix texture analysis, and color histogram extraction to obtain production parameter feature data sets, equipment operation feature data sets, and production image feature data sets; The production quality assessment module builds a production quality assessment model, inputs the comprehensive feature data set of the production line into the trained production quality assessment model, and predicts the production quality assessment label; The automatic sorting module determines whether the industrial cameras produced are qualified according to the production quality assessment labels; if the industrial cameras produced are qualified, the quality grade of the industrial cameras is obtained through analysis; sorting instructions are issued through the production intelligent control terminal, and the industrial cameras produced are automatically sorted according to the quality grade of the industrial cameras; The production line collaborative control module, if the industrial cameras produced are unqualified, will use the hybrid intelligent optimization algorithm to optimize the collaborative work between production equipment, collaboratively perform production scheduling optimization and equipment operation status optimization, use the genetic algorithm to optimize production scheduling, and use the comprehensive quality score of the industrial camera as the fitness function to guide the optimization; use the particle swarm optimization algorithm to adjust the equipment operating parameters, and control the constraints through the dynamic penalty coefficient to adjust and control the production of industrial cameras until the produced industrial cameras are qualified.

Images