Image processing analysis method for computer vision technology

Abstract

The invention relates to the field of image processing analysis of a visual technology, in particular to an image processing analysis method for a computer visual technology. The method comprises the following steps: acquiring various types of images according to each high-definition camera, transmitting the images to a data analysis module, analyzing according to the data acquired images, sending the analyzed data to an early warning module, constructing a convolutional neural network model, processing annotated image data to obtain a prediction result and an error value, and performing early warning according to the data of the data analysis. The method comprises the following steps: respectively taking corresponding measures for the sports equipment, collecting sports information of athletes, matching the sports information with the equipment, creating a virtual model to manufacture initial equipment, adjusting parameters according to feedback and measurement, and reproducing. And the accuracy and efficiency of quality and performance evaluation of the sports equipment are improved.

Description

Technical Field

[0001] The present invention relates to the field of image processing and analysis, and in particular to an image processing and analysis method used in computer vision technology. Background Art

[0002] In today's sports field, the requirements for the performance, quality and applicability of sports equipment are increasing day by day. The emergence of computer vision technology provides a new analytical method for the design, production, testing and use of sports equipment. With the professionalization and commercialization of sports, athletes' personalized needs for equipment are increasing. In order to meet the physical characteristics and sports styles of different athletes, sports equipment needs to be customized and optimized more accurately. Traditional sports equipment detection and evaluation methods often rely on manual operation, which is not only inefficient, but also easily affected by subjective factors, resulting in inaccurate results. At the same time, manual detection may be difficult to detect subtle defects and potential problems of equipment. The development of convolutional neural networks provides a powerful tool for automatic extraction of image features and pattern recognition.

[0003] By using computer vision technology to capture the interaction between athletes and equipment, we can gain a deeper understanding of the impact of equipment on athletic performance and provide a scientific basis for equipment improvement and innovation. Image processing and analysis methods using computer vision technology have broad application prospects and urgent needs in the field of sports equipment, and can drive the sports equipment industry towards a more intelligent, personalized and high-quality direction. Summary of the invention

[0004] In order to solve the technical problems raised by the above background technology, the present invention provides an image processing and analysis method for computer vision technology.

[0005] The purpose of the present invention can be achieved through the following technical solutions: Step 1, data collection: collect images of various sports equipment through high-definition cameras installed in the factory, obtain images of various sports equipment, label each sports equipment, including the type, function and quality grade of each sports equipment, and perform image enhancement processing on the images of each sports equipment to obtain the enhanced images of each sports equipment.

[0006] High-definition cameras are installed in multiple designated areas in the training venue to obtain the time information of the current monitoring period, and screenshots of the athletes' various action states corresponding to the current monitoring period are collected. The athletes' movement types are analyzed and labeled based on the screenshots. Specifically: if the athlete leaves the ground with one foot and moves forward quickly with his arms swinging regularly, it can be labeled as a running action; if the athlete jumps with both feet, his body rises into the air, and he throws the object forward with both hands, it is labeled as a throwing action; and each labeled movement type image of the corresponding athlete is obtained.

[0007] The key joint positions are marked, including the shoulder joint, elbow joint, wrist joint, hip joint, knee joint and ankle joint. When marking, the coordinate positions of all joints in the image need to be accurately marked.

[0008] The input end of the data center is connected to the athlete management system, which monitors the athlete's motion data in real time through high-definition cameras. The motion data includes force output, speed, and motion trajectory, etc. The motion data is converted into images and labeled.

[0009] Step 2: Data analysis: Focus the enhanced images of each sports equipment on the outline to obtain the outline corresponding to each sports equipment, divide the outline into outline area and other areas, extract the preset outline area corresponding to each sports equipment in the database, match and overlap the obtained actual outline area with the preset outline area, check whether there are abnormal points in the outline area corresponding to each sports equipment, and the abnormal points include protrusions, depressions, cracks and scratches. If there are abnormal points in the outline area, mark the outline area as an abnormal outline area, count all the abnormal outline areas of the sports equipment, and obtain the number of outline areas, where the number of each outline area can be marked as a, b, c, ..., z, obtain the corresponding contour lines of the equipment through the acquired contour area, check whether the contour lines corresponding to each sports equipment are smooth, if smooth, mark them as smooth contour lines, measure the actual length of each smooth contour line through the instrument, obtain the actual length of each smooth contour line, extract the preset standard length and allowable error range of the data center, compare the measured actual length data with the preset standard length, if the actual length is within the allowable error range of the standard length, it is determined that this part of the contour line of the sports equipment meets the quality requirements; if the actual length exceeds the allowable error range, it is necessary to further analyze the cause and generate a compensation signal to send to the early warning processing module.

[0010] Feature extraction of the bending degree of some sports equipment is performed, specifically: S1: Bend the sports equipment to the maximum value to obtain a curved arc, align the scale edge of the ruler with the two ends of the chord of the arc, ensure that the ruler and the chord completely overlap, and directly read the scale values ​​on the ruler corresponding to the two ends of the chord. The difference between the two is the chord length, marked as L; S2: Place a ruler perpendicular to the chord with one end touching the vertex of the arc. Place one right-angled side of the ruler against the ruler and the other right-angled side touching the arc. The scale value at the point where the ruler touches the arc is the chord height, marked as H. S3: Calculate using formula The curvature radius R of the corresponding sports equipment is obtained. The smaller the curvature radius, the greater the curvature degree, and the larger the curvature radius, the smaller the curvature degree; S4: Extract the preset standard curvature radii of the data center, and compare the preset standard curvature radii with the actual curvature of the corresponding sports equipment. If the preset standard curvature radii are greater than the actual curvature of the corresponding sports equipment, it means that the equipment is unqualified, and an unqualified signal is generated and sent to the early warning module. Otherwise, it means that the sports equipment is qualified.

[0011] The texture analysis of each sports equipment is as follows: the enhanced images of each sports equipment are obtained through data acquisition, marked as initial images, and the grayscale values ​​of pixels in the initial images are obtained. Select the grayscale of the grayscale co-occurrence matrix by grayscale value, convert the obtained grayscale image into 4, 8 and 16 levels, etc., to reduce the matrix size and calculation amount, set the sliding window size and select the calculation direction, set the step size d, which means that the center pixel is compared with its adjacent pixels. For each window matrix, count the grayscale co-occurrence matrix according to the selected direction, count the number of pixel pairs with specific grayscale value combinations, count the number of simultaneous occurrences of pixels with grayscale value i and pixels with grayscale value j to its right, and add them to the (i, j) and (j, i) positions of the grayscale co-occurrence matrix. Perform this step for other directions as well. Divide all elements of the obtained grayscale co-occurrence matrix by the sum of all elements in the matrix to obtain a probability matrix, and then calculate it using the formula P The contrast GH is obtained, the preset contrast threshold of the data center is extracted, and the actually obtained contrast is compared with the preset contrast threshold. If the actually obtained contrast is greater than the preset contrast threshold, it means that the grayscale difference is obvious. Otherwise, it means that the grayscale difference is not significant.

[0012] Step 3: Construct a convolutional neural network model. The architecture of the model includes multiple convolutional layers, pooling layers, and fully connected layers. Obtain the labeled image data obtained by data collection, input the labeled image data into the training model, and send the labeled image data to the convolutional layer. The convolutional layer multiplies the convolution kernel with the elements at the corresponding position of the local area of ​​the input image according to the convolution calculation, and then sums them to obtain the corresponding feature map. Each feature map is obtained through each convolution kernel. The convolutional layer sends the obtained feature maps to the pooling layer. The pooling layer selects the maximum value in each pooling window as the output to obtain the final feature map. The final feature map is flattened into a vector and sent to the fully connected layer. The neurons in the fully connected layer are connected to all neurons in the previous layer through weights. The weighted sum of the input vector and the weight is calculated and the bias is added to obtain the final prediction result. The prediction result is compared with the real image label, and the formula is used. Get the image error value, where n is the number of images, yn is the true value of the nth image, and ui is the model's predicted value for the nth image.

[0013] Step 4: Early warning processing, receiving the compensation signal sent by the data analysis, extracting the deviation information in the compensation signal, and then extracting the historical adjustment parameters of the corresponding sports equipment from the data center. The early warning processing sends the deviation information and historical adjustment parameters to the factory central control system. The factory central control system generates equipment adjustment values ​​based on the deviation information and the corresponding sports equipment historical adjustment parameters, and sends the equipment adjustment values ​​to the control terminal of the corresponding production equipment in real time, and refreshes the parameters of the real-time application in operation through the production equipment terminal.

[0014] By receiving the unqualified signals sent by data analysis, the early warning processing sends the unqualified signals to the factory central control system. The factory central control system receives the signals and sends a stop command to the equipment control terminal on the corresponding production line. The equipment control terminal stops the current production line movement state. The factory central control system extracts the production batches of the corresponding sports equipment and notifies the mobile terminals of the on-site staff. The batches of sports equipment that have been produced with problems are isolated and marked, the emergency troubleshooting process is initiated, and technical personnel are organized to conduct a comprehensive inspection of the production equipment, processes and raw materials.

[0015] Step 5: Equipment customization. Sports equipment is customized according to the physical data of each athlete. The sport type, joint image and motion data of each athlete are obtained through data collection. The joint image and motion data constitute the body motion information of each athlete. The sports equipment used by each athlete in the sport type is extracted. The body motion information of each athlete is matched with the sports equipment of each athlete to obtain the equipment matching value of each athlete. A virtual model of the equipment is created by computer modeling. The equipment matching value is input into the virtual model. The virtual model manufactures initial equipment according to the athletes and effects. The athletes train the initial equipment and collect feedback from the athletes to the factory central control module. At the same time, professional equipment is used to measure the performance indicators of the equipment and the force distribution of the equipment during use. The parameter adjustment values ​​are recalculated, including changing the size, material properties and structure of the equipment. The adjustment values ​​are sent to the factory central control system. The factory central control system re-produces through the production line, manufactures improved prototypes and tests them again until the equipment reaches the best performance and fit.

[0016] Compared with the prior art, the beneficial effects of the present invention are: improving the accuracy of quality inspection, being able to detect tiny defects, flaws and inconsistencies that are difficult to detect manually, and ensuring the high quality and reliability of sports equipment. Enhance customization accuracy, more accurately capture the physical characteristics and sports needs of athletes, provide accurate data support for customized personalized sports equipment, and improve athletes' usage experience and performance. Improve production efficiency, realize automated online detection and quality control, reduce the time and cost of manual inspection, and speed up the production of sports equipment. Real-time monitoring and feedback, real-time monitoring during the use of equipment by athletes, timely detection of potential problems and provision of feedback, help adjust training strategies and prevent injuries. Promote the development of sports science, provide more accurate and detailed equipment and athlete interaction data for sports science research, and promote academic research and practical applications in related fields. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. The following drawings are not intentionally scaled to the actual size, and the focus is on illustrating the main purpose of the present invention.

[0018] Figure 1 The present invention is a flow chart of the method. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work also fall within the scope of protection of the present invention.

[0020] Reference Figure 1 As shown, the present invention is an image processing and analysis method for computer vision technology, and the specific steps are as follows: Step 1, data collection: collect images of various sports equipment through high-definition cameras installed in the factory, obtain images of various sports equipment, label each sports equipment, including the type, function and quality grade of each sports equipment, and perform image enhancement processing on the images of each sports equipment to obtain the enhanced images of each sports equipment.

[0021] High-definition cameras are installed in multiple designated areas in the training venue to obtain the time information of the current monitoring period, and screenshots of the athletes' various action states corresponding to the current monitoring period are collected. The athletes' movement types are analyzed and labeled based on the screenshots. Specifically: if the athlete leaves the ground with one foot and moves forward quickly with his arms swinging regularly, it can be labeled as a running action; if the athlete jumps with both feet, his body rises into the air, and he throws the object forward with both hands, it is labeled as a throwing action; and each labeled movement type image of the corresponding athlete is obtained.

[0022] The key joint positions are annotated. The key joints include shoulder joints, elbow joints, wrist joints, hip joints, knee joints and ankle joints. When annotating, the coordinate positions of all joints in the image need to be accurately marked.

[0023] The input end of the data center is connected to the athlete management system, which monitors the athlete's motion data in real time through high-definition cameras. The motion data includes body size, force output, speed, and motion trajectory, etc. The motion data is converted into images and labeled.

[0024] Step 2: Data analysis: Focus the enhanced images of each sports equipment on the outline to obtain the outline corresponding to each sports equipment, divide the outline into outline area and other areas, extract the preset outline area corresponding to each sports equipment in the database, match and overlap the obtained actual outline area with the preset outline area, check whether there are abnormal points in the outline area corresponding to each sports equipment, and the abnormal points include protrusions, depressions, cracks and scratches. If there are abnormal points in the outline area, mark the outline area as an abnormal outline area, count all the abnormal outline areas of the sports equipment, obtain the number of outline areas, and obtain the outline area through the obtained outline area. Check the corresponding contour lines of each sports equipment to see if they are smooth. If they are smooth, mark them as smooth contour lines. Measure the actual length of each smooth contour line through the instrument to obtain the actual length of each smooth contour line. Extract the preset standard length and allowable error range of the data center, and compare the measured actual length data with the preset standard length. If the actual length is within the allowable error range of the standard length, it is determined that this part of the contour line of the sports equipment meets the quality requirements; if the actual length exceeds the allowable error range, calculate the required compensation amount according to the degree of deviation between the actual length and the standard length, generate a compensation signal and send it to the early warning processing.

[0025] Feature extraction of the bending degree of some sports equipment is performed, specifically: S1: Bend the sports equipment to the maximum value to obtain a curved arc, align the scale edge of the ruler with the two ends of the chord of the arc, ensure that the ruler and the chord completely overlap, and directly read the scale values ​​on the ruler corresponding to the two ends of the chord. The difference between the two is the chord length, marked as L; S2: Place a ruler perpendicular to the chord with one end touching the vertex of the arc. Place one right-angled side of the ruler against the ruler and the other right-angled side touching the arc. The scale value at the point where the ruler touches the arc is the chord height, marked as H. S3: Calculate using formula The curvature radius R of the corresponding sports equipment is obtained. The smaller the curvature radius, the greater the curvature degree, and the larger the curvature radius, the smaller the curvature degree; S4: Extract the preset standard curvature radii of the data center, and compare the preset standard curvature radii with the actual curvature of the corresponding sports equipment. If the preset standard curvature radii are greater than the actual curvature of the corresponding sports equipment, it means that the equipment is unqualified, and an unqualified signal is generated and sent to the early warning module. Otherwise, it means that the sports equipment is qualified.

[0026] The texture analysis of each sports equipment is as follows: the enhanced images of each sports equipment are obtained through data acquisition, marked as initial images, and the grayscale values ​​of pixels in the initial images are obtained. Select the grayscale of the grayscale co-occurrence matrix by grayscale value, convert the obtained grayscale image into 4, 8 and 16 levels, etc., to reduce the matrix size and calculation amount, set the sliding window size and select the calculation direction, set the step size d, which means that the center pixel is compared with its adjacent pixels. For each window matrix, count the grayscale co-occurrence matrix according to the selected direction, count the number of pixel pairs with specific grayscale value combinations, count the number of simultaneous occurrences of pixels with grayscale value i and pixels with grayscale value j to its right, and add them to the (i, j) and (j, i) positions of the grayscale co-occurrence matrix. Perform this step for other directions as well. Divide all elements of the obtained grayscale co-occurrence matrix by the sum of all elements in the matrix to obtain a probability matrix, and then calculate it using the formula P The contrast GH is obtained, the preset contrast threshold of the data center is extracted, and the actually obtained contrast is compared with the preset contrast threshold. If the actually obtained contrast is greater than the preset contrast threshold, it means that the grayscale difference is obvious. Otherwise, it means that the grayscale difference is not significant.

[0027] Step 3: Construct a convolutional neural network model. The architecture of the model includes multiple convolutional layers, pooling layers, and fully connected layers. Obtain the labeled image data obtained from data collection, input the labeled image data into the training model, and send the labeled image data to the convolutional layer. The convolutional layer multiplies the convolution kernel with the elements at the corresponding position of the local area of ​​the input image according to the convolution calculation, and then sums them to obtain the corresponding feature map. Each feature map is obtained through each convolution kernel. The convolutional layer sends the obtained feature maps to the pooling layer. The pooling layer selects the maximum value in each pooling window as the output to obtain the final feature map. The final feature map is flattened into a vector, and the vector is sent to the fully connected layer. The neurons in the fully connected layer are connected to all neurons in the previous layer through weights. The weighted sum of the input vector and the weight is calculated and the bias is added to obtain the prediction result. The prediction result is compared with the real image label, and the formula is used. The image error value L is obtained, where n is the number of images, yn is the true value of the nth image, and ui is the model's predicted value for the nth image. The preset error value of the data center is extracted and compared with the image error value. If the preset error value is greater than the image error value, it means that the error is normal. Otherwise, it means that the image error is large.

[0028] Step 4: Early warning processing, receiving the compensation signal sent by the data analysis, extracting the deviation information in the compensation signal, and then extracting the historical adjustment parameters of the corresponding sports equipment from the data center. The early warning processing sends the deviation information and historical adjustment parameters to the factory central control system. The factory central control system generates equipment adjustment values ​​based on the deviation information and the corresponding sports equipment historical adjustment parameters, and sends the equipment adjustment values ​​to the control terminal of the corresponding production equipment in real time, and refreshes the parameters of the real-time application in operation through the production equipment terminal.

[0029] By receiving the unqualified signals sent by data analysis, the early warning processing sends the unqualified signals to the factory central control system. The factory central control system receives the signals and sends a stop command to the equipment control terminal on the corresponding production line. The equipment control terminal stops the current production line movement state. The factory central control system extracts the production batches of the corresponding sports equipment and notifies the mobile terminals of the on-site staff. The batches of sports equipment that have been produced with problems are isolated and marked, the emergency troubleshooting process is initiated, and technical personnel are organized to conduct a comprehensive inspection of the production equipment, processes and raw materials.

[0030] Step 5: Equipment customization. Sports equipment is customized according to the physical data of each athlete. The sport type, joint image and motion data of each athlete are obtained through data collection. The joint image and motion data constitute the body motion information of each athlete. The sports equipment used by each athlete in the sport type is extracted. The body motion information of each athlete is matched with the sports equipment of each athlete to obtain the equipment matching value of each athlete. A virtual model of the equipment is created by computer modeling. The equipment matching value is input into the virtual model. The virtual model manufactures initial equipment according to the athletes and effects. The athletes train the initial equipment and collect feedback from the athletes to the factory central control module. At the same time, professional equipment is used to measure the performance indicators of the equipment and the force distribution of the equipment during use. The parameter adjustment values ​​are recalculated, including changing the size, material properties and structure of the equipment. The adjustment values ​​are sent to the factory central control system. The factory central control system re-produces through the production line, manufactures improved prototypes and tests them again until the equipment reaches the best performance and fit.

[0031] The above is an explanation of the present invention and should not be considered as a limitation thereof. Although several exemplary embodiments of the present invention have been described, it will be readily appreciated by those skilled in the art that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present invention. Therefore, all such modifications are intended to be included within the scope of the present invention as defined in the claims. It should be understood that the above is an explanation of the present invention and should not be considered as being limited to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present invention is defined by the claims and their equivalents.

Claims

1. An image processing and analysis method for computer vision technology, characterized in that: The steps include: Collect various types of images according to each high-definition camera and transmit the images to data analysis; Analyze the images collected by the data and send the analyzed data to the early warning module; Construct a convolutional neural network model, process the labeled image data to obtain the prediction results and error values, input the labeled image data into the training model according to the labeled image data obtained by data collection, and the training model sends the labeled image data to the convolution layer. The convolution layer multiplies the convolution kernel with the elements of the corresponding position of the local area of ​​the input image according to the convolution calculation, and then sums them to obtain the corresponding feature map. Each feature map is obtained through each convolution kernel. The convolution layer sends each feature map to the pooling layer. The pooling layer selects the maximum value in each pooling window as the output to obtain the final feature map. The final feature map is flattened into a vector and sent to the fully connected layer. The neurons in the fully connected layer are connected to all the neurons in the previous layer through weights. The weighted sum of the input vector and the weight is calculated and the bias is added to obtain the final prediction result. The prediction result is compared with the real image label, and the formula is used. Get the image error value, where n is the number of images, yn is the true value of the nth image, and ui is the model's predicted value for the nth image; According to the data analyzed, corresponding measures are taken respectively; Collect athlete movement information and match it with equipment, create virtual models to manufacture initial equipment, adjust parameters based on feedback and measurements and re-produce.

2. The image processing and analysis method for computer vision technology according to claim 1, characterized in that: Analyze the collected images and send the analyzed data to the early warning module. The specific method is as follows: Focus the enhanced images of each sports equipment on the contour to obtain the contour corresponding to each sports equipment, divide the contour into contour area and other areas, extract the preset contour area corresponding to each sports equipment in the database, match and overlap the obtained actual contour area with the preset contour area, check whether there are abnormal points in the contour area corresponding to each sports equipment, and the abnormal points include protrusions, depressions, cracks and scratches. If there are abnormal points in the contour area, mark the contour area as an abnormal contour area, count all the abnormal contour areas of the sports equipment, obtain the number of contour areas, obtain the corresponding contour lines of the equipment through the obtained contour area, check whether the contour lines corresponding to each sports equipment are smooth, if smooth, mark them as smooth contour lines, measure the actual length of each smooth contour line by an instrument, obtain the actual length of each smooth contour line, extract the preset standard length and allowable error range of the data center, compare the measured actual length data with the preset standard length, if the actual length is within the allowable error range of the standard length, it is determined that this part of the contour line of the sports equipment meets the quality requirements; if the actual length exceeds the allowable error range, generate a compensation signal and send it to the early warning processing module; Feature extraction of the bending degree of some sports equipment is performed, specifically: S1: Bend the sports equipment to the maximum value to obtain a curved arc, align the scale edge of the ruler with the two ends of the chord of the arc, ensure that the ruler and the chord completely overlap, and directly read the scale values ​​on the ruler corresponding to the two ends of the chord. The difference between the two is the chord length, marked as L; S2: Place a ruler perpendicular to the chord with one end touching the vertex of the arc. Place one right-angled side of the ruler against the ruler and the other right-angled side touching the arc. The scale value at the point where the ruler touches the arc is the chord height, marked as H. S3: Calculate using formula Get the curvature radius R of the corresponding sports equipment; S4: extracting the preset standard curvature radii of the data center, and comparing the preset standard curvature radii with the actual curvature of the corresponding sports equipment. If the preset standard curvature radii are greater than the actual curvature of the corresponding sports equipment, it means that the equipment is unqualified, and an unqualified signal is generated and sent to the early warning module. Otherwise, it means that the sports equipment is qualified. The texture of each sports equipment is analyzed, specifically: the enhanced images of each sports equipment are obtained through data acquisition, marked as the initial image, the grayscale value of the pixel in the initial image is obtained, the grayscale level of the grayscale co-occurrence matrix is ​​selected by the grayscale value, and the obtained grayscale image volume is converted into 4, 8 and 16 levels to reduce the matrix size and calculation amount, set the sliding window size and select the calculation direction, set the step size d, which means that the center pixel is compared with its adjacent pixels, and for each window matrix, the grayscale co-occurrence matrix is ​​counted according to the selected direction, the number of pixel pairs with a specific grayscale value combination is counted, and the number of simultaneous occurrences of the pixel with grayscale value i and the pixel with grayscale value j on its right is counted, and it is accumulated to the (i, j) and (j, i) positions of the grayscale co-occurrence matrix. Similarly, this step is performed for other directions, and all elements of the grayscale co-occurrence matrix are divided by the sum of all elements in the matrix to obtain a probability matrix, and then the formula is used to calculate P The contrast GH is obtained, the preset contrast threshold of the data center is extracted, and the actually obtained contrast is compared with the preset contrast threshold. If the actually obtained contrast is greater than the preset contrast threshold, it means that the grayscale difference is obvious. Otherwise, it means that the grayscale difference is not significant.

3. The image processing and analysis method for computer vision technology according to claim 1, characterized in that: According to the data analyzed, corresponding measures are taken respectively, the specific methods are as follows: The early warning processing receives the compensation signal sent by the data analysis, extracts the deviation information in the compensation signal, and then extracts the historical adjustment parameters of the corresponding sports equipment from the data center. The early warning processing sends the deviation information and the historical adjustment parameters to the factory central control system. The factory central control system generates the equipment adjustment value according to the deviation information and the historical adjustment parameters of the corresponding sports equipment, and sends the equipment adjustment value to the control terminal of the corresponding production equipment in real time, and refreshes the parameters of the real-time application in the running through the production equipment terminal; By receiving the unqualified signal sent by data analysis, the early warning processing sends the unqualified signal to the factory central control system. The factory central control system receives the signal and sends a stop command to the equipment control terminal on the corresponding production line. The equipment control terminal stops the current production line movement state. The factory central control system extracts the production batch of the corresponding sports equipment and sends it to the mobile phone terminal of the on-site staff to isolate and mark the batches of sports equipment that have been produced with problems.

4. The image processing and analysis method for computer vision technology according to claim 1, characterized in that: Collect athlete movement information and match it with equipment, create a virtual model to manufacture the initial equipment, adjust parameters based on feedback and measurements and re-produce, the specific method is as follows: Through data collection, the movement type, joint image and movement data of each athlete are obtained. The joint image and movement data constitute the body movement information of each athlete. The sports equipment used by each athlete in the sport type is extracted, and the body movement information of each athlete and the sports equipment of each athlete are matched to obtain the equipment matching value of each athlete. A virtual model of the equipment is created by computer modeling, and the equipment matching value is input into the virtual model. The virtual model manufactures initial equipment according to the athletes and effects. The athletes train the initial equipment, and the athletes' feedback is collected to the factory central control module. At the same time, professional equipment is used to measure the performance indicators of the equipment and the force distribution of the equipment during use, and the parameter adjustment values ​​are recalculated, including changing the size, material properties and structure of the equipment. The adjustment values ​​are sent to the factory central control system, and the factory central control system re-produces through the production line.

5. The image processing and analysis method for computer vision technology according to claim 1, characterized in that: The specific method of collecting various types of images according to each high-definition camera and transmitting the images to data analysis is as follows: The images of various sports equipment are collected by high-definition cameras installed in the factory, and the images of various sports equipment are obtained, and the images of various sports equipment are marked, including the type, function and quality grade of each sports equipment, and the images of various sports equipment are enhanced to obtain the images of various sports equipment after the enhancement processing; Install high-definition cameras in multiple designated areas of the training venue to obtain the time information of the current monitoring period, and take screenshots of the athlete's various action states corresponding to the current monitoring period. Analyze the athlete's movement type based on the screenshots and mark them. Specifically, if you see an athlete with one foot off the ground moving forward quickly and his arms swinging regularly, it can be marked as a running action; if the athlete jumps with both feet, the body rises into the air, and both hands throw the object forward with force, it is marked as a throwing action; obtain the corresponding athlete's various marked movement type images; Label the key joints, including the shoulder, elbow, wrist, hip, knee and ankle joints. The coordinates of all joints in the image need to be accurately marked during labeling. The input end of the data center is connected to the athlete management system, which monitors the athlete's motion data in real time through high-definition cameras. The motion data includes force output, speed and motion trajectory, and converts the motion data into image form and annotates it.

Images