High-definition visual inspection and identification system based on image processing industrial personal computer

Abstract

The invention belongs to the field of visual inspection, and relates to a high-definition visual inspection and identification system based on an image processing industrial personal computer, which comprises a historical data acquisition module for acquiring image data of historical records of the same product; the feature point extraction module is used for constructing a convolutional neural network in deep learning and extracting key feature points and other feature points of the image data; the three-dimensional model construction module is used for calculating space coordinate points of other feature points in the three-dimensional prediction model; the error threshold setting module is used for setting a first error threshold and a second error threshold; the sorting scheme selection module is used for selecting a proper sorting scheme; the qualified threshold setting module is used for setting a first qualified threshold and a second qualified threshold; and the production scheme optimization module is used for selecting a proper optimization scheme. The problem that the accuracy and the consistency of detection results are difficult to guarantee due to the fact that a visual detection method is influenced by human factors is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of visual detection and relates to a high-definition visual detection and recognition system based on an image processing industrial computer. Background Art

[0002] Traditional visual inspection methods mainly rely on manual labeling and empirical judgment. This method is not only inefficient and difficult to cope with large-scale, high-intensity production tasks, but also the inspection results are affected by human factors, such as operator fatigue and mood swings, which may make it difficult to ensure the accuracy and consistency of the inspection results. In modern industrial production, this uncertain visual inspection method is unacceptable because any slight error may have a significant impact on product quality and production efficiency.

[0003] With the rapid development of industrial automation and intelligent technology, modern production has higher and higher requirements for product quality inspection. Uncertain visual inspection methods can no longer meet this demand for efficient, accurate and stable inspection. There is an urgent need for a solution that can automatically, quickly and accurately complete the inspection task to ensure the stability and consistency of product quality.

[0004] In response to the above problems, traditional visual inspection methods are affected by human factors, which easily leads to difficulties in ensuring the accuracy and consistency of the inspection results. Summary of the invention

[0005] In order to solve the problem that traditional visual inspection methods are affected by human factors, which easily leads to the difficulty in ensuring the accuracy and consistency of inspection results, the present invention provides a high-definition visual inspection and recognition system based on an image processing industrial computer.

[0006] A high-definition visual inspection and recognition system based on an image processing industrial computer includes the following modules: The historical data collection module obtains image data of the same product history records at different angles, under different lighting conditions, and in different background environments; Feature point extraction module, used to build convolutional neural networks in deep learning and extract key feature points and other feature points of image data; A three-dimensional model building module is used to build a three-dimensional prediction model of feature points and calculate the spatial coordinates of other feature points in the three-dimensional prediction model; An error threshold setting module is used to obtain the error values ​​between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of different pictures, and to set the first error threshold and the second error threshold; A sorting scheme selection module is used to calculate the error parameters between the spatial coordinate points of other feature points of the actual product and the spatial coordinate points of other feature points of the three-dimensional prediction model, determine the relationship between the error parameters and the first error threshold and the second error threshold, and select a suitable sorting scheme; A qualified threshold setting module is used to monitor the qualified rate achieved during the product production process and set a first qualified threshold and a second qualified threshold; The production plan optimization module compares the relationship between the product qualification rate and the first qualification threshold and the second qualification threshold, and selects a suitable optimization plan based on the judgment result.

[0007] Preferably, obtaining the image data comprises the following steps: Collect multiple image data of the same product history, including image data with different lighting conditions, angles, and background environments; The image data is preprocessed using filtering technology to reduce image noise, contrast adjustment and sharpening, normalization, and correction processing steps.

[0008] Preferably, constructing a convolutional neural network in deep learning includes the following steps: Label the key feature points and other feature points of the image data, build a convolutional neural network in deep learning, and divide the labeled image data into training sets, validation sets, and test sets; The training set trains the convolutional neural network to extract key feature points and other feature points of image data with different angles, different lighting conditions and different background environments; the validation set verifies whether the convolutional neural network can extract key feature points and other feature points of the image data; the test set tests the convolutional neural network and finally evaluates the effect of feature point extraction.

[0009] Preferably, constructing the three-dimensional model construction module comprises the following steps: Extract key feature points of image data. The types of key feature points include corner points of products, edge points of products, and specific marking points. Filter out key feature points of identifiability and stability in different image data. A three-dimensional spatial coordinate system is constructed according to the key feature points, the three-dimensional position of the camera, and the shooting direction of the lens, a three-dimensional prediction model of the feature points is obtained, and the spatial coordinate points of other feature points in the three-dimensional spatial coordinate system are calculated.

[0010] Preferably, constructing the three-dimensional space coordinate system comprises the following steps: Use the camera tracking algorithm to track the motion trajectory of the camera and calculate the shooting direction of the lens; use the laser rangefinder to measure the three-dimensional position of the camera; obtain the internal parameters of the camera's focal length, sensor size, and distortion coefficient; One or more key feature points are selected as the coordinate origin, and the direction of the coordinate axis is determined using the camera's shooting direction and the camera's internal parameters; a three-dimensional space coordinate system is established using the coordinate origin and the direction of the coordinate axis; the X-axis and the Y-axis are parallel to the two-dimensional image plane, and the Z-axis is consistent with the camera's shooting direction.

[0011] Preferably, constructing the three-dimensional prediction model comprises the following steps: By utilizing the internal and external parameters of the camera and combining the correspondence between key feature points, the projection matrix used to describe the mapping relationship from three-dimensional space to two-dimensional image plane is solved. By combining the establishment of a three-dimensional space coordinate system and the solution of the projection matrix, the three-dimensional prediction model is constructed. The coordinate points of other feature points of the two-dimensional image plane in three-dimensional space are calculated using the back-projection process of the projection matrix.

[0012] Preferably, setting the first error threshold and the second error threshold comprises the following steps: Obtain key feature points and other feature points of different images, input the three-dimensional prediction model, match the key feature points of different images with the three-dimensional spatial coordinate system of the three-dimensional prediction model, use the least squares method to approximate the real three-dimensional coordinate points, there are errors between the three-dimensional coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points actually measured, combine the calculated errors to form a data set to set a first error threshold and a second error threshold; For each feature point in the three-dimensional prediction model, calculate the spatial coordinate errors between other feature points in the three-dimensional prediction model and the corresponding feature points actually measured, and calculate the average error and maximum error of the spatial coordinate errors of other feature points; The spatial coordinate error is calculated to satisfy the following formula: ; in, , , It is the first The x, y, z coordinates of the feature points; , , The actual measurement The x, y, z coordinates of the feature points; is the spatial coordinate error; The average error is calculated to satisfy the following formula: ; in, is the average error, represents the number of feature points, It is the sum of the spatial coordinate errors of all feature points; The maximum error is calculated to satisfy the following formula: ; in, It means finding the largest one among all feature point spatial coordinate errors.

[0013] Preferably, selecting a suitable sorting scheme includes the following steps: Extracting feature points of image data of small batch products in the same batch, inputting them into the three-dimensional prediction model, matching key feature points of the products with the three-dimensional spatial coordinate system of the three-dimensional prediction model, calculating error parameters between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of the real products, judging the relationship between the error parameters and the first error threshold and the second error threshold, and selecting a suitable sorting scheme according to the judgment result; If the error parameter is less than the first error threshold, it means that the accuracy of the three-dimensional prediction model is acceptable; If the error parameter is greater than or equal to the first error threshold and less than the second error threshold, it means that the accuracy of the three-dimensional prediction model needs to be further fine-tuned; If the error parameter is greater than or equal to the second error threshold, it is determined that the error parameters between the spatial coordinate points of other feature points of the 3D prediction model and the spatial coordinate points of other feature points of the actual product are beyond the prediction range, indicating that the 3D prediction model needs to be rebuilt.

[0014] Preferably, setting the first qualified threshold and the second qualified threshold comprises the following steps: Real-time monitoring of the qualified rate of products in the same batch during the production process. The products meet the qualified standards: the error parameters between the spatial coordinate points of other feature points of the trained three-dimensional prediction model and the spatial coordinate points of other real feature points of the product are less than the first error threshold; the first qualified threshold and the second qualified threshold are set according to the qualified rate of the product.

[0015] Preferably, selecting a suitable optimization scheme comprises the following steps: If the qualified rate of the product is greater than the second qualified threshold, it means that the product can directly pass the quality inspection; If the qualified rate of the product is greater than the first qualified threshold and less than or equal to the second qualified threshold, it means that the enterprise needs to further check whether there are errors in the production process and procedures; If the qualified rate of the product is less than or equal to the first qualified threshold, it means that the company needs to further check whether there are errors in the shape and size of the batch of products.

[0016] In summary, the present invention includes the following beneficial technical effects: 1. By using the convolutional neural network in deep learning to extract key feature points and other feature points of image data and constructing a three-dimensional prediction model, the precise coordinates of other feature points in three-dimensional space can be calculated. This high-precision feature point extraction and spatial coordinate calculation greatly improves the accuracy of detection and recognition; at the same time, because the system can automatically process and analyze image data, it also significantly improves the efficiency of detection and recognition and reduces manual intervention and errors; 2. By setting the first qualified threshold and the second qualified threshold, it is possible to determine in real time whether the product meets the qualified standards and select the appropriate optimization plan according to the change in the qualified rate; this dynamic and real-time quality monitoring method enables enterprises to find problems in time and take corresponding measures to ensure the stability and reliability of product quality. The system continuously optimizes and adjusts the detection and recognition algorithm based on historical data and real-time data to adapt to changes in different products and production environments; 3. Through automated and intelligent image processing and data analysis, the system can provide enterprises with rich production data and information, helping them to better understand the production process and product quality. These data and information can provide strong support for enterprises to formulate production plans and optimize production processes. The system integrates and links with other production management systems to achieve sharing and collaborative management of production data, further improving the efficiency and intelligence level of production management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The invention discloses a framework schematic diagram of a high-definition visual detection and recognition system based on an image processing industrial computer. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.

[0019] The following is combined with Figure 1 The preferred embodiments of the present invention are described in detail.

[0020] See attached Figure 1 As shown, the present invention proposes a high-definition visual inspection and recognition system based on an image processing industrial computer, including: a historical data acquisition module, a feature point extraction module, a three-dimensional model construction module, an error threshold setting module, a sorting scheme selection module, a qualified threshold setting module, and a production scheme optimization module.

[0021] The historical data acquisition module is used to obtain image data of the same product history records at different angles, under different lighting conditions, and under different background environments, and to pre-process the image data.

[0022] Specifically, multiple image data of the same product's historical records are collected based on multiple channels such as historical records and the company's database; multiple image data, including image data with different angles, different lighting conditions, and different background environments; Image data is preprocessed using filtering technology to reduce image noise, contrast adjustment and sharpening, normalization, correction and other steps.

[0023] The feature point extraction module is used to construct a convolutional neural network (CNN) in deep learning. The preprocessed image data is input into the convolutional neural network (CNN) in deep learning to extract key feature points and other feature points of the image data.

[0024] Specifically, the key feature points and other feature points of the image data are annotated through manual annotation methods, and a convolutional neural network (CNN) in deep learning is constructed. The annotated image data is divided into a training set, a validation set, and a test set. The training set trains the convolutional neural network (CNN) to extract the key feature points and other feature points of the image data, the validation set verifies whether the convolutional neural network (CNN) can extract the key feature points and other feature points of the image data, and the test set tests the effect of the convolutional neural network (CNN) in extracting the key feature points and other feature points of the image data.

[0025] Exemplarily, image data of a product history record is obtained, and the unique key feature point and multiple other feature points in the image are manually annotated. The training set, the validation set, and the test set are allocated in a ratio of 7:2:1. A convolutional neural network (CNN) in deep learning including multiple convolutional layers, pooling layers, and fully connected layers is constructed, and the multiple convolutional layers are used to extract the unique key feature point and multiple other feature points of the image data; The training set trains the convolutional neural network (CNN) to extract key feature points and other feature points of image data with different angles, different lighting conditions, and different background environments; the validation set verifies whether the convolutional neural network (CNN) can extract key feature points and other feature points of image data; the test set finally evaluates the effect of feature point extraction.

[0026] The 3D model building module builds a 3D space coordinate system based on key feature points, camera position, and lens shooting direction, obtains the 3D prediction model of the feature points, and calculates the spatial coordinate points of other feature points in the 3D space coordinate system.

[0027] Specifically, extract key feature points of image data, and the types of key feature points include corner points of products, edge points of products, and specific marking points. Filter key feature points of identifiability and stability in different image data as the basis for building a three-dimensional coordinate system; build a three-dimensional space coordinate system based on key feature points, the three-dimensional position of the camera, and the shooting direction of the lens, and obtain a three-dimensional prediction model of the feature points; calculate the spatial coordinate points of other feature points in the three-dimensional space coordinate system based on the obtained three-dimensional prediction model.

[0028] In one embodiment of the present invention, establishing a three-dimensional space coordinate system includes the following steps: Use the camera tracking algorithm to track the motion trajectory of the camera, analyze the camera's posture, the camera's shooting direction is consistent with the optical axis direction, and calculate the shooting direction of the lens; use the laser rangefinder to measure the three-dimensional position of the camera; obtain the camera's focal length, sensor size, distortion coefficient and other internal parameters; One or more key feature points are selected as the coordinate origin. The selection of the coordinate origin facilitates subsequent three-dimensional reconstruction and data analysis. The direction of the coordinate axis is determined using the camera's shooting direction and the camera's internal parameters. A three-dimensional space coordinate system is established using the coordinate origin and the direction of the coordinate axis. The X-axis and Y-axis are parallel to the two-dimensional image plane, and the Z-axis is consistent with the camera's shooting direction.

[0029] In one embodiment of the present invention, constructing the three-dimensional prediction model includes the following steps: The projection matrix used to describe the mapping relationship from three-dimensional space to two-dimensional image plane is solved by using the internal and external parameters of the camera and the corresponding relationship of key feature points. The three-dimensional prediction model is constructed by combining the establishment of the three-dimensional space coordinate system and the solution of the projection matrix, and the coordinate points of other feature points of the two-dimensional image plane in three-dimensional space are calculated by the back-projection process of the projection matrix.

[0030] For example, assuming that the camera has been calibrated, obtain The projection matrix is ​​M: ; in, and is the focal length of the camera, and are the coordinates of the light center, It is obtained through the camera calibration process.

[0031] Extract the two-dimensional coordinate point P(100,200) from the captured two-dimensional image and convert it into homogeneous coordinate form (100,200,1), if it is not feasible to find the inverse projection matrix, the three-dimensional coordinate points are continuously adjusted through the iterative optimization method; the iterative optimization method of the least squares method is used to approximate the true three-dimensional coordinate points, and the three-dimensional space coordinate points corresponding to the two-dimensional coordinate points P(u,v) of other feature points are obtained (x,y,z).

[0032] The error threshold setting module obtains the key feature points and other feature points of different pictures, inputs the three-dimensional prediction model of the feature points, matches the key feature points of different pictures with the three-dimensional spatial coordinate system of the three-dimensional prediction model, calculates the error values ​​between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of different pictures, and sets the first error threshold and the second error threshold based on the data set composed of the error values.

[0033] Specifically, the least squares method is used to approximate the true three-dimensional coordinate points. There are errors between the three-dimensional coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points actually measured. The first error threshold and the second error threshold are set in combination with the calculated error composition data set.

[0034] Among them, for each feature point in the three-dimensional prediction model, calculating the spatial coordinate errors of other feature points in the three-dimensional prediction model and the corresponding feature points actually measured, and calculating the average error and maximum error of the spatial coordinate errors of other feature points, including the following steps: The spatial coordinate error is calculated to satisfy the following formula: ; in, , , It is the first The x, y, z coordinates of the feature points; , , The actual measurement The x, y, z coordinates of the feature points; is the spatial coordinate error; The average error is calculated to satisfy the following formula: ; in, is the average error, represents the number of feature points, It is the sum of the spatial coordinate errors of all feature points; The maximum error is calculated to satisfy the following formula: ; in, It means finding the largest one among all feature point spatial coordinate errors.

[0035] For example, assuming that the spatial coordinate errors of the other feature points of the two image data are 3 mm and 5 mm, we know that: the average error mm, maximum error mm; The first error threshold and the second error threshold are set according to the calculated average error and maximum error; The first error threshold is set to 80% of the maximum error, and the first error threshold is 4 mm; The second error threshold is set to twice the average error, and the second error threshold is 8 mm.

[0036] The sorting scheme selection module is used to calculate the error parameters between the spatial coordinate points of other actual feature points of the product and the spatial coordinate points of other feature points of the three-dimensional prediction model, determine the relationship between the error parameters and the first error threshold and the second error threshold, and select a suitable sorting scheme.

[0037] Specifically, extract the key feature points and other feature points of the image data of the small batch of products in the same batch, input them into the 3D prediction model, and match the key feature points of the products with the 3D spatial coordinate system of the 3D prediction model. Calculate the error parameters between the spatial coordinate points of the other feature points of the 3D prediction model and the spatial coordinate points of the other feature points of the real products; judge the relationship between the error parameters and the first error threshold and the second error threshold, and select a suitable sorting scheme according to the judgment result; If the error parameter is less than the first error threshold, it means that the accuracy of the three-dimensional prediction model is acceptable; If the error parameter is greater than or equal to the first error threshold and less than the second error threshold, it means that the accuracy of the three-dimensional prediction model needs to be further fine-tuned; If the error parameter is greater than or equal to the second error threshold, it can be determined that the error parameters between the spatial coordinate points of other feature points of the 3D prediction model and the spatial coordinate points of other feature points of the actual product are beyond the prediction range, indicating that the 3D prediction model needs to be rebuilt.

[0038] For example, the error parameter is set , indicating that the 3D prediction model is acceptable and that the 3D prediction model can be used for high-definition visual inspection and recognition of products; set up , indicating that the three-dimensional prediction model is further fine-tuned, calculating the error parameters between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of the real product, and inputting them into the three-dimensional prediction model to fine-tune the accuracy of the three-dimensional prediction model; set up , indicating that the 3D prediction model needs to be rebuilt and multiple image data from historical records needs to be reselected for training.

[0039] The qualified threshold setting module is used to monitor the qualified rate achieved in the product production process and set the first qualified threshold and the second qualified threshold according to the qualified rate of the product.

[0040] Specifically, the qualified rate of products in the same batch during the production process is monitored in real time, and the products meet the qualified standards: the error parameters between the spatial coordinate points of other feature points of the trained three-dimensional prediction model and the spatial coordinate points of other real feature points of the product are less than the first error threshold; the first qualified threshold and the second qualified threshold are set according to the qualified rate of the product.

[0041] For example, assume that certain precision mechanical parts need to undergo strict quality inspection to ensure that they meet the design requirements, the first qualified threshold is set to 90%, and the second qualified threshold is set to 95%.

[0042] The production plan optimization module sets the first qualified threshold and the second qualified threshold, compares the relationship between the qualified rate of the product and the first qualified threshold and the second qualified threshold, and selects a suitable optimization plan based on the judgment result.

[0043] Specifically, the qualified rate of products in the same batch is monitored in real time, the relationship between the qualified rate of the products and the first qualified threshold and the second qualified threshold is compared, and a suitable optimization scheme is selected according to the judgment result, including: If the qualified rate of the product is greater than the second qualified threshold, it means that the product can directly pass the quality inspection; If the qualified rate of the product is greater than the first qualified threshold and less than or equal to the second qualified threshold, it means that the enterprise needs to further check whether there are errors in the production process and procedures; If the qualified rate of the product is less than or equal to the first qualified threshold, it means that the company needs to further check whether there are errors in the shape and size of the batch of products.

[0044] For example, if the product qualification rate is , indicating that the product has passed the quality inspection directly; if Product qualification rate , further check and correct errors in the product's production process and procedures; If the product's qualification rate , further detect errors in the shape and size of the batch of products.

[0045] The interface definitions and communication protocols between modules clarify the interface specifications between modules, including: interface name, parameter type, return value, etc., and formulate corresponding communication protocols to ensure that data and information can be transmitted correctly and efficiently between modules.

[0046] The above embodiments are only used to illustrate the technical solutions of the present invention. Although the present invention is described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A high-definition visual detection and recognition system based on an image processing industrial computer, characterized in that: Includes the following modules: The historical data collection module obtains image data of the same product history records at different angles, under different lighting conditions, and in different background environments; Feature point extraction module, used to build convolutional neural networks in deep learning and extract key feature points and other feature points of image data; A three-dimensional model building module is used to build a three-dimensional prediction model of feature points and calculate the spatial coordinates of other feature points in the three-dimensional prediction model; An error threshold setting module is used to obtain the error values ​​between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of different pictures, and to set the first error threshold and the second error threshold; A sorting scheme selection module is used to calculate the error parameters between the spatial coordinate points of other feature points of the actual product and the spatial coordinate points of other feature points of the three-dimensional prediction model, determine the relationship between the error parameters and the first error threshold and the second error threshold, and select a suitable sorting scheme; A qualified threshold setting module is used to monitor the qualified rate achieved during the product production process and set a first qualified threshold and a second qualified threshold; The production plan optimization module compares the relationship between the product qualification rate and the first qualification threshold and the second qualification threshold, and selects a suitable optimization plan based on the judgment result.

2. According to claim 1, a high-definition visual detection and recognition system based on an image processing industrial computer is characterized in that: Acquiring the image data comprises the following steps: Collect multiple image data of the same product history, including image data with different lighting conditions, angles, and background environments; The image data is preprocessed using filtering technology to reduce image noise, contrast adjustment and sharpening, normalization, and correction processing steps.

3. According to claim 1, a high-definition visual detection and recognition system based on an image processing industrial computer is characterized in that: Constructing a convolutional neural network for deep learning includes the following steps: Label the key feature points and other feature points of the image data, build a convolutional neural network in deep learning, and divide the labeled image data into training sets, validation sets, and test sets; The training set trains the convolutional neural network to extract key feature points and other feature points of image data with different angles, different lighting conditions and different background environments; the validation set verifies whether the convolutional neural network can extract key feature points and other feature points of the image data; the test set tests the convolutional neural network and finally evaluates the effect of feature point extraction.

4. According to claim 3, a high-definition visual detection and recognition system based on an image processing industrial computer is characterized in that: Constructing the three-dimensional model construction module includes the following steps: Extract key feature points of image data. The types of key feature points include corner points of products, edge points of products, and specific marking points. Filter out key feature points of identifiability and stability in different image data. A three-dimensional spatial coordinate system is constructed according to the key feature points, the three-dimensional position of the camera, and the shooting direction of the lens, a three-dimensional prediction model of the feature points is obtained, and the spatial coordinate points of other feature points in the three-dimensional spatial coordinate system are calculated.

5. A high-definition visual detection and recognition system based on an image processing industrial computer according to claim 4, characterized in that: Constructing the three-dimensional space coordinate system includes the following steps: Use the camera tracking algorithm to track the motion trajectory of the camera and calculate the shooting direction of the lens; use the laser rangefinder to measure the three-dimensional position of the camera; obtain the internal parameters of the camera's focal length, sensor size, and distortion coefficient; One or more key feature points are selected as the coordinate origin, and the direction of the coordinate axis is determined using the camera's shooting direction and the camera's internal parameters; a three-dimensional space coordinate system is established using the coordinate origin and the direction of the coordinate axis; the X-axis and the Y-axis are parallel to the two-dimensional image plane, and the Z-axis is consistent with the camera's shooting direction.

6. The high-definition visual detection and recognition system based on an image processing industrial computer according to claim 5 is characterized in that: Constructing the three-dimensional prediction model includes the following steps: By utilizing the internal and external parameters of the camera and combining the correspondence between key feature points, the projection matrix used to describe the mapping relationship from three-dimensional space to two-dimensional image plane is solved. By combining the establishment of a three-dimensional space coordinate system and the solution of the projection matrix, the three-dimensional prediction model is constructed. The coordinate points of other feature points of the two-dimensional image plane in three-dimensional space are calculated using the back-projection process of the projection matrix.

7. The high-definition visual detection and recognition system based on an image processing industrial computer according to claim 5 is characterized in that: Setting the first error threshold and the second error threshold comprises the following steps: Obtain key feature points and other feature points of different images, input the three-dimensional prediction model, match the key feature points of different images with the three-dimensional spatial coordinate system of the three-dimensional prediction model, use the least squares method to approximate the real three-dimensional coordinate points, there are errors between the three-dimensional coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points actually measured, combine the calculated errors to form a data set to set a first error threshold and a second error threshold; For each feature point in the three-dimensional prediction model, calculate the spatial coordinate errors between other feature points in the three-dimensional prediction model and the corresponding feature points actually measured, and calculate the average error and maximum error of the spatial coordinate errors of other feature points; The spatial coordinate error is calculated to satisfy the following formula: ； in, , , It is the first The x, y, z coordinates of the feature points; , , The actual measurement The x, y, z coordinates of the feature points; is the spatial coordinate error; The average error is calculated to satisfy the following formula: ； in, is the average error, represents the number of feature points, It is the sum of the spatial coordinate errors of all feature points; The maximum error is calculated to satisfy the following formula: ； in, It means finding the largest one among all feature point spatial coordinate errors.

8. The high-definition visual detection and recognition system based on an image processing industrial computer according to claim 7 is characterized in that: Selecting a suitable sorting solution includes the following steps: Extracting feature points of image data of small batch products in the same batch, inputting them into the three-dimensional prediction model, matching key feature points of the products with the three-dimensional spatial coordinate system of the three-dimensional prediction model, calculating error parameters between the spatial coordinate points of other feature points of the three-dimensional prediction model and the spatial coordinate points of other feature points of the real products, judging the relationship between the error parameters and the first error threshold and the second error threshold, and selecting a suitable sorting scheme according to the judgment result; If the error parameter is less than the first error threshold, it means that the accuracy of the three-dimensional prediction model is acceptable; If the error parameter is greater than or equal to the first error threshold and less than the second error threshold, it means that the accuracy of the three-dimensional prediction model needs to be further fine-tuned; If the error parameter is greater than or equal to the second error threshold, it is determined that the error parameters between the spatial coordinate points of other feature points of the 3D prediction model and the spatial coordinate points of other feature points of the actual product are beyond the prediction range, indicating that the 3D prediction model needs to be rebuilt.

9. The high-definition visual detection and recognition system based on an image processing industrial computer according to claim 1 is characterized in that: Setting the first qualified threshold and the second qualified threshold comprises the following steps: Real-time monitoring of the qualified rate of products in the same batch during the production process. The products meet the qualified standards: the error parameters between the spatial coordinate points of other feature points of the trained three-dimensional prediction model and the spatial coordinate points of other real feature points of the product are less than the first error threshold; the first qualified threshold and the second qualified threshold are set according to the qualified rate of the product.

10. A high-definition visual detection and recognition system based on an image processing industrial computer according to claim 9, characterized in that: Selecting a suitable optimization solution includes the following steps: If the qualified rate of the product is greater than the second qualified threshold, it means that the product can directly pass the quality inspection; If the qualified rate of the product is greater than the first qualified threshold and less than or equal to the second qualified threshold, it means that the enterprise needs to further check whether there are errors in the production process and procedures; If the qualified rate of the product is less than or equal to the first qualified threshold, it means that the company needs to further check whether there are errors in the shape and size of the batch of products.

Images