Visual inspection system for surface quality of sand blasting piece

Abstract

The invention discloses a visual detection system for the surface quality of a sand blasting piece, relates to the technical field of sand blasting quality detection, and realizes dynamic optimization of a sand blasting process. According to the generation sequence of the multispectral sand blasting image data, each multispectral sand blasting image data is sequentially compared with pixels in the subsequently generated multispectral sand blasting image data, and the pixels in the multispectral sand blasting image data are divided into product appearance pixels and sand blasting result pixels according to the comparison result. Establishing a standard product sand blasting result image model according to the product appearance pixels and the sand blasting result pixels, setting three-dimensional point cloud distribution in the standard product sand blasting result image model according to the sand blasting work events, and comparing the three-dimensional point cloud distribution with the local three-dimensional point location distribution diagrams in sequence according to the time sequence; and generating and executing a corresponding spray gun adjustment decision according to the comparison result.

Description

Technical Field

[0001] The invention relates to the technical field of sandblasting quality detection, in particular to a visual detection system for the surface quality of a sandblasted part. Background Art

[0002] In modern manufacturing, sandblasting is a common surface treatment process used to improve the surface quality and appearance of parts. However, the traditional sandblasting process often relies on manual inspection to evaluate the surface quality of sandblasted parts. This method is not only time-consuming and labor-intensive, but also greatly affected by subjective factors, making it difficult to ensure quality consistency.

[0003] The existing technology usually uses a camera to capture images during the sandblasting process and identifies surface defects through image processing algorithms. However, these systems often have problems such as low detection accuracy and poor adaptability, especially when faced with complex sandblasting work events and changeable sandblasting results. Therefore, a visual inspection system for the surface quality of sandblasted parts is provided. Summary of the invention

[0004] In order to solve the above technical problems, the object of the present invention is to provide a visual inspection system for the surface quality of sandblasted parts.

[0005] In order to achieve the above object, the present invention provides the following technical solutions: A visual inspection system for the surface quality of sandblasted parts, comprising a cloud computing platform, wherein the cloud computing platform is communicatively connected with a sandblasting video acquisition module, a standard part feature analysis module, and a characterization defect detection module; The sandblasting video acquisition module is used to set several sandblasting work events, and then collect multi-spectral sandblasting video data and overall displacement curves when various sandblasting work events are executed several times through a gravity sensing device and a multi-spectral imager, and collect real-time laser video data and real-time overall displacement curves during the execution of real-time sandblasting work events through a laser imager and a gravity sensing device; The standard feature analysis module is used to establish a three-dimensional coordinate system, and divide the multi-spectral sandblasting video data into a plurality of multi-spectral sandblasting image data by frame, and then map the multi-spectral sandblasting image data onto the three-dimensional coordinate system in sequence according to the overall displacement curve; According to the generation order of the multispectral sandblasting image data, the pixels of each multispectral sandblasting image data are compared with the pixels in the subsequently generated multispectral sandblasting image data in turn, and the pixels in the multispectral sandblasting image data are divided into product appearance pixels and sandblasting result pixels according to the comparison result, and then a standard product sandblasting result image model is established according to the product appearance pixels and the sandblasting result pixels, and at the same time, a three-dimensional point cloud distribution is set in the standard product sandblasting result image model according to the sandblasting work events; The characterization defect detection module is used to generate a local three-dimensional point distribution map based on real-time laser video data and real-time overall displacement curve and map it on the standard product sandblasting result image model, compare the local three-dimensional point distribution map with the three-dimensional point cloud distribution in chronological order, and generate and execute corresponding spray gun adjustment decisions based on the comparison results.

[0006] Furthermore, the acquisition process of the multi-spectral sandblasting video data, the overall displacement curve and the real-time laser video data includes: The sandblasting video acquisition module pre-stores a product sandblasting plan and is communicatively connected to n sandblasting machines, wherein the product sandblasting plan includes the name of the sandblasting product, the sandblasting pattern, and the expected sandblasting thickness; The sandblasting machine is provided with a multi-spectral imager, a laser imager, a gravity sensing device and a spray gun, and n is a natural number greater than 0; According to the product sandblasting plan, several sandblasting work events are set, and each sandblasting work event is executed m times. During the execution of each sandblasting work event, the multi-spectral imager collects multi-spectral sandblasting video data of the interaction between the spray gun and the outer surface of the target product in real time. At the same time, when the spray gun moves as a whole, the gravity sensing device records the overall displacement curve of the spray gun, where m is a natural number greater than 0; When each sandblasting machine performs a real-time sandblasting work event, real-time laser video data of the interaction between the spray gun and the outer surface of the target product is collected, and the corresponding n real-time overall displacement curves are synchronously collected.

[0007] Furthermore, according to the overall displacement curve, the process of sequentially mapping the multi-spectral sandblasting image data onto a three-dimensional coordinate system includes: The multispectral sandblasting video data is divided into several multispectral sandblasting image data by frame, and a three-dimensional coordinate system is established. The overall displacement curves generated by each sandblasting machine in the same sandblasting working event and the same execution number are matched and spliced. Then, according to the changing trend of each overall displacement curve at each time node, the multispectral sandblasting image data generated by each sandblasting machine are overlapped and mapped in the three-dimensional coordinate system.

[0008] Furthermore, the process of dividing the pixels in the multispectral sandblasting image data into product appearance pixels and sandblasting result pixels includes: Obtain the pixel value of each pixel in each multispectral sandblasting image data, and start from the multispectral sandblasting image data generated at the first time node, and according to the overlapping trend of the multispectral sandblasting image data, extract the image fragments that exist in part with the multispectral sandblasting image data generated at the first time node from the k multispectral sandblasting image data after the multispectral sandblasting image data generated at the first time node, and record them as overlapping image fragments, where k is any integer between (0, 10); A first pixel difference threshold and a second pixel difference threshold are set to determine the magnitude relationship between the pixel difference values ​​between pixels and the first pixel difference threshold and the second pixel difference threshold in the overlapping image segments of k multispectral sandblasting image data after the multispectral sandblasting image data generated at each time node; According to the judgment result, each pixel is marked as a product appearance pixel and a sandblasting result pixel, and the product appearance pixels are divided into texture product appearance pixels and plane product appearance pixels.

[0009] Furthermore, the process of establishing the standard product sandblasting result image model includes: Count the product sandblasting result image models generated under the same sandblasting work event but different execution times, the number of occurrences of the image model parts containing sandblasting result pixels at various positions, and set the frequency threshold; The image model portion of the sandblasting result pixels whose occurrence times are less than the frequency threshold is converted into the image model portion of the corresponding product appearance pixels, otherwise no operation is performed; When all image model parts corresponding to the sandblasting result pixels are determined to be converted, the standard product sandblasting result image model is obtained, and then the texture curve is set in the standard product sandblasting result image model according to the texture product appearance pixels.

[0010] Furthermore, the process of setting the three-dimensional point cloud distribution in the standard product sandblasting result image model includes: Then, according to the degree of reflection of the sandblasting material to the laser signal and the expected sandblasting thickness, a three-dimensional point cloud distribution is set on the image model portion corresponding to the sandblasting result pixels on the standard product sandblasting result image model, and the horizontal point cloud height is set, and then the relative sandblasting height of each three-dimensional point cloud distribution is marked according to the horizontal point cloud height and the texture curve distribution.

[0011] Furthermore, the process of comparing the local three-dimensional point distribution map with the three-dimensional point cloud distribution includes: The real-time laser video data is divided into a number of real-time laser image data by frame, and the real-time laser image data are overlapped and distributed in a three-dimensional coordinate system in sequence according to the change trend of each real-time overall displacement curve; Generate a corresponding local 3D point distribution map based on the real-time laser image data generated at each time node, and map each local 3D point distribution map onto the standard product sandblasting result image model, set a point cloud height difference threshold, and compare the total point cloud height difference between the real-time point cloud height of the local 3D point distribution map generated at each time node and the corresponding relative sandblasting height in turn; And judge whether the local three-dimensional point distribution map generated at each time node overlaps with the model part corresponding to the product appearance pixels in the standard product sandblasting result image model, and judge whether the sandblasting operation of the corresponding sandblasting machine at the current time node meets the sandblasting requirements based on the judgment result.

[0012] Furthermore, the generation process of the spray gun adjustment decision includes: If the reason for the judgment is that the total point cloud height difference is greater than the point cloud height difference threshold, and half of the real-time point cloud heights are greater than the corresponding relative sandblasting height, or half of the real-time point cloud heights are less than or equal to the corresponding relative sandblasting height, the spray gun adjustment decision is used to increase or decrease the sandblasting intensity of the corresponding spray gun at the next time node; If the reason for the judgment is that the local three-dimensional point distribution map generated at each time node has overlapped with the model portion corresponding to the product appearance pixels in the standard product sandblasting result image model, a spray gun adjustment decision is generated according to the opposite direction of the relative direction of the existing portion; The spray gun adjustment decision is sent to the corresponding sandblasting machine, and then the sandblasting machine adjusts the sandblasting intensity or sandblasting direction according to the spray gun adjustment decision.

[0013] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention divides the multispectral sandblasting video data into multispectral sandblasting image data by frame, and then maps these image data to a three-dimensional coordinate system according to the overall displacement curve. By sequentially comparing the multispectral sandblasting image data, the product appearance pixels and the sandblasting result pixels are distinguished, and a standard product sandblasting result image model is established. At the same time, a three-dimensional point cloud distribution is set on the model, thereby accurately capturing the surface details of the sandblasted parts and providing a data basis for the accuracy of subsequent sandblasting defect detection.

[0014] 2. The present invention generates a local three-dimensional point distribution map based on real-time laser video data and real-time overall displacement curve, and maps it to the standard product sandblasting result image model. By comparing the local three-dimensional point distribution map with the three-dimensional point cloud distribution in time sequence, based on the real-time collected data and comparison results, the spray gun adjustment decision is instantly generated, thereby realizing dynamic optimization of the sandblasting process and improving the quality of the sandblasting product. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic diagram of the present invention. DETAILED DESCRIPTION

[0016] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0017] like Figure 1As shown, a visual inspection system for the surface quality of sandblasted parts includes a cloud computing platform, wherein the cloud computing platform is communicatively connected with a sandblasting video acquisition module, a standard part feature analysis module, and a characterization defect detection module; The sandblasting video acquisition module is used to set several sandblasting work events, and then collect multi-spectral sandblasting video data and overall displacement curves when various sandblasting work events are executed several times through a gravity sensing device and a multi-spectral imager, and collect real-time laser video data and real-time overall displacement curves during the execution of real-time sandblasting work events through a laser imager and a gravity sensing device; The standard feature analysis module is used to establish a three-dimensional coordinate system, and divide the multi-spectral sandblasting video data into a plurality of multi-spectral sandblasting image data by frame, and then map the multi-spectral sandblasting image data onto the three-dimensional coordinate system in sequence according to the overall displacement curve; According to the generation order of the multispectral sandblasting image data, the pixels of each multispectral sandblasting image data are compared with the pixels in the subsequently generated multispectral sandblasting image data in turn, and the pixels in the multispectral sandblasting image data are divided into product appearance pixels and sandblasting result pixels according to the comparison result, and then a standard product sandblasting result image model is established according to the product appearance pixels and the sandblasting result pixels, and at the same time, a three-dimensional point cloud distribution is set in the standard product sandblasting result image model according to the sandblasting work events; The characterization defect detection module is used to generate a local three-dimensional point distribution map based on real-time laser video data and real-time overall displacement curve and map it on the standard product sandblasting result image model, compare the local three-dimensional point distribution map with the three-dimensional point cloud distribution in chronological order, and generate and execute corresponding spray gun adjustment decisions based on the comparison results.

[0018] Further, the working principle of the present invention is specifically described below through embodiments: The staff uploads several product sandblasting plans to the sandblasting video acquisition module and the standard part feature analysis module, wherein the product sandblasting plan includes the sandblasting product name, sandblasting pattern and expected sandblasting thickness; The sandblasting video acquisition module is communicatively connected to n sandblasting machines, wherein the sandblasting machines are provided with a multi-spectral imager, a laser imager, a gravity sensing device and a spray gun, and n is a natural number greater than 0; Each sandblasting machine is numbered a1, a2, a3, ..., a n , and set several sandblasting work events according to the product sandblasting plan, and there is any difference in the sandblasting product names or sandblasting patterns corresponding to different sandblasting work events; Execute each sandblasting work event m times respectively. During the execution of each sandblasting work event, call the spray gun of the sandblasting machine to perform sandblasting operation on the outer surface of the target product according to the sandblasting pattern. At the same time, the multi-spectral imager of the sandblasting machine collects multi-spectral sandblasting video data of the interaction between the spray gun and the outer surface of the target product in real time. At the same time, when the spray gun moves as a whole, the gravity sensing device records the overall displacement curve of the spray gun, wherein m is a natural number greater than 0; It should be noted that the video acquisition direction of the multi-spectral imager and the laser imager is always consistent with the direction of the sandblasting operation of the associated spray gun; During the real-time sandblasting process of each sandblasting machine, the laser imager located on the sandblasting machine uses the same method as the multispectral imager to collect real-time laser video data during the interaction between the spray gun and the outer surface of the target product, and synchronously collects the corresponding n real-time overall displacement curves.

[0019] Furthermore, when all the sandblasting work events are executed m times, the sandblasting video data acquisition module will mark the corresponding sandblasting machine number and sandblasting work event for each multi-spectral sandblasting video data and the overall displacement curve, and then send all the multi-spectral sandblasting video data and the overall displacement curve to the standard feature analysis module, and send the real-time laser video data and the real-time overall displacement curve to the characterization defect detection module; The standard part feature analysis module divides the multispectral sandblasting video data into several multispectral sandblasting image data by frame. Since the shooting angle of the multispectral imager is always consistent with that of the spray gun, the outer surface position of the target product is always included in the shooting range of the multispectral imager. At the same time, since the spectral absorption rate of the outer surface of the target product to the same frequency band is different before and after sandblasting, the multispectral sandblasting image data generated by the corresponding position at the two time nodes before and after are significantly different. Establish a three-dimensional coordinate system to match and splice the overall displacement curves generated by each sandblasting machine at the same sandblasting working event and the same execution times; Then, according to the change trend of each overall displacement curve at each time node, the multispectral sandblasting image data correspondingly generated by each sandblasting machine is overlapped and mapped in a three-dimensional coordinate system, wherein the change trend of each multispectral sandblasting image data in the three-dimensional coordinate system is the same as that of the overall displacement curve; Obtain the pixel value of each pixel in each multispectral sandblasting image data, and start from the multispectral sandblasting image data generated at the first time node, and according to the overlapping trend of the multispectral sandblasting image data, extract the image fragments that exist in part with the multispectral sandblasting image data generated at the first time node from the k multispectral sandblasting image data after the multispectral sandblasting image data generated at the first time node, and record them as overlapping image fragments, where k is any integer between (0, 10); It should be noted that, since the target product generally has uneven textures on its surface, such as the groove curve on the front cover of a vehicle, the multispectral imager acquires images of the corresponding parts. Due to lighting reasons, the acquisition effect of the multispectral imager on the groove part of the texture is darker than that on the convex part of the texture, resulting in differences in the pixel values ​​of the corresponding pixels, but the difference is smaller than that before and after sandblasting. Setting a first pixel difference threshold and a second pixel difference threshold, wherein the first pixel difference threshold is less than the second pixel difference threshold, and determining whether a pixel difference between each pixel in each overlapping image segment and a corresponding portion of pixels in the multispectral sandblasting image data generated at the first time node is greater than or equal to the second pixel difference threshold; If the pixel difference is greater than or equal to the second pixel difference threshold, the corresponding pixel in the multispectral sandblasting image data generated at the first time node is marked as a sandblasting result pixel; if the pixel difference is less than the second pixel difference threshold, the corresponding pixel is marked as a product appearance pixel; Then, k multispectral sandblasting image data after the multispectral sandblasting image data generated at the second time node are selected, and the process of labeling each pixel in the multispectral sandblasting image data generated at the first time node is repeated, and so on until the pixels in the multispectral sandblasting image data generated at the last time node are labeled and stopped; The process of marking the pixels of the sandblasting results is adopted to again judge the size relationship between the pixel difference between the product appearance pixels and the first pixel difference threshold in the overlapping image segments of k multispectral sandblasting image data after the multispectral sandblasting image data generated at each time node, and then divide the product appearance pixels into texture product appearance pixels and plane product appearance pixels.

[0020] Further, according to the distribution of sandblasting result pixels and product appearance pixels in each multispectral sandblasting image data under the same number of executions, as well as the distribution and overlap of each multispectral sandblasting image data in a three-dimensional coordinate system, the product appearance pixels and sandblasting result pixels in each multispectral sandblasting image data are spliced ​​respectively, so as to obtain a product appearance image model and a product sandblasting result image model of the target product before and after the sandblasting work event under the corresponding number of executions. For the product sandblasting result image model, a texture curve is set on the surface of the product sandblasting result image model according to the distribution of texture product appearance pixels; It should be noted that the product appearance image model and the product sandblasting result image model have the same overall structure, but the product sandblasting result image model includes both the sandblasting result pixels and the image model parts corresponding to the product appearance pixels, while the product appearance image model is generated only based on the product appearance pixels; Count the product sandblasting result image models generated under the same sandblasting work event but different execution times, the number of occurrences of the image model parts containing sandblasting result pixels at each position, and set a frequency threshold, where the frequency threshold is less than m; The image model portion of the sandblasting result pixels whose occurrence times are less than the frequency threshold is converted into the image model portion of the corresponding product appearance pixels, otherwise no operation is performed; When all image model parts corresponding to the sandblasting result pixels are determined to be converted, a standard product sandblasting result image model is obtained, and then the texture curve on the surface of the product sandblasting result image model is mapped onto the standard product sandblasting result image model; During the sandblasting process on the outer surface of the target product, as the sandblasting thickness changes continuously, the reflection degree of the laser signal emitted by the laser imager by the outer surface of the target product also changes accordingly; Then, according to the degree of reflection of the sandblasting material to the laser signal and the expected sandblasting thickness, a three-dimensional point cloud distribution is set on the image model portion corresponding to the sandblasting result pixels on the standard product sandblasting result image model, and the horizontal point cloud height is set, and then the relative sandblasting height of each three-dimensional point cloud distribution is marked according to the horizontal point cloud height and the texture curve distribution.

[0021] Further, after the characterization defect detection module receives the real-time laser video data and the real-time overall displacement curve from the sandblasting image acquisition module, the real-time laser video data is divided into a plurality of real-time laser image data by frame; By mapping the multi-spectral sandblasting image data to a three-dimensional coordinate system, the real-time laser image data is sequentially overlapped and distributed in the three-dimensional coordinate system according to the change trend of each real-time overall displacement curve, and the real-time laser image data distribution in the three-dimensional coordinate system is updated in real time according to the subsequently generated real-time laser video data and real-time overall displacement curve; At the same time, due to the different reflection degrees of laser signals by sandblasting results of different thicknesses, the corresponding local three-dimensional point distribution map is generated according to the real-time laser image data generated at each time node, and each local three-dimensional point distribution map is mapped onto the standard product sandblasting result image model; Set the point cloud height difference threshold, and compare the total point cloud height difference between the real-time point cloud height of the local 3D point distribution map generated at each time node and the corresponding relative sandblasting height; If the total point cloud height difference value is less than or equal to the point cloud height difference threshold, it is judged that the sandblasting operation of the corresponding sandblasting machine at the current time node meets the sandblasting requirements; If the total point cloud height difference value is greater than the point cloud height difference threshold, it is determined that the sandblasting operation of the corresponding sandblasting machine at the current time node does not meet the sandblasting requirements; At the same time, it is determined whether the local three-dimensional point distribution map generated at each time node overlaps with the model corresponding to the product appearance pixels in the standard product sandblasting result image model. If so, it is determined that the sandblasting operation of the corresponding sandblasting machine at the current time node does not meet the sandblasting requirements; If it does not exist, it is determined that the sandblasting operation of the corresponding sandblasting machine at the current time node meets the sandblasting requirements.

[0022] Further, if it is determined that the sandblasting operation of the sandblasting machine at the current time node meets the sandblasting requirements, no operation is performed; If it is determined that the sandblasting operation of the sandblasting machine at the current time node does not meet the sandblasting requirements, a spray gun adjustment decision is set for the corresponding sandblasting machine according to the determination reason; If the reason for the judgment is that the total point cloud height difference is greater than the point cloud height difference threshold, and half of the real-time point cloud heights are greater than the corresponding relative sandblasting height, or half of the real-time point cloud heights are less than or equal to the corresponding relative sandblasting height, the spray gun adjustment decision is used to increase or decrease the sandblasting intensity of the corresponding spray gun at the next time node; If the reason for the judgment is that the local three-dimensional point distribution map generated at each time node has overlapped with the model portion corresponding to the product appearance pixels in the standard product sandblasting result image model, a spray gun adjustment decision is generated according to the opposite direction of the relative direction of the existing portion; The spray gun adjustment decision is sent to the corresponding sandblasting machine, and then the sandblasting machine adjusts the sandblasting intensity or sandblasting direction according to the spray gun adjustment decision; The above process of generating spray gun adjustment decisions is repeated until the corresponding real-time sandblasting work event ends.

[0023] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A visual inspection system for the surface quality of sandblasted parts, comprising a cloud computing platform, characterized in that: The cloud computing platform is communicatively connected with a sandblasting video acquisition module, a standard part feature analysis module, and a characterization defect detection module; the sandblasting video acquisition module is used to set several sandblasting work events, and then collect multispectral sandblasting video data and overall displacement curves when various sandblasting work events are executed several times through a gravity sensing device and a multispectral imager, and collect real-time laser video data and real-time overall displacement curves during the execution of real-time sandblasting work events through a laser imager and a gravity sensing device; the standard part feature analysis module is used to establish a three-dimensional coordinate system, and divide the multispectral sandblasting video data into several multispectral sandblasting image data by frame, and then map the multispectral sandblasting image data to the three-dimensional coordinate system in sequence according to the overall displacement curve; according to the multispectral sandblasting The image data is generated in an order that each multispectral sandblasting image data is compared with the pixels in the subsequently generated multispectral sandblasting image data in turn, and the pixels in the multispectral sandblasting image data are divided into product appearance pixels and sandblasting result pixels according to the comparison results, and then a standard product sandblasting result image model is established according to the product appearance pixels and the sandblasting result pixels, and at the same time, a three-dimensional point cloud distribution is set in the standard product sandblasting result image model according to the sandblasting work event; the characterization defect detection module is used to generate a local three-dimensional point distribution map according to the real-time laser video data and the real-time overall displacement curve and map it on the standard product sandblasting result image model, and the local three-dimensional point distribution map is compared with the three-dimensional point cloud distribution in chronological order, and the corresponding spray gun adjustment decision is generated and executed according to the comparison result.

2. A visual inspection system for the surface quality of sandblasted parts according to claim 1, characterized in that: The acquisition process of the multi-spectral sandblasting video data, the overall displacement curve and the real-time laser video data includes: The sandblasting video acquisition module pre-stores a product sandblasting plan and is communicatively connected to n sandblasting machines, wherein the product sandblasting plan includes the name of the sandblasting product, the sandblasting pattern, and the expected sandblasting thickness; The sandblasting machine is provided with a multi-spectral imager, a laser imager, a gravity sensing device and a spray gun, and n is a natural number greater than 0; According to the product sandblasting plan, several sandblasting work events are set, and each sandblasting work event is executed m times. During the execution of each sandblasting work event, the multi-spectral imager collects multi-spectral sandblasting video data of the interaction between the spray gun and the outer surface of the target product in real time. At the same time, when the spray gun moves as a whole, the gravity sensing device records the overall displacement curve of the spray gun, where m is a natural number greater than 0; During the real-time sandblasting work event of each sandblasting machine, real-time laser video data of the interaction between the spray gun and the outer surface of the target product is collected, and the corresponding real-time overall displacement curve is synchronously collected.

3. A visual inspection system for the surface quality of sandblasted parts according to claim 2, characterized in that: According to the overall displacement curve, the process of mapping the multi-spectral sandblasting image data onto the three-dimensional coordinate system in sequence includes: The multispectral sandblasting video data is divided into several multispectral sandblasting image data by frame, and a three-dimensional coordinate system is established. The overall displacement curves generated by each sandblasting machine in the same sandblasting working event and the same execution number are matched and spliced. Then, according to the changing trend of each overall displacement curve at each time node, each multispectral sandblasting image data is overlapped and mapped in the three-dimensional coordinate system.

4. A visual inspection system for the surface quality of sandblasted parts according to claim 3, characterized in that: The process of dividing the pixels in the multispectral sandblasting image data into product appearance pixels and sandblasting result pixels includes: Obtain the pixel value of each pixel in each multispectral sandblasting image data, and starting from the multispectral sandblasting image data generated at the first time node, extract the image fragments that exist in part with the multispectral sandblasting image data generated at the first time node from the k multispectral sandblasting image data after the multispectral sandblasting image data generated at the first time node, and record them as overlapping image fragments, where k is any integer between (0, 10); Then, a first pixel difference threshold and a second pixel difference threshold are set, and the magnitude relationship between the pixel difference values ​​between pixels and the first pixel difference threshold and the second pixel difference threshold in the overlapping image segments of k multispectral sandblasting image data after the multispectral sandblasting image data generated at each time node is determined in sequence; According to the judgment result, each pixel is marked as a product appearance pixel and a sandblasting result pixel, and the product appearance pixels are divided into texture product appearance pixels and plane product appearance pixels.

5. A visual inspection system for the surface quality of sandblasted parts according to claim 4, characterized in that: The process of establishing the standard product sandblasting result image model includes: Count the product sandblasting result image models generated under the same sandblasting work event but different execution times, the number of occurrences of the image model parts containing sandblasting result pixels at each position, and set a frequency threshold. The image model parts of sandblasting result pixels with a number of occurrences less than the frequency threshold are converted into the image model parts corresponding to the product appearance pixels, otherwise no operation is performed; When all image model parts corresponding to the sandblasting result pixels are determined to be converted, a standard product sandblasting result image model is obtained, and a texture curve is set on the standard product sandblasting result image model according to the texture product appearance pixels.

6. A visual inspection system for the surface quality of sandblasted parts according to claim 5, characterized in that: The process of setting up the 3D point cloud distribution in the standard product sandblasting result image model includes: According to the degree of reflection of the sandblasting material to the laser signal and the expected sandblasting thickness, a three-dimensional point cloud distribution is set on the image model portion corresponding to the sandblasting result pixels on the standard product sandblasting result image model, and the horizontal point cloud height is set, and then the relative sandblasting height is marked for each three-dimensional point cloud distribution according to the horizontal point cloud height and the texture curve distribution.

7. A visual inspection system for the surface quality of sandblasted parts according to claim 6, characterized in that: The process of comparing the local three-dimensional point distribution map with the three-dimensional point cloud distribution includes: The real-time laser video data is divided into a number of real-time laser image data by frame, and the real-time laser image data are overlapped and distributed in a three-dimensional coordinate system in sequence according to the change trend of each real-time overall displacement curve; Generate a corresponding local 3D point distribution map based on the real-time laser image data generated at each time node, and map each local 3D point distribution map onto the standard product sandblasting result image model, set a point cloud height difference threshold, and compare the total point cloud height difference between the real-time point cloud height of the local 3D point distribution map generated at each time node and the corresponding relative sandblasting height in turn; And judge whether the local three-dimensional point distribution map generated at each time node overlaps with the model part corresponding to the product appearance pixels in the standard product sandblasting result image model, and judge whether the sandblasting operation of the corresponding sandblasting machine at the current time node meets the sandblasting requirements based on the judgment result.

8. A visual inspection system for the surface quality of sandblasted parts according to claim 7, characterized in that: The generation process of the spray gun adjustment decision includes: If the reason for judgment is that the total point cloud height difference is greater than the point cloud height difference threshold, and half of the real-time point cloud heights are greater than the corresponding relative sandblasting height, or half of the real-time point cloud heights are less than or equal to the corresponding relative sandblasting height, then the spray gun adjustment decision is used to increase or decrease the sandblasting intensity of the corresponding spray gun at the next time node; if the reason for judgment is that the local three-dimensional point distribution map generated at each time node has an overlapped model part corresponding to the product appearance pixels in the standard product sandblasting result image model, then the spray gun adjustment decision is generated according to the opposite direction of the relative direction of the existing part.

Images