Injection molding part surface treatment quality inspection process and device based on visual tracking

Abstract

The invention discloses an injection molding part surface treatment quality inspection process and device based on visual tracking, and relates to the technical field of injection molding part surface treatment visual quality inspection. The quality inspection mode does not interfere with the process rhythm of each injection molding part surface treatment production line, visual analysis original materials are collected in a video set mode, different-angle pictures of the target injection molding part can be more comprehensively obtained, and the probability of quality inspection failure caused by unsuccessful single-angle picture feature matching is reduced; based on a machine learning model, whether injection molding part target areas of different injection molding part surface treatment production line process nodes are qualified or not is judged, a quality inspection result is output according to a qualification judgment result, and a worker analyzes whether unqualified semi-finished products need to be repaired or directly scrapped or not; and the quality inspection accuracy is high, the production takt is not affected, and the surface treatment consumable consumption caused by defective products is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of visual quality inspection of injection molded part surface treatment, and in particular to a process and device for quality inspection of injection molded part surface treatment based on visual tracking. Background Art

[0002] With the diversification and personalization of automotive interior and exterior trims, automotive exterior trims have been given more changes. Different surface treatments of injection molded parts can bring different visual effects. In order to pursue better visual effects, complex surface treatment processes are often required, such as injection molding, spraying, chrome plating, thermal transfer, screen printing, etc.

[0003] In conventional injection molding surface treatment production lines, the final finished automotive interior parts are generally inspected for quality. This quality inspection may be visual inspection or manual inspection. The surface structural defects and visual defects can be found. Although this quality inspection process can realize semi-automatic or fully automatic finished product inspection, it cannot automatically inspect all links of the surface treatment production line. Such defects may occur after the injection molding is completed, such as wavy lines and cracks in injection molding defects, and may also occur in chrome plating, such as coating shedding, burning, yellowing, white fog, bubbling, rough coating, pitting, pinholes, etc. during the chrome plating process of injection molding parts. If these defects are not discovered and handled in time, it will lead to a waste of resources required for subsequent surface treatment process steps.

[0004] Therefore, it is urgent to develop processes and related equipment that can realize automated quality inspection of each link of the injection molding surface treatment production line. Summary of the invention

[0005] In order to solve the technical problem of automatic quality inspection of each link of the above-mentioned injection molded parts surface treatment production line, the present invention provides an injection molded parts surface treatment quality inspection process and device based on visual tracking. The following technical solutions are adopted:

[0006] The surface treatment quality inspection process of injection molded parts based on visual tracking includes the following steps:

[0007] Step 1, using an injection molded part surface treatment quality inspection device to collect an image set of the injection molded part during the surface treatment process in real time;

[0008] Step 2, the visual analysis module of the injection molded part surface treatment quality inspection device uses a background subtraction method to detect the injection molded part targets in the image set based on the time sequence;

[0009] Step 3, using a tracking algorithm to track the target area of ​​the injection molded part at the process node of the injection molded part surface treatment production line;

[0010] Step 4, for the target area of ​​the injection molded part detected by tracking, extracting shape features and texture features based on a multi-threaded fast feature extraction algorithm;

[0011] Step 5, using a machine learning model to determine whether the target area of ​​the injection molded parts at different injection molded parts surface treatment production line process nodes is qualified;

[0012] Step 6: If the judgment results of the target area of ​​the injection molded parts at each production line process node are qualified, the surface treatment of the injection molded parts is judged to be qualified. If the judgment results of the target area of ​​the injection molded parts are unqualified, the injection molded parts are rejected at the rejection station set after the current production line process node.

[0013] By adopting the above technical scheme, firstly, an injection molded part surface treatment quality inspection device is used to collect a set of images of the injection molded parts in the surface treatment process in real time. The image set essentially refers to the visual picture of a section of the production line behind the process node of the injection molded part surface treatment production line. This quality inspection method will not interfere with the process rhythm of each injection molded part surface treatment production line. For example, the process node of the injection molded part surface treatment production line generally refers to a section of the production line after each process step is completed, such as a section of the production line behind the injection molded part outlet, a section of the production line after the primer spraying treatment is completed, a section of the production line after the chrome plating treatment is completed, etc. The tracking shooting length of this section of the production line can be determined according to the specific situation, for example, it can be one meter.

[0014] The use of video collection to collect visual analysis raw materials can more comprehensively obtain different angles of the target injection molded parts. The principle is that during the tracking process, the speed is inconsistent with the speed of the assembly line, which will lead to changes in shooting angle and light, so more angles can be captured. Shooting at different angles makes subsequent feature matching easier to succeed, reducing the probability of quality inspection failure due to unsuccessful feature matching of single-angle photos.

[0015] Subsequently, the background subtraction method is used to detect the injection molded parts targets in the image set based on the time sequence, so as to reduce the interference of the background miscellaneous items in the visual analysis process of the injection molded parts targets. The shape features and texture features are extracted based on the multi-threaded fast feature extraction algorithm. The multi-threaded fast feature extraction algorithm refers to the contour segmentation of the injection molded parts target area, and the feature extraction of multiple contour segmentation areas is performed simultaneously by multi-threading. Finally, the machine learning model is used to judge whether the injection molded parts target areas of different injection molded parts surface treatment production line process nodes are qualified. Finally, the quality inspection results are output according to the qualified judgment results. If the judgment results of the injection molded parts target areas of each production line process node are qualified, the results of judging the qualified surface treatment of the injection molded parts can be output. If the judgment results of the injection molded parts target areas are unqualified, the injection molded parts are rejected at the rejection station set after the current production line process node. The unqualified rejection of semi-finished products can avoid the waste of materials caused by the continuation of subsequent processes. The staff can analyze whether the unqualified semi-finished products need to be repaired or directly scrapped. The quality inspection accuracy is high, the production rhythm is not affected, and the consumption of surface treatment consumables caused by defective products is reduced.

[0016] Optionally, step 2 further includes a preprocessing step before the background subtraction method, and the preprocessing step includes grayscale processing and inter-frame difference processing.

[0017] Optionally, the frame difference processing is implemented using the following formula:

[0018] D(x, y, t) = |I g (x, y, t)-I g (x, y, t-1)|;

[0019] Where D(x, y, t) refers to the change in pixel intensity at the image position (x, y) at time t, I g (x, y, t) represents the grayscale value of the image position (x, y) at time t, I g (x, y, t-1) represents the grayscale value of the image position (x, y) at time t-1.

[0020] By adopting the above technical solution, usually due to the color of the surface process treatment, it is a color image, which usually needs to be converted into a grayscale image;

[0021] In order to detect motion, the use of inter-frame difference processing methods can improve the success rate of subsequent feature recognition.

[0022] Optionally, the specific method of the background subtraction method in step 2 is: create a background model, and implement background subtraction by taking the average or median value of the pixels in the video sequence. Let B(x, y) be the pixel value of the background model at the position (x, y), and the background model is updated by the following formula:

[0023] B(x, y)=α·B(x, y)+(1-α)·I(x, y);

[0024] Where I(x, y) is the pixel value at position (x, y) in the current frame, and α is a weight coefficient between 0 and 1, which is used to control the speed of background update;

[0025] Compare the current frame with the background model to detect the target area of ​​the injection molded part. Let F(x, y) be the value of the foreground mask at position (x, y). Then the injection molded part target area detection is achieved by the following formula:

[0026]

[0027] Where T is the threshold value. If the pixel value difference is greater than T, it is classified as the target area of ​​the injection molded part.

[0028] By adopting the above technical solution, a background model is created, which is usually achieved by taking the average or median of the pixels in the video sequence. Injection molded part target area detection: The current frame is compared with the background model to detect the injection molded part target area detection.

[0029] Where T is a threshold used to determine whether the pixel value difference is large enough to classify it as the target area of ​​the injection molded part.

[0030] Optionally, in step 3, an optical flow method is used to track the target area of ​​the injection molded part. The optical flow method is implemented using the following formula:

[0031] I(x,y,t)=I(x+Δx,y+Δy,t+Δt);

[0032] Where I(x, y, t) represents the pixel brightness at the image position (x, y) at time t, x represents the horizontal coordinate of the image, y represents the vertical coordinate of the image, Δx and Δy are the components of the optical flow vector in the x and yy directions respectively, and Δt represents the displacement of the optical flow vector in time.

[0033] By adopting the above technical solution and using the optical flow method or other tracking algorithms to track the target area of ​​the injection molded parts, the complete extraction of the target area of ​​the injection molded parts can be achieved, and the target area of ​​the injection molded parts can be provided for subsequent visual quality inspection.

[0034] Optionally, in step 4, the multi-threaded fast feature extraction algorithm includes the following steps:

[0035] Step 41, extracting the contour of the target area of ​​the injection molded part successfully tracked in step 3, and dividing it into blocks based on the extracted closed complete contour to obtain a plurality of feature items to be determined;

[0036] Step 42: Match multiple feature items to be determined based on a preset Haar-like feature database.

[0037] Optionally, the Haar-like feature of the feature item to be determined is calculated using the following formula:

[0038]

[0039] Where II(x, y) is the value of the integral image at point (x, y), and I(i, j) is the pixel value of the original image at point (i, j);

[0040] Using the integral image, the pixels that enclose the complete contour are quickly calculated using the following formula:

[0041] sum=II(x2, y2)-II(x1, y1)-II(x2, y1)+II(x1, y2);

[0042] Among them, (x1, y1) and (x2, y2) are the endpoint coordinates of the longest line that closes the complete contour;

[0043] If the difference between the calculation result of the pixels of the closed complete contour and the calculation result of the pixels in the feature database is within the set difference threshold, the match is judged to be successful.

[0044] Optionally, the difference threshold is 5%.

[0045] By adopting the above technical solution, Haar-like features are extracted by calculating the difference in pixel intensity between adjacent areas in the image, which can quickly realize feature extraction and matching. Multi-threading is used to simultaneously process multiple feature items to be determined, which greatly improves the feature extraction efficiency and can match the surface treatment production rhythm to avoid reducing the production rhythm due to quality inspection.

[0046] The device for quality inspection of surface treatment of injection molded parts based on visual tracking comprises a plurality of visual tracking shooting devices and a visual analysis module. The plurality of visual tracking shooting devices are respectively installed at process nodes of the plastic parts surface treatment production line. The visual tracking shooting device comprises a bracket, an electric linear slide and a visual camera. The track of the electric linear slide is installed on the rear side of the process node through the bracket. The visual camera is installed on the slider of the electric linear slide and follows the lateral movement of the slider to realize follow-up shooting. The visual analysis module is respectively communicated with the visual cameras of the plurality of visual tracking shooting devices to exchange shooting picture data. The visual analysis module runs a quality inspection program designed based on the visual tracking quality inspection process for surface treatment of injection molded parts and outputs quality inspection results.

[0047] By adopting the above technical solution, an electric linear slide is used to dynamically track and shoot the original picture. The electric linear slide runs smoothly and the speed is controllable. A smaller differential can be achieved for tracking and shooting. The captured picture can also meet the shooting requirements of multiple angles and different light conditions brought by the differential, greatly improving the success rate of visual quality inspection.

[0048] Optionally, the visual analysis module includes a memory, a visual analysis computer and a display, the memory being communicatively connected to visual cameras of multiple visual tracking and shooting devices respectively, the visual analysis computer running a quality inspection program designed for injection molded parts surface treatment quality inspection process based on visual tracking, displaying the quality inspection results, and controlling the execution actions of rejection stations set after multiple production line process nodes respectively according to the quality inspection results.

[0049] In summary, the present invention includes at least one of the following beneficial technical effects:

[0050] The present invention can provide a quality inspection process and device for surface treatment of injection molded parts based on visual tracking. First, the quality inspection device for surface treatment of injection molded parts is used to collect image sets of injection molded parts in the surface treatment process in real time. This quality inspection method will not interfere with the process rhythm of each injection molded part surface treatment production line. The original visual analysis material is collected in the form of video sets, and different angle images of the target injection molded parts can be obtained more comprehensively. The principle is that during the tracking process, the speed is inconsistent with the speed of the assembly line, which will cause changes in the shooting angle and light, so more angles can be shot, and shooting at different angles makes subsequent feature matching easier to succeed, reducing the probability of quality inspection failure due to unsuccessful feature matching of single-angle photos.

[0051] Subsequently, the background subtraction method is used to detect the injection molded parts targets in the image set based on the time sequence, so as to reduce the interference of the background miscellaneous items in the visual analysis process of the injection molded parts targets. The shape features and texture features are extracted based on the multi-threaded fast feature extraction algorithm. The multi-threaded fast feature extraction algorithm refers to the contour segmentation of the injection molded parts target area, and the feature extraction of multiple contour segmentation areas is performed simultaneously by multi-threading. Finally, the machine learning model is used to judge whether the injection molded parts target areas of different injection molded parts surface treatment production line process nodes are qualified. Finally, the quality inspection results are output according to the qualified judgment results. If the judgment results of the injection molded parts target areas of each production line process node are qualified, the results of judging the qualified surface treatment of the injection molded parts can be output. If the judgment results of the injection molded parts target areas are unqualified, the injection molded parts are rejected at the rejection station set after the current production line process node. The unqualified rejection of semi-finished products can avoid the waste of materials caused by the continuation of subsequent processes. The staff can analyze whether the unqualified semi-finished products need to be repaired or directly scrapped. The quality inspection accuracy is high, the production rhythm is not affected, and the consumption of surface treatment consumables caused by defective products is reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a schematic diagram of the process flow of the injection molded parts surface treatment quality inspection process based on visual tracking of the present invention;

[0053] Figure 2 It is a schematic diagram of the component installation and connection principle of the injection molded parts surface treatment quality inspection device based on visual tracking of the present invention.

[0054] Explanation of the accompanying drawings: 1. Visual analysis module; 11. Memory; 12. Visual analysis computer; 13. Display; 2. Electric linear slide; 3. Visual camera. DETAILED DESCRIPTION

[0055] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0056] The embodiment of the present invention discloses an injection molded part surface treatment quality inspection process and a device based on visual tracking.

[0057] Reference Figure 1 and Figure 2 , Embodiment 1, the surface treatment quality inspection process of injection molded parts based on visual tracking, comprises the following steps:

[0058] Step 1, using an injection molded part surface treatment quality inspection device to collect an image set of the injection molded part during the surface treatment process in real time;

[0059] Step 2, the visual analysis module 1 of the injection molded part surface treatment quality inspection device uses a background subtraction method to detect the injection molded part targets in the image set based on the time sequence;

[0060] Step 3, using a tracking algorithm to track the target area of ​​the injection molded part at the process node of the injection molded part surface treatment production line;

[0061] Step 4, for the target area of ​​the injection molded part detected by tracking, extracting shape features and texture features based on a multi-threaded fast feature extraction algorithm;

[0062] Step 5, using a machine learning model to determine whether the target area of ​​the injection molded parts at different injection molded parts surface treatment production line process nodes is qualified;

[0063] Step 6: If the judgment results of the target area of ​​the injection molded parts at each production line process node are qualified, the surface treatment of the injection molded parts is judged to be qualified. If the judgment results of the target area of ​​the injection molded parts are unqualified, the injection molded parts are rejected at the rejection station set after the current production line process node.

[0064] First, the injection molded parts surface treatment quality inspection device is used to collect real-time image sets of injection molded parts during the surface treatment process. The image set essentially refers to the visual picture of a section of the assembly line behind the process node of the injection molded parts surface treatment production line. This quality inspection method will not interfere with the process rhythm of each injection molded parts surface treatment production line. For example, the process node of the injection molded parts surface treatment production line generally refers to a section of the assembly line after each process step is completed, such as a section of the assembly line behind the injection molded parts outlet, a section of the assembly line after the primer spraying treatment is completed, a section of the assembly line after the chrome plating treatment is completed, etc. The tracking shooting length of this section of the assembly line can be determined according to the specific situation, for example, it can be one meter.

[0065] The use of video collection to collect visual analysis raw materials can more comprehensively obtain different angles of the target injection molded parts. The principle is that during the tracking process, the speed is inconsistent with the speed of the assembly line, which will cause changes in shooting angle and light, so more angles can be captured. Shooting at different angles makes subsequent feature matching easier to succeed, reducing the probability of quality inspection failure due to unsuccessful feature matching of single-angle photos.

[0066] Subsequently, the background subtraction method is used to detect the injection molded parts targets in the image set based on the time sequence, so as to reduce the interference of the background miscellaneous items in the visual analysis process of the injection molded parts targets. The shape features and texture features are extracted based on the multi-threaded fast feature extraction algorithm. The multi-threaded fast feature extraction algorithm refers to the contour segmentation of the injection molded parts target area, and the use of multi-threading to extract features from multiple contour segmentation areas at the same time. Finally, the machine learning model is used to judge whether the injection molded parts target areas at different injection molded parts surface treatment production line process nodes are qualified.

[0067] Use machine learning models (such as SVM, neural network, etc.) to classify features and determine whether the quality is qualified. The decision function of SVM is:

[0068]

[0069] Among them, α i is the Lagrange multiplier, y i is the category label, including qualified and unqualified for different reasons, K is the kernel function, and b is the bias term;

[0070] Finally, the quality inspection results are output according to the qualified judgment results. If the judgment results of the target areas of the injection molded parts at each production line process node are qualified, the results of judging the qualified surface treatment of the injection molded parts can be output. If the judgment results of the target areas of the injection molded parts are unqualified, the reasons for the unqualified can also be judged, such as defects, surface effect deviations, etc. Different rejection stations can be set according to different unqualified reasons. According to different unqualified reasons, the corresponding rejection stations set after the current production line process node will reject the injection molded parts. The unqualified rejection of semi-finished products can avoid material waste caused by the continuation of subsequent processes. The staff can analyze whether the unqualified semi-finished products need to be repaired or directly scrapped. The quality inspection accuracy is high, the production rhythm is not affected, and the consumption of surface treatment consumables caused by defective products is reduced.

[0071] In embodiment 2, step 2 also includes a preprocessing step before the background subtraction method, and the preprocessing step includes grayscale processing and inter-frame difference processing.

[0072] In embodiment 3, the frame difference processing is implemented using the following formula:

[0073] D(x, y, t) = |I g (x, y, t)-I g (x, y, t-1)|;

[0074] Where D(x, y, t) refers to the change in pixel intensity at the image position (x, y) at time t, I g (x, y, t) represents the grayscale value of the image position (x, y) at time t, I g (x, y, t-1) represents the grayscale value of the image position (x, y) at time t-1.

[0075] Usually, since the surface processing has color, it is a color image and usually needs to be converted into a grayscale image;

[0076] In order to detect motion, the use of inter-frame difference processing methods can improve the success rate of subsequent feature recognition.

[0077] Embodiment 4, the specific method of the background subtraction method in step 2 is: create a background model, and implement background subtraction by taking the average value or median value of the pixels in the video sequence. Let B(x, y) be the pixel value of the background model at position (x, y), and then the background model is updated by the following formula:

[0078] B(x, y)=α·B(x, y)+(1-α)·I(x, y);

[0079] Where I(x, y) is the pixel value at position (x, y) in the current frame, and α is a weight coefficient between 0 and 1, which is used to control the speed of background update;

[0080] Compare the current frame with the background model to detect the target area of ​​the injection molded part. Let F(x, y) be the value of the foreground mask at position (x, y). Then the injection molded part target area detection is achieved by the following formula:

[0081]

[0082] Where T is the threshold value. If the pixel value difference is greater than T, it is classified as the target area of ​​the injection molded part.

[0083] Create a background model, usually by taking the average or median of the pixels in the video sequence. Injection molding target area detection: Compare the current frame with the background model to detect the injection molding target area detection.

[0084] Where T is a threshold used to determine whether the pixel value difference is large enough to classify it as the target area of ​​the injection molded part.

[0085] In Example 5, in step 3, the optical flow method is used to track the target area of ​​the injection molded part. The optical flow method is implemented using the following formula:

[0086] I(x,y,t)=I(x+Δx,y+Δy,t+Δt);

[0087] Where I(x, y, t) represents the pixel brightness at the image position (x, y) at time t, x represents the horizontal coordinate of the image, y represents the vertical coordinate of the image, Δx and Δy are the components of the optical flow vector in the x and yy directions respectively, and Δt represents the displacement of the optical flow vector in time.

[0088] Using optical flow method or other tracking algorithms to track the target area of ​​the injection molded parts can achieve complete extraction of the target area of ​​the injection molded parts and provide the target area of ​​the injection molded parts for subsequent visual quality inspection.

[0089] In Example 6, in step 4, the multi-threaded fast feature extraction algorithm includes the following steps:

[0090] Step 41, extracting the contour of the target area of ​​the injection molded part successfully tracked in step 3, and dividing it into blocks based on the extracted closed complete contour to obtain a plurality of feature items to be determined;

[0091] Step 42: Match multiple feature items to be determined based on a preset Haar-like feature database.

[0092] In Example 7, the Haar-like feature of the feature item to be determined is calculated using the following formula:

[0093]

[0094] Where II(x, y) is the value of the integral image at point (x, y), and I(i, j) is the pixel value of the original image at point (i, j);

[0095] Using the integral image, the pixels that enclose the complete contour are quickly calculated using the following formula:

[0096] sum=II(x2, y2)-II(x1, y1)-II(x2, y1)+II(x1, y2);

[0097] Among them, (x1, y1) and (x2, y2) are the endpoint coordinates of the longest line that closes the complete contour;

[0098] If the difference between the calculation result of the pixels of the closed complete contour and the calculation result of the pixels in the feature database is within the set difference threshold, the match is judged to be successful.

[0099] In Example 8, the difference threshold is 5%.

[0100] Haar-like features are extracted by calculating the difference in pixel intensity between adjacent areas in an image. This allows for rapid feature extraction and matching. It uses multi-threading to simultaneously process multiple feature items to be determined, greatly improving feature extraction efficiency. It can match the production rhythm of surface treatment and avoid reducing the production rhythm due to quality inspection.

[0101] Example 9, an injection molded part surface treatment quality inspection device based on visual tracking, includes multiple visual tracking shooting devices and a visual analysis module 1, the multiple visual tracking shooting devices are respectively installed at the process nodes of the plastic part surface treatment production line, the visual tracking shooting device includes a bracket, an electric linear slide 2 and a visual camera 3, the track of the electric linear slide 2 is installed on the rear side of the process node through the bracket, the visual camera 3 is installed on the slider of the electric linear slide 2, and follows the lateral movement of the slider to achieve follow-up shooting, the visual analysis module 1 is respectively communicated with the visual cameras 3 of the multiple visual tracking shooting devices, and the picture data is shot interactively, the visual analysis module 1 runs the quality inspection program designed based on the visual tracking of the injection molded part surface treatment quality inspection process, and outputs the quality inspection results.

[0102] An electric linear slide 2 is used to dynamically track and shoot the original image. The electric linear slide 2 runs smoothly and has a controllable speed. A smaller differential can be achieved for tracking and shooting. The shot image can also meet the shooting requirements of multiple angles and under different light conditions brought about by the differential, greatly improving the success rate of visual quality inspection.

[0103] In embodiment 10, the visual analysis module 1 includes a memory 11, a visual analysis computer 12 and a display 13. The memory 11 is respectively connected to the visual cameras 3 of multiple visual tracking and shooting devices for communication. The visual analysis computer 12 runs a quality inspection program designed based on visual tracking for the quality inspection process of surface treatment of injection molded parts, displays the quality inspection results through the display 13, and controls the execution actions of the rejection stations set after the process nodes of multiple production lines according to the quality inspection results.

[0104] The implementation principle of the present invention is explained below through a specific visual quality inspection case:

[0105] A car dashboard, the product is made of PC+ABS injection molding and surface spraying technology. The product surface must not have burrs, shrinkage, weld lines, ejector marks, flashing, deformation, pitting, cracking, oil accumulation and other appearance defects;

[0106] The injection molded parts surface treatment quality inspection device is used to collect the image set of the injection molded parts in the surface treatment process in real time;

[0107] The visual analysis module 1 of the injection molded part surface treatment quality inspection device uses a background subtraction method to detect the injection molded part targets in the image set based on the time sequence;

[0108] Use tracking algorithms to track target areas of injection molded parts at process nodes of the injection molded parts surface treatment production line;

[0109] For the target area of ​​the injection molded part detected by tracking, shape features and texture features are extracted based on a multi-threaded fast feature extraction algorithm;

[0110] Use machine learning models to determine whether the target area of ​​the injection molded parts at different injection molded parts surface treatment production line process nodes is qualified;

[0111] During the quality inspection of a batch of car dashboards, the machine learning model found that the visual inspection results after the primer spraying station were unqualified due to uneven primer. The corresponding rejection station set after the primer spraying station node rejected the batch of car dashboards. The rejection of unqualified semi-finished products can avoid material waste caused by the subsequent process. The staff analyzed the unqualified semi-finished products and determined that the primer spraying equipment had a fault. After maintenance, the production line was reopened and the subsequent visual quality inspections were all qualified. The consumption of subsequent surface treatment consumables such as topcoat caused by defective products is reduced.

[0112] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. The surface treatment quality inspection process of injection molded parts based on visual tracking is characterized by: The following steps are involved: Step 1, using an injection molded part surface treatment quality inspection device to collect an image set of the injection molded part during the surface treatment process in real time; Step 2, the visual analysis module (1) of the injection molded part surface treatment quality inspection device uses a background subtraction method to detect the injection molded part targets in the image set based on the time sequence; Step 3, using a tracking algorithm to track the target area of ​​the injection molded part at the process node of the injection molded part surface treatment production line; Step 4, for the target area of ​​the injection molded part detected by tracking, extracting shape features and texture features based on a multi-threaded fast feature extraction algorithm; Step 5, using a machine learning model to determine whether the target area of ​​the injection molded parts at different injection molded parts surface treatment production line process nodes is qualified; Step 6: If the judgment results of the target area of ​​the injection molded parts at each production line process node are qualified, the surface treatment of the injection molded parts is judged to be qualified. If the judgment results of the target area of ​​the injection molded parts are unqualified, the injection molded parts are rejected at the rejection station set after the current production line process node.

2. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 1 is characterized in that: In step 2, a preprocessing step is also included before the background subtraction method, and the preprocessing step includes grayscale processing and inter-frame difference processing.

3. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 2 is characterized in that: The frame difference processing is implemented using the following formula: D ( x,y,t)=|I g (x,y,t)-I g (x,y,t-1)|; Where D(x, y, t) refers to the image position at time t ( x, y) pixel intensity change, I g (x, y, t) represents the grayscale value of the image position (x, y) at time t, I g (x, y, t-1) represents the image position at time t-1 ( x, y) grayscale value.

4. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 3 is characterized in that: The specific method of background subtraction in step 2 is: create a background model, and implement background subtraction by taking the average or median value of the pixels in the video sequence. Let B(x, y) be the pixel value of the background model at position (x, y), then the background model is updated by the following formula: B(x, y)=α·B(x, y)+(1-α)·I(x, y); Where I(x, y) is the pixel value at position (x, y) in the current frame, and α is a weight coefficient between 0 and 1, which is used to control the speed of background update; Compare the current frame with the background model to detect the target area of ​​the injection molded part. Let F(x, y) be the value of the foreground mask at position (x, y). Then the injection molded part target area detection is achieved by the following formula: Where T is the threshold value. If the pixel value difference is greater than T, it is classified as the target area of ​​the injection molded part.

5. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 4 is characterized in that: In step 3, the optical flow method is used to track the target area of ​​the injection molded part. The optical flow method is implemented using the following formula: I(x,y,t)=I(x+Δx,y+Δy,t+Δt); Where I(x, y, t) represents the pixel brightness at the image position (x, y) at time t, x represents the horizontal coordinate of the image, y represents the vertical coordinate of the image, Δx and Δy are the components of the optical flow vector in the x and yy directions respectively, and Δt represents the displacement of the optical flow vector in time.

6. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 5 is characterized in that: In step 4, the multi-threaded fast feature extraction algorithm includes the following steps: Step 41, extracting the contour of the target area of ​​the injection molded part successfully tracked in step 3, and dividing it into blocks based on the extracted closed complete contour to obtain a plurality of feature items to be determined; Step 42: Match multiple feature items to be determined based on a preset Haar-like feature database.

7. The process for quality inspection of surface treatment of injection molded parts based on visual tracking according to claim 6, characterized in that: The following formula is used to calculate the Haar-like features of the feature items to be determined: Where II(x, y) is the value of the integral image at point (x, y), and I(i, j) is the pixel value of the original image at point (i, j); Using the integral image, the pixels that enclose the complete contour are quickly calculated using the following formula: sum=II(x2, y2)-II(x1, y1)-II(x2, y1)+II(x1, y2); Among them, (x1, y1) and (x2, y2) are the endpoint coordinates of the longest line that closes the complete contour; If the difference between the calculation result of the pixels of the closed complete contour and the calculation result of the pixels in the feature database is within the set difference threshold, the match is judged to be successful.

8. The injection molded parts surface treatment quality inspection process based on visual tracking according to claim 7 is characterized in that: The difference threshold is 5%.

9. The surface treatment quality inspection device for injection molded parts based on visual tracking is characterized by: The invention comprises a plurality of visual tracking and shooting devices and a visual analysis module (1), wherein the plurality of visual tracking and shooting devices are respectively installed at process nodes of a plastic part surface treatment production line, the visual tracking and shooting devices comprise a bracket, an electric linear slide (2) and a visual camera (3), the track of the electric linear slide (2) is installed on the rear side of the process node through the bracket, the visual camera (3) is installed on the slider of the electric linear slide (2), and follows the lateral movement of the slider to realize follow-up shooting, the visual analysis module (1) is respectively communicated with the visual cameras (3) of the plurality of visual tracking and shooting devices, and interactively shoots picture data, the visual analysis module (1) runs a quality inspection program designed for the quality inspection process of injection molded part surface treatment based on visual tracking as described in claim 8, and outputs quality inspection results.

10. The device for quality inspection of surface treatment of injection molded parts based on visual tracking according to claim 9, characterized in that: The visual analysis module (1) comprises a memory (11), a visual analysis computer (12) and a display (13), wherein the memory (11) is respectively connected to the visual cameras (3) of the plurality of visual tracking shooting devices for communication, and the visual analysis computer (12) runs a quality inspection program for the injection molded parts surface treatment quality inspection process design based on visual tracking as described in claim 8, displays the quality inspection results through the display (13), and controls the execution actions of the rejection stations arranged after the process nodes of the plurality of production lines according to the quality inspection results.

Images