Injection molding defect detection device of TV shell and detection and control method of injection molding defect detection device

Abstract

The invention relates to the technical field of injection molding defect detection, in particular to an injection molding defect detection device of a TV shell and a detection and control method of the injection molding defect detection device. Acquiring defect detection image information about a lost model detection area on the TV casing sample through a defect detection control scheme, and performing pixel variation calculation of local image structure tensor on the defect detection image information to obtain a first defect detection result; and based on the first defect monitoring result, extracting a pixel particle variation region of the detection loss model region, and mapping a source pixel saturated color gamut space color of the pixel particle variation region to a target pixel saturated color gamut space of a qualified injection molding product so as to analyze the pixel contrast of the injection molding defect type, thereby obtaining a second defect detection result. According to the invention, multi-directional injection molding finished product detection can be carried out on the TV casing through the injection molding defect detection device, so as to identify whether the injection molding finished product has injection molding defects, provide an optimization basis for injection molding production of the TV casing, and improve the injection molding detection accuracy of the TV casing.

Description

Technical Field

[0001] The present invention relates to the technical field of injection molding defect detection, and particularly to an injection molding defect detection device for a TV case and its detection and control method. Background Art

[0002] In the current TV industry, with the continuous progress of technology and the increasing requirements of consumers for product quality, the injection molding quality of TV cases has received more and more attention. As one of the most critical processes in the production of TV cases, injection molding often faces various local subtle defect problems, such as bubbles, cracks, shrinkage, and surface unevenness. These subtle defects not only affect the appearance of the product but may also, to a certain extent, affect the structural strength and durability of the TV, thus affecting the quality of the entire TV.

[0003] The existing defect detection equipment has a low detection accuracy in the visual system it sets, making it difficult to achieve efficient and accurate detection of subtle defects in TV cases. Moreover, the detection of existing injection molding defect detection equipment usually relies on manual-assisted inspection or simple machine vision technology to identify injection molding defects. However, manual detection has problems such as strong subjectivity, low efficiency, and easy omission of inspection, resulting in inaccurate detection results. Especially in large-scale production, defective products cannot be discovered and excluded in a timely manner, increasing the production cost and rework rate of TV cases. Some machine vision detection technologies have been applied to the automatic identification of injection molding defects, but the existing detection systems still face problems such as low accuracy, poor adaptability, and insufficient anti-interference ability. Especially in the detection of complex appearance features and small defects on TV cases, it is often difficult to achieve accurate judgment, reducing the reliability of injection molding defect detection for TV cases. Therefore, it is necessary to develop an injection molding defect detection device and control method that is more intelligent, precise, and real-time in dealing with the phenomena of missed inspection of TV cases, unique appearance features, and subtle injection molding defects, so as to improve the accuracy of detection, reduce manual intervention, and optimize production efficiency. Summary of the Invention

[0004] The present invention overcomes the deficiencies of the prior art and provides an injection molding defect detection device for a TV case and its detection and control method.

[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] The first aspect of the present invention provides a detection and control method for an injection molding defect detection device for a TV case, including the following steps:

[0007] S102: Construct a standard injection molding grid model of the TV case. Based on the smooth update of the standard injection molding grid model, generate a spatially element smooth grid model. By variational fitting the local process defect detection model extracted from the image data on the spatially element smooth grid model to the spatially element smooth grid model and performing a fitting degree analysis, obtain the detected missing model area;

[0008] S104: Construct an element scalar embedding field for the specified detection area of the TV case. Determine whether the grid scalar value is greater than the isosurface threshold to segment and embed the detected missing model area into the element scalar embedding field, obtain a spatial element coordinate matrix, and based on the spatial element coordinate matrix, control the injection molding defect detection device to perform detection and obtain a defect detection control scheme;

[0009] S106: Obtain defect detection image information about the detected missing model area on the TV case sample through the defect detection control scheme. Perform pixel variation calculation and analysis of the local image structure tensor on the defect monitoring image information to obtain a first defect detection result;

[0010] S108: Based on the first defect monitoring result, extract the pixel particle variation area of the detected missing model area. Map the source pixel saturated color gamut space color of the pixel particle variation area to the target pixel saturated color gamut space of the qualified injection molded product to analyze the pixel contrast of the injection molding defect types and obtain a second defect detection result.

[0011] More specifically, the step S102 specifically includes the following steps:

[0012] Obtain the standardized design drawing of the TV case. According to the standardized design drawing, construct a standard injection molding grid model of the TV case. Based on the analysis of the injection molding rework rate, perform smooth update on the interwoven grid vertices of the standard injection molding grid model to obtain a spatially element smooth grid model of the TV case;

[0013] Obtain the preset detection process of the injection molding defect detection device for the standardized injection molding of the TV case. Extract the process shooting decision points of the vision camera through the preset detection process, and based on the process shooting decision points, divide the spatially element smooth model into several sub-smooth models;

[0014] Extract the multi-frame defect detection image data output by the vision camera of the injection molding defect detection device for inspecting the TV case in each sub-smooth model after the vision camera has completely executed the preset detection process through the detection log. Introduce the Sobel operator to perform feature extraction and modeling on each frame of defect detection image data to obtain a local process defect detection model;

[0015] Obtain the spatial element distribution on each sub-smooth model corresponding to the local process defect detection model, and construct a variational lower bound of the model log marginal likelihood based on the spatial element distribution;

[0016] Preset the benchmark model variational distribution of each sub - smoothing model according to the spatial element smoothing model, maximize the variational lower bound, generate the posterior model variational distribution of the local defect detection model, calculate the hash misalignment function between the posterior model variational distribution and the benchmark model variational distribution, and determine the fitting degree of each local defect detection model relative to each sub - smoothing model according to the hash misalignment function;

[0017] If the fitting degree is greater than the preset fitting degree, then fit the local defect detection model corresponding to this fitting degree to the current sub - smoothing model, and finally strip out the un - fitted model area in the spatial element smoothing model, which is marked as the detected missing model area.

[0018] More specifically, the steps of obtaining the standardized design drawing of the TV case, constructing the standard injection - molding grid model of the TV case according to the standardized design drawing, and performing smooth update on the interwoven grid vertices of the standard injection - molding grid model based on the analysis of the injection - molding rework rate to obtain the spatial element smooth grid model of the TV case are as follows:

[0019] Obtain the standardized design drawing of the TV case, and obtain multiple groups of standardized three - dimensional design parameters of the TV case according to the standardized design drawing;

[0020] Input multiple groups of the standardized three - dimensional design parameters into the SolidWorks model design software for modeling operations, generate the standard injection - molding grid model of the TV case, and extract the interwoven grid vertices of the standard injection - molding grid model and the smoothing parameters between adjacent interwoven grid vertices;

[0021] Introduce the Laplace coordinate operator, construct the neighborhood coordinate platform of the standard injection - molding grid model by assigning values in the Laplace coordinate operator based on the smoothing parameters between adjacent interwoven grid vertices, calculate the neighborhood information of the interwoven grid vertices through the neighborhood coordinate platform, and obtain the Laplace neighborhood coordinates of each interwoven grid vertex;

[0022] Perform forward smoothing and reverse smoothing processing with weighted average on each interwoven grid vertex according to the Laplace neighborhood coordinates, so as to update and iterate the current neighborhood position of each interwoven grid vertex in the standard injection - molding grid model, make each interwoven grid vertex closer to the average position of its affiliated neighborhood vertices, and obtain the positive smoothing weight value and the negative smoothing weight value;

[0023] Obtain the detection log of the injection - molding defect detection device and the injection - molding production requirements of the TV case, extract the historical injection - molding rework rate of the TV case at a preset time sequence through the detection log, and preset the allowable injection - molding rework rate according to the injection - molding production requirements;

[0024] Calculate the injection molding rework rate difference between the historical injection molding rework rate and the allowable injection molding rework rate, preset the smoothing update iteration frequency based on the injection molding rework rate difference, and repeat the above update iteration steps of forward smoothing and reverse smoothing of the weighted average interleaved grid vertices until the smoothing update iteration frequency is reached, and generate positive and negative smoothing weight matrices;

[0025] Reconstruct the standard injection molding grid model according to the positive and negative smoothing weight matrices to obtain the spatial element smoothing grid model of the TV case.

[0026] More specifically, the step S104 specifically includes the following steps:

[0027] Obtain the specified detection area of the TV case photographed and detected by the vision camera in the injection molding defect detection device, and obtain the spatial specification parameters of the specified detection area;

[0028] Based on the spatial specification parameters, construct an element scalar embedding field for the specified detection area, only plan the field area corresponding to the detection missing model area in the element scalar embedding field, mark it as a sub-detection missing scalar field, obtain the fixed scalar value of the sub-detection missing scalar field, and preset the isosurface embedding threshold according to the fixed scalar value;

[0029] Divide the detection missing model area into N affiliated model bodies, obtain the vertex grid scalar values of each model vertex on each affiliated model body, and check whether the vertex grid scalar value of each model vertex is greater than the isosurface threshold;

[0030] If the vertex grid scalar value is greater than the isosurface embedding threshold, input a binary value of 1 for this model vertex; if the vertex grid scalar value is greater than the isosurface embedding threshold, input a binary value of 0 for this model vertex, and generate a binary vertex status code;

[0031] According to the programming translation of the binary vertex status code, determine the embedding boundaries of each affiliated model body embedded in the sub-detection missing scalar field, obtain multiple scalar embedding boundaries, and perform linear interpolation calculation on each scalar embedding boundary based on the grid scalar value to obtain the corresponding element scalar embedding isosurface of each affiliated model body;

[0032] Connect the corresponding element scalar embedding isosurfaces of all affiliated model bodies, and finally generate the spatial element coordinate matrix of the detection missing model area located in the specified detection area, and control the flipping mechanism to perform fixed-point shooting defect detection on the detection missing model area of the TV case in the specified detection area according to the spatial element coordinate matrix to obtain the defect detection control scheme.

[0033] More specifically, the step S106 specifically includes the following steps:

[0034] Obtain a TV set shell sample that has completed injection molding production, and perform the defect detection control scheme on the TV set shell sample through an injection molding defect detection device to obtain defect detection image information about the detected missing model area on the TV set shell sample;

[0035] Use the Sobel operator to perform pixel-level edge detection on the defect detection image information to obtain the actual horizontal vector gradient and the actual vertical vector gradient of the detected missing model area. Calculate the second moment of the defect detection image information based on the actual horizontal vector gradient and the actual vertical vector gradient, and generate a local image structure tensor for the detected missing model area;

[0036] Obtain the actual image resolution of the defect detection image information. Based on the actual image resolution, perform eigenvalue decomposition of the local pixel structure on the local image structure tensor to obtain a series of structure eigenvalues and structure eigenvectors of the local graphic structure tensor, and determine the actual pixel structure flatness of the detected missing model area according to the series of structure eigenvalues and structure eigenvectors;

[0037] Obtain the standard image information of the TV set shell captured by the injection molding defect detection device when executing the defect detection control scheme according to the standard injection molding grid model of the TV set shell, calculate and obtain the pixel structure flatness of the standard graphic information, which is defined as the reference pixel structure flatness;

[0038] If the actual pixel structure flatness is higher than the reference pixel structure flatness, then calibrate the TV set shell in this detected missing model area as a qualified injection molding product; if the actual pixel structure flatness is lower than the reference pixel structure flatness, then calibrate the TV set shell in this detected missing model area as a defective injection molding product, and obtain the first defect detection result.

[0039] More specifically, the step S108 specifically includes the following steps:

[0040] If the first defect detection result shows that the TV set shell is a defective injection molding product, then extract the pixel particle variation area of the detected missing model area from the defect detection image information of the defective injection molding product based on the actual pixel structure flatness;

[0041] Obtain the color saturation interval of the pixel particle variation area, which is defined as the first color saturation interval. Construct a pixel saturation color gamut space for the pixel particle variation area in the defective injection molding product according to the first color saturation interval, which is defined as the source pixel saturation color gamut space;

[0042] Obtain the color saturation interval of the pixel particle variation area expressed in the qualified injection molding product, which is defined as the second color saturation interval. Construct a pixel saturation color gamut space for the pixel particle variation area in the defective injection molding product according to the second color saturation interval, which is defined as the target pixel saturation color gamut space;

[0043] Obtain a color mapping function table for the color saturation expression of different types of injection molding defects and the established Gamma correction interval for each type of injection molding defect through a big data network, establish a color mapping rule according to the color mapping function table,

[0044] Introduce a linear transformation algorithm, and map each saturated color in the source pixel saturation color gamut space to the target pixel saturation color gamut space based on the color mapping rule in the linear transformation algorithm to generate a linear color saturation mapping equation;

[0045] Solve the linear color saturation mapping equation to obtain a series of mapped chromaticity values, calculate and determine the restored pixel contrast of the pixel particle mutation area according to the series of mapped chromaticity values, and obtain the pixel contrast of the pixel particle mutation area expressed in the qualified injection molded product, which is defined as the initial pixel contrast;

[0046] Introduce a Gamma correction formula to calculate the correction of the restored pixel contrast and the initial pixel contrast to obtain the actual Gamma correction value. If the actual Gamma correction value is within the established Gamma correction interval, directly output the type of injection molding defect corresponding to the established Gamma correction interval to obtain the second defect detection result.

[0047] The second aspect of the present invention provides an injection molding defect detection device for a TV case, and the injection molding defect detection device includes:

[0048] An image acquisition module, which is used to capture defect image data of the injection molded product of the TV case from multiple directions;

[0049] An image processing module, which is responsible for introducing a Sobel operator to calculate and extract the feature information of the defect detection image data of different frames;

[0050] A flipping module, which is used to clamp the TV case within the specified detection area and flip it in multiple directions so that the image acquisition module can comprehensively photograph the detection missing model area of the TV case;

[0051] A model construction module, which is used to perform modeling calculation processing on the defect detection image data extracted by the image processing module to output a local process defect detection model;

[0052] A binary coding module, and the data processing module is used to check whether the vertex grid scalar value of each model vertex on the attached model body is greater than the isosurface embedding threshold of the sub-detection missing scalar field, and program a binary status code of 0 or 1 for the model vertex of the attached model body.

[0053] The present invention solves the technical defects existing in the background art, and the beneficial technical effects of the present invention are as follows:

[0054] Construct a standard injection molding grid model of the TV case. Based on the model smoothing update of the standard injection molding grid model, generate a spatial element smoothed grid model. By variationally fitting the local process defect detection model extracted from the image data on the spatial element smoothed grid model to the spatial element smoothed grid model and performing a fitting degree analysis, obtain the detected missing model area. Construct an element scalar embedding field for the specified detection area of the TV case. Determine whether the grid scalar value is greater than the isosurface threshold to segment and embed the detected missing model area into the element scalar embedding field, obtain a spatial element coordinate matrix, and control the injection molding defect detection device based on the spatial element coordinate matrix for detection to obtain a defect detection control scheme. Obtain defect detection image information about the detected missing model area on the TV case sample through the defect detection control scheme, perform pixel variation calculation and analysis of the local image structure tensor on the defect monitoring image information to obtain a first defect detection result. Extract the pixel particle variation area of the detected missing model area based on the first defect monitoring result, and map the source pixel saturated color gamut space color of the pixel particle variation area to the target pixel saturated color gamut space of the qualified injection molded product to analyze the pixel contrast of the injection molding defect types, thereby obtaining a second defect detection result. The present invention can perform multi-faceted injection molding finished product detection on the TV case through the injection molding defect detection device to identify whether there are injection molding defects in the injection molding finished product, provide an optimization basis for the injection molding production of the TV case, and effectively improve the injection molding quality and efficiency of the TV case. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0056] Figure 1 Shows the overall structural schematic diagram of the present device;

[0057] Figure 2 Shows the internal structural schematic diagram of the present device;

[0058] Figure 3 Shows the bottom view structural schematic diagram of the present device. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] In order to more clearly understand the above objects, features and advantages of the present invention, the present invention will be further described in detail below in conjunction with the drawings and specific embodiments. It should be noted that, without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other.

[0060] In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited by the specific embodiments disclosed below.

[0061] The first aspect of the present invention provides a detection and control method for an injection molding defect detection device of a TV case, as Figure 1 shown, including the following steps:

[0062] S102: Construct a standard injection molding grid model of the TV case, generate a spatial element smooth grid model based on the model smooth update of the standard injection molding grid model, and obtain a detected missing model area by variationally fitting a local process defect detection model extracted from the image data on the spatial element smooth grid model and performing a fitting degree analysis;

[0063] S104: Construct an element scalar embedding field for a specified detection area of the TV case, determine whether the grid scalar value is greater than the isosurface threshold, segment and embed the detected missing model area into the element scalar embedding field to obtain a spatial element coordinate matrix, and control the injection molding defect detection device to perform detection based on the spatial element coordinate matrix to obtain a defect detection control scheme;

[0064] S106: Obtain defect detection image information about the detected missing model area on the TV case sample through the defect detection control scheme, perform pixel variation calculation and analysis on the local image structure tensor of the defect monitoring image information to obtain a first defect detection result;

[0065] S108: Extract the pixel particle variation area of the detected missing model area based on the first defect monitoring result, map the source pixel saturated color gamut space color of the pixel particle variation area to the target pixel saturated color gamut space of a qualified injection molded product, and analyze the pixel contrast of the injection molding defect types to obtain a second defect detection result.

[0066] More specifically, the step S102 specifically includes the following steps:

[0067] Obtain the standardized design drawing of the TV case, construct a standard injection molding grid model of the TV case according to the standardized design drawing, and perform smooth update on the interlaced grid vertices of the standard injection molding grid model based on the injection molding rework rate analysis to obtain the spatial element smooth grid model of the TV case;

[0068] Obtain the preset detection process of the injection molding defect detection device for the standardized injection molding of TV casings. Extract the process shooting decision points of the vision camera through the preset detection process, and divide the spatial element smoothing model into several sub-smoothing models based on the process shooting decision points;

[0069] Extract the multi-frame defect detection image data output by the vision camera of the injection molding defect detection device for inspecting the TV casing in each sub-smoothing model after the vision camera has completely executed the preset detection process through the detection log. Introduce the Sobel operator to extract features and model each frame of defect detection image data to obtain a local process defect detection model;

[0070] Obtain the spatial element distribution of each sub-smoothing model on the corresponding local process defect detection model, and construct a variational lower bound of the model log marginal likelihood based on the spatial element distribution;

[0071] According to the spatial element smoothing model, preset the variational distribution of the reference model of each sub-smoothing model, maximize the variational lower bound, generate the posterior model variational distribution of the local defect detection model, calculate the hash misalignment function between the posterior model variational distribution and the reference model variational distribution, and determine the degree of fit of each local defect detection model relative to each sub-smoothing model according to the hash misalignment function;

[0072] If the degree of fit is greater than the preset degree of fit, then fit the local defect detection model corresponding to the degree of fit to the current sub-smoothing model, and finally peel off the unfit model area in the spatial element smoothing model and mark it as the detected missing model area.

[0073] It should be noted that the appearance shapes of TV casings are diverse. Some TV casings have abnormal surfaces, internal grooves, fine fixing holes, and patterns in multiple parts. Due to the relatively unique positions of these shape features on the TV casing, it is difficult for the vision camera module of some existing injection molding defect detection devices to globally capture the defect images of these shape features according to the preset control decision, resulting in the omission of the injection molding defects of these shape features after injection molding and the phenomenon of missed detection in local areas, which greatly affects the use of the TV casing. Therefore, it is particularly important to exclude the areas of missed detection. For this reason, this method constructs a spatial element smooth grid model of the TV casing, aiming to more smoothly display the spatial distribution of elements such as points, lines, and surfaces of the entity after the TV casing is injection molded, so as to maximize the restoration of the detection missed state of the above-mentioned unique appearance features by the injection molding defect detection device in terms of spatial position. Then, the local process defect detection model variation in the TV casing defect image captured by the injection molding defect detection device after executing the preset detection process is variationally fitted to the spatial element smooth grid model, and the un-fitted model area is the missed area that is difficult to be globally detected by the current injection molding defect detection device for some unique appearance features on the TV casing after reasonable control operation, that is, the detection missed model area. This may be caused by the low control accuracy or improper control of the injection molding defect detection device. Therefore, the control analysis and optimization of the subsequent injection molding defect detection device can be carried out based on the extracted detection missed model area, so as to eliminate the phenomenon of missed detection of unique appearance features by the injection molding defect detection device and improve the richness and accuracy of multi-dimensional detection of injection molding defects.

[0074] It should be noted that for the regional fitting of the local process defect detection model on the spatial element smoothing grid model, this method constructs a variational lower bound of the model log marginal likelihood using the spatial element distribution on the local process defect detection model, providing a matching degree inference reference for the precise alignment and fitting of the two, making the fitting of the model in space more accurate. Maximizing this variational lower bound means maximizing the matching degree between the point, line, and plane element distributions on the local process defect detection model and the spatial element smoothing grid model. The larger the variational lower bound, the closer the point, line, and plane element distributions on the local process defect detection model are to the element distributions in a certain area of the spatial element smoothing grid model. Therefore, after maximizing the variational lower bound, calculating the hash misalignment function between the posterior model variational distribution and the benchmark model variational distribution of the local defect detection model can determine the fitting degree between the two. If the fitting degree is greater than the preset fitting degree, it means that the local defect detection model belongs to the specified spatial area on the spatial element smoothing grid model, so the local defect detection model should be fitted to the current sub-smoothing model. Finally, the unfitted model area is the area missed by the device during the inspection of the TV case. Through this method, it is possible to trace the missed inspection area of the TV case based on the existing control inspection results of the injection molding defect detection device, providing a reliable analysis basis for the subsequent image vision detection control of the injection molding defect detection device for the TV case, avoiding the situation of dead angle detection errors in the device, eliminating the phenomenon of global missed inspections, and improving the detection control quality of the vision camera for globally exploring the defects of the TV case.

[0075] More specifically, the steps for obtaining the standardized design drawing of the TV case, constructing the standard injection molding grid model of the TV case according to the standardized design drawing, and performing smooth update on the interwoven grid vertices of the standard injection molding grid model based on the injection molding rework rate analysis to obtain the spatial element smoothing grid model of the TV case are as follows:

[0076] Obtain the standardized design drawing of the TV case, and obtain multiple groups of standardized three-dimensional design parameters of the TV case according to the standardized design drawing;

[0077] Input multiple groups of the standardized three-dimensional design parameters into the SolidWorks model design software for modeling operations to generate the standard injection molding grid model of the TV case, and extract the interwoven grid vertices of the standard injection molding grid model and the smoothing parameters between adjacent interwoven grid vertices;

[0078] Introduce the Laplace coordinate operator, construct the neighborhood coordinate platform of the standard injection molding grid model by assigning values based on the smoothing parameters between adjacent interwoven grid vertices in the Laplace coordinate operator, calculate the neighborhood information of the interwoven grid vertices through the neighborhood coordinate platform, and obtain the Laplace neighborhood coordinates of each interwoven grid vertex;

[0079] Perform forward and backward smoothing processing of weighted averaging on each interwoven grid vertex according to the Laplace neighborhood coordinates, thereby updating and iterating the current neighborhood position of each interwoven grid vertex in the standard injection molding grid model, making each interwoven grid vertex closer to the average position of its affiliated neighborhood vertices, and obtaining positive smoothing weight values and negative smoothing weight values;

[0080] Obtain the detection log of the injection molding defect detection device and the injection molding production requirements of the TV case. Extract the historical injection molding rework rate of the TV case at a preset time sequence from the detection log, and preset the allowable injection molding rework rate according to the injection molding production requirements;

[0081] Calculate the injection molding rework rate difference between the historical injection molding rework rate and the allowable injection molding rework rate, preset the smoothing update iteration frequency based on the injection molding rework rate difference, and repeat the above update iteration steps of forward and backward smoothing of weighted averaging of interwoven grid vertices until the smoothing update iteration frequency is reached, and generate positive and negative smoothing weight matrices;

[0082] Reconstruct the standard injection molding grid model according to the positive and negative smoothing weight matrices to obtain the spatial element smoothing grid model of the TV case.

[0083] It should be noted that due to the different appearance shapes of TV cases and the poor visual inspection control performance of some injection molding defect detection devices, some special-shaped surfaces or smooth areas such as inner grooves are difficult to be detected by the visual camera of the device, resulting in detection blind spots. The contour texture of the image of the TV case captured by the visual camera in space shows smooth abnormalities such as distortion and twisting, which leads to large deviations in the subsequent tracing of the lost areas detected on the TV case, affecting the optimization of the device control accuracy, and reducing the device's injection molding defect detection and identification quality for the TV case. Therefore, it is necessary to provide a model that maintains and reflects the highly smooth standard contour texture of the TV case as much as possible to further accurately trace the missed areas when the injection molding defect detection device detects the TV case, thereby ensuring the traceability accuracy and global rationality of the lost model area. To this end, this method first constructs a standard injection-molded mesh model of a TV case. Using the smoothing parameters between adjacent interwoven mesh vertices on this standard injection-molded mesh model, coordinates are assigned using the Laplace coordinate operator. This assigns Laplace neighborhood coordinates to each interwoven mesh vertex. These Laplace neighborhood coordinates reflect the neighborhood relationship between the interwoven mesh vertex and its surrounding neighboring vertices. This determines the smoothing trend of each vertex corresponding to each model detail in the standard model, revealing the general direction of smoothing optimization. Then, based on the Laplace neighborhood coordinates, a weighted average forward and reverse smoothing process is performed on each interwoven mesh vertex. This iteratively updates the current neighborhood position of each interwoven mesh vertex within the standard injection-molded mesh model. By updating each vertex position, it moves toward the average position of its neighboring vertices, achieving a smoothing effect. Specifically, forward smoothing adjusts each vertex according to a positive weight value so that its vertex position moves closer to the center of the neighboring vertices, while reverse smoothing slightly withdraws the over-smoothing effect in forward smoothing according to the negative weight value. This helps to reduce the over-smoothing phenomenon that may be caused by forward smoothing and retain the details of the original shape, thereby achieving gradual smoothing optimization in the general direction of a smooth change trend without destroying the overall structural display of the original standard model. The final output positive smoothing weight value and negative smoothing weight value are the specific values for optimizing the smoothness of the standard model, which can further make the surface of the mesh model smoother.

[0084] It should be noted that since the injection defect detection device uses a non - smooth TV case model as its control basis, there are defect detection errors in the TV case, which causes a significant increase in the rework rate of the injection production of the TV case. It can be seen that the injection rework rate depends on the accuracy of the smoothness of the TV case model. Therefore, this method limits the update iteration frequency of forward smoothing and reverse smoothing based on the difference in injection rework rate between the historical injection rework rate and the allowable injection rework rate, so as to make the smooth optimization of the model tend to the state constraint of the allowable injection rework rate to the greatest extent. This significantly improves the smoothness of the visual camera of the injection defect detection device in detecting elements such as points, lines, and planes on the TV case in space, ensures the traceability accuracy of detection dead corners such as missing model areas in detection, and reduces the increase in rework rate caused by subsequent injection defect detection errors.

[0085] More specifically, step S104 specifically includes the following steps:

[0086] Obtain the specified detection area of the TV case detected by the visual camera in the injection defect detection device, and obtain the spatial specification parameters of the specified detection area;

[0087] Based on the spatial specification parameters, construct an element scalar embedding field for the specified detection area, only plan the field area corresponding to the missing model area in the element scalar embedding field, mark it as the sub - missing scalar field, obtain the fixed scalar value of the sub - missing scalar field, and preset the isosurface embedding threshold according to the fixed scalar value;

[0088] Divide the missing model area into N affiliated model bodies, obtain the vertex grid scalar values of each model vertex on each affiliated model body, and check whether the vertex grid scalar value of each model vertex is greater than the isosurface threshold;

[0089] If the vertex grid scalar value is greater than the isosurface embedding threshold, input a binary value of 1 for this model vertex; if the vertex grid scalar value is greater than the isosurface embedding threshold, input a binary value of 0 for this model vertex, and generate a binary vertex status code;

[0090] According to the programming translation of the binary vertex status code, determine the embedding boundary of each affiliated model body embedded in the sub - missing scalar field, obtain multiple scalar embedding boundaries, and perform linear interpolation calculation on each scalar embedding boundary based on the grid scalar value to obtain the corresponding element scalar embedding isosurface of each affiliated model body;

[0091] Connect the corresponding element scalar embedding isosurfaces of all affiliated model bodies, finally generate the spatial element coordinate matrix of the missing model area located in the specified detection area, and control the flipping mechanism to perform fixed - point shooting defect detection on the missing model area of the TV case in the specified detection area according to the spatial element coordinate matrix, and obtain the defect detection control scheme.

[0092] It should be noted that for some special-shaped surfaces, inner grooves, fine fixing holes, and textures on the appearance of TV casings, high-precision and high-resolution vision control functions are required to meet the high-precision defect detection of these unique shape features. However, some injection defect detection devices still lack the vision detection control accuracy for TV casings and the rationality of flipping control, making it difficult for the vision camera to accurately explore and shoot along the point-line-plane spatial elements of these features under the full-range flipping of the flipping mechanism. As a result, the output image cannot display and identify more subtle hidden injection defects, reducing the detection accuracy of the injection defect detection device, leading to a sharp increase in the detail detection error rate. Therefore, it is particularly important to control the subtle detection of these unique shape features by the injection defect detection device, which directly affects the injection production quality of TV casings. In response to this, this method constructs an element scalar embedding field for the specified detection area of the TV casing, maps and embeds the spatial elements such as points, lines, and surfaces in the detection missing model area into this element scalar embedding field, and finally generates the coordinate matrix of these spatial elements, so that the vision camera can perform a detailed traversal shooting of the detection missing model area of the TV casing in the specified detection area according to the position distribution described by this spatial element coordinate matrix, enabling the vision camera to obtain a more detailed hidden defect image of the detection missing model area, achieving a refined defect detection control effect. Compared with traditional technologies, it improves the phenomenon that the vision camera is prone to global detection omissions or unable to detect unique appearance features during shooting, effectively improving the defect detection accuracy and reliability of TV casings.

[0093] It should be noted that the detection missing model area is divided into N subsidiary model bodies, and the vertex grid scalar values of each model vertex on each subsidiary model body are used for mapping. On the one hand, it can ensure that the detection missing model area can be individually detected, improving the subdivision of the mapping of the detection missing model area. On the other hand, it can reduce the error rate of mapping embedding and improve the fineness of detail defect detection. If the vertex grid scalar value is greater than the isosurface embedding threshold, it means that the model vertex is located inside the surface of the sub-detection missing scalar field. Therefore, each vertex is set to a binary value of 1. If the vertex grid scalar value is less than the isosurface embedding threshold, it means that the model vertex is located outside the surface of the sub-detection missing scalar field. So it is set to a binary value of 0. This describes the specific mapping landing points of the detection missing model area corresponding to the specified detection area. Thus, based on the programming translation of these binary vertex status codes, the scalar embedding boundary can be quickly determined, and linear interpolation is performed on each of the scalar embedding boundaries based on the vertex grid scalar values of the subsidiary model bodies and connected, thereby forming a spatial element coordinate matrix of the detection missing model area located in the specified detection area. Through this method, the spatial elements such as points, lines, and surfaces included in the detected missing model area can be quickly and accurately mapped and embedded into the corresponding specified detection area of the visual camera to detect the TV case, so that these spatial elements can be spatially represented by a coordinate matrix in the specified detection area, providing a more authoritative flip operation control basis for the defect detection of the TV case. Thus, the flip mechanism of the injection molding defect detection device accurately flips the TV case so that the visual camera can perform detailed detection on the hidden defects of its unique appearance features, improving the meticulousness and reliability of the defect detection of the TV case and avoiding defect detection errors and omissions.

[0094] More specifically, the step S106 specifically includes the following steps:

[0095] Obtain a TV case sample after injection molding production, and use the injection molding defect detection device to execute the defect detection control plan to detect the TV case sample to obtain defect detection image information about the detection missing model area on the TV case sample;

[0096] Use the Sobel operator to perform pixel-level edge detection on the defect detection image information to obtain the actual horizontal vector gradient and actual vertical vector gradient of the detection missing model area, and calculate the second moment of the defect detection image information based on the actual horizontal vector gradient and actual vertical vector gradient to generate the local image structure tensor of the detection missing model area;

[0097] Obtain the actual image resolution of the defect detection image information, perform eigen - decomposition of the local pixel structure on the local image structure tensor based on the actual image resolution, obtain a series of structure eigenvalues and structure eigenvectors of the local graphic structure tensor, and determine the actual pixel structure flatness of the detected missing model area according to the series of structure eigenvalues and structure eigenvectors.

[0098] According to the standard injection - molding grid model of the TV shell, the injection - molding defect detection device executes the defect detection control scheme to capture and shoot the standard image information of the TV shell, calculate and obtain the pixel structure flatness of the standard graphic information, which is defined as the reference pixel structure flatness.

[0099] If the actual pixel structure flatness is higher than the reference pixel structure flatness, calibrate the TV shell in the detected missing model area as a qualified injection - molded product; if the actual pixel structure flatness is lower than the reference pixel structure flatness, calibrate the TV shell in the detected missing model area as a defective injection - molded product, and obtain the first defect detection result.

[0100] It should be noted that during the injection - molding process of the TV shell, there will be some relatively subtle injection - molding defects that are difficult to be captured and detected by the camera. For example, an extremely small crack appears on a certain special - shaped surface, and the existing injection - molding defect detection device fails to properly control the shooting resolution of the vision camera, resulting in the failure to capture this small crack, thus ignoring this injection - molding defect. This may affect the damage and quality performance during the installation and use of the TV shell, increasing the injection - molding rework rate and cost output of the TV shell. Therefore, it is necessary to accurately detect whether there are detailed injection - molding defects on the TV shell to determine whether the TV shell product is qualified. For this reason, this method first obtains the defect image data of the detected missing model area of the TV shell through the injection - molding defect detection device executing the defect detection control scheme. Since pixel features such as edges, textures, and contours in the image are the gradients of local brightness changes, the Sobel operator is used to extract the actual horizontal vector gradient and the actual vertical vector gradient of the detected missing model area at the pixel level of the defect detection image information; the image gradient is the basis of the structure tensor, and the gradient represents the degree of local area change. Its gradient value is usually large at edges, corners, etc., and the second - order moment reflects the change and mutual relationship of the gradient. Constructing the second - order moment of the gradient can effectively capture the change pattern of the local area of the image. Therefore, further calculate the second - order moment of the defect detection image information according to the gradient to generate the local image structure tensor of the detected missing model area. The structure tensor can further describe the local geometric features of the defect detection image information.

[0101] It should be noted that the capture of the defect detection image is based on the resolution of the vision camera. Therefore, the decomposition of the local features of the image should be carried out based on the resolution as the benchmark. Among them, the structure eigenvalue reveals the main direction and variation amplitude of the local image area. The size and difference of the eigenvalues can help distinguish different types of regional changes, that is, whether the local image area is flat; the structure eigenvector provides the direction information of the local structure. Therefore, based on this series of structure eigenvalues and structure eigenvectors, the actual pixel structure flatness of the detected missing model area can be accurately determined, and the actual pixel structure flatness reflects the pixel variation degree of the local image of the detected missing model area. If the actual pixel structure flatness is higher than the reference pixel structure flatness, it means that the pixel structure is relatively stable and no variation has occurred, that is, there are no subtle injection defects on the TV case to change the pixel structure of the image, so the TV case is a qualified injection product. If the actual pixel structure flatness is lower than the reference pixel structure flatness, it means that there is a pixel variation phenomenon in the defect detection image of the TV case, that is, there are subtle injection defects on the TV case, so the TV case is an unqualified injection product. Through this method, it is possible to calculate the pixel variation of the structure tensor based on the captured defect detection image, so as to quickly identify and detect whether there are relatively subtle injection defects on the TV case, and then achieve the accurate injection defect detection effect of the TV case product, and improve the credibility and accuracy of the TV case defect detection.

[0102] More specifically, the step S108 specifically includes the following steps:

[0103] If the first defect detection result shows that the TV case is a defective injection product, then based on the actual pixel structure flatness, extract the pixel particle variation area of the detected missing model area from the defect detection image information of the defective injection product;

[0104] Obtain the color saturation interval of the pixel particle variation area, define it as the first color saturation interval, and construct the pixel saturation color gamut space of the pixel particle variation area in the defective injection product, define it as the source pixel saturation color gamut space;

[0105] Obtain the color saturation interval of the pixel particle variation area expressed in the qualified injection product, define it as the second color saturation interval, and construct the pixel saturation color gamut space of the pixel particle variation area in the defective injection product, define it as the target pixel saturation color gamut space;

[0106] Through the big data network, obtain the color mapping function table of the color saturation expressions of different injection defect types and the established Gamma correction interval for each injection defect type, and establish a color mapping rule according to the color mapping function table,

[0107] Introduce a linear transformation algorithm, and map each saturated color in the source pixel saturated color gamut space to the target pixel saturated color gamut space in the linear transformation algorithm based on the color mapping rule, and generate a linear color saturation mapping equation;

[0108] Solve the linear color saturation mapping equation to obtain a series of mapped chromaticity values, calculate and determine the restored pixel contrast of the pixel particle mutation area according to the series of mapped chromaticity values, and obtain the pixel contrast of the pixel particle mutation area expressed in the qualified injection-molded product, which is defined as the initial pixel contrast;

[0109] Introduce the Gamma correction formula to calculate the correction of the restored pixel contrast and the initial pixel contrast, and obtain the actual Gamma correction value. If the actual Gamma correction value is within the established Gamma correction interval, directly output the type of injection molding defect corresponding to the established Gamma correction interval to obtain the second defect detection result.

[0110] It should be noted that if the first defect detection result shows that the TV case is a defective injection-molded product, it is necessary to further determine the types of injection molding defects that exist, so as to provide a strong improvement basis for the optimization of the injection molding production of the subsequent TV case. Therefore, this method constructs a color mapping rule of pixel color saturation through a color mapping function for the color saturation expression of different injection molding defect types, and then linearly transforms and maps each saturated color in the source pixel saturated color gamut space of the pixel particle mutation area to the target pixel saturated color gamut space, so as to achieve the effect that the color saturation of the mutated pixel area on the defective injection-molded product is restored to the corresponding color saturation that the qualified injection-molded product should have, and further determine the types of injection molding defects that cause the change of its color saturation according to the mapped chromaticity expression of the color saturation restoration. Among them, the Chinese name of the Gamma correction formula is the gamma correction formula, which is used to adjust the contrast of the image, so that the display of the image is more in line with the visual perception of the human eye. Through Gamma correction, the pixel chromaticity display of the pixel particle mutation area expressed in the qualified injection-molded product can be output as the restored pixel display. Therefore, the Gamma correction value can be used as a key feature for tracing the types of injection molding defects. Through this method, the types of fine injection molding defects on the TV case can be traced, and the function of quickly and reliably identifying injection molding defects can be realized, which helps to improve the optimization accuracy of the injection molding process of the TV case.

[0111] The second aspect of the present invention provides an injection molding defect detection device for a TV case, as Figures 1-3 shown, the injection molding defect detection device includes:

[0112] An image acquisition module 1011, which is used to capture defect image data of the injection-molded finished product of the TV case from multiple directions;

[0113] The image processing module 1012 is responsible for introducing the Sobel operator to calculate and extract the feature information of the defect detection image data of different frames.

[0114] The flipping module 1013 is used to clamp the TV cabinet within the specified detection area and flip it in multiple directions, so that the image acquisition module can comprehensively photograph the detection missing model area of the TV cabinet.

[0115] The model construction module 1014 is used to perform modeling calculation processing on the defect detection image data extracted by the image processing module, so as to output a local process defect detection model.

[0116] The binary coding module 1015 is used to check whether the vertex grid scalar value of each model vertex on the attached model body is greater than the isosurface embedding threshold of the sub-detection missing scalar field, and program the binary status code of 0 or 1 for the model vertex of the attached model body accordingly.

[0117] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all of them should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A detection and control method for an injection molding defect detection device for a TV casing, characterized in that: The following steps are involved: S102: Constructing a standard injection molding mesh model for the TV housing, generating a spatial element smooth mesh model based on a model smoothing update of the standard injection molding mesh model, and performing a fit analysis by variationally fitting a local process defect detection model extracted from image data on the spatial element smooth mesh model to the spatial element smooth mesh model to obtain a detection missing model region; S104: constructing an element scalar embedding field for a designated inspection area of a TV casing, determining whether a grid scalar value is greater than an isosurface threshold, segmenting the inspection missing model area and embedding the embedded element scalar field, obtaining a spatial element coordinate matrix, and controlling an injection molding defect detection device to perform inspection based on the spatial element coordinate matrix to obtain a defect detection control solution; S106: Obtaining defect detection image information about the missing model area on the TV casing sample through the defect detection control solution, performing pixel variation calculation and analysis of the local image structure tensor on the defect monitoring image information to obtain a first defect detection result; S108: Based on the first defect monitoring result, the pixel particle mutation area of the lost model area is extracted and detected, and the source pixel saturation color space of the pixel particle mutation area is mapped to the target pixel saturation color space of the qualified injection molded product to analyze the pixel contrast of the injection molding defect type and obtain the second defect detection result.

2. The method for detecting and controlling a device for detecting defects in injection molding of a TV casing according to claim 1, wherein: The step S102 specifically includes the following steps: Obtaining a standardized design drawing of a TV case, constructing a standard injection molding mesh model of the TV case based on the standardized design drawing, and smoothly updating interwoven mesh vertices of the standard injection molding mesh model based on an injection molding rework rate analysis to obtain a spatial element smoothed mesh model of the TV case; Obtaining a preset inspection process for standardized injection molding of a TV housing by an injection molding defect detection device, extracting process shooting decision points of a visual camera through the preset inspection process, and segmenting the spatial element smoothing model into a plurality of sub-smoothing models based on the process shooting decision points; The visual camera of the injection molding defect detection device completely executes the preset inspection process through the inspection log. After that, the multi-frame defect detection image data output by the TV case is inspected in each sub-smoothing model. The Sobel operator is introduced to extract features from each frame of defect detection image data and model it, thus obtaining a local process defect detection model. Obtaining the spatial element distribution of each sub-smoothing model corresponding to the local process defect detection model, and constructing a variational lower bound of the model logarithmic marginal likelihood based on the spatial element distribution; Presetting a baseline model variational distribution for each sub-smoothing model based on the spatial element smoothing model, maximizing the variational lower bound, generating a posterior model variational distribution for the local defect detection model, calculating a hash dislocation function between the posterior model variational distribution and the baseline model variational distribution, and determining the degree of fit of each local defect detection model relative to each sub-smoothing model based on the hash dislocation function; If the fit is greater than the preset fit, the local defect detection model corresponding to the fit is fitted to the current sub-smoothing model, and finally the unfitted model area in the spatial element smoothing model is stripped out and marked as the detection missing model area.

3. The method for detecting and controlling a device for detecting injection defects of a TV casing according to claim 2, wherein: The method of obtaining a standardized design drawing of a TV case, constructing a standard injection molding mesh model of the TV case according to the standardized design drawing, and smoothly updating the interwoven mesh vertices of the standard injection molding mesh model based on an injection molding rework rate analysis to obtain a spatial element smooth mesh model of the TV case specifically includes the following steps: Obtaining a standardized design drawing of a TV casing, and obtaining multiple sets of standardized three-dimensional design parameters of the TV casing based on the standardized design drawing; Inputting multiple sets of the standardized three-dimensional design parameters into SolidWorks model design software for modeling operations to generate a standard injection molding mesh model of the TV case, and extracting interwoven mesh vertices and smoothing parameters between adjacent interwoven mesh vertices of the standard injection molding mesh model; A Laplace coordinate operator is introduced, and a neighborhood coordinate platform of a standard injection molding mesh model is constructed by assigning a smoothing parameter between adjacent interwoven mesh vertices in the Laplace coordinate operator. The neighborhood information of the interwoven mesh vertices is calculated using the neighborhood coordinate platform to obtain the Laplace neighborhood coordinates of each interwoven mesh vertex. Performing weighted average forward smoothing and reverse smoothing on each interwoven mesh vertex according to the Laplace neighborhood coordinates, thereby iteratively updating the current neighborhood position of each interwoven mesh vertex in the standard injection molding mesh model, so that each interwoven mesh vertex is closer to the average position of its own neighborhood vertices, and obtaining a positive smoothing weight value and a negative smoothing weight value; Obtaining a test log of an injection molding defect detection device and injection molding production requirements of a TV casing, extracting a historical injection molding rework rate of the TV casing at a preset time sequence through the test log, and presetting an allowable injection molding rework rate based on the injection molding production requirements; Calculating the injection molding rework rate difference between the historical injection molding rework rate and the allowable injection molding rework rate, presetting a smoothing update iteration frequency based on the injection molding rework rate difference, repeating the update iteration steps of forward smoothing and reverse smoothing of the weighted average interwoven mesh vertices until the smoothing update iteration frequency is reached, and generating a positive and negative smoothing weight matrix; The standard injection molding mesh model is reconstructed according to the positive and negative smoothing weight matrices to obtain the spatial element smoothing mesh model of the TV case.

4. The method for detecting and controlling a device for detecting defects in injection molding of a TV casing according to claim 1, wherein: The step S104 specifically includes the following steps: Obtaining a designated inspection area for a TV casing captured and inspected by a visual camera within an injection molding defect inspection device, and obtaining spatial specification parameters of the designated inspection area; constructing an element scalar embedding field of a designated detection area based on the spatial specification parameters, planning only a field area of the detection missing model area corresponding to the element scalar embedding field, marking it as a sub-detection missing scalar field, obtaining a fixed scalar value of the sub-detection missing scalar field, and presetting an isosurface embedding threshold according to the fixed scalar value; Divide the detected missing model area into N subsidiary model bodies, obtain a vertex mesh scalar value of each model vertex on each subsidiary model body, and check whether the vertex mesh scalar value of each model vertex is greater than an isosurface threshold; If the vertex mesh scalar value is greater than the isosurface embedding threshold, a binary value of 1 is input for the model vertex; if the vertex mesh scalar value is greater than the isosurface embedding threshold, a binary value of 0 is input for the model vertex to generate a binary vertex state code; Determining the embedding boundary of the missing scalar field of the embedding sub-detection of each subsidiary model body according to the programming translation of the binary vertex state code to obtain a plurality of scalar embedding boundaries, and performing linear interpolation calculation on each of the scalar embedding boundaries based on the grid scalar value to obtain an element scalar embedding isosurface corresponding to each subsidiary model body; Connect the corresponding element scalars of all the subsidiary model bodies and embed them into the isosurface, and finally generate a spatial element coordinate matrix in which the detection missing model area is located in the specified detection area. Control the flip mechanism to perform fixed-point shooting defect detection on the detection missing model area of the TV case in the specified detection area according to the spatial element coordinate matrix, and obtain a defect detection control plan.

5. The method for detecting and controlling an injection molding defect detection device for a TV casing according to claim 1, wherein: The step S106 specifically includes the following steps: Obtain a TV case sample that has been produced by injection molding, and inspect the TV case sample by executing the defect detection control scheme using an injection molding defect detection device to obtain defect detection image information about a missing model area on the TV case sample; Performing pixel-level edge detection on the defect detection image information using a Sobel operator to obtain an actual horizontal vector gradient and an actual vertical vector gradient in a detection missing model area, calculating a second-order moment of the defect detection image information based on the actual horizontal vector gradient and the actual vertical vector gradient, and generating a local image structure tensor for the detection missing model area; obtaining an actual image resolution of the defect detection image information, performing eigendecomposition of a local pixel structure on the local image structure tensor based on the actual image resolution to obtain a series of structural eigenvalues and structural eigenvectors of the local image structure tensor, and determining the actual pixel structure flatness of the detection missing model area based on the series of structural eigenvalues and structural eigenvectors; The injection molding defect detection device obtains the standard injection molding mesh model of the TV case and executes the defect detection control scheme to capture standard image information of the TV case, and calculates and obtains the pixel structure flatness of the standard image information, which is defined as the reference pixel structure flatness; If the actual pixel structure flatness is higher than the reference pixel structure flatness, the TV case in the detection missing model area is calibrated as a qualified injection molded product; if the actual pixel structure flatness is lower than the reference pixel structure flatness, the TV case in the detection missing model area is calibrated as a defective injection molded product, and the first defect detection result is obtained.

6. The method for detecting and controlling an injection molding defect detection device for a TV casing according to claim 1, wherein: The step S108 specifically includes the following steps: If the first defect detection result shows that the TV housing is a defective injection molded product, extracting a pixel particle variation region for detecting a missing model region from the defect detection image information of the defective injection molded product based on the actual pixel structure flatness; Obtaining a color saturation interval of the pixel particle mutation region, defined as a first color saturation interval; constructing a pixel saturation color gamut space of the pixel particle mutation region in the defective injection molded product based on the first color saturation interval, defined as a source pixel saturation color gamut space; Obtaining a color saturation interval of the pixel particle mutation region expressed in the qualified injection molded product, defined as a second color saturation interval; constructing a pixel saturation color gamut space of the pixel particle mutation region in the defective injection molded product based on the second color saturation interval, defined as a target pixel saturation color gamut space; The color mapping function table for the color saturation expression of different types of injection molding defects and the established Gamma correction interval for each type of injection molding defect are obtained through the big data network, and the color mapping rules are established according to the color mapping function table. Introducing a linear transformation algorithm, mapping each saturated color in the source pixel saturated color gamut space to the target pixel saturated color gamut space in the linear transformation algorithm based on the color mapping rule, and generating a linear color saturation mapping equation; Solving the linear color saturation mapping equation to obtain a series of mapped chromaticity values, calculating and determining the restored pixel contrast of the pixel particle mutation region based on the series of mapped chromaticity values, and obtaining the pixel contrast of the pixel particle mutation region expressed in the qualified injection molded product, which is defined as the initial pixel contrast; A gamma correction formula is introduced to calculate the correction of the restored pixel contrast and the initial pixel contrast to obtain an actual gamma correction value. If the actual gamma correction value is within a predetermined gamma correction interval, the type of injection molding defect corresponding to the predetermined gamma correction interval is directly output to obtain a second defect detection result.

7. A device for detecting defects in injection molding of a TV casing, characterized in that: The injection molding defect detection device comprises: An image acquisition module, the image acquisition module is used to capture defect image data of the finished injection-molded TV casing from multiple angles; An image processing module is responsible for introducing a Sobel operator to calculate and extract feature information of defect detection image data of different frames; A flip module is used to hold the TV casing in the designated detection area and flip it in multiple directions, so that the image acquisition module can fully capture the detection lost model area of the TV casing; A model building module, the model building module is used to perform modeling and calculation processing on the defect detection image data extracted by the image processing module to output a local process defect detection model; A binary encoding module, wherein the data processing module is used to check whether the vertex mesh scalar value of each model vertex on the attached model body is greater than the isosurface embedding threshold of the sub-detection missing scalar field, thereby programming the model vertex of the attached model body to input a binary state code of 0 or 1.

Images