Injection molding quality detection method and system for automobile injection molding part
By employing an anomaly response mechanism based on local structural anisotropy and neighborhood consistency analysis, combined with high-precision imaging equipment and image threshold segmentation algorithms, the problem of low detection accuracy in automotive injection molded parts inspection has been solved, achieving high-precision, automated, and real-time quality inspection of injection molded parts.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-28
- Publication Date
- 2026-03-10
AI Technical Summary
In existing technologies, the quality inspection of automotive injection molded parts relies on manual optical inspection, which is inefficient and highly subjective. Traditional image processing methods are sensitive to changes in lighting and texture complexity, while deep learning methods have high training costs and insufficient generalization ability, resulting in low detection accuracy.
An anomaly response mechanism based on local structural anisotropy and neighborhood consistency analysis is adopted. Combined with high-precision imaging equipment and image thresholding segmentation algorithm, the anomaly response mechanism is calculated by acquiring images of automotive injection molded parts, calculating local structural anisotropy and anomaly degree, generating an anomaly degree feature map, and using image thresholding segmentation and connected component analysis for automated detection.
It achieves highly sensitive feature extraction of surface defects in injection molded parts, significantly improving detection accuracy and stability. It can accurately identify defects under different lighting and texture conditions, reducing false detections and missed detections, and has automated and real-time detection capabilities.
Smart Images
Figure CN121639619A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology. More specifically, this invention relates to a method and system for inspecting the injection molding quality of automotive injection molded parts. Background Technology
[0002] As a crucial component of vehicle exterior, interior, and structural parts, automotive injection-molded parts directly impact the assembly precision, appearance consistency, and mechanical performance stability of the entire vehicle. While injection molding technology is relatively mature in modern automotive manufacturing, factors such as material shrinkage, uneven mold temperature, fluctuating melt flow characteristics, insufficient cooling efficiency, and variations in equipment stability still contribute to various quality defects on the surface of injection-molded parts. These defects include scorching, flow marks, weld lines, short shots, dents, surface particles, and warpage. These defects not only reduce the appearance quality of the parts but can also negatively affect subsequent painting, assembly, and the overall durability of the vehicle. Therefore, high-precision, repeatable, and automated quality inspection of injection-molded parts is of paramount importance.
[0003] Currently, the quality inspection of injection molded parts still relies primarily on manual optical inspection, depending on the inspectors' visual observation and judgment of the parts' surfaces. Manual inspection suffers from low efficiency, high subjectivity, poor consistency, and high labor intensity. Especially in mass production scenarios, it is difficult to maintain high precision manually for extended periods, and fatigue, experience differences, or environmental changes can easily lead to missed or false positives. Furthermore, the complex optical characteristics of most injection molded parts, such as textures, curved surfaces, localized reflections, or uneven coloring, further increase the difficulty of manual inspection.
[0004] To improve detection efficiency and accuracy, the industry has gradually adopted computer vision-based automated detection methods, such as edge detection, threshold segmentation, color analysis, and template matching, which are based on traditional image processing techniques. However, these methods generally rely on simple local features and are highly sensitive to changes in lighting, surface reflection, texture complexity, and workpiece deformation, leading to unstable detection results. When the surface of the injection molded part contains reflective areas, weak texture regions, or local noise, traditional algorithms often struggle to accurately separate the true defect areas, frequently resulting in oversegmentation, undersegmentation, or sensitivity to noise. Furthermore, while deep learning-based detection methods can improve detection accuracy, model training relies on a large number of labeled samples, and when faced with injection molded parts of different materials, colors, or shapes, they are prone to limitations such as insufficient generalization ability, high deployment costs, and long training cycles, leading to low accuracy in quality detection. Summary of the Invention
[0005] To address the problem of low accuracy in quality inspection mentioned in the background art, the present invention provides solutions in the following aspects.
[0006] In a first aspect, the present invention provides a method for injection molding quality inspection of automotive injection molded parts, comprising: acquiring an image of an automotive injection molded part to be inspected; calculating the local structural anisotropy of a target pixel in the image of the automotive injection molded part to be inspected, wherein the target pixel is any pixel in the image of the automotive injection molded part to be inspected; calculating the degree of anomaly of the target pixel in the image of the automotive injection molded part to be inspected, wherein the degree of anomaly is positively correlated with local structural anisotropy and the Euclidean distance between a pixel in a set neighborhood window centered on the target pixel and the target pixel, and is inversely correlated with the absolute value of the difference in local structural anisotropy between a pixel in a set neighborhood window centered on the target pixel and the target pixel; mapping the degree of anomaly value to a preset two-dimensional coordinate space to generate a corresponding degree of anomaly feature map; applying an image thresholding segmentation algorithm to the degree of anomaly feature map to obtain a binarized defect mask map; and performing injection molding quality inspection judgment on the image of the automotive injection molded part to be inspected based on the binarized defect mask map.
[0007] The above technical solution constructs an anomaly response mechanism that can simultaneously characterize local structural orientation features and the degree of abrupt changes in neighborhood consistency. This enables the real defect area in the image to show a significantly enhanced response in the anomaly degree feature map, while the responses of normal areas and random noise areas are effectively suppressed. Furthermore, the quality judgment process executed through a binarized mask achieves automated and real-time detection, significantly reducing human error and enhancing adaptability under different surface textures, lighting conditions, and imaging quality variations. This, in turn, improves the overall accuracy, stability, and engineering practicality of automotive injection molded parts appearance quality inspection.
[0008] Furthermore, target pixels Local structural anisotropy for: , For target pixel Within the set neighborhood window centered on Gradient magnitude of each pixel For target pixel Within the set neighborhood window centered on Gradient direction of each pixel The total number of pixels within the defined neighborhood window. These are the preset hyperparameters.
[0009] The above technical solution utilizes the gradient magnitude and direction information of each pixel in the neighborhood around the target pixel to construct a structural feature quantity that can reflect the concentration of local texture direction. This results in regions with obvious texture directionality, clear edge contours, or significant structural abrupt changes exhibiting higher responses in the calculation results, while regions with random textures, more noise, or more uniform structures show significantly reduced responses. It can maintain stable performance under conditions of uneven illumination, surface reflection, or weak texture, significantly improving the accuracy and reliability of defect detection in complex real-world scenarios.
[0010] Furthermore, target pixels abnormality for: , For target pixel Local structural anisotropy, For target pixel Within the set neighborhood window centered on Local structural anisotropy of individual pixels The total number of pixels within the defined neighborhood window. For target pixel With target pixel Within the set neighborhood window centered on Euclidean distance between pixels For the natural constant An exponential function with base 0.
[0011] The above technical solution constructs an anomaly response mechanism that is highly sensitive to real defects by simultaneously combining the local structural features of the target pixel with the structural differences and spatial distance relationships between it and its surrounding pixels. When the target pixel and its neighboring pixels maintain the same structural performance, the neighborhood difference term tends to be small, and its anomaly response mainly reflects its own structural features, resulting in a low overall performance. However, when there are local abrupt changes around the target pixel caused by flow marks, weld lines, pits, particles, etc., the neighborhood difference term increases significantly, causing the exponential term to decay rapidly. This leads to a sharp increase in the anomaly response at the defect location, achieving a strong highlighting of the structurally abnormal region.
[0012] Furthermore, CCD or CMOS cameras are used to acquire images of the automotive injection molded parts to be inspected.
[0013] Furthermore, it also includes denoising the image of the automotive injection molded part to be detected.
[0014] Furthermore, a bilateral filtering denoising algorithm is used to denoise the image of the automotive injection molded part to be detected.
[0015] The above technical solution uses bilateral filtering to denoise the image to be tested, which can effectively smooth out random noise and weak texture interference in the image, while maintaining the surface edge contour, fine structural changes and local abrupt changes that may represent defects of the injection molded part.
[0016] Furthermore, the image thresholding algorithm is the maximum inter-class variance method.
[0017] Furthermore, based on the binarized defect mask image, injection molding quality inspection and judgment are performed on the automotive injection molded part image to be inspected. Specifically, connected component analysis is performed on the binarized defect mask image to extract multiple suspected defect regions. If the area of the suspected defect region exceeds a preset area threshold, the automotive injection molded part to be inspected is determined to have quality defects.
[0018] The aforementioned technical solution, through connected component analysis of the binarized defect mask image, can automatically aggregate spatially adjacent and morphologically continuous abnormal pixels into independent candidate defect regions. This effectively restores the true shape, range, and distribution characteristics of potential defects while simultaneously filtering out pseudo-defect interference from isolated noise points. Furthermore, by combining this with area thresholding, the size of defects can be quantitatively determined according to actual process requirements and quality standards. This avoids misidentifying minute, irregular areas with minimal impact on product performance and appearance as defects, thus improving the engineering applicability of the judgment results.
[0019] Furthermore, the set neighborhood window is 3. 3 or 5 5.
[0020] In a second aspect, the present invention provides an injection molding quality inspection system for automotive injection molded parts, comprising a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the injection molding quality inspection method for automotive injection molded parts described in any of the above claims is implemented.
[0021] The beneficial effects of this invention are as follows: This invention achieves highly sensitive feature extraction of surface defects in automotive injection molded parts by introducing an anomaly calculation method that combines local structural directionality and neighborhood consistency analysis based on acquired images of the parts. It quantifies the concentration of texture direction using local structural anisotropy and combines the structural differences and spatial distance relationships between pixels in the neighborhood to generate a significantly enhanced response to defect areas, while suppressing interference from noise and normal textures. This clearly highlights potential defects in the anomaly feature map. An adaptive threshold segmentation algorithm generates a binary defect mask, and connected component analysis and area determination are used to achieve precise defect localization and reliable quality assessment, making the inspection process automated, real-time, and highly accurate. This significantly improves the accuracy, reliability, and engineering practicality of automotive injection molded part appearance quality inspection. Attached Figure Description
[0022] Figure 1 This is a flowchart schematically illustrating a method for inspecting the injection molding quality of automotive injection molded parts according to an embodiment of the present invention; Figure 2 This is a schematic block diagram illustrating the structure of an injection molding quality inspection system for automotive injection molded parts according to an embodiment of the present invention. Detailed Implementation
[0023] An embodiment of a method for testing the injection molding quality of automotive injection molded parts.
[0024] like Figure 1 As shown, a flowchart of an injection molding quality inspection method for automotive injection molded parts according to an embodiment of the present invention includes the following steps: S1: Obtain the image of the automotive injection molded part to be inspected.
[0025] In a preferred embodiment, the raw surface image of the automotive injection molded part to be inspected is acquired using a high-resolution CCD camera or a high-speed CMOS camera. CCD cameras offer advantages such as low imaging noise, wide dynamic range, and stable image quality, enabling them to obtain high-fidelity image data even under fluctuating lighting conditions or when the injection molded part has complex textures. CMOS cameras, on the other hand, feature higher frame rates and lower power consumption, making them suitable for scenarios involving high-speed movement or frequent changes in inspection positions on production lines, effectively improving the response speed and real-time performance of the inspection system.
[0026] By using the aforementioned imaging equipment, it is possible to ensure that the acquired images have high clarity and structural information integrity, thereby providing a high-quality data foundation for subsequent defect identification algorithms and significantly reducing the risk of false detection and missed detection caused by insufficient image quality.
[0027] Furthermore, to improve the separability of effective features in the image, the image of the automotive injection molded part to be detected undergoes denoising processing that emphasizes both noise suppression and detail preservation. Denoising methods may include, but are not limited to: edge-preserving denoising algorithms based on bilateral filtering, used to reduce random noise while maintaining the edge contours and structural details of the injection molded part; multi-scale denoising algorithms based on wavelet thresholding, which can suppress high-frequency noise components at different scales, maintaining high detectability of fine surface textures and potential micro-defects; and texture enhancement denoising strategies based on adaptive filtering, which can dynamically adjust the filtering weights according to the noise statistical characteristics of local image regions, thereby improving the signal-to-noise ratio and texture consistency of complex structural regions.
[0028] The above denoising process can effectively eliminate thermal noise, transmission noise, and random texture interference caused by surface reflection of the injection molded part, making the target defect area in the image more prominent, and its grayscale changes, contour boundaries and structural features clearer.
[0029] S2: Calculate the local structural anisotropy and anomaly degree of the target pixel in the image of the automotive injection molded part to be detected.
[0030] In a preferred embodiment, the target pixel Local structural anisotropy for: , For target pixel Within the set neighborhood window centered on Gradient magnitude of each pixel For target pixel Within the set neighborhood window centered on Gradient direction of each pixel The total number of pixels within the defined neighborhood window. The preset hyperparameters are used; the set neighborhood window is 3. 3 or 5 5. Of course, you can also set it according to the actual situation.
[0031] By constructing a local neighborhood around the target pixel and weighting and aggregating the gradient magnitude and direction of all pixels within that neighborhood, a structural feature quantity reflecting the consistency of local texture orientation is obtained. If there are obvious directional textures or defect boundaries within the neighborhood, the gradient directions of each pixel will be concentrated, resulting in a larger vector synthesis after the accumulation of cosine and sine direction components, significantly improving the feature quantity. Conversely, when the regional texture is random, noisy, or lacks obvious directionality, the gradient directions of each pixel cancel each other out, resulting in a smaller vector magnitude. In this way, the scheme effectively characterizes the anisotropy of the local structure, making true defect edges or structural anomalies show higher responses in the feature map, while flat areas and random noise areas show lower responses, thereby improving the accuracy and robustness of subsequent defect detection.
[0032] target pixel abnormality for: , For target pixel Local structural anisotropy, For target pixel Within the set neighborhood window centered on Local structural anisotropy of individual pixels The total number of pixels within the defined neighborhood window. For target pixel With target pixel Within the set neighborhood window centered on Euclidean distance between pixels For the natural constant An exponential function with base 0.
[0033] Building upon the establishment of local structural anisotropy, a neighborhood consistency constraint is further introduced. By weighting the structural differences between the target pixel and its neighboring pixels according to distance, and then suppressing the degree of these differences exponentially, an anomaly response value that simultaneously reflects the strength of local directionality and the degree of abrupt changes in neighborhood structure is obtained. When the target pixel and its surrounding pixels have similar structural features, the weighted difference is small, and the exponential term tends to be large, making the anomaly response mainly determined by its own structural features. Conversely, when there is a significant structural deviation or abrupt change between the target pixel and its neighborhood, the weighted difference increases, and the exponential term rapidly decays, significantly improving the final anomaly response. Through this construction method, the scheme can not only highlight defect features such as abrupt changes in local texture direction, edge breaks, and abrupt structural expansion, but also effectively suppress erroneous responses caused by non-defect factors such as noise and weak texture changes, thereby significantly enhancing the sensitivity and discriminative power of the feature map for real defect areas.
[0034] S3: Apply an image thresholding algorithm to the anomaly feature map to obtain a binary defect mask map, and perform injection molding quality inspection and judgment on the image of the automotive injection molded part to be inspected based on the binary defect mask map.
[0035] In a preferred embodiment, the image thresholding process employs the maximum inter-class variance method to perform globally adaptive threshold calculation on the anomaly feature map. This method analyzes the inter-class differences in pixel grayscale values between the foreground and background classes in the entire feature map, automatically selecting the optimal threshold that maximizes inter-class separability, thereby generating a highly discriminative binary defect mask. Compared to manual or empirical thresholding, this method can dynamically adjust the threshold based on variations in the surface material of different injection molded parts, lighting conditions, and imaging noise. Therefore, it exhibits higher stability and generalization ability in complex production environments, effectively avoiding missed and false detections caused by improper threshold selection.
[0036] The injection molding quality inspection and judgment are performed on the image of the automotive injection molded part to be inspected based on the binarized defect mask. Specifically, connected component analysis is performed on the binarized defect mask to extract multiple suspected defect regions. If the area of the suspected defect region exceeds a preset area threshold, the automotive injection molded part to be inspected is judged to have quality defects.
[0037] This invention achieves highly sensitive detection of minute defects and local structural abrupt changes by introducing an anomaly response mechanism that combines local structural anisotropy with neighborhood consistency analysis into injection molded part images. This effectively suppresses false responses from noise and weakly textured areas. High-precision imaging equipment is used to acquire images, combined with bilateral filtering for denoising, preserving defect edges and detailed features while improving the stability of feature extraction. Anomaly response mapping generates feature maps, and the maximum inter-class variance method is applied for adaptive threshold segmentation, achieving accurate separation of defect regions. Connected component analysis combined with area determination is then used to quantify and identify true defects, ensuring the reliability and engineering applicability of the detection results. The overall method exhibits high robustness, high automation, and strong adaptability to variations in texture, lighting, and injection molded part materials. It enables rapid, accurate, and real-time injection molding quality inspection on a systematic hardware platform, significantly improving the accuracy, stability, and practicality of detecting appearance and structural defects in automotive injection molded parts.
[0038] An embodiment of an injection molding quality inspection system for automotive injection molded parts: like Figure 2 As shown, a structural block diagram of an injection molding quality inspection system for automotive injection molded parts according to an embodiment of the present invention includes a processor and a memory.
[0039] This invention also provides an injection molding quality inspection system for automotive injection molded parts. For example... Figure 2As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the injection molding quality inspection method for automotive injection molded parts according to the present invention.
[0040] The injection molding quality inspection system for automotive injection molded parts also includes other components well known to those skilled in the art, such as communication interfaces. Their settings and functions are known in the art and will not be described in detail here.
[0041] In this invention, the aforementioned memory can be any tangible medium containing or storing a program that can be used or combined with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc., or any other medium that can be used to store desired information and can be accessed by an application, module, or both. Any such computer storage medium can be part of a device or accessible to or connected to a device. Any application or module described in this invention can be implemented using computer-readable / executable instructions stored or otherwise maintained by such a computer-readable medium.
[0042] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.
[0043] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.
Claims
1. A method for detecting the injection quality of an injection-molded part of an automobile, characterized in that The method comprises the following steps: acquiring an image of an automobile injection molded part to be detected; calculating local structure anisotropy of a target pixel point in the image of the automobile injection molded part to be detected, the target pixel point being any pixel point in the image of the automobile injection molded part to be detected; calculating an abnormality degree of the target pixel point, the abnormality degree being positively correlated with the local structure anisotropy and the Euclidean distance between the target pixel point and the pixel points in a set neighborhood window centered on the target pixel point, and being inversely correlated with the absolute value of the difference between the local structure anisotropy of the target pixel point and the local structure anisotropy of the pixel points in the set neighborhood window centered on the target pixel point; mapping the abnormality degree value to a preset two-dimensional coordinate space to generate a corresponding abnormality degree feature map; applying an image threshold segmentation algorithm to the abnormality degree feature map to obtain a binary defect mask image; and performing injection quality detection and judgment on the image of the automobile injection molded part to be detected according to the binary defect mask image.
2. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, target pixel Local structural anisotropy for: , For target pixel Within the set neighborhood window centered on Gradient magnitude of each pixel For target pixel Within the set neighborhood window centered on Gradient direction of each pixel The total number of pixels within the defined neighborhood window. These are the preset hyperparameters.
3. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, target pixel abnormality for: , For target pixel Local structural anisotropy, For target pixel Within the set neighborhood window centered on Local structural anisotropy of individual pixels The total number of pixels within the defined neighborhood window. For target pixel With target pixel Within the set neighborhood window centered on Euclidean distance between pixels For the natural constant An exponential function with base 0.
4. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, The image of the automobile injection molded part to be detected is acquired by using a CCD camera or a CMOS camera.
5. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, The image of the automobile injection molded part to be detected is also subjected to denoising processing.
6. The method for detecting the injection molding quality of an automobile injection molding part according to claim 5, characterized in that, The image of the automobile injection molded part to be detected is subjected to denoising processing by using a bilateral filter denoising algorithm.
7. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, The image threshold segmentation algorithm is the maximum inter-class variance method.
8. The method for detecting the injection molding quality of an automobile injection molding part according to claim 1, characterized in that, The injection quality detection and judgment on the image of the automobile injection molded part to be detected according to the binary defect mask image is specifically performed as follows: connected component analysis is performed on the binary defect mask image to extract a plurality of suspected defect regions, and if the area of a suspected defect region exceeds a preset area threshold, the automobile injection molded part to be detected is determined to have a quality defect.
9. The method of claim 1, wherein the method further comprises: determining a quality of the injection molded part based on the at least one of the first and second images. The set neighborhood window is 3 3 or 5 5.
10. A system for inspecting the injection molding quality of automotive injection molded parts, characterized in that, The method comprises a memory and a processor, and the memory stores computer program instructions, which, when executed by the processor, realize the injection quality detection method of the automobile injection molded part according to any one of claims 1-9.