Adhesive sticking quality detection method and device for automatic adhesive sticking equipment

Abstract

The invention relates to the technical field of image data processing, in particular to a rubberizing quality detection method and device for automatic rubberizing equipment, and the method comprises the steps: obtaining an original image of a to-be-detected rubberizing region; performing filtering processing on the original image by using an improved adaptive bilateral filtering algorithm to obtain a filtered image, performing differential operation on the filtered image and the original image to obtain a residual image, and performing adaptive threshold segmentation on the residual image to extract a defect region, and carrying out rubberizing quality detection based on the geometrical characteristics of the defect area. The problem that the accuracy of rubberizing quality detection is not high is solved.

Description

Technical Field

[0001] This invention relates to the field of image data processing technology. More specifically, this invention relates to a method and apparatus for detecting the adhesive application quality of automated adhesive application equipment. Background Technology

[0002] With the continuous development of automated manufacturing technology, automatic adhesive application equipment has been applied in industrial fields such as electronics manufacturing, automotive parts, and precision machinery assembly to accurately apply tapes, sealants, or functional adhesives to workpiece surfaces. The quality of adhesive application directly affects the sealing performance, structural strength, and appearance consistency of products. Defects such as adhesive breaks, stringing, edge lifting, misalignment, or partial missing areas often lead to decreased product performance or even failure. Therefore, how to quickly, accurately, and consistently inspect the applied area after application has become one of the key technologies for the intelligent development of automatic adhesive application equipment.

[0003] In existing technologies, machine vision-based adhesive application quality inspection methods are gradually becoming the mainstream approach. These methods typically acquire images of the adhesive application area using industrial cameras and combine them with algorithms such as image filtering, edge detection, and threshold segmentation to identify adhesive application defects. Among these, bilateral filtering algorithms are widely used in the preprocessing of adhesive application images or background modeling processes due to their ability to smooth noise while preserving edge features. To improve the adaptability of the algorithm in complex industrial environments, some solutions introduce so-called adaptive bilateral filtering methods, adjusting the filtering parameters to a certain extent to accommodate the processing needs of different image regions.

[0004] However, existing adaptive bilateral filtering algorithms still have significant shortcomings in the application of adhesive bonding quality inspection. On the one hand, fixed value range kernel parameters are usually constrained only by pixel grayscale differences, making it difficult to distinguish between normal grayscale changes caused by uneven lighting, metal surface reflection, or differences in adhesive material and actual adhesive bonding defects. As a result, when the defect is minor or in a transitional state, defect features are easily filtered out or weakened. On the other hand, when there are obvious abnormal structures such as breaks or stringing in the adhesive bonding area, fixed value range kernel parameters may over-respond to noise or local extreme grayscale changes, thereby amplifying irrelevant interference and affecting the stability of subsequent defect segmentation, resulting in low accuracy in adhesive bonding quality inspection. Summary of the Invention

[0005] To address the problem of low accuracy in adhesive quality testing 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 detecting the adhesive application quality of an automatic adhesive application device, comprising: acquiring an original image of an adhesive application area to be detected; filtering the original image using an improved adaptive bilateral filtering algorithm to obtain a filtered image; performing a difference operation between the filtered image and the original image to obtain a residual image; performing adaptive threshold segmentation on the residual image to extract defect areas; and performing adhesive application quality detection based on the geometric features of the defect areas; wherein the improved adaptive bilateral filtering algorithm includes a domain kernel parameter and a spatial domain kernel function, the spatial domain kernel function being inversely correlated with the degree of defect, and the domain kernel parameter being positively correlated with the degree of defect; the degree of defect is the product of the degree of homing and the directional curvature, the degree of homing characterizes the directional consistency of the target pixel, the directional curvature being positively correlated with the degree of homing of each pixel in a set second neighborhood centered on the target pixel, and inversely correlated with the gradient direction angle difference between each pixel in the set second neighborhood centered on the target pixel and the target pixel, and the target pixel being any pixel in the original image.

[0007] The above technical solution achieves a high degree of coupling between filtering and defect detection processes, making it easier to extract and segment adhesive defects in the residual image, improving the accuracy and stability of defect identification, and providing reliable and engineerable technical support for online quality inspection of automatic adhesive application equipment under complex working conditions.

[0008] Furthermore, pixels The range of kernel parameters at the location for: , Kernel parameters of the basic range, As a regulating factor, For pixels The degree of defect, This is the normalization function.

[0009] The above technical solution incorporates the degree of defect of the pixel into the adaptive adjustment of the value domain kernel parameter, so that the weight of gray-level similarity during the filtering process can dynamically change with the intensity of local structural anomalies, thereby enhancing the normal adhesive area and smooth background, while keeping the gray-level features from being overly smoothed in the defect area.

[0010] Furthermore, pixels Spatial domain kernel parameters for: , Based on the kernel parameters of the fundamental space domain, For the natural constant An exponential function with base 0. For pixels The degree of defect.

[0011] The above technical solution introduces the degree of defect of the pixel into the adaptive adjustment of the spatial domain kernel parameter, so that the smoothing contribution of the neighboring pixels to the target pixel during the filtering process can automatically change according to the intensity of local defects. This maintains a strong spatial smoothing effect in the normal adhesive area, effectively suppressing noise and illumination fluctuations. At the same time, it significantly weakens the neighborhood smoothing effect in areas with abnormal structures such as stringing, adhesive breakage, edge lifting or local offset, preventing abnormal features from being smoothed out by the surrounding normal area.

[0012] Furthermore, pixels same degree of direction for: , In pixels The center is set to the first neighboring pixel. gradient magnitude, In pixels The center is set to the first neighboring pixel. The gradient direction angle, The total number of pixels within the first neighborhood is set. These are the preset hyperparameters.

[0013] The above technical solution can accurately distinguish between real adhesive structure and non-structural interference from the perspective of local directional consistency, providing a reliable and physically meaningful characteristic basis for subsequent defect degree assessment, parameter adaptive adjustment and adhesive quality judgment, thereby significantly improving the accuracy and stability of detection.

[0014] Furthermore, pixels directional curvature for: , For pixels The degree of similarity at the same location In pixels The center is set to the second neighborhood of the pixel. The degree of similarity at each location The total number of pixels within the second neighborhood is set. For pixels The gradient direction angle, In pixels The center is set to the second neighborhood of the pixel. The gradient direction angle, wherein the second neighborhood range is set as an integer multiple of the first neighborhood range.

[0015] The above technical solution combines the local directional consistency of a pixel with the directional consistency and differences of pixels in its larger neighborhood for comprehensive calculation, effectively characterizing the spatial directional curvature and structural complexity of the adhesive application area. When the target area has good directional continuity and high directional consistency within its neighborhood, the directional curvature value is low, indicating that the area is a normal adhesive application or a regularly extended area. However, in cases of tape stretching, breakage, edge curling, or local displacement, the directional curvature value increases significantly due to obvious directional changes and disordered directional patterns within the neighborhood, thus highlighting the abnormal structure.

[0016] Furthermore, pixels First set neighborhood range for: , This is a rounding function. , These are the preset minimum and maximum neighborhood windows, respectively. For pixels gradient magnitude, In pixels The average gradient magnitude of all pixels within the initial neighborhood of the center is set. , Each in pixels The initial neighborhood of the center is set to the middle pixel. The gradient magnitude and gradient direction angle, To find the modulus function, The total number of pixels in the initial neighborhood range is defined as follows. To preset the first hyperparameter, This is the normalization function.

[0017] The above technical solution enables adaptive matching of the neighborhood window to different adhesive application states, effectively improving the filtering and subsequent adhesive application quality detection process's ability to retain minor defects and suppress interference from complex backgrounds, thereby enhancing the stability and reliability of the detection results.

[0018] Furthermore, adhesive application quality inspection is performed based on the geometric features of the defective areas, including: fitting each defective area with a minimum bounding rectangle to obtain a minimum bounding rectangle, and obtaining the aspect ratio of the minimum bounding rectangle. When the aspect ratio is greater than a set threshold, it is determined that there is an adhesive application defect.

[0019] Furthermore, a CCD camera is used to acquire the original image of the area to be detected where the adhesive is applied.

[0020] Furthermore, it also includes: performing grayscale conversion and homomorphic filtering on the original image.

[0021] In a second aspect, the present invention provides an adhesive application quality detection device for an automatic adhesive application equipment, 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 adhesive application quality detection method for an automatic adhesive application equipment described in any one of the above embodiments is implemented.

[0022] The beneficial effects of this invention are as follows: This invention achieves high-precision detection of adhesive application quality in automated adhesive application equipment by combining an improved adaptive bilateral filtering algorithm with residual image analysis and adaptive threshold segmentation. The invention adaptively adjusts the spatial smoothing intensity and gray-level similarity weights during the filtering process based on the local structural features and defect severity of pixels. This effectively smooths normal adhesive application areas and suppresses noise, while maintaining the integrity of gray-level and directional features in abnormal areas such as stringing, adhesive breakage, edge lifting, or local offset, enhancing defect separability. Furthermore, the introduction of a neighborhood adaptive adjustment mechanism allows the filtering window to dynamically change in different regions, ensuring smoothness consistency while preserving subtle defect information. This significantly improves detection accuracy, robustness, and engineering applicability, providing reliable and efficient online quality monitoring capabilities for automated adhesive application equipment under complex operating conditions. Attached Figure Description

[0023] Figure 1 This is a flowchart schematically illustrating a method for detecting the adhesive application quality of an automatic adhesive application device according to an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the comparison of the effects before and after algorithm improvement in the adhesive application quality detection method for an automatic adhesive application device according to an embodiment of the present invention; Figure 3 This is a schematic block diagram illustrating the structure of an adhesive application quality inspection device for an automatic adhesive application equipment according to an embodiment of the present invention. Detailed Implementation

[0024] Example of a method for detecting adhesive application quality in automated adhesive application equipment.

[0025] like Figure 1 The flowchart shown is a method for detecting the adhesive application quality of an automatic adhesive application device according to an embodiment of the present invention, which includes the following steps: S1: Obtain the original image of the area to be detected where the adhesive is applied.

[0026] In a preferred embodiment, an industrial-grade CCD camera acquires the original image of the adhesive application area to be inspected. The CCD camera is fixedly mounted above or to the side of the inspection station of the automated adhesive application equipment to ensure real-time and stable acquisition of surface images of the adhesive application area during the application process. By employing CCD imaging, image blurring or distortion caused by factors such as mechanical vibration and lighting fluctuations can be effectively avoided, improving the stability and consistency of adhesive application quality inspection.

[0027] Furthermore, after acquiring the original image, the original image is sequentially subjected to grayscale conversion and homomorphic filtering. Grayscale conversion transforms a multi-channel color image into a single-channel grayscale image, thereby reducing the dimensionality of the image data, decreasing computational complexity, and highlighting the structural information of the adhesive-coated area in terms of brightness. Since adhesive defects (such as stringing, breakage, lifting, or uneven edges) typically exhibit local abrupt changes or discontinuous features in the grayscale distribution, grayscale conversion helps enhance the recognizability of these features in subsequent gradient calculations and orientation analysis.

[0028] After grayscale conversion, homomorphic filtering is applied to the grayscale image. This homomorphic filtering separates the image's luminance and reflection components, suppresses low-frequency illumination components, and enhances high-frequency detail components. This effectively reduces overall brightness drift caused by surface reflections of the adhesive material, uneven ambient lighting, or attenuation of the device's light source. Through homomorphic filtering, the impact of background lighting variations on image quality is significantly reduced while preserving details of the adhesive edges, tape texture, and local defects, resulting in a more balanced spatial structure and grayscale distribution in the adhesive area.

[0029] S2: The original image is filtered using an improved adaptive bilateral filtering algorithm to obtain a filtered image.

[0030] like Figure 2 The figure shows a comparison of the effects before and after the algorithm improvement of the adhesive application quality detection method for automatic adhesive application equipment according to an embodiment of the present invention.

[0031] In a preferred embodiment, the improved adaptive bilateral filtering algorithm includes a range kernel parameter and a spatial domain kernel function, and the pixel... The range of kernel parameters at the location for: , Kernel parameters of the basic range, As a regulating factor, For pixels The degree of defect, This is the normalization function.

[0032] The defect characterization results of pixels are incorporated into the gray-level similarity constraint of bilateral filtering. A nonlinear normalization function is used to smoothly map the defect severity, allowing the kernel parameter to adaptively adjust according to the degree of local structural anomalies. This avoids the insufficient adaptability caused by using fixed parameters under different defect intensities. Specifically, the mapping method appropriately amplifies the kernel parameter in regions with low defect severity or in a transitional state to enhance sensitivity to gray-level differences and highlight potential anomalies. In regions with high defect severity, the parameter variation gradually saturates, preventing excessive amplification of noise or local extreme values. This gradual, adaptive parameter adjustment mechanism effectively improves the response to gray-level anomalies in adhesive-defect areas while maintaining a smooth effect in normal adhesive application areas, achieving a good balance between defect enhancement and noise suppression in the filtering results.

[0033] pixel Spatial domain kernel parameters for: , Based on the kernel parameters of the fundamental space domain, For the natural constant An exponential function with base 0. For pixels The degree of defect.

[0034] By directly incorporating the defect characterization results of pixels into the adjustment process of spatial smoothing weights, the spatial domain constraints can adaptively change with the degree of local structural anomalies, thus overcoming the oversmoothing problem caused by the fixed spatial kernel parameter in traditional bilateral filtering. Through an exponential mapping relationship, when a pixel is in a structurally stable region with low defect levels, the spatial kernel parameter maintains a large value, allowing surrounding pixels to fully participate in the smoothing calculation, thereby effectively suppressing noise and improving overall image consistency. Conversely, when a pixel is located in anomaly areas such as adhesive edges, stringing, or breaks, the spatial kernel parameter decreases rapidly with increasing defect levels, significantly weakening the smoothing effect of neighboring pixels on that pixel and preventing abnormal structures from being masked by surrounding background information. This adjustment mechanism, where spatial smoothing intensity and defect level are inversely linked, effectively protects defective regions while ensuring smoothing effects in normal adhesive areas. This makes defect features clearer and more stable in subsequent residual calculations and segmentation, thereby improving the accuracy and reliability of adhesive quality detection.

[0035] The degree of defect is the product of the degree of unidirectionality and the directional curvature, per pixel. same degree of direction for: , In pixels The center is set to the first neighboring pixel. gradient magnitude, In pixels The center is set to the first neighboring pixel. The gradient direction angle, The total number of pixels within the first neighborhood is set. These are the preset hyperparameters.

[0036] The gradient information of each pixel in the neighborhood of the target pixel is uniformly mapped into the direction vector space. By weighting and superimposing different directional components, the consistency of gradient directions within a local region is characterized as a whole. The gradient magnitude is used as a weight in the calculation, giving significant edges or tape contours a higher influence in the statistical process, while the contribution of weak textures or noisy regions is naturally weakened. By normalizing the direction vector synthesis results, the influence of neighborhood scale changes or overall brightness differences on the calculation results can be effectively eliminated, allowing the degree of homogeneity to stably reflect whether the local structure exhibits obvious directional characteristics. Thus, in normal adhesive areas or at the edges of regular tape, the degree of homogeneity is higher, representing structural continuity and consistent direction. However, in abnormal areas such as stringing, adhesive breakage, and peeling, the degree of homogeneity is significantly reduced due to disordered gradient direction distribution or multi-directional competition. This provides a criterion for subsequent defect severity modeling that is highly sensitive to directional disorder features and has the ability to suppress random noise.

[0037] pixel First set neighborhood range for: , This is a rounding function. , These are the preset minimum and maximum neighborhood windows, respectively. For pixels gradient magnitude, In pixels The mean of the gradient magnitudes of all pixels within the initial neighborhood of the center is set, and in this embodiment, it excludes... The extreme case where it is 0. , Each in pixels The initial neighborhood of the center is set to the middle pixel. The gradient magnitude and gradient direction angle, To find the modulus function, The total number of pixels in the initial neighborhood range is defined as follows. To preset the first hyperparameter, This is a normalization function. In this embodiment, the initial neighborhood range is set to 3. 3. Of course, you can also set it according to the actual situation.

[0038] This mechanism maps the local structural complexity of a pixel location to the adjustment of the neighborhood scale. By comprehensively considering the differences between the target pixel and its surrounding pixels in terms of intensity variation amplitude and directional consistency, the neighborhood range is dynamically adjusted according to the local characteristics of the image. When the gradient change in the target region is drastic or the directional distribution is significantly discrete, it indicates that the region may be in an unstable structural position such as an adhesive edge, break, or stringing. In this case, the response value is compressed through a nonlinear normalization function, causing the neighborhood range to adaptively decrease. This avoids excessive smoothing of abnormal structures by an overly large neighborhood, thus improving the sensitivity to small defects and directional anomalies. Conversely, when the gradient change in the target region is gentle and the directional distribution tends to be consistent, it indicates that the region is more likely to belong to normal adhesive or background areas. The neighborhood range is expanded through a corresponding mapping relationship, allowing more surrounding pixels to participate in statistical calculations, thereby enhancing noise resistance and improving the stability of feature estimation. Thus, this neighborhood adaptive mechanism effectively suppresses the interference of noise and illumination fluctuations on feature calculations while ensuring that local abnormal structures are not weakened. It provides a reasonable scale, stable response, and engineering applicability for subsequent adhesive defect feature extraction and quality judgment.

[0039] pixel directional curvature for: , For pixels The degree of similarity at each location In pixels The center is set to the second neighborhood of the pixel. The degree of similarity at each location The total number of pixels within the second neighborhood is set. For pixels The gradient direction angle, In pixels The center is set to the second neighborhood of the pixel. The gradient direction angle, wherein the second neighborhood range is set as an integer multiple of the first neighborhood range.

[0040] By jointly modeling the directional consistency of the target location with the directional consistency and directional difference magnitude of locations within the extended neighborhood, directional changes are effectively amplified only in regions with well-defined structures and stable directional characteristics, while being naturally suppressed in areas with random noise or weak texture. Specifically, by introducing a sinusoidal mapping of directional differences, angular changes can be smoothly converted into continuous variations, thus avoiding unstable responses caused by abrupt directional changes. Therefore, when abnormalities such as tape bending, stringing, or edge undulations occur during adhesive application, this directional curvature accurately reflects the degree of disruption to directional continuity, while maintaining a lower value in normal straight lines or regular adhesive application areas. This ensures high sensitivity to real structural anomalies while effectively reducing the impact of environmental noise and local texture interference on defect detection.

[0041] S3: Perform a difference operation between the filtered image and the original image to obtain a residual image, perform adaptive threshold segmentation on the residual image to extract the defect region, and perform adhesive quality detection based on the geometric features of the defect region.

[0042] In a preferred embodiment, the adhesive application quality is detected and judged based on the extracted geometric features of the defective areas. Specifically, after segmenting the defective areas, geometric analysis is performed on each individual defective area. By fitting a minimum bounding rectangle with rotation, a minimum bounding rectangle matching the actual shape and direction of the defective area is constructed. Unlike using an axis-aligned rectangle, the rotation minimum bounding rectangle can adaptively adjust the orientation of the rectangle, ensuring that its long side is consistent with the main direction of the defective area, thereby more accurately reflecting the true spatial extension of the adhesive application anomaly.

[0043] After obtaining the minimum bounding rectangle, its length and width information are further extracted, and the corresponding aspect ratio parameters are calculated. Typical defects generated during the adhesive application process, such as tape stringing, edge lifting, localized adhesive breaks, or misalignment, typically exhibit slender, directional, or irregularly extending geometric features in spatial morphology. Therefore, the aspect ratio of such defects in the rotated minimum bounding rectangle is often significantly larger than that of the normal adhesive application area or scattered anomalies caused by noise. By comparing the aspect ratio with a pre-set threshold, when the aspect ratio of a defective area exceeds the threshold, the defective area can be determined to be a valid defect affecting the adhesive application quality, thus outputting the adhesive application defect judgment result.

[0044] This invention introduces an adaptive bilateral filtering mechanism based on defect severity into automated adhesive applicator equipment. By dynamically adjusting filtering parameters based on pixel local gradient magnitude, directional consistency, and directional curvature, it achieves smooth enhancement of normal adhesive application areas and feature preservation of abnormal areas, effectively highlighting defect information during filtering and residual calculation. Through adaptive threshold segmentation of the residual image, local defect areas of the adhesive tape can be accurately extracted, and quantitative analysis of the geometric features of these defect areas enables precise identification of various adhesive application anomalies such as stringing, adhesive breakage, edge lifting, and local offset. Simultaneously, the solution incorporates adaptive adjustment of the neighborhood range and comprehensive evaluation of directional consistency and curvature, ensuring that the filtering and defect detection processes maintain high sensitivity to subtle anomalies while suppressing noise interference. Furthermore, geometric feature analysis methods such as minimum bounding rectangle fitting are used to quantitatively determine defects, significantly improving the accuracy, stability, and engineering feasibility of adhesive application quality inspection. This provides a reliable and efficient technical guarantee for online monitoring and quality control of automated adhesive applicator equipment in complex production environments.

[0045] Example of an adhesive application quality inspection device for automated adhesive application equipment: like Figure 3 As shown in the figure, the structural block diagram of the adhesive quality detection device for an automatic adhesive applicator according to an embodiment of the present invention includes a processor and a memory.

[0046] This invention also provides a device for detecting the adhesive application quality of an automatic adhesive application machine. For example... Figure 3 As shown, the system includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the adhesive quality detection method for an automatic adhesive applicator according to the present invention.

[0047] The adhesive quality detection device for automatic adhesive applicator 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.

[0048] 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.

[0049] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0050] 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 adhesive application quality of an automatic adhesive application equipment, characterized in that, include: Acquire the original image of the adhesive-covered area to be detected; The original image is filtered using an improved adaptive bilateral filtering algorithm to obtain a filtered image. The filtered image and the original image are then differentially processed to obtain a residual image. The residual image is then subjected to adaptive threshold segmentation to extract defect regions. Finally, the adhesive bonding quality is detected based on the geometric features of the defect regions. The improved adaptive bilateral filtering algorithm includes a range kernel parameter and a spatial kernel function. The spatial kernel function is inversely correlated with the degree of defect, while the range kernel parameter is positively correlated with the degree of defect. The degree of defect is the product of the degree of homing and the directional curvature. The degree of homing characterizes the directional consistency of the target pixel. The directional curvature is positively correlated with the degree of homing of each pixel in a set second neighborhood centered on the target pixel, and inversely correlated with the gradient direction angle difference between each pixel in the set second neighborhood centered on the target pixel and the target pixel. The target pixel is any pixel in the original image.

2. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, pixel The range of kernel parameters at the location for: , Kernel parameters of the basic range, As a regulating factor, For pixels The degree of defect, This is the normalization function.

3. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, pixel Spatial domain kernel parameters for: , Based on the kernel parameters of the fundamental space domain, For the natural constant An exponential function with base 0. For pixels The degree of defect.

4. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, pixel same degree of direction for: , In pixels The center is set to the first neighboring pixel. gradient magnitude, In pixels The center is set to the first neighboring pixel. The gradient direction angle, The total number of pixels within the first neighborhood is set. These are the preset hyperparameters.

5. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, pixel directional curvature for: , For pixels The degree of similarity at each location In pixels The center is set to the second neighborhood of the pixel. The degree of similarity at the same location The total number of pixels within the second neighborhood is set. For pixels The gradient direction angle, In pixels The center is set to the second neighborhood of the pixel. The gradient direction angle, wherein the second neighborhood range is set as an integer multiple of the first neighborhood range.

6. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 4, characterized in that, pixel First set neighborhood range for: , This is a rounding function. , These are the preset minimum and maximum neighborhood windows, respectively. For pixels gradient magnitude, In pixels The average gradient magnitude of all pixels within the initial neighborhood of the center is set. , Each in pixels The initial neighborhood of the center is set to the middle pixel. The gradient magnitude and gradient direction angle, To find the modulus function, The total number of pixels in the initial neighborhood range is defined as follows. To preset the first hyperparameter, This is the normalization function.

7. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, Adhesive application quality inspection based on the geometric features of defective areas includes: fitting each defective area with a minimum bounding rectangle to obtain a minimum bounding rectangle, and obtaining the aspect ratio of the minimum bounding rectangle. When the aspect ratio is greater than a set threshold, it is determined that there is an adhesive application defect.

8. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, Use a CCD camera to acquire the original image of the area to be inspected for adhesive application.

9. The method for detecting the adhesive application quality of an automatic adhesive application device according to claim 1, characterized in that, Also includes: The original image is then subjected to grayscale conversion and homomorphic filtering.

10. A device for detecting the adhesive application quality of an automatic adhesive application machine, characterized in that, It includes a memory and a processor, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the adhesive quality detection method for an automatic adhesive applicator as described in any one of claims 1 to 9 is implemented.

Images