Intelligent detection method for bending processing of aluminum alloy plate

Abstract

The invention relates to the technical field of image processing, in particular to an intelligent detection method for aluminum alloy plate bending processing, and the method comprises the steps: collecting a bending image sequence; separating each frame of image into a substrate image containing material texture information and a reflection image containing highlight information to obtain a substrate image sequence; carrying out optical flow analysis and texture analysis on the substrate image sequence so as to identify processing defects generated in the bending process of the aluminum alloy plate; and based on the processing defect and the reflection image, generating a visual result graph marked with a processing defect position and a highlight position at the same time so as to realize detection of the bending defect of the aluminum alloy plate. By means of the technical scheme, illumination interference can be separated from the source, and finally stable and reliable detection of the bending defects of the aluminum alloy plate is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, in particular to an intelligent detection method for aluminum alloy plate bending processing. BACKGROUND

[0002] Bending forming is one of the key processing links of aluminum alloy plate, wherein the bending R angle, i.e. the inner radius of the bending die, usually affects the shape and size after bending, so the bending R angle plays a decisive role in the final assembly precision and quality of the product. However, in actual production, such as natural wear of the stamping die, uneven thickness of the plate itself, and internal residual stress generated in the cold working process of the material, etc., all of which may cause various surface defects in the R angle area during forming.

[0003] At present, the light flow method and the texture analysis method are generally used in the prior art to monitor the defects randomly generated in the dynamic processing process in real time. Among them, the light flow method constructs a motion vector field by tracking the motion trajectory of the pixel points in the image sequence, so as to quantitatively analyze the deformation of the object surface; the texture analysis method describes the gray pattern change of the local area of the image by mathematical method, to identify whether the state of the material surface changes.

[0004] However, the common physical assumption of the light flow method and the texture analysis method is that the brightness is constant, that is, the brightness of the corresponding pixels in the images taken at different times should remain unchanged, or at most undergo weak linear change, for the same physical point on the observed object. However, the aluminum alloy plate, especially the plate after surface treatment such as wire drawing, polishing or anodic oxidation, will show strong mirror or semi-mirror reflection characteristics, under the action of the bending machine, the surface curvature and normal direction of the R angle area will change continuously and dramatically to form a highlight area, resulting in a high false alarm rate of the prior art. Therefore, how to effectively eliminate or suppress this dynamic highlight reflection interference under the condition of conventional industrial lighting, so that the light flow and texture analysis technology can be stably and reliably applied to the defect detection of high-reflective plate, is a technical problem to be solved. SUMMARY

[0005] In order to solve the technical problem of high light reflection interference in the bending detection of aluminum alloy plate, the present application provides an intelligent detection method for aluminum alloy plate bending processing, which can effectively filter out false defect signals caused by high light movement, and improve the accuracy of bending detection.

[0006] The present application provides an intelligent detection method for bending processing of aluminum alloy plates, and the prevention and control method includes: collecting images of bent aluminum alloy plates and preprocessing them to obtain a bending image sequence; performing image decomposition on each frame of the bending image sequence, separating each frame of the image into a base image containing material texture information and a reflection image containing highlight information to obtain a base image sequence; performing optical flow analysis and texture analysis on the base image sequence to identify processing defects generated by the aluminum alloy plates during the bending process; based on the processing defects and the reflection images, generating a visual result image that simultaneously marks the processing defect position and the highlight position to realize the detection of bending defects of the aluminum alloy plates.

[0007] In one embodiment, the image decomposition of each frame in the bent image sequence includes: calculating a basis confidence weight for each pixel in each frame in the bent image sequence, the basis confidence weight being used to represent the probability that the pixel belongs to a basis image; and obtaining a basis image through iterative optimization based on the basis confidence weight.

[0008] By calculating a base confidence for each pixel, the possibility that the pixel belongs to a stable material texture can be quantified, which can effectively distinguish the inherent texture of the material from transient highlight reflections, ensuring the accuracy of subsequent separation.

[0009] In one embodiment, the base confidence weight satisfies the relationship: ;in, is the basis confidence weight of pixel (x,y), V(x,y) and D(x,y) represent the window intrinsic variability and total window variability at pixel (x,y), respectively. V(x,y) and D(x,y) are standard components of the relative total variation filtering algorithm. V(x,y) is used to measure the magnitude of the gradient within the window, and D(x,y) is used to comprehensively measure the magnitude and directional consistency of the gradient within the window. is the control parameter of the Gaussian function.

[0010] In one embodiment, the optical flow analysis of the base image sequence includes: calculating an optical flow field based on the base image sequence; calculating a motion discontinuity index based on the optical flow field, wherein the motion discontinuity index is used to indicate the degree of abnormality of the optical flow field; and comparing the motion discontinuity index with a preset motion threshold to identify processing defects.

[0011] Calculating the optical flow field on a base image with high light interference removed can more accurately reflect the real physical deformation of the material and improve the accuracy and sensitivity of detection.

[0012] In one embodiment, the motion discontinuity indicator satisfies the relationship: ;in, The coordinates are The motion discontinuity of the pixel at the tth moment, and is the preset weight coefficient, Indicates the pixel at time t The divergence of Indicates the pixel at time t The curl of is the pixel at time t The optical flow vector, is the default derivative operator.

[0013] In one embodiment, the texture analysis includes: extracting texture features of each frame of the base image based on the base image sequence to obtain a texture feature sequence; calculating the difference between the texture features of two consecutive frames in the texture feature sequence to obtain a texture mutation index; and comparing the texture mutation index with a preset texture threshold to identify processing defects.

[0014] Analyzing texture changes on the substrate image can more realistically characterize the surface state of the material, effectively avoiding misjudging moving highlights as texture mutations, thereby improving the accuracy of surface defect detection.

[0015] In one embodiment, the texture mutation index satisfies the relationship: ;in, is the texture mutation index at time t, represents the chi-square distance used to measure the difference between two histograms, and Represent the local binary pattern feature maps of the base image at time t and time t-1 respectively, Represents a function that computes a feature histogram.

[0016] In one embodiment, the identification of processing defects generated in the bending process of aluminum alloy plates includes: weighted fusion of the motion discontinuity index obtained by optical flow analysis and the texture mutation index obtained by texture analysis to generate a defect confidence map; thresholding the defect confidence map to obtain a pure defect mask map that marks the location of the processing defect, and the pure defect mask map represents the processing defect.

[0017] By integrating these two indicators, we can not only determine whether there are defects, but also make a preliminary distinction between the types of defects by analyzing which indicator dominates the occurrence of defects, providing a reference for subsequent process improvements and fault tracing.

[0018] In one embodiment, the generation process of the visualization result map is as follows: the original image of the current frame is used as the background layer; the pure defect mask map corresponding to the processing defect is superimposed on the background layer as the defect layer; the reflection image is thresholded to obtain a highlight interference mask map, and the highlight interference mask map is superimposed on the background layer as the highlight verification layer to obtain a visualization result map.

[0019] By separating and displaying defects and highlights on the same original image, it can be visually demonstrated that the detected defects are not caused by highlight artifacts, thereby improving the credibility of the detection results.

[0020] The technical solution of this application has the following beneficial technical effects: This application decomposes the reflection and base images, transferring the analysis object from the original image that is severely interfered with by light to the base image that only reflects the information of the material itself, thereby ensuring the constant brightness premise on which the subsequent optical flow and texture analysis algorithms are based. Therefore, it can effectively filter out false defect signals caused by highlight movement, and realize stable and reliable detection of bending defects of aluminum alloy plates. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of an intelligent detection method for aluminum alloy plate bending processing according to an embodiment of the present application.

[0022] Figure 2 Schematic diagram of an aluminum alloy plate to be tested according to an embodiment of the present application.

[0023] Figure 3 Schematic diagram of a visualization result diagram according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0025] Figure 1 This is a flow chart of an intelligent detection method for aluminum alloy plate bending processing according to an embodiment of the present application. Figure 1 As shown, the intelligent detection method 100 for aluminum alloy plate bending processing includes steps S101 to S104, which are described in detail below.

[0026] S101: Collect and pre-process images of the bent aluminum alloy sheet to obtain a bending image sequence.

[0027] In one embodiment, an industrial camera can be deployed to photograph the aluminum alloy sheet being bent. It is worth noting that to address the large variations in light intensity in a workshop environment, a high dynamic range (HDR) camera is preferred. This ensures that details in both the highlight areas and the dark areas of the sheet can be clearly captured in a single exposure, preventing information loss.

[0028] Furthermore, the camera can continuously capture the bending processing images of the aluminum alloy plate at a constant frame rate, for example, 200 frames per second, such as Figure 2 , which is a schematic diagram of the aluminum alloy plate to be tested collected in the embodiment of the present application.

[0029] Because the captured raw images may contain random noise generated by the camera sensor, preprocessing the captured images of the bent aluminum alloy sheet is necessary to reduce the impact of this noise on subsequent calculations. For example, this preprocessing method can be to smooth each image frame using a Gaussian filter with a kernel size of 3x3. This effectively suppresses Gaussian noise while effectively preserving image edges and detail. In this application, all preprocessed images are arranged in chronological order of capture to form a bent image sequence.

[0030] In this way, by deploying appropriate cameras and performing preprocessing, high-quality, low-noise bending image sequences can be obtained.

[0031] S102 , performing image decomposition on each frame of the bent image sequence, separating each frame of the image into a base image containing material texture information and a reflection image containing highlight information, to obtain a base image sequence.

[0032] In one embodiment, the relative total variation (RTV) filtering technique can be used to decompose each frame in a warped image sequence. RTV is a structure-preserving image smoothing filter that differs fundamentally from traditional Gaussian and median filters. Traditional filters struggle to distinguish between structural details that should be preserved, such as object edges or material texture, and non-structural details that should be removed, such as noise or transient highlights. While the inherent texture of a material appears locally as high-frequency detail, its local gradient is typically coherent. Meanwhile, the edges of specular reflections, while also high-frequency and strong, are relatively isolated in appearance and spatial distribution, and their gradient direction lacks consistency with that of the surrounding texture. The RTV algorithm exploits the statistical differences in local gradient magnitude and direction to identify and separate these two distinct types of details.

[0033] Specifically, for each frame in the warped image sequence, the intrinsic variability V(x,y) and total variability D(x,y) of each pixel (x,y) in the image within its neighborhood window are calculated. For example, the window can be 7x7. V(x,y) and D(x,y) are standard components of the RTV algorithm. V(x,y) primarily measures the magnitude of the gradient within the window, while D(x,y) comprehensively measures the magnitude and directional consistency of the gradient.

[0034] Furthermore, a base confidence weight can be constructed to guide the work of the filter. Its function is to assign a weight to each pixel. The weight represents the possibility that the pixel belongs to the stable base layer texture that should be retained. The base confidence weight satisfies the relationship:

[0035] in, is the base confidence weight of pixel (x, y), ranging from 0 to 1. The closer the value is to 1, the more likely the pixel belongs to the base layer. V(x, y) and D(x, y) represent the window intrinsic variability and total window variability at pixel (x, y), respectively. is the control parameter of the Gaussian function.

[0036] Furthermore, the gradient inconsistency of the highlight edge is high and should be given a low weight; the gradient consistency of the material texture is high and should be given a high weight; and The ratio is close to 1 at isolated strong edge locations and is much smaller than 1 in areas with coherent texture or smooth grayscale. It is a hyperparameter used to control the width of the Gaussian function, which determines the sensitivity of the weight function to gradient inconsistency. For example, the value can be 0.3.

[0037] Then, after calculating the basis confidence weights for all pixels, a basis confidence weight map can be obtained, which can be applied to an iterative optimization solver, such as weighted least squares. The goal of the solution process is to find a smooth image such that The difference between high-weighted areas and the original image is minimized, while larger differences are allowed in low-weighted areas, i.e., strong smoothing is performed. In this way, through iterative optimization, the highlight reflections in the original image are effectively smoothed and suppressed, ultimately obtaining a base image that only contains the stable texture information of the plate. Furthermore, by subtracting the base image from the original image, a reflection image containing all the separated highlight information is obtained, ultimately obtaining a base image sequence.

[0038] In this way, through image decomposition based on relative total variation, dynamic highlights can be removed from the original image, providing a pure data source without interference from lighting for subsequent precise defect analysis.

[0039] S103, performing optical flow analysis and texture analysis on the base image sequence to identify processing defects generated during the bending process of the aluminum alloy plate.

[0040] In one embodiment, a dense optical flow algorithm can be applied to the base image sequence to calculate the optical flow field (also called the velocity field) between adjacent frames. Since the base image sequence used as input already satisfies the basic assumption of constant brightness, the calculated optical flow field can accurately reflect the actual physical motion speed and direction of each point on the plate surface.

[0041] Furthermore, in order to quantify defects from the obtained optical flow field, a motion discontinuity index can be constructed to capture abnormal patterns in the optical flow field. These patterns are usually directly related to physical discontinuities of the material, such as cracks and tears. The motion discontinuity index satisfies the relationship:

[0042] in, The coordinates are The motion discontinuity of the pixel at the tth moment, and is the preset weight coefficient, Indicates the pixel at time t The divergence of Indicates the pixel at time t The curl of is the pixel at time t The optical flow vector, is the default derivative operator.

[0043] Further, It can be Nabla operator (Naples operator) or Del operator, As a mathematical operator, it represents the derivative operation in all directions of space. For example, in a two-dimensional In space, it can be expressed as a vector form:

[0044] in, represents the rate of change in the x direction, Indicates the rate of change in the y direction.

[0045] It can be understood that the divergence describes the expansion or compression strength of the velocity field. The opening of a microcrack will produce a large positive divergence value. The absolute value is used in this scheme because we only need to care about the severity of the expansion or compression of the velocity field without considering its direction; the curl measures the local rotation strength of a point. Non-zero curl indicates that the flow is rotational, which corresponds to the occurrence of shear or tearing, that is, adjacent material layers are offset from each other. Zero curl indicates that the flow is non-rotational, which corresponds to the smooth, tear-free stretching or compression of the workpiece.

[0046] Further, and The weight coefficients can all be set to 0.5. During normal, uniform material deformation, the divergence and rotation of the optical flow field are very close to zero. Any processing defect that causes physical discontinuity in the material will cause the divergence or rotation to increase sharply at the defect location. Therefore, when the MDI value exceeds the preset motion threshold, it can be considered that a processing defect exists at that point. For example, the motion threshold can be set to 0.05.

[0047] In an optional embodiment, on the same base image sequence, the local binary pattern (LBP) feature of each pixel can be calculated to obtain the texture feature of each frame of the base image, where LBP is an efficient texture description operator that encodes local texture patterns by comparing the grayscale relationship between a pixel and its neighboring pixels.

[0048] In order to characterize abnormal changes in surface conditions, such as wrinkles and other defects, the difference between the texture features of two consecutive frames in the texture feature sequence can be calculated to obtain the texture mutation index, which is used to measure the intensity of the texture change in the entire R angle area between consecutive time frames. The texture mutation index satisfies the relationship:

[0049] in, is the texture mutation index at time t, Represents the chi-square distance used to measure the difference between two histograms, which is used to measure the difference between two histograms. and Represent the local binary pattern feature maps of the base image at time t and time t-1 respectively, Represents a function that computes a feature histogram.

[0050] During a normal bending process, surface texture changes gradually, with the LBP histograms of two consecutive frames being very similar and the TAI value being very small. However, when a processing defect, such as wrinkling, suddenly occurs on the surface, a new LBP pattern is introduced, causing a dramatic change in the histogram and a noticeable spike in the TAI value. Therefore, when the TAI value exceeds a preset texture threshold, it is considered a processing defect. For example, the texture threshold can be 2.0.

[0051] Furthermore, by fusing the analysis results of MDI and TAI, for example, by thresholding and normalizing them separately and then taking the union, or performing weighted fusion after normalization and then thresholding, a pure defect mask map that comprehensively marks the locations of all identified processing defects can be obtained.

[0052] In this way, by calculating the two indicators MDI and TAI in parallel, the deformation state of the sheet can be comprehensively monitored from two orthogonal dimensions: motion and texture, thereby reliably identifying various processing defects.

[0053] S104: Based on the processing defects and the reflection image, a visualization result image is generated, which simultaneously marks the processing defect position and the highlight position, so as to realize the detection of bending defects of the aluminum alloy plate.

[0054] In one embodiment, pixels with a value of 1 in the pure defect mask are the locations of identified processing defects. Because the reflected image captures all dynamic highlight information in the original image, a brightness threshold is applied to the reflected image. For example, by extracting the pixel area with brightness values ​​in the top 5%, a binary highlight interference mask is generated. Pixels with a value of 1 in this mask correspond to the locations of dynamic highlights in the original image.

[0055] Finally, a visualization result map with both the processing defect position and the highlight position marked can be generated to integrate all key information. The generation process is as follows: the original image of the current frame is used as the background layer; the pure defect mask map corresponding to the processing defect is superimposed on the background layer as the defect layer; the highlight interference mask map is superimposed on the background layer as the highlight verification layer to obtain the visualization result map. Figure 3 , which is a schematic diagram of a visualization result diagram according to an embodiment of the present application, wherein the highlight area is light gray, the shadow area is dark gray, and the potential defect area is pure white.

[0056] In this way, by generating visual results, it not only provides workers with clear defect positioning, but also effectively enhances the reliability of the detection system.

[0057] It should be noted that a person skilled in the art may make a number of modifications and improvements without departing from the concept of the present application, and these modifications and improvements are all within the scope of protection of the present application. Therefore, the scope of protection of the patent application shall be based on the appended claims.

Claims

1. An intelligent detection method for aluminum alloy plate bending processing, characterized in that: include: Collecting and preprocessing images of bent aluminum alloy sheets to obtain bending image sequences; performing image decomposition on each frame of the bent image sequence, separating each frame of the image into a base image containing material texture information and a reflection image containing highlight information, to obtain a base image sequence; performing optical flow analysis and texture analysis on the base image sequence to identify processing defects generated during the bending process of the aluminum alloy plate; Based on the processing defect and the reflected image, a visualization result image is generated that simultaneously marks the processing defect position and the highlight position, so as to realize the detection of bending defects of the aluminum alloy plate.

2. The intelligent detection method for aluminum alloy plate bending according to claim 1, characterized in that: The performing image decomposition on each frame of the bent image sequence includes: Calculating a basis confidence weight for each pixel in each frame of the warped image sequence, wherein the basis confidence weight is used to represent the probability that the pixel belongs to the basis image; Based on the base confidence weight, a base image is obtained through iterative optimization.

3. The intelligent detection method for aluminum alloy plate bending processing according to claim 2, characterized in that: The base confidence weight satisfies the relationship: in, is the basis confidence weight of pixel (x, y), V(x, y) and D(x, y) represent the window intrinsic variability and total variability of the window at pixel (x, y), respectively. V(x, y) and D(x, y) are standard components of the relative total variation filtering algorithm. V(x, y) is used to measure the magnitude of the gradient within the window, and D(x, y) is used to comprehensively measure the magnitude and direction consistency of the gradient within the window. is the control parameter of the Gaussian function.

4. The intelligent detection method for aluminum alloy plate bending according to claim 1, characterized in that: The performing optical flow analysis on the base image sequence includes: Calculating an optical flow field based on the base image sequence; Calculating a motion discontinuity index based on the optical flow field, wherein the motion discontinuity index is used to indicate a degree of abnormality of the optical flow field; The motion discontinuity indicator is compared with a preset motion threshold to identify processing defects.

5. The intelligent detection method for aluminum alloy plate bending processing according to claim 4, characterized in that: The motion discontinuity index satisfies the relationship: in, The coordinates are The motion discontinuity of the pixel at the tth moment, and is the preset weight coefficient, Indicates the pixel at time t The divergence of Indicates the pixel at time t The curl of is the pixel at time t The optical flow vector, is the default derivative operator.

6. The intelligent detection method for aluminum alloy plate bending according to claim 1, characterized in that: The texture analysis includes: Extracting texture features of each frame of the base image based on the base image sequence to obtain a texture feature sequence; Calculating the difference between the texture features of two consecutive frames in the texture feature sequence to obtain a texture mutation index; The texture mutation index is compared with a preset texture threshold to identify processing defects.

7. The intelligent detection method for aluminum alloy plate bending processing according to claim 6, characterized in that: The texture mutation index satisfies the relationship: in, is the texture mutation index at time t, represents the chi-square distance used to measure the difference between two histograms, and Represent the local binary pattern feature maps of the base image at time t and time t-1 respectively, Represents a function that computes a feature histogram.

8. The intelligent detection method for aluminum alloy plate bending according to claim 1, characterized in that: The identification of processing defects generated during the bending process of aluminum alloy plates includes: The motion discontinuity index obtained by optical flow analysis and the texture mutation index obtained by texture analysis are weightedly fused to generate a defect confidence map; Thresholding processing is performed on the defect confidence map to obtain a pure defect mask map marking the processing defect position, wherein the pure defect mask map represents the processing defect.

9. The intelligent detection method for aluminum alloy plate bending according to claim 1, characterized in that: The generation process of the visualization result graph is as follows: Use the original image of the current frame as the background layer; Overlaying a pure defect mask image corresponding to the processing defect on the background layer as a defect layer; Thresholding processing is performed on the reflection image to obtain a highlight interference mask image, and the highlight interference mask image is superimposed on the background layer as a highlight verification layer to obtain a visualization result image.

Images