Aluminum foil surface defect detection system based on visual recognition
By combining the data acquisition, processing, and defect identification modules of the visual recognition system with various environmental data matching and Hough transform, the environmental impact problem in aluminum foil surface defect detection has been solved, achieving high accuracy and reliability in detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SHAANXI CHUCHUANG ZESHENG NEW MATERIAL CO LTD
- Filing Date
- 2026-02-05
- Publication Date
- 2026-05-15
AI Technical Summary
Microscopic and macroscopic defects are easily generated on the surface of aluminum foil during high-speed rolling or slitting, and the high reflectivity makes the imaging quality susceptible to environmental influences. Existing detection systems lack sufficient accuracy and reliability.
A visual recognition-based aluminum foil surface defect detection system is adopted, including data acquisition, processing, layer optimization, and defect recognition modules. By setting multiple standard environmental data, dynamic matching and layer fusion, combined with fusion center point alignment, reference point matching and Hough transform method to optimize the boundary, adaptive image analysis is achieved.
It significantly improves the accuracy and reliability of testing in variable industrial environments, accurately determines whether the size and shape of aluminum foil meet the standards, and reduces the impact of changes in ambient light.
Smart Images
Figure CN121633124B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of aluminum foil inspection technology, specifically to a visual recognition-based aluminum foil surface defect detection system. Background Technology
[0002] Aluminum foil, as an important industrial material, is widely used in packaging, electronics, power, and aerospace fields due to its excellent barrier properties, conductivity, and ductility. Its surface quality is a key indicator determining product performance and reliability. However, during the high-speed rolling or slitting process of aluminum foil, due to factors such as process fluctuations, environmental interference, and the material itself, the surface is prone to various microscopic and macroscopic defects such as pinholes, scratches, oil stains, bright spots, dark spots, and wrinkles.
[0003] In existing technologies, the high reflectivity of aluminum foil surfaces makes image quality highly susceptible to environmental influences. Therefore, a visual recognition-based aluminum foil surface defect detection system is provided. Summary of the Invention
[0004] The purpose of this invention is to provide a visual recognition-based aluminum foil surface defect detection system.
[0005] The objective of this invention can be achieved through the following technical solution: a visual recognition-based aluminum foil surface defect detection system, including a monitoring center, and further including components that communicate with and / or are electrically connected to the monitoring center.
[0006] The data acquisition module is used to acquire image data and environmental data within the detection area;
[0007] The data processing module is used to process the acquired image data to obtain a region fusion layer;
[0008] The layer optimization module is used to optimize the layer boundaries of the obtained region blending layer and split the region blending layer after boundary optimization to obtain the aluminum foil feature layer.
[0009] The defect identification module is used to perform surface defect analysis on the feature layer of aluminum foil and output the surface defect analysis results.
[0010] Furthermore, the process by which the data acquisition module acquires image data and environmental data within the detection area includes:
[0011] The data acquisition module includes a visual acquisition unit and an environmental acquisition unit;
[0012] Set up a detection area, and deploy visual acquisition units at designated locations within the detection area. The shooting range of the deployed visual acquisition units completely covers the detection area.
[0013] The deployed visual acquisition units acquire image data within the detection area in real time.
[0014] Each corner of the detection area is equipped with a fixed marker point, and a corresponding environmental acquisition unit is deployed at the marker point. The environmental data of the corresponding location is acquired in real time through the deployed environmental acquisition unit. The environmental data includes light intensity and light angle.
[0015] Furthermore, before formally conducting aluminum foil defect detection, different standard environmental data are set for the detection area. Under different standard environmental data, image data is collected from the detection area where no aluminum foil is placed through the vision acquisition unit.
[0016] Image data collected under different standard environmental data are recorded as the reference image layer of the corresponding standard environmental data, and a reference center and several reference points distributed in a circle with the reference center as the center are set in the reference image layer.
[0017] Then, place the qualified aluminum foil in the detection area, and collect image data of the detection area through the vision acquisition unit;
[0018] Image data collected under different standard environmental data are recorded as the corresponding standard environmental data reference image layer. A reference center and several reference points are set in the reference image layer, and the reference center and several reference points are all within the coverage area of the aluminum foil.
[0019] Furthermore, the data processing module processes the acquired image data to obtain the region fusion layer, including the following steps:
[0020] The obtained environmental data is matched with standard environmental data, and the baseline image layer and control image layer corresponding to the standard environmental data that is closest to the current environmental data are called.
[0021] Construct a planar coordinate system and map the reference image layer and the comparison image layer to the planar coordinate system in sequence;
[0022] Align the reference center of the reference image layer with the reference center of the control image layer, and each reference point corresponds to a control point to obtain a reference fusion layer. Record the positions of the corresponding reference center and control center as fusion center points, and the positions of the reference points and control points as fusion reference points.
[0023] The obtained image data is used as the layer to be identified and mapped into a planar coordinate system;
[0024] Align the center point of the layer to be identified with the fusion center point, and generate corresponding marker points based on the center point of the layer to be identified according to the distribution of the fusion reference points;
[0025] After aligning each marker point with a blending reference point, a region blending layer is obtained.
[0026] Furthermore, the layer optimization module optimizes the layer boundaries of the obtained region blending layer and splits the boundary-optimized region blending layer to obtain the aluminum foil feature layer. The process includes:
[0027] The layers to be identified, the reference image layer, and the comparison image layer in the region fusion layer are rasterized to obtain the corresponding grayscale images;
[0028] The obtained layer to be identified is binarized, and the aluminum foil area is labeled according to the binarization result.
[0029] Using the boundary of the aluminum foil coverage area in the reference image layer as a benchmark, the boundary of the marked aluminum foil area is optimized by the Hough transform method.
[0030] Image features are extracted from the grayscale image corresponding to the optimized aluminum foil area to obtain the corresponding feature values. The aluminum foil area after image feature extraction is then split to obtain the corresponding aluminum foil feature layer.
[0031] Furthermore, using the boundary of the aluminum foil coverage area in the reference image layer as a benchmark, the process of optimizing the boundary of the marked aluminum foil area using the Hough transform method includes:
[0032] Overlay the aluminum foil area of the layer to be identified with the aluminum foil area of the reference image layer;
[0033] Obtain the overlap area of the aluminum foil region in the layer to be identified and the reference image layer;
[0034] The layer to be identified is rotated in rotation so that each rotation aligns the marker points in the layer to be identified with a new fusion reference point, until all fusion reference points are traversed;
[0035] The state with the maximum overlap area corresponding to each rotation is recorded as the state to be optimized;
[0036] Continue rotating the layer to be recognized in both directions while it is in the state of being optimized;
[0037] Obtain the trend of overlapping area changes after rotation on both sides. If the overlapping area decreases after rotation on both sides, mark the state to be optimized as the final optimized state.
[0038] If the overlapping area increases, continue rotating until the overlapping area begins to decrease. Then, record the state with the largest overlapping area as the final optimized state.
[0039] The boundary of the reference image layer in the final optimized state is used as the reference for the boundary of the layer to be identified, and the Hough transform method is used to optimize the boundary of the layer to be identified.
[0040] Furthermore, the defect identification module performs surface defect analysis on the aluminum foil feature layer and outputs the surface defect analysis results, including:
[0041] The area of the layer to be identified after boundary optimization is compared with the coverage area and boundary shape of the aluminum foil in the reference image layer. If they are consistent, it means that the size and shape of the aluminum foil corresponding to the layer to be identified are qualified; otherwise, it means that it is unqualified.
[0042] Based on the premise that the size and shape are qualified, the layer to be identified is divided into several unit areas;
[0043] The extracted feature values in each unit region are normalized to obtain the normalized feature values corresponding to that unit region.
[0044] Similarly, the normalized feature values of the corresponding regions of the comparison image layer are obtained, and the two sets of normalized feature values are compared to obtain the feature difference.
[0045] The obtained feature difference is compared with a preset threshold range. If the feature difference is within the preset threshold range, it means that there is no defect in the corresponding unit area; otherwise, the corresponding unit area is marked as an abnormal area.
[0046] If there are adjacent abnormal regions, the corresponding abnormal regions will be merged. The region formed by merging at least two abnormal regions will be called the extended abnormal region.
[0047] Furthermore, the process of normalizing the feature values extracted from each unit region includes:
[0048] The maximum, minimum, and mean values of the feature values corresponding to the identified features within a unit region are obtained and denoted as follows: , as well as ;
[0049] The characteristic rate of change of the unit region is obtained, denoted as Tb, where:
[0050] ;
[0051] Set the threshold range for the rate of change to [1, a];
[0052] When Tb∈[1, a], then As the normalized characteristic value of this unit region;
[0053] When Tb [1, a], then if - > - Then As the normalized characteristic value of this unit region, if - < - Then As the normalized characteristic value of this unit region, if - = - Then , It also serves as the normalized characteristic value of this unit region.
[0054] Compared with the prior art, the beneficial effects of the present invention are:
[0055] 1. By setting up a variety of standard environmental data and performing dynamic matching and layer fusion during detection, an adaptive image analysis benchmark is constructed, thereby reducing the fluctuation of imaging quality caused by changes in ambient lighting and significantly improving the detection accuracy and reliability in variable industrial environments.
[0056] 2. By integrating center point alignment, reference point matching, and boundary optimization algorithms based on Hough transform and maximizing overlapping area, the aluminum foil contour within the detection area can be extracted, thereby determining whether the size and shape of the aluminum foil meet the standards. Attached Figure Description
[0057] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.
[0058] Figure 1 This is a schematic diagram of the present invention. Detailed Implementation
[0059] like Figure 1 As shown, the vision-based aluminum foil surface defect detection system includes a monitoring center and components that communicate and / or are electrically connected to the monitoring center.
[0060] The data acquisition module is used to acquire image data and environmental data within the detection area;
[0061] The data processing module is used to process the acquired image data to obtain a region fusion layer;
[0062] The layer optimization module is used to optimize the layer boundaries of the obtained region blending layer and split the region blending layer after boundary optimization to obtain the aluminum foil feature layer.
[0063] The defect identification module is used to perform surface defect analysis on the feature layer of aluminum foil and output the surface defect analysis results.
[0064] It should be further explained that, in the specific implementation process, the process by which the data acquisition module acquires image data and environmental data within the detection area includes:
[0065] The data acquisition module includes a visual acquisition unit and an environmental acquisition unit;
[0066] Set up a detection area, and deploy visual acquisition units at designated locations within the detection area. The shooting range of the deployed visual acquisition units completely covers the detection area.
[0067] The deployed visual acquisition units acquire image data within the detection area in real time.
[0068] Each corner of the detection area is equipped with a fixed marker point, and a corresponding environmental acquisition unit is deployed at the marker point. The environmental data of the corresponding location is acquired in real time through the deployed environmental acquisition unit. The environmental data includes light intensity and light angle.
[0069] In the specific implementation process, before the formal aluminum foil defect detection is carried out, different standard environmental data are set for the detection area. Under different standard environmental data, the visual acquisition unit collects image data of the detection area where no aluminum foil is placed.
[0070] Image data collected under different standard environmental data are recorded as the reference image layer of the corresponding standard environmental data, and a reference center and several reference points distributed in a circle with the reference center as the center are set in the reference image layer.
[0071] Then, place the qualified aluminum foil in the detection area, and collect image data of the detection area through the vision acquisition unit;
[0072] Image data collected under different standard environmental conditions are recorded as the corresponding standard environmental data control image layer. A control center and several control points are set within the control image layer. It should be noted that the control center and several control points are all within the coverage area of the aluminum foil.
[0073] It should be further explained that, in the specific implementation process, the data processing module processes the acquired image data to obtain the region fusion layer, including the following steps:
[0074] The obtained environmental data is matched with standard environmental data, and the baseline image layer and control image layer corresponding to the standard environmental data that is closest to the current environmental data are called.
[0075] Construct a planar coordinate system and map the reference image layer and the comparison image layer to the planar coordinate system in sequence;
[0076] Align the reference center of the reference image layer with the reference center of the control image layer, and each reference point corresponds to a control point to obtain a reference fusion layer. Record the positions of the corresponding reference center and control center as fusion center points, and the positions of the reference points and control points as fusion reference points.
[0077] The obtained image data is used as the layer to be identified and mapped into a planar coordinate system;
[0078] Align the center point of the layer to be identified with the fusion center point, and generate corresponding marker points based on the center point of the layer to be identified according to the distribution of the fusion reference points;
[0079] After aligning each marker point with a blending reference point, a region blending layer is obtained.
[0080] It should be further explained that, in the specific implementation process, the layer optimization module optimizes the layer boundaries of the obtained region blending layer, and then splits the boundary-optimized region blending layer to obtain the aluminum foil feature layer. The process includes:
[0081] The layers to be identified, the reference image layer, and the comparison image layer in the region fusion layer are rasterized to obtain the corresponding grayscale images;
[0082] The obtained layer to be identified is binarized, and the aluminum foil area is labeled according to the binarization result.
[0083] Using the boundary of the aluminum foil coverage area in the reference image layer as a benchmark, the boundary of the marked aluminum foil area is optimized by the Hough transform method.
[0084] Image features are extracted from the grayscale image corresponding to the optimized aluminum foil area to obtain the corresponding feature values. The aluminum foil area after image feature extraction is then split to obtain the corresponding aluminum foil feature layer. It should be further noted that image feature extraction of grayscale images is a common technique used by those skilled in the art, and will not be elaborated here. In the specific implementation of this invention, the MobileNet algorithm is used, but those skilled in the art can also use other feasible algorithms according to the actual situation.
[0085] It should be further explained that, in the specific implementation process, the process of optimizing the boundary of the marked aluminum foil area using the Hough transform method, based on the boundary of the aluminum foil coverage area in the reference image layer, includes:
[0086] Overlay the aluminum foil area of the layer to be identified with the aluminum foil area of the reference image layer;
[0087] Obtain the overlap area of the aluminum foil region in the layer to be identified and the reference image layer;
[0088] The layer to be identified is rotated in rotation so that each rotation aligns the marker points in the layer to be identified with a new fusion reference point, until all fusion reference points are traversed;
[0089] The state with the maximum overlap area corresponding to each rotation is recorded as the state to be optimized;
[0090] While the layer to be recognized is in the unoptimized state, continue to rotate it in both directions; it should be noted that rotating in both directions refers to rotating the layer to be recognized clockwise and counterclockwise.
[0091] Obtain the trend of overlapping area changes after rotation on both sides. If the overlapping area decreases after rotation on both sides, mark the state to be optimized as the final optimized state.
[0092] If the overlapping area increases, continue rotating until the overlapping area begins to decrease. Then, record the state with the largest overlapping area as the final optimized state.
[0093] The boundary of the reference image layer in the final optimized state is used as the reference for the boundary of the layer to be identified, and the Hough transform method is used to optimize the boundary of the layer to be identified.
[0094] It should be further explained that, in the specific implementation process, the defect identification module performs surface defect analysis on the aluminum foil feature layer and outputs the surface defect analysis results, including:
[0095] The area of the layer to be identified after boundary optimization is compared with the coverage area and boundary shape of the aluminum foil in the reference image layer. If they are consistent, it means that the size and shape of the aluminum foil corresponding to the layer to be identified are qualified; otherwise, it means that it is unqualified.
[0096] Based on the premise that the size and shape are qualified, the layer to be identified is divided into several unit areas;
[0097] The extracted feature values in each unit region are normalized to obtain the normalized feature values corresponding to that unit region.
[0098] Similarly, the normalized feature values of the corresponding regions of the comparison image layer are obtained, and the two sets of normalized feature values are compared to obtain the feature difference.
[0099] The obtained feature difference is compared with a preset threshold range. If the feature difference is within the preset threshold range, it means that there is no defect in the corresponding unit area; otherwise, the corresponding unit area is marked as an abnormal area.
[0100] If there are adjacent abnormal regions, the corresponding abnormal regions will be merged, and the region formed by merging at least two abnormal regions will be called the extended abnormal region.
[0101] The image portions corresponding to the abnormal area and the extended abnormal area are bounded and selected, then enlarged and uploaded to the monitoring center for technicians to confirm the specific defect type.
[0102] It should be further explained that, in the specific implementation process, the normalization process for the feature values extracted from each unit region includes:
[0103] The maximum, minimum, and mean values of the feature values corresponding to the identified features within a unit region are obtained and denoted as follows: , as well as ;
[0104] The characteristic rate of change of the unit region is obtained, denoted as Tb, where:
[0105] ;
[0106] Set the threshold range for the rate of change to [1, a];
[0107] When Tb∈[1, a], then As the normalized characteristic value of this unit region;
[0108] When Tb [1, a], then if - > - Then As the normalized characteristic value of this unit region, if - < - Then As the normalized characteristic value of this unit region, if - = - Then , It also serves as the normalized characteristic value of this unit region.
[0109] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications or equivalent substitutions made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.
Claims
1. A visual recognition-based aluminum foil surface defect detection system, characterized in that, This includes the monitoring center, as well as components that communicate and / or are electrically connected to the monitoring center. The data acquisition module is used to acquire image data and environmental data within the detection area; The data processing module is used to process the acquired image data to obtain a region fusion layer; The layer optimization module is used to optimize the layer boundaries of the obtained region blending layer and split the region blending layer after boundary optimization to obtain the aluminum foil feature layer. The defect identification module is used to perform surface defect analysis on the feature layer of aluminum foil and output the surface defect analysis results. The data processing module processes the acquired image data to obtain the region fusion layer. The process includes: The obtained environmental data is matched with standard environmental data, and the baseline image layer and control image layer corresponding to the standard environmental data that is closest to the current environmental data are called. Construct a planar coordinate system and map the reference image layer and the comparison image layer to the planar coordinate system in sequence; Align the reference center of the reference image layer with the reference center of the control image layer, and each reference point corresponds to a control point to obtain a reference fusion layer. Record the positions of the corresponding reference center and control center as fusion center points, and the positions of the reference points and control points as fusion reference points. The obtained image data is used as the layer to be identified and mapped into a planar coordinate system; Align the center point of the layer to be identified with the fusion center point, and generate corresponding marker points based on the center point of the layer to be identified according to the distribution of the fusion reference points; After aligning each marker point with a blending reference point, a region blending layer is obtained; The layer optimization module optimizes the layer boundaries of the obtained region blending layer and then splits the region blending layer after boundary optimization to obtain the aluminum foil feature layer. The process includes: The layers to be identified, the reference image layer, and the comparison image layer in the region fusion layer are rasterized to obtain the corresponding grayscale images; The obtained layer to be identified is binarized, and the aluminum foil area is labeled according to the binarization result. Using the boundary of the aluminum foil coverage area in the reference image layer as a benchmark, the boundary of the marked aluminum foil area is optimized by the Hough transform method. Image features are extracted from the grayscale image corresponding to the optimized aluminum foil area to obtain the corresponding feature values. The aluminum foil area after image feature extraction is then split to obtain the corresponding aluminum foil feature layer.
2. The visual recognition-based aluminum foil surface defect detection system according to claim 1, characterized in that, The process by which the data acquisition module acquires image data and environmental data within the detection area includes: The data acquisition module includes a visual acquisition unit and an environmental acquisition unit; Set up a detection area, and deploy visual acquisition units at designated locations within the detection area. The shooting range of the deployed visual acquisition units completely covers the detection area. The deployed visual acquisition units acquire image data within the detection area in real time. Each corner of the detection area is equipped with a fixed marker point, and a corresponding environmental acquisition unit is deployed at the marker point. The environmental data of the corresponding location is acquired in real time through the deployed environmental acquisition unit. The environmental data includes light intensity and light angle.
3. The visual recognition-based aluminum foil surface defect detection system according to claim 2, characterized in that, Before formally conducting aluminum foil defect detection, different standard environmental data are set for the detection area. Under different standard environmental data, image data is collected from the detection area where no aluminum foil is placed through the vision acquisition unit. Image data collected under different standard environmental data are recorded as the reference image layer of the corresponding standard environmental data, and a reference center and several reference points distributed in a circle with the reference center as the center are set in the reference image layer. Then, place the qualified aluminum foil in the detection area, and collect image data of the detection area through the vision acquisition unit; Image data collected under different standard environmental data are recorded as the corresponding standard environmental data reference image layer. A reference center and several reference points are set in the reference image layer, and the reference center and several reference points are all within the coverage area of the aluminum foil.
4. The visual recognition-based aluminum foil surface defect detection system according to claim 3, characterized in that, The process of optimizing the boundary of the marked aluminum foil region using the Hough transform method, based on the boundary of the aluminum foil coverage area in the reference image layer, includes: Overlay the aluminum foil area of the layer to be identified with the aluminum foil area of the reference image layer; Obtain the overlap area of the aluminum foil region in the layer to be identified and the reference image layer; The layer to be identified is rotated in rotation so that each rotation aligns the marker points in the layer to be identified with a new fusion reference point, until all fusion reference points are traversed; The state with the maximum overlap area corresponding to each rotation is recorded as the state to be optimized; Continue rotating the layer to be recognized in both directions while it is in the state of being optimized; Obtain the trend of overlapping area changes after rotation on both sides. If the overlapping area decreases after rotation on both sides, mark the state to be optimized as the final optimized state. If the overlapping area increases, continue rotating until the overlapping area begins to decrease. Then, record the state with the largest overlapping area as the final optimized state. The boundary of the reference image layer in the final optimized state is used as the reference for the boundary of the layer to be identified, and the Hough transform method is used to optimize the boundary of the layer to be identified.
5. The visual recognition-based aluminum foil surface defect detection system according to claim 4, characterized in that, The process by which the defect identification module performs surface defect analysis on the aluminum foil feature layer and outputs the surface defect analysis results includes: The area of the layer to be identified after boundary optimization is compared with the coverage area and boundary shape of the aluminum foil in the reference image layer. If they are consistent, it means that the size and shape of the aluminum foil corresponding to the layer to be identified are qualified; otherwise, it means that it is unqualified. Based on the premise that the size and shape are qualified, the layer to be identified is divided into several unit areas; The extracted feature values in each unit region are normalized to obtain the normalized feature values corresponding to that unit region. Similarly, the normalized feature values of the corresponding regions of the comparison image layer are obtained, and the two sets of normalized feature values are compared to obtain the feature difference. The obtained feature difference is compared with a preset threshold range. If the feature difference is within the preset threshold range, it means that there is no defect in the corresponding unit area; otherwise, the corresponding unit area is marked as an abnormal area. If there are adjacent abnormal regions, the corresponding abnormal regions will be merged. The region formed by merging at least two abnormal regions will be called the extended abnormal region.
6. The visual recognition-based aluminum foil surface defect detection system according to claim 5, characterized in that, The process of normalizing the feature values extracted from each unit region includes: The maximum, minimum, and mean values of the feature values corresponding to the identified features within a unit region are obtained and denoted as follows: , as well as ; The characteristic rate of change of the unit region is obtained, denoted as Tb, where: ； Set the threshold range for the rate of change to [1, a]; When Tb∈[1, a], then As the normalized characteristic value of this unit region; When Tb [1, a], then if - > - Then As the normalized characteristic value of this unit region, if - < - Then As the normalized characteristic value of this unit region, if - = - Then , It also serves as the normalized characteristic value of this unit region.