Scanning electron microscope image contour extraction method based on morphological processing
Through the morphological processing method, the problem of incomplete, fractured or blurred contour extraction in scanning electron microscope images is solved, efficient and accurate contour extraction is achieved, and image processing accuracy and efficiency in integrated circuit manufacturing is improved.
Patent Information
- Application Number
- CN202411547168.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-01
- Publication Date
- 2024-12-03
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to accurately extract contour information from scanning electron microscope images, especially in integrated circuit manufacturing, which is difficult to meet the high precision requirements due to background noise, imaging conditions limitations and sample characteristics.
Using morphological processing methods, including image gradient extraction, expansion, corrosion, region growth algorithm and user interaction adjustment, combined with Gaussian filtering, non-maximum suppression and area threshold screening, gradually improving the contour continuity and integrity, output A clear final profile.
It realizes efficient and accurate extraction of scanning electron microscope image profiles, improves the accuracy and efficiency of contour extraction, enhances the continuity and integrity of contours, and provides support for process analysis of micro-nano structures such as integrated circuits.
Smart Images

Figure CN119068006A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of image processing technology, in particular to the subdivision field of scanning electron microscope (SEM) image processing, and specifically to a SEM image contour extraction method based on morphological processing. Background Art
[0002] With the rapid development of semiconductor technology, the integration of integrated circuits has been continuously improved, and the complexity of their internal structures has also increased. As an important tool for observing and analyzing the microstructure of materials, scanning electron microscopes play a vital role in the manufacturing process of integrated circuits. SEM images can provide high-resolution surface morphology information, which is an important basis for evaluating the quality of integrated circuit manufacturing processes, identifying potential defects and optimizing process parameters.
[0003] However, it is not easy to extract accurate contour information directly from SEM images. First, SEM images often contain a lot of complex background noise and interference signals, such as electron beam scattering effects, shadows caused by uneven sample surfaces, etc. These factors will seriously affect the accurate recognition of contours. Secondly, due to imaging conditions (such as scanning voltage, beam size, etc.) or sample characteristics (such as differences in material conductivity), the contours in SEM images may be broken, blurred or overlapped, further increasing the difficulty of contour extraction.
[0004] Traditional contour extraction methods, such as threshold segmentation and edge detection, can achieve preliminary contour recognition to a certain extent, but they are often difficult to effectively cope with the above challenges. Especially when dealing with high-precision scenes such as integrated circuits, these methods often have problems such as incomplete contour extraction and high false recognition rate, which are difficult to meet the needs of practical applications.
[0005] Therefore, developing a method that can automatically, accurately and efficiently extract contour information from scanning electron microscope images is of great significance for improving the level of integrated circuit manufacturing process analysis and promoting the progress of the semiconductor industry. Morphological operations are an important branch in the field of image processing, which analyzes and processes images based on their shape and structural features. These operations are usually used for binary images, and a tool called a structural element is required to define the nature and scope of the operation. Morphological operations include erosion, dilation, opening, closing, morphological gradient, top hat and black hat transformations. However, at present, it has not been possible to realize scanning electron microscope contour extraction using morphological related operations. Summary of the invention
[0006] The first purpose of the present invention is to address the deficiencies of the prior art and propose a scanning electron microscope contour extraction method based on morphological processing, which aims to efficiently and accurately extract contour information from scanning electron microscope images to support process analysis and detection of micro-nano structures such as integrated circuits. The specific invention content is as follows:
[0007] First, in step S10, the present invention receives and inputs a scanning electron microscope image as a processing object, and reads it in the form of a grayscale image to obtain two-dimensional grayscale value array information.
[0008] Subsequently, in step S20, the present invention uses a contour extraction algorithm based on image gradient to directly extract the contour of the input scanning electron microscope image to obtain a preliminary contour as the first figure. The purpose of this step is to fully extract the original information in the image and lay the foundation for subsequent processing. However, the preliminary contour obtained in this step will contain many redundant erroneous contours, and due to the limitation of imaging, some boundaries of the real figure will be fuzzy, resulting in discontinuous extracted contours.
[0009] Next, in step S30, in order to enhance the continuity and integrity of the contour, the present invention performs an expansion operation on the first figure to obtain a second figure. The expansion operation can effectively fill the small gaps between the contours and enhance the continuity of the true contour. At this time, the background area is divided into a plurality of independent areas by the contour.
[0010] On this basis, the region growing algorithm is used to automatically select the regions to be filled, and seeds are set in these regions to fill these regions, as shown in step S40. This step can distinguish the graphic region from the blank region.
[0011] However, due to the complexity of the image and the limitations of the algorithm, the result of the automatic filling may still require manual intervention. Therefore, in step S50, the present invention allows the user to manually adjust the area filling result to obtain a more accurate third figure.
[0012] Then, in step S60, in order to eliminate the size deviation that may be introduced by the dilation operation, the present invention performs an erosion operation on the third figure to obtain a fourth figure. The erosion operation can restore the original shape of the contour and remove unnecessary background noise.
[0013] Finally, in step S70, in order to further improve the quality of contour extraction, the present invention sets an area threshold to eliminate small-area graphics in the contour extraction result. These small-area contours are often noise or irrelevant details, and they can be effectively filtered out by setting a reasonable threshold. Finally, contour extraction is performed on the graphics after filtering out the small-area graphics, and a clear and accurate final contour is finally output.
[0014] Further, step S20 includes:
[0015] In step S201, the present invention first performs Gaussian filtering on the input scanning electron microscope image to remove noise and details in the image while retaining the main features of the image.
[0016] Next, in step S202, the present invention performs gradient calculation on the filtered image to obtain a gradient graph.
[0017] Then, in step S203, the present invention performs non-maximum suppression processing on the gradient graph. By comparing the gradient magnitudes of each pixel with its neighboring pixels, only the local maximum value in the gradient direction is retained as the edge point, which effectively removes the non-edge pixel points in the gradient graph and retains the true contour information.
[0018] Meanwhile, in step S204, the present invention obtains the 20% quantile, i.e., the proportion of pixels with lower gray values, by sorting the pixel gray values of the SEM image. Using this statistical information, the present invention calculates the expected contour length in combination with the filtered image.
[0019] Finally, in step S205, the present invention determines the first threshold and the second threshold using a dichotomy method according to the expected contour length. Through these two thresholds, the present invention screens the contours in the gradient graph, removes contour segments that do not meet the conditions, and obtains the final contour result.
[0020] Further, step S40 includes:
[0021] In step S401, the present invention first traverses the second graph. By adopting a depth-first search algorithm, the present invention can accurately determine the number of independent regions in the image and determine a seed point for each region.
[0022] Next, in step S402, the present invention globally sorts the pixel grayscale values of the scanning electron microscope image to obtain the median of the overall pixel grayscale value.
[0023] Then, in step S403, the present invention sorts the pixel grayscale values of the scanning electron microscope image again for each independent region to obtain the median of the pixel grayscale value of each region.
[0024] Finally, in step S404, the present invention determines whether it is necessary to flip the pixels of the region by comparing the overall median with the median of each region. If the overall median is less than the median of a region, it means that the grayscale value of the region is generally high, indicating that the region is likely to be a graphic region. The present invention will flip the pixels of the region, that is, invert its grayscale value.
[0025] A second object of the present invention is to provide a scanning electron microscope image contour extraction and optimization system for realizing the above method, comprising:
[0026] Image input module, used to obtain images and convert them into grayscale images;
[0027] Contour extraction module, used to extract the contour of the grayscale image;
[0028] The expansion module is used to expand the contour of the grayscale image;
[0029] The area filling module is used to fill the area segmented by the expanded contour;
[0030] A user interaction module, used to manually adjust the area filling result;
[0031] An erosion module, used to erode the adjusted filling result;
[0032] The contour optimization module is used to optimize the eroded graphics and extract the final contour.
[0033] A third object of the present invention is to provide a computer-readable storage medium having a computer program stored thereon, which implements the above method when the program is executed by a processor.
[0034] A fourth object of the present invention is to provide an electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and the processor implements the above method when executing the computer program.
[0035] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:
[0036] The present invention proposes a method for extracting contours from scanning electron microscope images based on morphological processing, aiming to solve the deficiencies in the prior art and provide a more accurate and efficient image processing solution for the field of integrated circuit manufacturing. The contour extraction method based on morphological processing of the present invention realizes efficient and accurate extraction of the contours of scanning electron microscope images, removes redundant erroneous contours through morphological operations, and connects the disconnected real contours. This method not only improves the accuracy and efficiency of contour extraction, but also enhances the continuity and integrity of the contour, providing strong support for process analysis and detection of micro-nano structures such as integrated circuits. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figures 1 to 3 It is a flowchart of the SEM image contour extraction method based on morphological processing;
[0038] Figure 4 is a schematic diagram of a scanning electron microscope image;
[0039] Figure 5 is a schematic diagram of a first figure;
[0040] Figure 6 is a schematic diagram of a second figure;
[0041] Figure 7 is a schematic diagram of the third figure;
[0042] Figure 8 is a schematic diagram of a fourth figure;
[0043] Fig. 9 A schematic diagram of the final outline. DETAILED DESCRIPTION
[0044] The present invention is further described in detail below with reference to the accompanying drawings and specific implementation modes.
[0045] Figures 1 to 3 It is a flowchart of the SEM image contour extraction method based on morphological processing;
[0046] Please refer to Figure 1 , the scanning electron microscope contour extraction method based on morphological processing includes:
[0047] In step S10, the present invention receives and inputs a scanning electron microscope image as a processing object, and reads it in the form of a grayscale image to obtain two-dimensional grayscale value array information. Figure 4 Schematic diagram of a scanning electron microscope image. In this embodiment, the scanning electron microscope image is a photoresist pattern after development, the image resolution is 512 pixels*512 pixels, and the pixel bit depth is 8.
[0048] Then, in step S20, the present invention uses a contour extraction algorithm to directly extract the contour of the input scanning electron microscope image to obtain a preliminary contour as the first graph. Figure 5 Schematic diagram of the first figure. The purpose of this step is to fully extract the original information in the image, to lay the foundation for subsequent processing. In the present embodiment, due to the limitation of integrated circuit manufacturing process, a white edge region with a larger grayscale pixel value can be formed in the transition region of the photoresist region and the blank region, and both sides of the white edge may be identified as a profile, but only the profile close to the blank side is the profile that needs to be extracted. In addition, due to the higher surface roughness of the photoresist region, it is easy to identify the redundant profile of error in the photoresist region. In addition, due to the limitation of scanning electron microscope imaging, the profile identified is not continuous, and there may be a phenomenon of profile disconnection.
[0049] Next, in step S30, in order to enhance the continuity and integrity of the real contour, the present invention performs an expansion operation on the first figure to obtain a second figure. Figure 6is a schematic diagram of the second graphic. The dilation operation can expand the pixels of the contour outward so that the disconnected real contours are connected together. In this embodiment, the filter kernel size used in the dilation operation is 5*5. The contour pixels of the second graphic divide the background area into several independent areas, which are graphic areas or blank areas.
[0050] On this basis, the region growing algorithm is used to automatically select the regions to be filled, set seeds in these regions, and fill these regions, as shown in step S40. In this embodiment, the pixel value of the graphic region is set to 255, and the pixel value of the blank region is set to 0. This step can distinguish the graphic region from the blank region, so that the outline graphic is restored to the original graphic, and also lays the foundation for further graphic operations. At the same time, the automatic filling operation is conducive to reducing manual intervention and reducing workload.
[0051] However, due to image complexity and algorithm limitations, the result of automatic filling may be wrong and require manual intervention. Therefore, in step S50, the present invention allows the user to manually adjust the area filling result to obtain a more accurate third figure, thereby ensuring the reliability of the processing result. Figure 7 is a schematic diagram of the third graphic. In this embodiment, the white area is the graphic area, the gray area is the outline area, and the black area is the blank area. Since the graphic area will identify redundant outlines, the outline area is attributed to the graphic area in subsequent operations to eliminate redundant erroneous outlines.
[0052] Then, in step S60, in order to eliminate the size deviation that may be introduced by the dilation operation, the present invention uses a filter kernel of the same size to perform an erosion operation on the third image to obtain a fourth image. The erosion operation can restore the original size of the image. Figure 8 is a schematic diagram of the fourth figure. The fourth figure can be regarded as a simplified scanning electron microscope image, wherein the white area corresponds to the graphic area, and the black area corresponds to the blank area.
[0053] Finally, in step S70, in order to further improve the quality of contour extraction, the present invention sets an area threshold to eliminate small-area graphics in the contour extraction result. These small-area contours are often noise or irrelevant details, and they can be effectively filtered out by setting a reasonable threshold. In this embodiment, the area threshold is set to 400 pixels. Finally, the fourth figure after filtering out the small-area graphics is contour extracted (any contour extraction algorithm can be used, such as the Sobel operator, the Scharr operator, the Laplacian operator or the contour extraction algorithm described in step 20), and a clear and accurate final contour is output. Fig. 9 A schematic diagram of the final profile is provided as a clear and accurate SEM profile map that can be used for further analysis, measurement, or visualization.
[0054] Will Fig. 9 and Figure 4 , Figure 5 By comparison, it can be seen that the method of the present invention significantly improves the accuracy of contour extraction.
[0055] Further, such as Figure 2 As shown, step S20 includes:
[0056] Step S201: Gaussian filter processing. In the specific implementation, we first read the input SEM image into the memory. Then, the Gaussian filter is applied to smooth the image. The kernel size of the Gaussian filter is selected as 15*15, and the standard deviation σ is set to 10. This parameter combination has been verified by many experiments and can effectively remove small noise and unnecessary details in the image while keeping the main edge features of the image from being overly blurred. The processed image is used as the input for the next step.
[0057] Step S202: Gradient calculation. After obtaining the filtered image, we use the Sobel operator to calculate its horizontal and vertical gradients. The convolution kernels of the Sobel operator in the x and y directions are [-1 0 1; -2 0 2; -1 0 1] and [-1 -2 -1; 0 0 0; 1 2 1], respectively. By convolving these two convolution kernels with the filtered image, we obtain two gradient graphs: Gx and Gy, representing the gradient strength in the horizontal and vertical directions, respectively. Then, use G = sqrt(Gx^2 + Gy^2) to calculate the gradient amplitude G of each pixel, and θ = arctan2(Gy, Gx) to calculate the gradient direction θ, thereby generating a complete gradient graph.
[0058] Step S203: Non-maximum suppression. For each pixel in the gradient graph, we examine its 8-neighborhood (including adjacent pixels in the horizontal and vertical directions, and adjacent pixels on two diagonals). According to the gradient direction θ of the pixel, we determine the two adjacent pixels in its gradient direction. Then, compare the gradient amplitude of the pixel with the gradient amplitudes of the two adjacent pixels. If the gradient amplitude of the pixel is a local maximum, the point is retained as an edge point; otherwise, it is marked as a non-edge point and set to zero. This process traverses the entire gradient graph to complete the non-maximum suppression process.
[0059] Step S204: Calculate the expected contour length. In step S204, we first sort the grayscale values of the filtered image to obtain the distribution of the grayscale values of all pixels. Then, calculate the 20% quantile of the grayscale value, that is, the grayscale value limit of 20% pixels with lower grayscale values. Based on this limit, we assume that pixels below this limit are more likely to belong to the background rather than the edge. Combined with the size of the filtered image and the estimated proportion of edge pixels, we use geometric and statistical methods to estimate the expected contour length. The expected contour length calculated in this embodiment is 3477.
[0060] Step S205: Determine the threshold value and filter the contour by binary search. In order to filter out the expected contour from the gradient graph, we use binary search to determine two threshold values (first threshold value and second threshold value). Initially, a wider threshold range is set, and the contour lengths corresponding to multiple candidate threshold values within the range are calculated. By comparing the differences between these contour lengths and the expected contour lengths, we continuously adjust the threshold range until we find two threshold values that make the contour length closest to the expected one. In this embodiment, the first threshold value is set to half of the second threshold value, and the second threshold value is determined by binary search in the range of 0 to 20000 according to the expected contour length calculated in step S204, and the calculated second threshold value is 10078.
[0061] Finally, these two thresholds are used to screen the contours in the gradient graph. Specifically, only when the gradient amplitude of the contour segment is higher than the first threshold and the gradient amplitude of at least one continuous pixel is higher than the second threshold, the contour segment is retained as the final edge contour. After this step, we obtained the preliminary edge contour results of the SEM image.
[0062] Further, such as Figure 3 As shown, step S40 includes:
[0063] Step S401: traverse the second graph and determine the independent region and seed point. Load the SEM image to be processed into the system memory, define the image size as M×N pixels, where M is the image height and N is the image width. Create an M×N Boolean matrix visited to mark whether each pixel has been visited, initialized to all false (unvisited). Create a list regions to store the seed points and corresponding pixel sets of the independent region. Start traversing from the upper left corner pixel (0,0) of the image, and for each unvisited pixel (i,j), perform a depth-first search: mark (i,j) as visited (true) and add it to the new entry of regions as the seed point of the current region. Check the four adjacent pixels above, below, left and right of the current pixel. If they have not been visited and are within the image boundary, recursively perform a depth-first search. Repeat the above process until all pixels have been visited, and the seed point uses the first traversed pixel point of each independent region, that is, the identification of all independent regions and the determination of seed points are completed.
[0064] Step S402: Global pixel grayscale value sorting to obtain the overall median. Extract the grayscale value of each pixel of the SEM image into a one-dimensional array grayValues. Sort grayValues. Since the grayscale value range of the SEM image is usually 0-255, if the total number of pixels in the image is P, the overall median is the grayscale value at the position P / 2 (for an even number P, it is the average of P / 2 and P / 2+1) after sorting.
[0065] Step S403: Calculate the median gray value for each independent region. For each region in regions, extract the gray values of all pixels in the region into a new array regionGrayValues. Each regionGrayValues is sorted, and the median is calculated based on the number of pixels in the region, in the same way as the global median calculation.
[0066] Step S404: Compare and flip the regional pixels. Traverse regions, and for each region, compare its grayscale median with the overall median. If the median of a region is greater than the overall median, the region is considered to be a possible highlight graphic region, and a grayscale flip operation is performed on each pixel in the region. The flip operation is defined as: the original grayscale value gray becomes 255 - gray. Update all flipped regions back to the original image, and save or display the processed image.
[0067] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.
Claims
1. A method for extracting contours from scanning electron microscope images based on morphological processing, characterized in that: The following steps are involved: Step S10, using a scanning electron microscope to acquire an image, and converting the image into a grayscale image; Step S20, extracting the contour of the grayscale image using a contour extraction algorithm based on image gradient to obtain a first image; Step S30, performing an expansion operation on the first graphic to obtain a second graphic; the second graphic divides the background area of the grayscale image into a plurality of independent areas, wherein the independent areas include a blank area and a graphic area; Step S40, using a region growing algorithm to fill the independent region, and distinguishing the blank region from the graphic region; Step S50, manually adjusting the filling result of the independent area to obtain a third figure; Step S60, performing an erosion operation on the third graphic to restore the size to the same as the first graphic, thereby obtaining a fourth graphic; Step S70: setting an area threshold, eliminating small-area graphics with an area lower than the area threshold in the fourth graphic, and then performing contour extraction to output a final contour.
2. The method for extracting contours from scanning electron microscope images based on morphological processing according to claim 1, characterized in that: The contour extraction algorithm based on image gradient in step S20 includes the following steps: Gaussian filtering, gradient calculation, non-maximum suppression, and automatic threshold determination.
3. The method for extracting contours from scanning electron microscope images based on morphological processing according to claim 2, characterized in that: The contour extraction algorithm based on image gradient described in step S20 is specifically: Step S201: Gaussian filtering: using a Gaussian filter to smooth the grayscale image, remove fine noise in the image, and obtain a filtered image; Step S202: gradient calculation: use the Sobel operator to calculate the horizontal and vertical gradients of the filtered image, obtain the horizontal gradient map Gx and the vertical gradient map Gy, then use G = sqrt(Gx^2 + Gy^2) to calculate the gradient amplitude G of each pixel in the image, and use θ = arctan2(Gy, Gx) to calculate the gradient direction θ, thereby generating a complete gradient graph; Step S203: non-maximum suppression: for each pixel point in the gradient graph, examine the pixels in its 8-neighborhood; Select a pixel point, determine the two adjacent pixels in its gradient direction according to the gradient direction θ of the selected pixel point, compare the gradient amplitude of the selected pixel point with the gradient amplitudes of the two adjacent pixels, and if the gradient amplitude of the selected pixel point is a local maximum, retain the selected pixel point as an edge point; Otherwise, mark it as a non-edge point and set it to zero; traverse each pixel in the entire gradient graph to complete the non-maximum suppression process; Step S204: Calculate the expected contour length; sort the grayscale values of the filtered image to obtain the distribution of the grayscale values of all pixels and calculate the grayscale value limit. Assuming that the pixels below the grayscale value limit belong to the background rather than the contour edge, the expected contour length is estimated by combining the size of the filtered image and the estimated proportion of the contour edge pixels; Step S205: Determine the first threshold and the second threshold that make the contour length closest to the expected contour length by the binary method, and then screen the contours in the gradient graph; when the gradient amplitude of the contour segment in the contour is higher than the first threshold at the same time and the gradient amplitude of at least one continuous pixel is higher than the second threshold, the contour segment is retained as the final edge contour; all retained contour segments are the first graph.
4. The method for extracting contours from scanning electron microscope images based on morphological processing according to claim 1, characterized in that: The region growing algorithm in step S40 includes the following steps: The number of independent regions and the seeds corresponding to each region are determined, the median of the pixel values of the scanning electron microscope image is determined, and whether to flip the pixels of the region is determined according to the median of the pixel values of the independent region and the median of the scanning electron microscope image.
5. The method for extracting contours from scanning electron microscope images based on morphological processing according to claim 4, characterized in that: The region growing algorithm in step S40 includes the following steps: Step S401: define the image size of the second graphic as M×N pixels, where M is the image height and N is the image width; Create an M×N Boolean matrix to mark whether each pixel has been visited, and initialize it to all unvisited; create a list to store the seed points and corresponding pixel sets of independent regions; Start traversing from the upper left corner pixel (0,0) of the second graph, and perform a depth-first search for each unvisited pixel (i,j): mark (i,j) as visited and add it to a new entry in the list as the seed point of the current region; check the four adjacent pixels above, below, left, and right of the current pixel. If they are unvisited and within the boundary of the second graph, recursively perform a depth-first search; repeat the above process until all pixels in the second graph are visited, and the seed point uses the first traversed pixel point of each independent region, that is, the identification of all independent regions and the determination of seed points are completed; Step S402: extracting the grayscale value of each pixel into a one-dimensional array, sorting the one-dimensional array, obtaining the grayscale value at the median position, and recording it as the global median grayscale value; Step S403: Calculate the median grayscale value for each independent region; for each region in the list, extract the grayscale values of all pixels in the region into a new array and obtain the grayscale value of the median position of each region respectively; Step S404: compare and flip the area pixels; traverse the entire list, for each area, compare its median grayscale value with the global median grayscale value, if it is greater than the global median grayscale value, then the area is considered to be a possible highlight graphic area, and a grayscale value flipping operation is performed on each pixel in the area, flipping the original grayscale value gray to 255-gary, otherwise it will not be executed; update all flipped areas back to the original image, and save or display the processed image.
6. The method for extracting contours from scanning electron microscope images based on morphological processing according to claim 1, characterized in that: The kernel size used in the erosion operation and the dilation operation is the same.
7. A scanning electron microscope image contour extraction and optimization system for implementing the method according to any one of claims 1 to 6, characterized in that: include: Image input module, used to obtain images and convert them into grayscale images; Contour extraction module, used to extract the contour of the grayscale image; The expansion module is used to expand the contour of the grayscale image; The region filling module is used to fill the region segmented by the expanded contour; A user interaction module, used to manually adjust the area filling result; An erosion module, used to erode the adjusted filling result; The contour optimization module is used to optimize the eroded graphics, eliminate small-area graphics, and extract the final contour.
8. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.
9. An electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and when the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.
Citation Information
Patent Citations
Frame line detection method based on connected domain analysis and morphological operation
CN110033471A
Morphological operation-based text region detection method and device
CN111476800A
Automobile mold plane crack detection method based on image processing
CN113240667A
Cited By
Rearview mirror mold precision detection method based on machine vision
CN120147309A
A method for detecting rearview mirror mold precision based on machine vision
CN120147309B
