A method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds

Abstract

This invention provides a method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds, belonging to the field of engineering measurement technology. The method combines two-dimensional and three-dimensional point cloud data fusion detection, overcoming the poor anti-interference capability of traditional two-dimensional image measurement methods. It can operate stably in complex welding environments, ensuring the reliability of the detection results. Simultaneously, by employing a spectral confocal sensor and three-dimensional point cloud technology, the three-dimensional morphology of the welding surface can be accurately captured. Noise is effectively eliminated through point cloud quantile filtering and MAD denoising technology, improving the accuracy of crater depth measurement. Furthermore, this invention automates the crater depth measurement process through various advanced algorithms (such as point cloud quantile filtering, percentile interpolation screening, and plane projection segmentation), reducing the need for manual intervention and improving detection efficiency.

Description

Technical Field

[0001] This invention relates to the field of engineering measurement technology, and in particular to a method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds. Background Technology

[0002] The integrated busbar (Cells Contact System, CCS) is a key component in new energy vehicle battery systems, and its performance directly affects the battery system's conductivity, thermal management, and overall performance. Traditional welding processes are used to connect battery cells, flexible printed circuits (FPCs), and aluminum busbars. However, this method has significant shortcomings in welding quality control, leading to frequent welding defects and affecting the overall system's efficiency and reliability. Especially in measuring weld pit depth, traditional manual measurement methods are inefficient and lack accuracy, failing to meet the high standards required by modern battery systems.

[0003] Currently, enterprises typically employ two main measurement methods: contact measurement and non-contact measurement. While contact measurement offers higher accuracy, it often damages parts. Non-contact measurement typically utilizes machine vision-based image processing technology, but two-dimensional image measurement suffers from poor interference resistance when applied to ultrasonic welding crater depth measurement. With the development of three-dimensional point cloud technology, this problem is expected to be solved.

[0004] 3D point cloud technology improves measurement accuracy and anti-interference capability by accurately capturing the three-dimensional morphology of the welding surface.

[0005] The general procedure for measuring the depth of ultrasonic welding craters using 3D point cloud data is as follows:

[0006] First, a pass-through filter is performed on the collected raw point cloud data to effectively separate the FPC point cloud from the Albar point cloud.

[0007] Next, voxel downsampling technology is used to reduce data density.

[0008] Based on the boundary detection results, the internal point cloud of the FPC is extracted. After obtaining the internal point cloud of the FPC, key features are extracted and normal vectors are calculated to lay the foundation for further analysis.

[0009] Next, the DBSCAN clustering algorithm was used to identify and extract independent pit bottom point cloud clusters from the internal point cloud of the FPC.

[0010] Finally, the depth of the weld pit is obtained by calculating the shortest distance between the point cloud at the bottom of each independent pit and the point cloud of the aluminum bar, that is, by quantifying the mean difference in the Z-axis direction.

[0011] However, this testing process is inefficient, requires multiple manual parameter modifications, and has a low degree of automation. Therefore, it is very important to design a more automated and efficient method. Summary of the Invention

[0012] The purpose of this invention is to address the problems of low automation and low efficiency in existing methods for measuring the depth of weld pits in ultrasonic welding FPCs and aluminum bars using two-dimensional images and three-dimensional point clouds. This invention provides a fusion detection method based on two-dimensional and three-dimensional images, which can automatically extract the weld pit point cloud data of the effective welding area and quickly and accurately measure the depth of the weld pit.

[0013] To achieve the above objectives, the present invention adopts the following technical solution:

[0014] A method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds includes the following steps:

[0015] S1: Collect raw point cloud data;

[0016] S2: Data preprocessing: Preprocess the raw point cloud data, filter out noise points with a Z value of 0, and obtain the preprocessed point cloud set denoted as Q1;

[0017] S3: Data filtering: Use point cloud quantile filtering to filter the point cloud set Q1 in S2 to obtain the FPC point cloud set denoted as Q2;

[0018] S4: Using the planar projection segmentation method, the effective solder pit area point set inside the FPC is segmented from the point set Q1 in S2 and denoted as Q3;

[0019] S5: Use percentile interpolation to filter the point cloud set Q3 in step S4 to obtain the overall pit bottom point cloud set, denoted as Q4.

[0020] S6: Using the planar projection segmentation method, segment out the independent pit bottom point set from Q4, denoted as D1, D2, D3, ..., D... n ;

[0021] S7: Apply MAD denoising method to D1, D2, D3, ..., D n Denoising is performed to obtain the denoised set of independent pit bottom points, denoted as E1, E2, E3, ..., E n ;

[0022] S8: Calculate the independent pit bottom point cloud set E1, E2, E3, ..., E n The Z-means are calculated and saved separately, denoted as Z1, Z2, Z3, ..., Zn. n ;

[0023] S9: Using the planar projection segmentation method, segment the aluminum bar planar point cloud (invalid solder pit area) from the point cloud Q1 in step S2, and denote it as Q5;

[0024] S10: Use point cloud quantile filtering to filter point cloud set Q5 in S9 to obtain reference surface point cloud set denoted as Q6;

[0025] S11: Calculate and save the Z-mean of the reference surface point set, denoted as Z0. b ;

[0026] S12: Sequentially calculate the Z-mean (Z1, Z2, Z3, ..., Z) of each independent pit bottom point cluster. n Subtract the Z-mean of the reference surface point set Z b This allows us to obtain the depth of each individual pit bottom.

[0027] Preferably, in step S1, raw point cloud data is acquired using a spectral confocal sensor.

[0028] Preferably, the specific steps of S3 are as follows:

[0029] S3-1: Sort the z-coordinates of the point cloud data Q1 after preliminary denoising in ascending order;

[0030] S3-2: Calculate the corresponding value Q based on the preset percentile P. p The Q of non-integer index i is estimated using a linear interpolation formula. p value;

[0031] S3-3: Using Q p The value is a threshold T, and points higher than this threshold are filtered out.

[0032] S3-4: Apply absolute median difference to the selected FPC point cloud for refined denoising, identify and remove outliers and noise.

[0033] Preferably, the preset percentile P value in S3 is 80.

[0034] Preferably, the specific steps of S4 are as follows:

[0035] S4-1: Project the 3D point cloud data onto a 2D plane (xy plane), set the mapping resolution conversion coefficient, and obtain a 2D image;

[0036] S4-2: Apply image processing algorithms (such as edge detection, threshold segmentation, etc.) to the two-dimensional image to extract the bounding rectangle coordinates of the ROI (effective solder pit area inside the FPC);

[0037] S4-3: Based on the coordinates of the 2D ROI rectangle, inversely map it to 3D space to determine its corresponding bounding box region in the point cloud Q1;

[0038] S4-4: Extract the point cloud subset within the bounding box to obtain the segmented point cloud corresponding to the ROI, i.e., the point cloud set of the effective solder pit area inside the FPC.

[0039] Preferably, step S6 is as follows:

[0040] S6-1: Project the point cloud data inside the FPC onto a two-dimensional plane (xy plane), set the mapping resolution conversion coefficient, and obtain a two-dimensional image;

[0041] S6-2: Apply image processing algorithms to eliminate discrete noise and connect adjacent noise points in a two-dimensional image;

[0042] S6-3: Calculate the area of ​​connected regions using contour extraction technology, and filter out connected regions within a specific area range, defining them as "connected domains" that represent independent pit bottoms;

[0043] S6-4: Calculate the centroid coordinates of each connected domain to determine the location information of each pit bottom;

[0044] S6-5: Inversely map these two-dimensional centroid coordinates back to the point cloud coordinate system in three-dimensional space to determine the corresponding bounding box region in the three-dimensional point cloud;

[0045] S6-6: Extract the subset of point cloud within the bounding box to obtain the segmented point cloud corresponding to each connected component, i.e., the independent pit bottom point cloud set.

[0046] Preferably, the resolution conversion factor is set to 50, and the image is subjected to dilation and erosion processing to obtain the ROI region.

[0047] Compared with the prior art, this application has the following beneficial effects:

[0048] 1. Strong anti-interference capability: The fusion detection method combining two-dimensional and three-dimensional point cloud data overcomes the problem of poor anti-interference capability of traditional two-dimensional image measurement methods, and can operate stably in complex welding environments, ensuring the reliability of detection results.

[0049] 2. High precision and high efficiency: Employing a spectral confocal sensor and 3D point cloud technology, it can accurately capture the 3D morphology of the weld surface. Through point cloud quantile filtering and MAD denoising technology, noise is effectively eliminated, improving the accuracy of weld pit depth measurement.

[0050] 3. High degree of automation: This invention automates the weld pit depth measurement process through a variety of advanced algorithms (such as point cloud quantile filtering, percentile interpolation screening, and plane projection segmentation), reducing the need for manual intervention and improving detection efficiency. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating a method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds in one embodiment of this application.

[0052] Figure 2 This is a point cloud data acquisition diagram in one embodiment of this application;

[0053] Figure 3 This is a visualization of the point cloud data to be segmented in one embodiment of this application;

[0054] Figure 4 This is a flowchart of the process for extracting FPC point clouds in one embodiment of this application;

[0055] Figure 5 This is a flowchart of segmenting the internal point cloud of an FPC according to one embodiment of this application;

[0056] Figure 6 This is a point cloud visualization of the effective solder pit area inside the FPC and the point cloud set (ineffective solder pit area) of the aluminum bar plane segmented in one embodiment of this application.

[0057] Figure 7 This is a flowchart illustrating the segmentation of the point cloud at the bottom of an independent pit in one embodiment of this application. Detailed Implementation

[0058] The present invention will be further described in detail below with reference to specific embodiments.

[0059] Please see Figure 1 A method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds includes the following steps:

[0060] S1: Data Collection

[0061] In one embodiment, raw point cloud data is acquired using a spectral confocal sensor;

[0062] For details, please refer to Figure 2 The data acquisition method is as follows:

[0063] S1-1: Sensor Setup and Calibration: Before the experiment begins, the sensor is properly set up and calibrated to ensure accurate capture of surface data of the welding area.

[0064] S1-2: Fixing the workpiece: Fix the finished workpiece in the appropriate position so that the sensor can completely capture the point cloud data of its surface.

[0065] S1-3: Data Acquisition: Use sensors to perform three-dimensional scanning of the surface of the finished workpiece to obtain its point cloud data.

[0066] Original point cloud data visualization images such as Figure 3As shown, the blue box represents the effective solder joint area inside the FPC, while the area outside the box is the point cloud of the aluminum busbar (ineffective solder joint area). In the visualized point cloud, the colors transition from blue to red based on the Z-axis values. Specifically, the minimum Z-axis value is displayed in blue, and the maximum value is displayed in red.

[0067] S2: Data Preprocessing

[0068] Specifically, in one implementation method, the original point cloud data is preprocessed to filter out noise points with a Z value of 0, and the preprocessed point cloud set is denoted as Q1.

[0069] S3: Data Filtering

[0070] In one embodiment, point cloud quantile filtering is used to filter the point cloud set Q1 in S2 to obtain the FPC point cloud set denoted as Q2;

[0071] The point cloud quantile filtering method is a filtering method based on percentiles and median absolute deviation (MAD). It sorts the Z-coordinates of the point cloud set in ascending order, and then uses linear interpolation techniques to accurately estimate the corresponding numerical value Q based on a preset percentile P. p As a screening threshold standard.

[0072] Next, the threshold T, calculated based on percentiles, is used to filter out FPC points with highly dissimilar characteristics from the point cloud set. Finally, MAD (Magnitude-Oriented Denoising) is introduced to perform outlier detection and denoising on the filtered FPC point cloud. For an ordered dataset Z = {z1, z2, ..., z...} n The dataset contains N data points. Given the target percentile p (0 < p < 1), the goal is to find its corresponding value Q. p Calculate the theoretical index position i:

[0073]

[0074] Given a dataset Z containing N data points, determine its median M. Then, for each data point z in the set... i Calculate the absolute difference between this difference and the median M of the entire dataset as |zi-M|. Take the median of the absolute differences for all data points, i.e.:

[0075] MAD=median(|z1-M|, |z2-M|,...,|z N -M|).

[0076] Please see Figure 4 The specific steps of S3 are as follows:

[0077] S3-1: Sort the z-coordinates of the point cloud data Q1 after preliminary denoising in ascending order;

[0078] S3-2: Calculate the corresponding value Q based on the preset percentile P. p The Q of non-integer index i is estimated using a linear interpolation formula. p value;

[0079] S3-3: Using Q p The value is a threshold T, and points (FPC point cloud) above this threshold are selected.

[0080] S3-4: Apply absolute median difference (MAD) to the selected FPC point cloud for refined denoising, and identify and remove outliers and noise.

[0081] The percentile P-value was set to 80 based on the experimental results.

[0082] S4: Using the planar projection segmentation method, the effective solder pit area point set inside the FPC is segmented from the point set Q1 in S2 and denoted as Q3;

[0083] In one embodiment, the planar projection segmentation method is a method of converting a three-dimensional point cloud into a two-dimensional image, processing the two-dimensional image to obtain the bounding rectangle of the ROI region (the effective solder pit region inside the FPC), and then inversely mapping it to three-dimensional space to segment out the required point cloud set.

[0084] Specifically, first, the minimum and maximum coordinates of point cloud set Q1 in the XY plane are obtained to ensure that all point cloud data can be completely represented in the image. Then, the coordinates of each point in point cloud set Q2 in 3D space are mapped to the coordinate system of the 2D image. The following formula is used here:

[0085]

[0086] Here, x and y represent the two-dimensional coordinates of the FPC point cloud portion projected onto the XY plane in 3D space, while pixel_x and pixel_y represent the corresponding pixel coordinates in the two-dimensional image. (x-MinX) and (MaxY-y) replace the coordinates (x, y) in the FPC point cloud data with offsets relative to the minimum value, ensuring that negative values ​​are not generated when mapped to the image. new_resolution is the resolution conversion factor used to control the size and precision of the point cloud in the image. MaxX-MinX and MaxY-MinY are the ranges of the original point cloud data along the horizontal and vertical axes in the XY plane, used to normalize the coordinate values ​​and ensure correct distribution when mapped to the image.

[0087] Next, image processing algorithms are applied to the two-dimensional image to extract the bounding rectangle coordinates of the region of interest (ROI, effective solder joint area inside the FPC);

[0088] Then, the coordinates of the obtained 2D ROI rectangle are inversely mapped to 3D space to determine the corresponding bounding box region in the original point cloud. The specific formula for 3D inverse mapping is as follows:

[0089]

[0090] Where, x corner and y corner These represent the pixel coordinates of the top-left and bottom-right corners of the bounding rectangle of the ROI region, respectively, while x and y represent the two-dimensional coordinates projected back into the XY plane in 3D space. MinX and MaxY are the minimum and maximum values ​​of the original point cloud data in the x and y directions, respectively, which are the origin positions of the point cloud coordinate system.

[0091] Finally, extract the point cloud subset within the bounding box to obtain the segmented point cloud corresponding to the ROI region, which is the point cloud set of the effective solder pit region inside the FPC.

[0092] like Figure 5 As shown, the specific steps of S4 are as follows:

[0093] S4-1: Project the 3D point cloud data onto a 2D plane (xy plane), set the mapping resolution conversion coefficient, and obtain a 2D image;

[0094] S4-2: Apply image processing algorithms (such as edge detection, threshold segmentation, etc.) to the two-dimensional image to extract the bounding rectangle coordinates of the ROI (effective solder pit area inside the FPC);

[0095] S4-3: Based on the coordinates of the 2D ROI rectangle, inversely map it to 3D space to determine its corresponding bounding box region in the point cloud Q1;

[0096] S4-4: Extract the point cloud subset within the bounding box to obtain the segmented point cloud corresponding to the ROI, i.e., the point cloud set of the effective solder pit area inside the FPC.

[0097] Please see Figure 6 , Figure 6 It is a point cloud visualization of the segmented effective solder pit area inside the FPC and the point cloud set (ineffective solder pit area) of the aluminum bar plane.

[0098] S5: Use percentile interpolation to filter the point cloud set Q3 in step S4 to obtain the overall pit bottom point cloud set, denoted as Q4.

[0099] S6: Using the planar projection segmentation method, segment out the independent pit bottom point set from Q4, denoted as D1, D2, D3, ..., D...n ;

[0100] Please see Figure 7 The specific steps of S6 are as follows:

[0101] S6-1: Project the point cloud data inside the FPC onto a two-dimensional plane (xy plane), set the mapping resolution conversion coefficient, and obtain a two-dimensional image;

[0102] S6-2: Apply image processing algorithms to eliminate discrete noise and connect adjacent noise points in a two-dimensional image;

[0103] S6-3: Calculate the area of ​​connected regions using contour extraction technology, and filter out connected regions within a specific area range, defining them as "connected domains" that represent independent pit bottoms;

[0104] S6-4: Calculate the centroid coordinates of each connected domain to determine the location information of each pit bottom;

[0105] S6-5: Inversely map these two-dimensional centroid coordinates back to the point cloud coordinate system in three-dimensional space to determine the corresponding bounding box region in the three-dimensional point cloud;

[0106] S6-6: Extract the subset of point cloud within the bounding box to obtain the segmented point cloud corresponding to each connected component, i.e., the independent pit bottom point cloud set.

[0107] To ensure reasonable computation time while maintaining sufficient image clarity for subsequent analysis and processing, a resolution conversion factor of 50 was set. Dilation and erosion processing was then applied to the image to obtain the Region of Interest (ROI).

[0108] S7: Apply MAD denoising method to D1, D2, D3, ..., D n Denoising is performed to obtain the denoised set of independent pit bottom points, denoted as E1, E2, E3, ..., E n ;

[0109] S8: Calculate the independent pit bottom point cloud set E1, E2, E3, ..., E n The Z-means are calculated and saved separately, denoted as Z1, Z2, Z3, ..., Zn. n ;

[0110] S9: Using the planar projection segmentation method, segment the aluminum bar planar point cloud (invalid solder pit area) from the point cloud Q1 in step S2, and denote it as Q5;

[0111] S10: Use point cloud quantile filtering to filter point cloud set Q5 in S9 to obtain reference surface point cloud set denoted as Q6;

[0112] S11: Calculate and save the Z-mean of the reference surface point set, denoted as Z0.b ;

[0113] S12: Sequentially calculate the Z-mean (Z1, Z2, Z3, ..., Z) of each independent pit bottom point cluster. n Subtract the Z-mean of the reference surface point set Z b This allows us to obtain the depth of each individual pit bottom.

[0114] In one embodiment, to ensure reasonable computation time while guaranteeing sufficient image clarity to support subsequent analysis and processing, the resolution conversion coefficient is set to 50 in steps S4 and S6. The image is then subjected to dilation and erosion processing to obtain the Region of Interest (ROI).

[0115] In summary, this application provides an ultrasonic welding crater depth measurement method based on a two-dimensional and three-dimensional fusion detection scheme, which aims to shorten the detection time and improve the detection accuracy. The above-mentioned measurement method integrates point cloud quantile filtering, percentile interpolation screening and plane projection segmentation technology, and realizes accurate positioning and depth measurement of the weld crater and reference surface of a workpiece with 50 effective weld craters.

[0116] Specifically, the ultrasonic welding crater depth measurement method based on three-dimensional point cloud in this application combines a fusion detection method of two-dimensional and three-dimensional point cloud data, which overcomes the problem of poor anti-interference ability of traditional two-dimensional image measurement methods, can operate stably in complex welding environments, and ensures the reliability of detection results.

[0117] Simultaneously employing a spectral confocal sensor and 3D point cloud technology, it can accurately capture the 3D morphology of the welding surface. Through point cloud filtering and MAD noise reduction technology, noise is effectively eliminated, improving the accuracy of weld pit depth measurement.

[0118] Furthermore, this application employs a variety of advanced algorithms, such as point cloud quantile filtering, percentile interpolation screening, and planar projection segmentation, to automate the weld pit depth measurement process, reduce the need for manual intervention, and thus improve detection efficiency.

Claims

1. A method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds, characterized in that: Includes the following steps: S1: Collect raw point cloud data; S2: Data preprocessing: Preprocess the raw point cloud data, filter out noise points with a Z value of 0, and obtain the preprocessed point cloud set denoted as Q1; S3: Data filtering: Use point cloud quantile filtering to filter the point cloud set Q1 in S2 to obtain the FPC point cloud set denoted as Q2; S4: Using the planar projection segmentation method, the effective solder pit area point set inside the FPC is segmented from the point set Q1 in S2 and denoted as Q3; S5: Use percentile interpolation to filter the point cloud set Q3 in step S4 to obtain the overall pit bottom point cloud set, denoted as Q4. S6: Using the planar projection segmentation method, segment out the independent pit bottom point set from Q4, denoted as D1, D2, D3, ..., D... n ; S7: Apply MAD denoising method to D1, D2, D3, ..., D n Denoising is performed to obtain the denoised set of independent pit bottom points, denoted as E1, E2, E3, ..., E n ; S8: Calculate the independent pit bottom point cloud set E1, E2, E3, ..., E n The Z-means are calculated and saved separately, denoted as Z1, Z2, Z3, ..., Zn. n ; S9: Using the planar projection segmentation method, the aluminum bar planar point cloud is segmented from the point cloud Q1 in step S2, which is the invalid weld pit area, denoted as Q5; S10: Use point cloud quantile filtering to filter point cloud set Q5 in S9 to obtain reference surface point cloud set denoted as Q6; S11: Calculate and save the Z-mean of the reference surface point set, denoted as Z0. b ; S12: Sequentially calculate the Z-mean (Z1, Z2, Z3, ..., Z) of each independent pit bottom point cluster. n Subtract the Z-mean of the reference surface point set Z b This allows us to obtain the depth of each individual pit bottom.

2. The method for measuring the depth of ultrasonic welding crater based on three-dimensional point clouds according to claim 1, characterized in that: In S1, raw point cloud data is acquired using a spectral confocal sensor.

3. The method for measuring the depth of ultrasonic welding crater based on three-dimensional point clouds according to claim 1, characterized in that: The specific steps of S3 are as follows: S3-1: Sort the z-coordinates of the point cloud data Q1 after preliminary denoising in ascending order; S3-2: Calculate the corresponding value Q based on the preset percentile P. p The Q at index position i is estimated using a linear interpolation formula. p value; S3-3: Using Q p The value is a threshold T, and points higher than this threshold are filtered out. S3-4: Apply absolute median difference to the selected FPC point cloud for refined denoising, identify and remove outliers and noise.

4. The method for measuring the depth of ultrasonic welding crater based on three-dimensional point clouds according to claim 3, characterized in that: The preset percentile P value in S3 is 80.

5. The method for measuring the depth of ultrasonic welding crater based on three-dimensional point clouds according to claim 1, characterized in that: The specific steps of S4 are as follows: S4-1: Project the 3D point cloud data onto the xy plane, set the mapping resolution conversion coefficient, and obtain a 2D image; S4-2: Apply edge detection and threshold segmentation to the two-dimensional image to extract the coordinates of the bounding rectangle of the ROI, where the ROI is the effective solder pit area inside the FPC; S4-3: Based on the coordinates of the 2D ROI rectangle, inversely map it to 3D space to determine its corresponding bounding box region in the point cloud Q1; S4-4: Extract the point cloud subset within the bounding box to obtain the segmented point cloud corresponding to the ROI, i.e., the point cloud set of the effective solder pit area inside the FPC.

6. The method for measuring the depth of ultrasonic welding crater based on three-dimensional point clouds according to claim 1, characterized in that: The steps in S6 are as follows: S6-1: Project the point cloud data inside the FPC onto the xy plane, set the mapping resolution conversion coefficient, and obtain a two-dimensional image; S6-2: Apply image processing algorithms to eliminate discrete noise and connect adjacent noise points in a two-dimensional image; S6-3: Calculate the area of ​​connected regions using contour extraction technology, and filter out connected regions within a specific area range, defining them as "connected domains" that represent independent pit bottoms; S6-4: Calculate the centroid coordinates of each connected domain to determine the location information of each pit bottom; S6-5: Inversely map these two-dimensional centroid coordinates back to the point cloud coordinate system in three-dimensional space to determine the corresponding bounding box region in the three-dimensional point cloud; S6-6: Extract the subset of point cloud within the bounding box to obtain the segmented point cloud corresponding to each connected component, i.e., the independent pit bottom point cloud set.

7. A method for measuring the depth of ultrasonic welding craters based on three-dimensional point clouds according to claim 5 or 6, characterized in that: The resolution conversion factor is set to 50, and the image is subjected to dilation and erosion processing to obtain the ROI region.

Images