A method for eliminating jitter interference in 3D visual online inspection of ship steel plates

Abstract

This invention discloses a method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates. This method employs 3D image acquisition sensors positioned above and below a roller bed to capture the coordinates of upper and lower surface points on the same cross-section of the steel plate being inspected. Low-pass filtering is then used to obtain a standard point cloud sequence, which is then calculated to obtain a static point cloud sequence. This method addresses the issue of inability to detect small-scale surface discontinuities in the collected 3D point cloud data due to jitter interference during online inspection of steel plate surface discontinuities using 3D machine vision technology, thereby improving the accuracy of surface defect detection.

Description

Technical Field

[0001] The present invention belongs to the technical field of shipbuilding, and in particular relates to a method for eliminating jitter interference in 3D visual online detection of ship steel plates. Background Art

[0002] Shipyards are major users of steel plates. They have many specifications and large quantities of steel plates. Steel plates of different varieties and specifications are generally stored in open-air stacks for a long time, which can easily lead to quality problems of the steel plates. This is manifested in a significant increase in the degree, number and incidence of pit defects on the surface of the steel plates. They need to be polished or repaired before they can be used. Therefore, a comprehensive inspection of the surface of marine steel plates is required to prevent surface defects from flowing into subsequent links and causing greater losses.

[0003] At present, the surface inspection of marine steel plates mainly relies on manual means. Manual visual inspection has the disadvantages of heavy workload, low efficiency, and low accuracy, and cannot guarantee the quality of steel plate inspection. In addition, the current main direction of online automatic detection of steel plate surface discontinuities is to use 2D image acquisition equipment, highlight the discontinuous features of the steel plate surface through special lighting, manually classify and annotate the defects of the collected images, use deep learning technology for model training, and finally perform inference calculations on the actual online collected images to realize the identification and classification of defects. The main problems of this 2D machine vision in inspection are:

[0004] 1. The color difference between the surface defects of steel plates and the normal steel plates is small, and there is a problem of missing collection;

[0005] 2. 2D machine vision online inspection technology is significantly affected by ambient light and requires complex shading devices to shield natural light, increasing system complexity and impacting the normal working process of the production line.

[0006] 3. Deep learning algorithm models based on 2D images require extensive training to achieve usable accuracy, which is labor-intensive and time-consuming to master.

[0007] 4. The results of 2D machine vision inspection only provide information on the type of defects, but cannot provide information on the severity of the defects, such as depth, area, location, etc. In fact, it is impossible to accurately assess the grade of the steel plate, and it is difficult to guide the repair work in subsequent processes.

[0008] On the other hand, although existing defect detection technology uses 3D machine vision technology to solve the detection problems of 2D machine vision technology, when shipbuilding steel plates move on the roller bed, the interaction between them and the rollers will produce unpredictable and uneven jitter. The jitter range of steel plates of different thicknesses and sizes in the normal direction of the roller bed transport surface is between 0.2mm and 0.4mm. At present, the detection accuracy requirement for the depth of discontinuous defects on the surface of shipbuilding steel plates is 0.1mm, that is, the detection system should be able to distinguish the depth difference of 0.1mm. When the jitter of the steel plate exceeds 0.2mm, the use of 3D machine vision technology for online measurement can no longer meet the needs of high-precision steel plate detection. Summary of the Invention

[0009] The purpose of the present invention is to overcome the deficiencies in the above-mentioned prior art and provide a method for eliminating jitter interference in 3D visual online detection of ship steel plates. The method of the present invention solves the problem that when using 3D machine vision technology for online detection of steel plate surface discontinuities, small-scale surface discontinuity defects cannot be found from the collected 3D point cloud data due to steel plate jitter interference, thereby improving the accuracy of steel plate surface defect detection.

[0010] In order to achieve the above invention purpose, the technical solution provided by the present invention is as follows:

[0011] A method for eliminating jitter interference in 3D visual online detection of ship steel plates, the method specifically comprising the following steps:

[0012] In the first step, 3D image acquisition sensors are placed above and below the roller bed. The steel plate to be inspected is placed on the roller bed and driven by the roller bed.

[0013] In the second step, a synchronous encoder is placed above the roller bed to measure the running speed of the steel plate to be tested on the roller bed. At the same time, the synchronous encoder sends synchronous image commands to the 3D image acquisition sensors above and below the roller bed.

[0014] In the third step, after receiving the synchronous image instruction, the 3D image acquisition sensors above and below the roller bed simultaneously acquire 3D point cloud images of the upper and lower surfaces of the inspection steel plate;

[0015] In the fourth step, a point cloud data set of the upper and lower surfaces of the steel plate to be inspected is obtained from the 3D point cloud image in the third step, and a low-pass filter is performed on the obtained point cloud data set to obtain a standard point cloud sequence;

[0016] Step 5: Select a reference point from the standard point cloud sequence obtained in the fourth step, and calculate the amplitude value through the standard point cloud sequence and the reference point;

[0017] Step 6: After the amplitude value is calculated, the amplitude difference of the corresponding standard point cloud sequence is calculated based on the amplitude value;

[0018] Step 7: After the amplitude difference calculation is completed, the stationary point cloud sequence U(y) corresponding to the point cloud sequence in the fourth step is calculated when the roller bed is stationary. n ,u n ), D(y n , d n );

[0019] Step 8: After the calculation of the static point cloud sequence is completed, the subsequent steel plate surface quality inspection process is carried out.

[0020] The 3D image acquisition sensors arranged above and below the above-mentioned roller bed are in the same vertical plane, and the synchronous encoder enables the 3D image acquisition sensors arranged above and below the roller bed to synchronously obtain 3D point cloud images of the upper and lower surfaces of the steel plate of the same steel plate cross-section.

[0021] The point cloud data set in the fourth step is specifically P{P0, P1, ...P N}, P′{P′ n , P′1,...P′ N}, where P is the coordinate of the upper surface point of the same steel plate section, P′ is the coordinate of the lower surface point of the same steel plate section, and P n (y n , z n ), P′ n (y n , z′ n ) are the coordinates of the upper surface point in the same steel plate section and the coordinates of the lower surface point in the same steel plate section corresponding to the coordinates of the upper surface point.

[0022] The specific steps for low-pass filtering the above-mentioned point cloud data set are as follows: comparing the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section in the acquired point cloud data set with the thickness of the steel plate to be detected; when the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section and the difference in thickness of the steel plate to be detected are in the range of 0.2mm to 0.4mm, the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section are a standard point cloud sequence.

[0023] The selection of reference points in the fourth step is specifically to select a reference point at the coordinates of the upper surface point of the same section of the steel plate to be tested and the coordinates of the lower surface point of the same section, and the Y coordinates of the two reference points are the same; the reference point of the same section of the steel plate to be tested is fixed and unchanged.

[0024] The specific steps for calculating the amplitude value through the standard point cloud sequence and the reference point in the above fifth step are: subtracting the Z coordinates of all point coordinates in the standard point cloud sequence from the Z coordinates of the reference point coordinates to obtain the amplitude values ​​of all point coordinates in the standard point cloud sequence.

[0025] The specific steps for calculating the amplitude difference of the corresponding standard point cloud sequence through the amplitude value in the above sixth step are: subtract the amplitude value of the upper surface point coordinate and the reference point in the standard point cloud sequence from the amplitude value of the lower surface coordinate with the same Y coordinate as the upper surface point coordinate in the standard point cloud sequence and the reference point to obtain the amplitude difference corresponding to the upper surface and lower surface in the same section of the steel plate.

[0026] The stationary point cloud sequence U(y n ,u n ), D(y n , d n The specific calculation formula is:

[0027]

[0028]

[0029] Wherein, U is the coordinate of the upper surface of the same section of the steel plate to be detected in a static state, D is the coordinate of the lower surface of the same section of the steel plate to be detected in a static state; Δ is the set threshold value, which is the detection accuracy value of the depth of the discontinuous defect on the surface of the steel plate to be detected, Z n -Z0 is the upper surface amplitude value of the same section of the steel plate to be tested, Z0 is the reference point, Z′ n -Z′0 is the amplitude value of the lower surface of the same section of the steel plate to be tested, and Z′0 is the reference point; (Z n -Z0)-(Z′ n -Z′0) is the amplitude difference.

[0030] Based on the above technical solution, the present invention's patented method for 3D visual online detection of ship steel plates and its elimination of jitter interference has achieved the following technical advantages through practical application:

[0031] 1. The present invention provides a method for eliminating jitter interference in 3D visual online detection of ship steel plates, which solves the problem of being unable to detect small-scale surface discontinuities from the collected 3D point cloud data due to jitter interference of the steel plates when using 3D machine vision technology for online detection of steel plate surface discontinuities, thereby improving the accuracy of steel plate surface defect detection.

[0032] 2. The present invention provides a method for online 3D visual detection of jitter interference and its elimination for ship steel plates. By utilizing 3D machine vision technology, the method is less affected by ambient light and can work normally in an indoor workshop environment, thereby improving applicability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a point cloud data set acquisition structure diagram in a method for 3D visual online detection of jitter interference and elimination of ship steel plates according to the present invention. DETAILED DESCRIPTION

[0034] To make the objectives, technical solutions, and advantages of the present invention more clearly apparent, the present invention is described below using specific examples shown in the accompanying drawings. However, it should be understood that these descriptions are merely illustrative and are not intended to limit the scope of the present invention. In addition, in the following description, descriptions of well-known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present invention.

[0035] like Figure 1 The present invention relates to a method for eliminating jitter interference in 3D visual online detection of ship steel plates, which specifically includes the following steps:

[0036] In the first step, 3D image acquisition sensors 1 are arranged above and below the roller bed respectively. The steel plate 3 to be inspected is placed on the roller bed and is driven by the roller bed to move the steel plate.

[0037] In the second step, a synchronous encoder 2 is arranged above the roller bed to measure the running speed of the steel plate 3 to be tested on the roller bed, and at the same time, the synchronous encoder 2 sends a synchronous image command to the 3D image acquisition sensor 1 above and below the roller bed;

[0038] In the third step, after the 3D image acquisition sensors 1 above and below the roller bed receive the synchronous image instruction, the 3D image acquisition sensors 1 above and below the roller bed simultaneously acquire 3D point cloud images of the upper and lower surfaces of the steel plate under inspection;

[0039] In the fourth step, a point cloud data set of the upper and lower surfaces of the steel plate 3 to be inspected is obtained from the 3D point cloud image in the third step, and a low-pass filter is performed on the obtained point cloud data set to obtain a standard point cloud sequence;

[0040] Step 5: Select a reference point from the standard point cloud sequence obtained in the fourth step, and calculate the amplitude value through the standard point cloud sequence and the reference point;

[0041] Step 6: After the amplitude value is calculated, the amplitude difference of the corresponding standard point cloud sequence is calculated based on the amplitude value;

[0042] Step 7: After the amplitude difference calculation is completed, the stationary point cloud sequence U(y) corresponding to the point cloud sequence in the fourth step is calculated when the roller bed is stationary. n ,u n ), D(y n , d n );

[0043] Step 8: After the static point cloud sequence calculation is completed, the subsequent steel plate surface quality inspection process is carried out.

[0044] The 3D image acquisition sensors 1 respectively arranged above and below the above-mentioned roller bed are in the same vertical plane, and the synchronous encoder 2 enables the 3D image acquisition sensors 1 respectively arranged above and below the roller bed to synchronously obtain 3D point cloud images of the upper and lower surfaces of the steel plate of the same steel plate cross-section.

[0045] The point cloud data set in the fourth step is specifically P{P0, P1, ...P N}, P{P′0,P′1,...,P′ N}, where P is the coordinate of the upper surface point of the same steel plate section, P′ is the coordinate of the lower surface point of the same steel plate section, and P n (y n , z n ), P′ n (y n , z′ n ) are the coordinates of the upper surface point in the same steel plate section and the coordinates of the lower surface point in the same steel plate section corresponding to the coordinates of the upper surface point.

[0046] The specific steps for low-pass filtering the above-mentioned point cloud data set are as follows: comparing the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section in the acquired point cloud data set with the thickness of the steel plate to be detected; when the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section and the difference in thickness of the steel plate to be detected are in the range of 0.2mm to 0.4mm, the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section are a standard point cloud sequence.

[0047] The selection of reference points in the above fourth step is specifically to select a reference point at the upper surface point coordinates and the lower surface point coordinates of the same section of the steel plate 3 to be tested, and the Y coordinates of the two reference points are the same; the reference point of the same section of the steel plate 3 to be tested is fixed and unchanged.

[0048] The specific steps for calculating the amplitude value through the standard point cloud sequence and the reference point in the above fifth step are: subtracting the Z coordinates of all point coordinates in the standard point cloud sequence from the Z coordinates of the reference point coordinates to obtain the amplitude values ​​of all point coordinates in the standard point cloud sequence.

[0049] The specific steps for calculating the amplitude difference of the corresponding standard point cloud sequence through the amplitude value in the above sixth step are: subtract the amplitude value of the upper surface point coordinate and the reference point in the standard point cloud sequence from the amplitude value of the lower surface coordinate with the same Y coordinate as the upper surface point coordinate in the standard point cloud sequence and the reference point to obtain the amplitude difference corresponding to the upper surface and lower surface in the same section of the steel plate.

[0050] The stationary point cloud sequence U(y n ,u n ), Dn(y n , d n The specific calculation formula is:

[0051]

[0052]

[0053] Wherein, U is the coordinate of the upper surface of the same cross section of the steel plate 3 to be detected in a static state, D is the coordinate of the lower surface of the same cross section of the steel plate 3 to be detected in a static state; Δ is the set threshold value, which is the detection accuracy value of the depth of the discontinuous defect on the surface of the steel plate to be detected, Z n -Z0 is the upper surface amplitude value of the same section of the steel plate 3 to be tested, Z0 is the reference point, Z′ n -Z′0 is the amplitude value of the lower surface of the same section of the steel plate 3 to be tested, and Z′0 is the reference point; (Z n -Z0)-(Z′ n -Z′0) is the amplitude difference; this solves the problem of being unable to detect small-scale surface discontinuities from the collected 3D point cloud data due to steel plate jitter interference when using 3D machine vision technology for online detection of steel plate surface discontinuities, thereby improving the accuracy of steel plate surface defect detection.

[0054] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solutions of the present invention. They should all be included in the scope of the technical solutions for which protection is sought in the present invention.

Claims

1. A method for eliminating jitter interference in 3D visual online detection of shipbuilding steel plates, characterized in that: The method specifically comprises the following steps: In the first step, 3D image acquisition sensors (1) are respectively arranged above and below the roller bed, and the steel plate (3) to be detected is placed on the roller bed and driven by the roller bed to move the steel plate; In the second step, a synchronous encoder (2) is arranged above the roller bed to measure the running speed of the steel plate (3) to be tested on the roller bed, and at the same time, the synchronous encoder (2) sends a synchronous image instruction to the 3D image acquisition sensor (1) above and below the roller bed; In the third step, after the 3D image acquisition sensors (1) above and below the roller bed receive the synchronous image instruction, the 3D image acquisition sensors (1) above and below the roller bed simultaneously acquire 3D point cloud images of the upper and lower surfaces of the detection steel plate; In the fourth step, a point cloud data set of the upper surface and the lower surface of the steel plate (3) to be inspected is obtained from the 3D point cloud image in the third step, and a low-pass filtering is performed on the obtained point cloud data set to obtain a standard point cloud sequence; Step 5: Select a reference point from the standard point cloud sequence obtained in the fourth step, and calculate the amplitude value through the standard point cloud sequence and the reference point; Step 6: After the amplitude value is calculated, the amplitude difference of the corresponding standard point cloud sequence is calculated based on the amplitude value; Step 7: After the amplitude difference calculation is completed, the stationary point cloud sequence U(y) corresponding to the point cloud sequence in the fourth step is calculated when the roller bed is stationary. n ,u n ), D(y n , d n ); Step 8: After the static point cloud sequence calculation is completed, the subsequent steel plate surface quality inspection process is carried out.

2. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The 3D image acquisition sensors (1) respectively arranged above and below the roller bed are located in the same vertical plane, and the synchronous encoder (2) enables the 3D image acquisition sensors (1) respectively arranged above and below the roller bed to synchronously acquire 3D point cloud images of the upper and lower surfaces of the steel plate of the same steel plate cross section.

3. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The point cloud data set in the fourth step is specifically P{P0, P1, ...P N }, P′{P′0, P′1,...P′ N }, where P is the coordinate of the upper surface point of the same steel plate section, P′ is the coordinate of the lower surface point of the same steel plate section, and P n (y n , z n ), P′ n (y n , z′ n ) are the coordinates of the upper surface point in the same steel plate section and the coordinates of the lower surface point in the same steel plate section corresponding to the coordinates of the upper surface point.

4. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 3 is characterized in that: The specific steps of low-pass filtering the point cloud data set are: comparing the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section in the acquired point cloud data set with the thickness of the steel plate to be detected; when the difference in Z coordinates of the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section is in the range of 0.2mm to 0.4mm from the thickness of the steel plate to be detected, the corresponding upper surface point coordinates and lower surface point coordinates in the same steel plate section are a standard point cloud sequence.

5. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The selection of the reference point in the fourth step is specifically to select a reference point at the coordinates of the upper surface point of the same cross section of the steel plate (3) to be tested and the coordinates of the lower surface point of the same cross section, and the Y coordinates of the two reference points are the same; the reference point of the same cross section of the steel plate (3) to be tested is fixed and unchanged.

6. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The specific steps of calculating the amplitude value through the standard point cloud sequence and the reference point in the fifth step are: subtracting the Z coordinates of all point coordinates in the standard point cloud sequence from the Z coordinates of the reference point coordinates to obtain the amplitude values ​​of all point coordinates in the standard point cloud sequence.

7. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The specific steps of calculating the amplitude difference of the corresponding standard point cloud sequence through the amplitude value in the sixth step are: subtracting the amplitude value of the lower surface coordinates with the same Y coordinate as the upper surface point coordinate in the standard point cloud sequence and the reference point from the amplitude value of the upper surface point coordinate in the standard point cloud sequence and the reference point to obtain the amplitude difference corresponding to the upper surface and lower surface in the same section of the steel plate.

8. The method for eliminating jitter interference during 3D visual online inspection of shipbuilding steel plates according to claim 1 is characterized in that: The seventh step is to use the stationary point cloud sequence U(y n ,u n ), D(y n , d n The specific calculation formula is: Wherein, U is the coordinate of the upper surface of the same cross section of the steel plate (3) to be detected in a static state, D is the coordinate of the lower surface of the same cross section of the steel plate (3) to be detected in a static state; Δ is a set threshold value, which is the detection accuracy value of the depth of the discontinuous defect on the surface of the steel plate to be detected, Z n -Z0 is the upper surface amplitude value of the same section of the steel plate (3) to be tested, Z0 is the reference point, Z′ n -Z′0 is the amplitude value of the lower surface of the same section of the steel plate (3) to be tested, and Z′0 is the reference point; (Z n -Z0)-(Z′ n -Z′0) is the amplitude difference.

Images