A method for measuring three-dimensional topography of a turbid water environment based on underwater binocular line structured light

Abstract

The application discloses a kind of muddy water environment three-dimensional topography measurement methods based on underwater binocular line structure light, which comprises the following steps: S1, build binocular vision three-dimensional topography measurement system based on line structure light scanning, and calibrate binocular camera using calibration plate in underwater environment;S2, line laser light source is placed on rotating or translating platform, so that line laser scanning covers the measured area.In the process of line laser scanning, left and right binocular cameras synchronously collect the sequence line structure light image of measured area;S3, the collected sequence line structure light image is processed by Gaussian band-pass filtering;And the contrast of the sequence line structure light image after filtering is enhanced by non-linear enhancement algorithm Processing;S4, the sequence line structure light image after enhancement is processed, and the sub-pixel center of laser light strip is extracted;S5, using binocular vision technology, based on camera calibration parameters and the sub-pixel center pixel coordinates of laser light strip extracted from the sequence binocular line structure light image, reconstruct the three-dimensional point cloud of measured area, complete three-dimensional topography reconstruction.

Description

Technical Field

[0001] This invention relates to the field of underwater three-dimensional visual measurement, specifically to a method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light. Background Technology

[0002] Underwater 3D topography measurement plays an indispensable role in marine resource exploration, underwater engineering inspection (such as dams, bridge piers, and pipelines), shipwreck archaeology, ecological environment monitoring, and safety. Acoustic-based underwater 3D topography measurement has limited resolution and low 3D reconstruction accuracy, making it difficult to meet the needs of refined underwater facility inspection. Vision-based methods using active structured light, with their significant advantages of non-contact, full-field measurement, and high resolution, show promise for high-precision underwater 3D topography measurement. However, natural aquatic environments, especially turbid waters such as inland rivers and near-shore ports, can degrade image quality, posing a serious challenge to vision-based 3D topography measurement.

[0003] In turbid water environments, due to the severe absorption and scattering of light by a large number of suspended particles, most active 3D vision measurement methods, such as structured light and speckle projection, struggle to achieve effective image projection and feature recognition. The incident structured light pattern is rapidly attenuated and blurred during propagation, resulting in extremely low contrast and feature degradation in the captured image, making reliable decoding and matching impossible.

[0004] Line structured light, due to its concentrated energy, strong anti-scattering ability, and ease of extraction and identification, has the potential for high-precision three-dimensional measurements in turbid water environments. However, traditional line structured light methods still face problems such as unstable fringe signal-to-noise ratio and severe scattering noise interference in turbid water, which limit their measurement accuracy and robustness. Summary of the Invention

[0005] Objective: This invention provides a method for measuring the three-dimensional topography of objects in turbid water environments based on underwater binocular line-structured light, enabling high-precision three-dimensional topography of the surface of the object under test in turbid water. The method effectively optimizes the imaging quality of line-structured light in turbid water, providing a systematic solution for line-structured light image filtering, enhancement, and reliable feature extraction, significantly improving the robustness and accuracy of three-dimensional topography measurement in turbid water.

[0006] Technical Solution: To achieve the above-mentioned objectives, this invention proposes a method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light. This method includes the following steps:

[0007] S1. Construct an underwater binocular vision measurement system. The system includes an underwater binocular camera, a line structured light source, and a rotating or translating platform. The line structured light source is a waterproof device. Place the line structured light source on the rotating or translating platform. Each of the two cameras is separately encapsulated in a waterproof shell. The observation window of the waterproof shell is made of a flat transparent material, and the optical axis of the camera is perpendicular to the plane of the observation window. Use a calibration board to calibrate the underwater binocular camera in the underwater environment, and then point the underwater binocular camera at the area to be measured.

[0008] S2. Use a light source to project a line laser onto the area to be measured, and rotate or translate the platform so that the line laser sweeps across the area to be measured; during the line laser scanning process, a binocular camera simultaneously acquires a sequence of line structured light images of the area to be measured;

[0009] S3. Perform Gaussian bandpass filtering on the acquired sequence of structured light images: Perform Fourier transform on each image to obtain the spectral centroid and frequency radius of the image. Calculate adaptive low-pass and high-pass frequency domain parameters using the spectral centroid and frequency radius. Use the Gaussian bandpass filtering algorithm to process the image and suppress background scattering and noise. Enhance the contrast of the filtered sequence of structured light images using a nonlinear enhancement algorithm.

[0010] S4. Use the Steger algorithm to process the enhanced sequence of structured light images to obtain the sub-pixel coordinates of the center of the laser stripe in the sequence image;

[0011] S5. Based on the camera calibration parameters and the sub-pixel coordinates of the laser light stripe center extracted from the sequence of binocular structured light images, the three-dimensional point cloud of the measured area is reconstructed to complete the three-dimensional topography reconstruction.

[0012] Furthermore, the image Gaussian bandpass filtering method and steps described in step S3 are as follows:

[0013] 3.1 Perform Fourier transform on the image:

[0014] ;

[0015] Wherein, the image resolution is M×N pixels; x, y represent the pixel coordinates in the horizontal and vertical directions of the image spatial domain, respectively; f(x,y) is the pixel gray value at coordinate (x,y); u, v represent the frequency components of the image in the horizontal and vertical directions, respectively; F(u,v) is a complex number representing the amplitude and phase information at frequency (u,v);

[0016] 3.2. Calculate the frequency domain characteristics based on the Fourier transform results. The distance from the center of the spectrum to any frequency point is the frequency radius D, which is:

[0017] ;

[0018] The spectral centroid C, which reflects the frequency distribution center of image energy, is:

[0019] ;

[0020] The spectral bandwidth B, which describes the degree of dispersion in the frequency distribution, is:

[0021] ;

[0022] 3.3. Design a Gaussian bandpass filter based on frequency domain characteristics. The low-pass frequency domain parameter σ1 is designed as a linear combination of the spectral centroid and the spectral bandwidth.

[0023] ;

[0024] Where the coefficient α=1, =1.2, the high-pass frequency domain parameter σ2 is designed as follows:

[0025] ;

[0026] Where the coefficient γ = 0.8, the Gaussian bandpass filter is in the form of:

[0027] ;

[0028] 3.4. Filter the image by multiplying the filter by the image spectrum:

[0029] ;

[0030] Performing a two-dimensional discrete inverse Fourier transform on G(u,v) yields the filtered image I(x,y):

[0031] ;

[0032] 3.5 The image nonlinear enhancement algorithm described in step S3 specifically involves: achieving differentiated enhancement of local regions with different brightness levels through adaptive gamma parameters related to grayscale values, such that the enhanced grayscale value I of a certain pixel in the image is... enhanced (x,y) is:

[0033] ;

[0034] Where I(x, y) is the pixel gray value at image coordinates (x, y) in the local region after filtering, and max(I) is the maximum gray value in the local region.

[0035] Furthermore, the method for extracting the sub-pixel center of the laser stripe in step S4 is the Steger method:

[0036] First, calculate the Hessian matrix of the line structured light image. The Hessian matrix is ​​obtained by convolving the grayscale image with a two-dimensional Gaussian differential kernel G:

[0037] ;

[0038] in, , , These represent the second-order partial derivative operation with respect to the x-direction of the image, the first-order partial derivative operation with respect to the y-direction after performing the first-order partial derivative operation with respect to the x-direction of the image, and the second-order partial derivative operation with respect to the y-direction of the image, respectively. , , These represent the results of taking the second-order partial derivative with respect to the x-direction of the image, the first-order partial derivative with respect to the y-direction after taking the first-order partial derivative with respect to the x-direction of the image, and the second-order partial derivative with respect to the y-direction of the image, respectively; the eigenvector corresponding to the largest eigenvalue of the Hessian matrix is ​​calculated as (n x n y The eigenvectors represent directions corresponding to the normal direction of the laser stripe. Let (x0, y0) be the integer pixel coordinates of a point in the laser stripe, and the sub-pixel coordinates (u, v) of this point be:

[0039] ;

[0040] in

[0041] ;

[0042] Where, r x r y Given the first-order partial derivatives of the image along x and y, to ensure that the sub-pixel position is within the selected integer pixel reference point, the following condition must be met:

[0043] .

[0044] Furthermore, the method for completing the three-dimensional topography reconstruction in step S5 is as follows:

[0045] Based on the binocular vision epipolar constraint criterion, the sub-pixel center coordinates of the laser beams in the sequence of binocular structured light images extracted in step S4 are matched; for any feature point on the laser beam, it is assumed that (u l ,v l ) represents the subpixel center coordinates of this point in the left camera image, (u r , v r () represents the center coordinates of the matching subpixel in the synchronously acquired image of the right camera.

[0046] Based on the principle of binocular vision triangulation, combined with camera calibration parameters and the corresponding sub-pixel center coordinates of a single set of matching, the three-dimensional spatial coordinates of the single feature point in the world coordinate system are calculated to complete the three-dimensional position reconstruction of the single point.

[0047] The three-dimensional coordinates of all feature points of the laser beam in a single frame image are calculated sequentially to obtain the three-dimensional point set corresponding to the laser beam in a single frame. The three-dimensional point sets of all frames in the full scanning time sequence are stitched and fused together. After removing outlier noise points and duplicate points, the complete three-dimensional point cloud of the measured area is obtained, and the three-dimensional morphology reconstruction of the muddy water environment is completed.

[0048] Furthermore, this invention proposes a three-dimensional topography measurement device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the aforementioned method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light.

[0049] Furthermore, this invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the aforementioned method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light. Attached Figure Description

[0050] Figure 1 This is a flowchart of the method.

[0051] Figure 2 This is a diagram showing the effects of filtering and enhancement processing on line structured light images.

[0052] Figure 3 This is a diagram showing the result of the 3D topographic reconstruction. Detailed Implementation

[0053] The invention will be further described below with reference to specific implementation examples.

[0054] like Figure 1 As shown, this invention proposes a method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light. This method includes the following steps:

[0055] S1. Construct an underwater binocular vision measurement system. The system includes an underwater binocular camera, a line structured light source, and a rotating or translating platform. The line structured light source is a waterproof device. Place the line structured light source on the rotating or translating platform. Each of the two cameras is separately encapsulated in a waterproof shell. The observation window of the waterproof shell is made of a flat transparent material, and the optical axis of the camera is perpendicular to the plane of the observation window. Use a calibration board to calibrate the underwater binocular camera in the underwater environment, and then point the underwater binocular camera at the area to be measured.

[0056] S2. Use a light source to project a line laser onto the area to be measured, and rotate or translate the platform so that the line laser sweeps across the area to be measured; during the line laser scanning process, a binocular camera simultaneously acquires a sequence of line structured light images of the area to be measured;

[0057] S3. Perform Gaussian bandpass filtering on the acquired sequence of structured light images: Perform Fourier transform on each image to obtain the spectral centroid and frequency radius of the image. Calculate adaptive low-pass and high-pass frequency domain parameters using the spectral centroid and frequency radius. Use the Gaussian bandpass filtering algorithm to process the image and suppress background scattering and noise. Enhance the contrast of the filtered sequence of structured light images using a nonlinear enhancement algorithm.

[0058] S4. Use the Steger algorithm to process the enhanced sequence of structured light images to obtain the sub-pixel coordinates of the center of the laser stripe in the sequence image;

[0059] S5. Based on the camera calibration parameters and the sub-pixel coordinates of the laser light stripe center extracted from the sequence of binocular structured light images, the three-dimensional point cloud of the measured area is reconstructed to complete the three-dimensional topography reconstruction.

[0060] The experiment was conducted in turbid water with a turbidity of 100-200 NTU. The original line structured light images are attached. Figure 2 As shown in the first image. Gaussian bandpass filtering is applied to the acquired sequence of structured light images. First, a Fourier transform is performed on the image:

[0061]

[0062] Wherein, the image resolution is M×N pixels; x, y represent the pixel coordinates in the horizontal and vertical directions of the image spatial domain, respectively; f(x,y) is the pixel gray value at coordinate (x,y); u, v represent the frequency components of the image in the horizontal and vertical directions, respectively; F(u,v) is a complex number representing the amplitude and phase information at frequency (u,v).

[0063] The frequency domain characteristics are calculated based on the Fourier transform results. First, the distance from the center of the spectrum to any frequency point is the frequency radius D, which is:

[0064]

[0065] The spectral centroid C, which reflects the frequency distribution center of image energy, is:

[0066]

[0067] The spectral bandwidth B, which describes the degree of dispersion in the frequency distribution, is:

[0068]

[0069] A Gaussian bandpass filter is designed based on its frequency domain characteristics. The low-pass frequency domain parameter σ1 is designed as a linear combination of the spectral centroid and the spectral bandwidth.

[0070]

[0071] Where the coefficients α = 1, β = 1.2, and the high-pass frequency domain parameter σ2 is designed as follows:

[0072]

[0073] Where the coefficient γ = 0.8. The Gaussian bandpass filter has the following form:

[0074]

[0075] To filter the image, multiply the filter by the image spectrum:

[0076] ;

[0077] Performing a two-dimensional discrete inverse Fourier transform on G(u,v) yields the filtered image I(x,y):

[0078] ;

[0079] The image of the line structured light after Gaussian bandpass filtering is attached. Figure 3 As shown in the second image.

[0080] Contrast enhancement is performed on the filtered sequence of structured light images using a nonlinear enhancement algorithm; differential enhancement of local regions with different brightness is achieved through an adaptive gamma parameter related to grayscale values, and the enhanced grayscale value I of a certain pixel in the image is obtained. enhanced (x, y) is:

[0081]

[0082] Where I(x, y) is the pixel grayscale value at coordinates (x, y) in the filtered local region, and max(I) is the maximum grayscale value in the local region. The nonlinearly enhanced line structured light image is shown in the attached figure. Figure 3 As shown in the third image.

[0083] The enhanced sequence of structured light images is further processed to extract the sub-pixel centers of the laser beams. The method for extracting the sub-pixel centers of the laser beams is the Steger method, specifically:

[0084] First, calculate the Hessian matrix of the line structured light image. The Hessian matrix is ​​obtained by convolving the grayscale image with a two-dimensional Gaussian differential kernel G:

[0085] ;

[0086] in, , , These represent the second-order partial derivative operation with respect to the x-direction of the image, the first-order partial derivative operation with respect to the y-direction after performing the first-order partial derivative operation with respect to the x-direction of the image, and the second-order partial derivative operation with respect to the y-direction of the image, respectively. , , These represent the results of taking the second-order partial derivative with respect to the x-direction of the image, the first-order partial derivative with respect to the y-direction after taking the first-order partial derivative with respect to the x-direction of the image, and the second-order partial derivative with respect to the y-direction of the image, respectively; the eigenvector corresponding to the largest eigenvalue of the Hessian matrix is ​​calculated as (n x n y The eigenvectors represent directions corresponding to the normal direction of the laser stripe. Let (x0, y0) be the integer pixel coordinates of a point in the laser stripe, and the sub-pixel coordinates (u, v) of this point be:

[0087] ;

[0088] in

[0089] ;

[0090] Where, r x r y Given the first-order partial derivatives of the image along x and y, to ensure that the sub-pixel position is within the selected integer pixel reference point, the following condition must be met:

[0091] .

[0092] The method for completing the three-dimensional topography reconstruction in step S5 is as follows:

[0093] Based on the binocular vision epipolar constraint criterion, the sub-pixel center coordinates of the laser beams in the sequence of binocular structured light images extracted in step S4 are matched; for any feature point on the laser beam, it is assumed that (u l ,v l ) represents the subpixel center coordinates of this point in the left camera image, (u r ,v r () represents the center coordinates of the matching subpixel in the synchronously acquired image of the right camera.

[0094] Based on the principle of binocular vision triangulation, combined with camera calibration parameters and the corresponding sub-pixel center coordinates of a single set of matching, the three-dimensional spatial coordinates of the single feature point in the world coordinate system are calculated to complete the three-dimensional position reconstruction of the single point.

[0095] For each feature point of the laser beam in a single frame image, the 3D coordinates of each point are calculated sequentially to obtain the 3D point set corresponding to the laser beam in that frame. The 3D point sets of all frames within the full scan sequence are then stitched together and fused. After removing outlier noise points and duplicate points, a complete 3D point cloud of the measured area is obtained, completing the high-precision 3D topography reconstruction of the turbid water environment. The reconstruction results are attached. Figure 3 As shown.

[0096] Furthermore, this invention proposes a three-dimensional topography measurement device, which includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the aforementioned high-precision three-dimensional topography measurement method for turbid water environments based on underwater binocular structured light.

[0097] Furthermore, this invention proposes a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the aforementioned method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light.

Claims

1. A method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light, characterized in that, The method includes the following steps: S1. Construct an underwater binocular vision measurement system, which includes an underwater binocular camera, a line structured light source, and a rotating or translating platform. Place the line structured light source on the rotating or translating platform. Each of the two cameras is separately encapsulated in a waterproof housing. The observation window of the waterproof housing is made of a flat transparent material, and the optical axis of the camera is perpendicular to the plane of the observation window. Use a calibration plate to calibrate the underwater binocular camera in an underwater environment, and then point the underwater binocular camera at the area to be measured. S2. Use a light source to project a line laser onto the area to be measured, and rotate or translate the platform so that the line laser sweeps across the area to be measured; during the line laser scanning process, a binocular camera simultaneously acquires a sequence of line structured light images of the area to be measured; S3. Perform Gaussian bandpass filtering on the acquired sequence of structured light images: Perform Fourier transform on each image to obtain the spectral centroid and frequency radius of the image. Calculate adaptive low-pass and high-pass frequency domain parameters using the spectral centroid and frequency radius. Use the Gaussian bandpass filtering algorithm to process the image and suppress background scattering and noise. Enhance the contrast of the filtered sequence of structured light images using a nonlinear enhancement algorithm. S4. Use the Steger algorithm to process the enhanced sequence of structured light images to obtain the sub-pixel coordinates of the center of the laser stripe in the sequence image; S5. Based on the camera calibration parameters and the sub-pixel coordinates of the laser light stripe center extracted from the sequence of binocular structured light images, the three-dimensional point cloud of the measured area is reconstructed to complete the three-dimensional topography reconstruction.

2. The method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light according to claim 1, characterized in that, The image Gaussian bandpass filtering method and steps described in step S3 are as follows: 3.1 Perform Fourier transform on the image: ； Wherein, the image resolution is M×N pixels; x, y represent the pixel coordinates in the horizontal and vertical directions of the image spatial domain, respectively; f(x,y) is the pixel gray value at coordinate (x,y); u, v represent the frequency components of the image in the horizontal and vertical directions, respectively; F(u,v) is a complex number representing the amplitude and phase information at frequency (u,v); 3.

2. Calculate the frequency domain characteristics based on the Fourier transform results. The distance from the center of the spectrum to any frequency point is the frequency radius D, which is: ； The spectral centroid C, which reflects the frequency distribution center of image energy, is: ； The spectral bandwidth B, which describes the degree of dispersion in the frequency distribution, is: ； 3.

3. Design a Gaussian bandpass filter based on frequency domain characteristics. The low-pass frequency domain parameter σ1 is designed as a linear combination of the spectral centroid and the spectral bandwidth. ； Where the coefficient α=1, =1.2, the high-pass frequency domain parameter σ2 is designed as follows: ； Where the coefficient γ = 0.8, the Gaussian bandpass filter is in the form of: ； 3.

4. Filter the image by multiplying the filter by the image spectrum: ； Performing a two-dimensional discrete inverse Fourier transform on G(u,v) yields the filtered image I(x,y): ； 3.5 The image nonlinear enhancement algorithm described in step S3 specifically involves: achieving differentiated enhancement of local regions with different brightness levels through adaptive gamma parameters related to grayscale values, such that the enhanced grayscale value I of a certain pixel in the image is... enhanced (x,y) is: ； Where I(x, y) is the pixel gray value at image coordinates (x, y) in the local region after filtering, and max(I) is the maximum gray value in the local region.

3. The method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light according to claim 2, characterized in that, The method for extracting the sub-pixel center of the laser stripe in step S4 is the Steger method: First, calculate the Hessian matrix of the line structured light image. The Hessian matrix is ​​obtained by convolving the grayscale image with a two-dimensional Gaussian differential kernel G: ； in, , , These represent the second-order partial derivative operation with respect to the x-direction of the image, the first-order partial derivative operation with respect to the y-direction after performing the first-order partial derivative operation with respect to the x-direction of the image, and the second-order partial derivative operation with respect to the y-direction of the image, respectively. , , These represent the results of taking the second-order partial derivative with respect to the x-direction of the image, the first-order partial derivative with respect to the y-direction after taking the first-order partial derivative with respect to the x-direction of the image, and the second-order partial derivative with respect to the y-direction of the image, respectively; the eigenvector corresponding to the largest eigenvalue of the Hessian matrix is ​​calculated as (n x n y The eigenvectors represent directions corresponding to the normal direction of the laser stripe. Let (x0, y0) be the integer pixel coordinates of a point in the laser stripe, and the sub-pixel coordinates (u, v) of this point be: ； in ； Where, r x r y Given the first-order partial derivatives of the image along x and y, to ensure that the sub-pixel position is within the selected integer pixel reference point, the following condition must be met: 。 4. The method for measuring the three-dimensional topography of turbid water environments based on underwater binocular structured light according to claim 3, characterized in that, The method for completing the three-dimensional topography reconstruction in step S5 is as follows: Based on the binocular vision epipolar constraint criterion, the sub-pixel center coordinates of the laser beams in the sequence of binocular structured light images extracted in step S4 are matched; for any feature point on the laser beam, it is assumed that (u l ,v l ) represents the subpixel center coordinates of this point in the left camera image, (u r , v r () represents the center coordinates of the matching subpixel in the synchronously acquired image of the right camera. Based on the principle of binocular vision triangulation, combined with camera calibration parameters and the corresponding sub-pixel center coordinates of a single set of matching, the three-dimensional spatial coordinates of the single feature point in the world coordinate system are calculated to complete the three-dimensional position reconstruction of the single point. The three-dimensional coordinates of all feature points of the laser beam in a single frame image are calculated sequentially to obtain the three-dimensional point set corresponding to the laser beam in a single frame. The three-dimensional point sets of all frames in the full scanning time sequence are stitched and fused together. After removing outlier noise points and duplicate points, the complete three-dimensional point cloud of the measured area is obtained, and the three-dimensional morphology reconstruction of the muddy water environment is completed.

5. A three-dimensional topography measurement device, characterized in that, The device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When executed by the processor, the computer program implements the steps of any of the underwater binocular structured light-based three-dimensional topography measurement methods for turbid water environments as described in claims 1-4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the three-dimensional topography measurement methods for turbid water environments based on underwater binocular structured light as described in claims 1-4.

Images