Underwater three-dimensional construction method based on fusion of binocular camera and multi-beam sonar

Abstract

This invention discloses an underwater 3D reconstruction method based on the fusion of a binocular camera and multibeam sonar, belonging to the field of underwater environment 3D reconstruction technology. The invention includes the following steps: Step 1, acquiring binocular optical images and multibeam acoustic images of an underwater target; Step 2, denoising and enhancing the binocular optical images, obtaining the underwater target's depth information through binocular correction and stereo matching, and calculating the underwater target's optical reconstruction point cloud; Step 3, denoising and enhancing the multibeam acoustic images, calculating the underwater target's acoustic reconstruction point cloud through feature point detection; Step 4, optimizing the sonar-camera coordinate transformation matrix based on feature points from the sonar and camera, and performing 3D reconstruction of the underwater target through depth information fusion. This invention can significantly improve the accuracy and coverage of the reconstructed point cloud, enabling the reconstructed point cloud to better restore the target object.

Description

Technical Field

[0001] This invention relates to the field of underwater environment 3D reconstruction technology, and in particular to an underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar. Background Technology

[0002] In recent years, the exploration and development of the ocean has received increasing attention, and underwater unmanned vehicles (UUVs) have been widely used. They play an important role in tasks such as polar scientific research, marine geology, resource extraction, underwater archaeology, ship inspection, and mineral mining, requiring path planning and obstacle avoidance based on perception of the external environment. Therefore, the three-dimensional reconstruction of the underwater environment by UUVs is crucial to the effectiveness of mission execution.

[0003] Underwater unmanned vehicles (UUVs) are often equipped with camera sensors to acquire optical information, enabling 3D reconstruction of the surrounding environment and objects. Traditional visual 3D reconstruction methods are widely used in terrestrial imagery, achieving millimeter-level accuracy in target depth calculation. 3D reconstruction based on terrestrial optical images relies on clear optical image features; however, light propagation in water is subject to interference such as scattering and absorption, leading to problems like blurriness, color distortion, and uneven brightness in camera images, affecting the accuracy of target depth calculation and posing challenges to 3D reconstruction based on underwater optical images. Many methods utilize active light sources for underwater 3D reconstruction; however, their performance is insufficient in turbid water environments.

[0004] Underwater acoustics is the most widely used measurement method in underwater applications, and sonar-based 3D reconstruction methods have been extensively studied. Sonar determines the distance to an object by analyzing the intensity and time of the echo received after emitting sound waves. It has excellent penetration and is unaffected by underwater lighting conditions. Imaging sonar obtains precise distance and orientation information of target objects by emitting sound beams in multiple directions. However, sonar imaging suffers from low resolution and lack of elevation information, making it difficult to capture the fine structure of targets and reconstruct their true shape. Therefore, underwater 3D reconstruction using sonar alone is extremely challenging.

[0005] With the continuous development of the carrying capacity of underwater unmanned vehicles (UUVs), modern UUVs are often equipped with both sonar and optical cameras. The combined use of these two technologies can effectively improve the 3D reconstruction capabilities of UUVs. Sonar can compensate for the inaccuracy of camera depth calculations, while the camera can help sonar obtain the elevation and angle of the target, supplementing the information lost due to the low resolution of sonar. The prerequisite for sonar-camera fusion calculation is obtaining the transformation matrix between the sonar and camera coordinate systems. Existing calibration methods have achieved centimeter-level errors; however, for underwater 3D reconstruction, this remains a significant source of error.

[0006] A search revealed a patent application (CN202310948944.0) entitled "An Underwater 3D Reconstruction Method Based on Binocular Multi-Line Structured Light," which proposes an underwater 3D reconstruction method based on binocular multi-line structured light. This method uses multi-line lasers to perform high-density scanning of underwater targets, combining binocular camera synchronous imaging with a time-series scanning mechanism to acquire high-density point cloud data. By fusing multiple laser information lines and multi-frame image data, efficient synchronous 3D reconstruction of the underwater environment is achieved. This method uses high-precision instruments for 3D reconstruction; however, its accuracy decreases in complex environments. Another patent application (CN202510145828.4) entitled "An Underwater 3D Reconstruction Method and System Based on UUV Cruise and 2D Sonar" discloses an underwater 3D reconstruction method and system based on UUV cruise and 2D sonar. It obtains pose information through a submersible cruise algorithm and fuses sonar measurements to achieve 3D reconstruction of the underwater scene. However, this method places high demands on the performance of the underwater unmanned vehicle and the accuracy of its sensors. The patent application CN202411277711.3, entitled "Underwater 3D Reconstruction Method, Device, Electronic Equipment, and Storage Medium for Single Sonar Images," utilizes simulated sonar images and ground truth depth maps to train a segmentation network. Combining sonar imaging principles and inverse depth characteristics, it improves the accuracy of pseudo-forward depth maps to reconstruct point cloud structures. However, this method is susceptible to noise. The patent application CN202411545762.X, entitled "A Harbor Basin 3D Reconstruction Method Based on Visual and Sonar Information Fusion," involves collecting harbor basin information, obtaining a pre-trained model through deep learning, deploying an underwater robot in the harbor basin, and dynamically selecting sonar, visual images, or a fusion of both for 3D reconstruction. A fuzzy logic fusion algorithm is used to optimize accuracy. However, this method relies on the pre-collection and training of environmental information. The patent application CN202410489594.0, entitled "An Underwater 3D Reconstruction Method for Fusion of Camera and Sonar Images," proposes an underwater 3D reconstruction method that fuses camera and sonar images. It aligns the sonar image and camera imaging styles through style transfer, associating acoustic and optical features within a unified coordinate framework. Geometric constraints are established based on cross-modal matching relationships to estimate the target's distance and orientation information, and the 3D spatial coordinates of the underwater target are obtained through weighted fusion and nonlinear optimization. However, this method relies on clear visual and sonar images and is difficult to adapt to complex underwater environments. Summary of the Invention

[0007] In view of the above-mentioned deficiencies of the prior art, the technical problem to be solved by the present invention is how to improve the coverage and accuracy of point clouds for three-dimensional reconstruction of underwater environments.

[0008] To achieve the above objectives, this invention provides an underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar, comprising the following steps: Step 1: Acquire binocular optical images and multibeam acoustic images of the underwater target; Step 2: Denoise and enhance the binocular optical image, obtain the underwater target depth information through binocular correction and stereo matching, and calculate the optical reconstruction point cloud of the underwater target; Step 3: Denoise and enhance the multibeam acoustic image, and calculate the underwater target acoustic reconstruction point cloud through feature point detection; Step 4: Optimize the sonar-camera coordinate transformation matrix based on the feature points in the sonar and camera, and perform three-dimensional reconstruction of the underwater target through depth information fusion.

[0009] Furthermore, step two also includes the following steps: Step 2.1 Binocular calibration; Step 2.2 Depth calculation; Step 2.3 Three-dimensional point cloud computing; Step 2.4 Threshold segmentation.

[0010] Furthermore, step three also includes the following steps: Step 3.1 Feature point detection; Step 3.2 Three-dimensional point cloud computing.

[0011] Furthermore, step four also includes the following steps: Step 4.1 Project the sonar points onto the left eye camera; Step 4.2 Optimize the sonar-camera coordinate transformation matrix; Step 4.3 Deep information fusion.

[0012] In some implementations, the binocular optical image and the multibeam acoustic image are obtained using a binocular camera and a multibeam sonar, respectively.

[0013] Furthermore, in step two, a semi-global block matching algorithm is used to perform stereo matching on the binocular camera, and a threshold segmentation method is used to filter noise from the binocular optical images.

[0014] Furthermore, in step three, a minimum unit average constant false alarm rate algorithm suitable for radar is used to detect feature points in the multibeam acoustic image, and outlier filtering is used to filter noise from the detected feature points.

[0015] Furthermore, in step four, the sonar and camera's three-dimensional point cloud are matched, and the depth information of the sonar point cloud is used to correct the camera's calculated depth.

[0016] Furthermore, in step four, based on the depth information of the sonar feature points, the depth calculated by the binocular camera for all corresponding image feature points is scaled down to calculate more accurate three-dimensional coordinates of the underwater target point cloud.

[0017] In some implementations, the scaling factor for the depth information of sonar feature points and the binocular camera's calculated depth for all corresponding image feature points is 1.

[0018] Compared with the prior art, the present invention has the following advantages: (1) The present invention can significantly improve the accuracy and coverage of reconstructed point clouds, so that the reconstructed point clouds can better restore the target object.

[0019] (2) The present invention significantly reduces the distance error of underwater 3D reconstruction point cloud and can restore the surface details of target object, thereby improving the coverage of reconstruction point cloud and improving the performance of underwater 3D reconstruction.

[0020] (3) When the error of the acousto-optic coordinate transformation matrix is ​​large, the present invention can significantly improve the coverage of the reconstructed point cloud and reduce the distance error with the target; when the error of the acousto-optic coordinate transformation matrix is ​​very small, the improvement of the distance error is small, but it can still improve the coverage of the reconstructed point cloud.

[0021] The following will further explain the concept, specific structure, and technical effects of the present invention in conjunction with the accompanying drawings, so as to fully understand the purpose, features, and effects of the present invention. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the algorithm flow in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the sonar imaging principle of an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the coordinate transformation principle from the sonar coordinate system to the camera coordinate system in an embodiment of the present invention. Detailed Implementation

[0023] The preferred embodiments of the present invention are described below with reference to the accompanying drawings to make the technical content clearer and easier to understand. The present invention can be embodied in many different forms, and the scope of protection of the present invention is not limited to the embodiments mentioned herein.

[0024] In the accompanying drawings, components with the same structure are indicated by the same numerical designation, and components with similar structures or functions are indicated by similar numerical designations. The dimensions and thicknesses of each component shown in the drawings are arbitrary, and the present invention does not limit the dimensions and thicknesses of each component. To make the illustrations clearer, the thickness of some components has been appropriately exaggerated in the drawings.

[0025] Currently, existing underwater 3D reconstruction methods suffer from the following problems: Underwater 3D reconstruction requires obtaining external environmental information through sensors such as cameras or sonar. Underwater optical imaging suffers from insufficient light, weak texture, and reflection / refraction, leading to significant errors in optical depth measurement. Sonar imaging, on the other hand, suffers from issues such as missing elevation angles and low resolution, making it difficult to reconstruct target structural details. Most existing methods rely solely on cameras or sonar for underwater 3D reconstruction, resulting in insufficient accuracy in the reconstructed point cloud. In the underwater environment, compared to the method of calculating target depth through parallax matching using optical binocular cameras, sonar depth measurement is unaffected by factors such as lighting conditions, light refraction, and color deviation, and its accuracy is far superior to optical measurements. However, due to the lack of elevation angles in sonar imaging and the problem of obstruction, sonar measurements cannot accurately pinpoint the spatial location of the target. Traditional methods use the depth information obtained through sonar as the depth of all corresponding optical image pixels, causing the reconstructed point cloud to lose some spatial information of the target object. Existing acousto-optic joint calibration methods can obtain relatively accurate acousto-optic coordinate transformation matrices, achieving centimeter-level errors. However, projecting sonar feature points onto optical images can still affect the accuracy of the 3D reconstructed point cloud. Furthermore, the fact that the calibration dataset for the same submersible is in a different environment and condition than the data used in actual applications can also cause some deviation in the originally accurate calibration parameters.

[0026] To address the aforementioned issues, this invention proposes an underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar. By fusing binocular camera and image sonar, and using the depth information of sonar feature points as a basis, the binocular visual calculation depth of all corresponding image feature points is scaled proportionally to obtain more accurate target point cloud 3D coordinates. By minimizing the depth error between sonar feature points and matched camera feature points, the transformation matrix from the camera coordinate system to the sonar coordinate system is optimized and applied to the underwater 3D reconstruction process to improve the coverage and accuracy of the reconstructed point cloud.

[0027] This invention proposes an underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar, comprising the following steps: Step 1: Acquire binocular optical images and multibeam acoustic images of the underwater target; Step 2: Denoise and enhance the binocular optical image, obtain the underwater target depth information through binocular correction and stereo matching, and calculate the optical reconstruction point cloud of the underwater target; Step 3: Denoise and enhance the multibeam acoustic image, and calculate the underwater target acoustic reconstruction point cloud through feature point detection; Step 4: Optimize the sonar-camera coordinate transformation matrix based on the feature points in the sonar and camera, and perform three-dimensional reconstruction of the underwater target through depth information fusion.

[0028] Furthermore, step two also includes the following steps: Step 2.1 Binocular calibration; Step 2.2 Depth calculation; Step 2.3 Three-dimensional point cloud computing; Step 2.4 Threshold segmentation.

[0029] Furthermore, step three also includes the following steps: Step 3.1 Feature point detection; Step 3.2 Three-dimensional point cloud computing.

[0030] Furthermore, step four also includes the following steps: Step 4.1 Project the sonar points onto the left eye camera; Step 4.2 Optimize the sonar-camera coordinate transformation matrix; Step 4.3 Deep information fusion.

[0031] In some implementations, the binocular optical image and the multibeam acoustic image are obtained using a binocular camera and a multibeam sonar, respectively.

[0032] Furthermore, in step two, a semi-global block matching algorithm is used to perform stereo matching on the binocular camera, and a threshold segmentation method is used to filter noise from the binocular optical images.

[0033] Furthermore, in step three, a minimum unit average constant false alarm rate algorithm suitable for radar is used to detect feature points in the multibeam acoustic image, and outlier filtering is used to filter noise from the detected feature points.

[0034] Furthermore, in step four, the sonar and camera's three-dimensional point cloud are matched, and the depth information of the sonar point cloud is used to correct the camera's calculated depth.

[0035] Furthermore, in step four, based on the depth information of the sonar feature points, the depth calculated by the binocular camera for all corresponding image feature points is scaled down to calculate more accurate three-dimensional coordinates of the underwater target point cloud.

[0036] In some implementations, the scaling factor for the depth information of sonar feature points and the binocular camera's calculated depth for all corresponding image feature points is 1.

[0037] The implementation and realization of the technical solution of the present invention will be further described below through specific embodiments. It should be noted that the implementation and realization of the technical solution of the present invention are not limited to the following embodiments.

[0038] like Figure 1 As shown, the present invention provides an underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar, comprising the following steps: Step 1: Acquire binocular optical images and multibeam acoustic images of underwater targets. Optical and acoustic images of underwater targets are obtained using binocular cameras and multibeam imaging sonar carried by the underwater unmanned submersible.

[0039] Step 2: Denoise and enhance the binocular optical image. Obtain the underwater target depth information through binocular correction and stereo matching, and calculate the optically reconstructed point cloud of the underwater target.

[0040] For binocular cameras, a semi-global block matching algorithm is used for binocular stereo matching. The depth value is calculated based on the disparity between the left and right cameras corresponding to the pixel, and a 3D point cloud is obtained based on the left camera.

[0041] Step 2.1 Binocular Correction

[0042] For stereo cameras, the first step is to obtain the extrinsic parameters of the stereo camera, namely the rotation matrix, through calibration. Translation matrix Then, the two imaging planes are calibrated and aligned to the same plane using the intrinsic and extrinsic parameter matrices of the binocular camera, and the optical focal length is calculated simultaneously. and Optical Center and the baseline distance of the binocular camera .

[0043] Step 2.2 Depth Calculation

[0044] A semi-global block matching algorithm is applied to perform stereo matching on the left and right eye images and to calculate the disparity of target pixels in the camera images. Pixel points are calculated based on the parameters of binocular correction. corresponding depth ,in and The coordinates of a pixel on the image are represented by the formula:

[0045] and all three-dimensional points The set is denoted as ,in , , These are the coordinates of the X-axis, Y-axis, and Z-axis of the 3D camera coordinate system.

[0046] Step 2.3 3D Point Cloud Computing

[0047] depth This is the camera's scale factor. Based on the perspective principle of pinhole cameras, the three-dimensional coordinates of a pixel in the camera coordinate system can be reconstructed, i.e.:

[0048] and all three-dimensional points The set is denoted as .

[0049] Step 2.4 Threshold Segmentation

[0050] Due to the limitations of underwater optical imaging, such as insufficient light, weak texture, and reflection and refraction, directly applying a semi-global block matching algorithm for binocular matching and depth calculation results in noise and mismatches, making it difficult to achieve the processing effect of terrestrial optical images. Therefore, further region segmentation processing of the underwater image is required. Through Gaussian filtering, thresholding, and morphological closing operations, a binary image of the effective portion of the optical image is obtained, i.e.:

[0051] in Represents the original optical image pixels The value, For the newly generated binary image pixel values, This is a region segmentation processing method. Indicates the region segmentation threshold, " The "" indicates that the latter is obtained as a condition for the former. Then, by combining the thresholded segmentation image with the stereo matching detection image, noise filtering of the binocular matching results is achieved, resulting in a 3D point cloud set reconstructed from the optical image. .

[0052] Step 3: Denoise and enhance the multibeam acoustic image, and calculate the underwater target acoustic reconstruction point cloud through feature point detection.

[0053] Based on the principle of sonar imaging, the minimum unit average constant false alarm rate algorithm is used to detect feature points in sonar images and obtain the acoustic reconstruction point cloud of underwater targets.

[0054] Step 3.1 Feature Point Detection

[0055] A minimum-cell average constant false alarm rate (CFAR) algorithm suitable for radar is used to detect feature points in sonar images. Outlier filtering is then applied to the detected feature points to remove falsely detected feature points caused by noise affecting the algorithm.

[0056] Because the imaging principle of sonar can be compared to shining a flashlight on an object, the object forms a "sound shadow" in the sonar image, such as... Figure 2 As shown, the object in front will block the object behind, so we take the first point that is closest to each sound wave.

[0057] Step 3.2 3D Point Cloud Computing

[0058] The resolution of the known sonar range is obtained from the acoustic images. The unit is m / pixel, representing the width of the sonar image. Angular resolution of sonar The unit is rad / pixel, which can be used to calculate a specific pixel. Corresponding physical points Distance to the sonar in the sonar coordinate system and relative sonar horizontal angle :

[0059] and Angle relative to the vertical direction of the sonar Unknown, as a range of The variables can be used to obtain the three-dimensional point cloud of the sonar:

[0060] in , , These are the coordinates of the X, Y, and Z axes of the sonar coordinate system, respectively. The entire sonar point cloud is denoted as... .

[0061] Step 4: Optimize the sonar-camera coordinate transformation matrix based on feature points from the sonar and camera, and perform 3D reconstruction of underwater targets through depth information fusion.

[0062] Based on feature points in the sonar and camera, the coordinate transformation matrix from the sonar coordinate system to the camera coordinate system is optimized, and the depth information of the two is fused to perform underwater 3D reconstruction by camera and sonar fusion.

[0063] Step 4.1 Project the sonar points onto the left eye camera

[0064] To fuse information from optical and sonar images, feature points detected in the sonar imaging are projected onto the optical image for matching. For a point visible in both the sonar and optical cameras... In the optical camera coordinate system, it can be represented as In the sonar coordinate system, it is represented as Based on the rotation transformation matrix from the sonar coordinate system to the optical camera coordinate system With translation transformation matrix :

[0065] The point cloud obtained by transforming the sonar point cloud into the camera coordinate system is then projected onto the camera imaging plane, using the following formula:

[0066] Point clouds calculated from sonar images contain unknown variables. .like Figure 3 As shown, all possible three-dimensional points recovered from feature points in the sonar imaging plane in the sonar coordinate system will form an arc. The points on this arc are transformed to the optical camera coordinate system, and then each point in the arc is projected onto the camera imaging plane, which will appear as a straight line.

[0067] Step 4.2 Optimize the sonar-camera coordinate transformation matrix

[0068] To address the issue of inaccurate reconstruction results caused by calibration parameter errors, the rotation transformation matrix from the sonar coordinate system to the optical camera coordinate system is used. With translation transformation matrix As optimization variables, a matching cost function is constructed.

[0069] First, the three-dimensional points recovered from each sonar feature point are set as a set, i.e., divided into... For each point cloud set Perform separately , Optimization. Constructing camera point depth information. With sonar point depth information The matching cost function is calculated, and the optimal value is obtained, as shown in the following formula:

[0070] Step 4.3 Deep Information Fusion

[0071] First, the point set is divided, meaning each sonar feature point and its corresponding line projection onto the optical image form a data set. For each data set, the depth information measured by the sonar is used... Using the baseline depth value, the matching depth value is the closest. pixels .by Using the reference pixel as the base pixel, the ratio of the depth values ​​of other pixels to the depth value of the base pixel is used as the weight, and... Multiplying them together gives the depth of each pixel, i.e.:

[0072] in Represents pixels Optical computational depth, This is a weighting factor, typically set to 1 as an adjustable parameter. Through the above steps, more accurate depth information of the target pixels in the optical image is obtained, leading to the underwater 3D reconstructed point cloud.

[0073] The method proposed in this invention can significantly improve the accuracy and coverage of reconstructed point clouds, enabling them to better reconstruct the target object. It significantly reduces the distance error of underwater 3D reconstructed point clouds and restores the surface details of the target object, thereby improving the coverage of the reconstructed point cloud and enhancing the performance of underwater 3D reconstruction. When the acousto-optic coordinate transformation matrix error is large, it can significantly improve the coverage of the reconstructed point cloud and reduce the distance error with the target. When the acousto-optic coordinate transformation matrix error is very small, the improvement in distance error is small, but it can still improve the coverage of the reconstructed point cloud.

[0074] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. An underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar, characterized in that, Includes the following steps: Step 1: Acquire binocular optical images and multibeam acoustic images of the underwater target; Step 2: Denoise and enhance the binocular optical image, obtain the underwater target depth information through binocular correction and stereo matching, and calculate the optical reconstruction point cloud of the underwater target; Step 3: Denoise and enhance the multibeam acoustic image, and calculate the underwater target acoustic reconstruction point cloud through feature point detection; Step 4: Optimize the sonar-camera coordinate transformation matrix based on the feature points in the sonar and camera, and perform three-dimensional reconstruction of the underwater target through depth information fusion.

2. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 1, characterized in that, Step two also includes the following steps: Step 2.1 Binocular calibration; Step 2.2 Depth calculation; Step 2.3 Three-dimensional point cloud computing; Step 2.4 Threshold segmentation.

3. The underwater 3D reconstruction method based on the fusion of binocular camera and multibeam sonar as described in claim 1, characterized in that, Step three also includes the following steps: Step 3.1 Feature point detection; Step 3.2 Three-dimensional point cloud computing.

4. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 1, characterized in that, Step four also includes the following steps: Step 4.1 Project the sonar points onto the left eye camera; Step 4.2 Optimize the sonar-camera coordinate transformation matrix; Step 4.3 Deep information fusion.

5. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 1, characterized in that, The binocular optical image and multibeam acoustic image were obtained using a binocular camera and a multibeam sonar, respectively.

6. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 5, characterized in that, In step two, a semi-global block matching algorithm is used to perform stereo matching on the binocular camera, and a threshold segmentation method is used to filter noise from the binocular optical image.

7. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 3, characterized in that, In step three, a minimum unit average constant false alarm rate algorithm suitable for radar is used to detect feature points in the multi-beam acoustic image, and outlier filtering is used to filter noise from the detected feature points.

8. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 4, characterized in that, In step four, the sonar and the camera's three-dimensional point cloud are matched, and the depth information of the sonar point cloud is used to correct the depth calculated by the camera.

9. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 8, characterized in that, In step four, based on the depth information of sonar feature points, the depth calculated by the binocular camera for all corresponding image feature points is scaled down to calculate more accurate three-dimensional coordinates of underwater target point clouds.

10. The underwater 3D construction method based on the fusion of binocular camera and multibeam sonar as described in claim 9, characterized in that, The scaling factor for the depth information of the sonar feature points and the depth calculated by the binocular camera for all corresponding image feature points is 1.

Images