Method for establishing ship model based on fusion of video image and lidar point cloud

Abstract

The present invention relates to a method for establishing a ship model based on the fusion of video image and laser radar point cloud, S1 respectively obtains the video image data and point cloud data of the ship passing by the port through a camera and a laser radar, and transmits them to the Alibaba Cloud data processing server through a wireless network after storage, and performs preliminary data processing; S2, the Alibaba Cloud data processing server performs bidirectional data matching on the point cloud data and the video image data according to the time difference, and obtains the matching data set at the same time node; S3, uses the weighted fusion method to fuse the point cloud data and the forward video image data, and obtains the point cloud data with spectral characteristics; S4, converts the format of the point cloud data with spectral characteristics into the modeling software, intercepts the point cloud data, starts the point cloud capture, and performs reverse modeling. The present invention uses the method for establishing a ship model by fusing the video image and the laser radar point cloud data, which improves the modeling accuracy of the ship model and reduces the workload of manual modeling.

Description

Technical Field

[0001] The present invention relates to the technical field of ship mapping images, and in particular to a method for establishing a ship model based on the fusion of video images and laser radar point clouds. Background Art

[0002] Lidar (Light Detection And Ranging) is a product of applying modern optical technology to remote sensing technology. Since the essence of laser is electromagnetic waves, it has the characteristics of high brightness, high directivity and high coherence. Therefore, compared with traditional radar technology, Lidar has certain advantages in angular resolution, distance resolution, speed resolution and speed measurement range. It can also capture image data of multiple targets and is not easily affected by external interference. The information of Lidar can be obtained by observing the amplitude, phase and polarization. Therefore, Lidar can not only measure distance accurately, but also reduce errors when measuring speed and tracking targets.

[0003] LiDAR has long been used in ship navigation detection to ensure safe navigation of ships. LiDAR point cloud information has high accuracy and contains multiple echo and reflection intensity information, making it an excellent data source for ship modeling. However, it does not have spectral feature information and cannot indicate ship characteristics. With the development of imaging technology, the pixels of image information are continuously improved, and the spectral information is rich, which can better obtain the continuous feature information of the observed object, making up for the lack of spectral information in modeling based solely on LiDAR.

[0004] With the continuous development of surveying and mapping technology, it is difficult for a single data model to cover all the information characteristics of a region and to meet the needs of ship information monitoring in many key monitoring waters. Under this premise, it is necessary to fuse the LiDAR point cloud data and image data, establish a 3D ship model with spectral characteristics, enrich the ship model database, and provide more basis for navigation supervision. Summary of the invention

[0005] The technical problem to be solved by the present invention is to provide a method for establishing a ship model based on the fusion of video images and lidar point clouds in response to the shortcomings of the above-mentioned prior art, which solves the problem that model construction with single data is difficult to cover all information characteristics of a region and is difficult to meet the ship information needs of many key monitored waters, enriches the ship model database, and provides more data for navigation supervision.

[0006] The technical solution adopted by the present invention to solve the above-mentioned technical problems is:

[0007] A method for establishing a ship model based on the fusion of video images and laser radar point clouds, comprising the following steps:

[0008] S1. Install a camera at a position facing the ship at the port. The camera faces the water surface and regularly takes pictures of ships passing by on the water surface, and stores the video image data. At the same time, deploy a laser radar at a fixed position in the port, regularly scan the ships passing by on the water surface to obtain point cloud data, and store the point cloud data. After the video image data and point cloud data are stored, they are transmitted to the Alibaba Cloud data processing server through the wireless network for preliminary data processing.

[0009] S2: Alibaba Cloud data processing server receives a frame of point cloud data from the LiDAR and a frame of video image data from the camera at regular intervals. After receiving the data, it performs bidirectional data matching between the point cloud data and the video image data based on the time difference to obtain a matching data set at the same time node.

[0010] S3. After the Alibaba Cloud data processing server creates the matching data set, it has obtained the forward video image data and point cloud data. It uses the weighted fusion method to fuse the lidar point cloud data and the forward video image data, and highlights the image results in terms of amplitude to achieve the effect of image enhancement, and obtain point cloud data with spectral characteristics.

[0011] S4. After the format conversion of the point cloud data with spectral characteristics, input it into the modeling software to intercept the point cloud data, start point cloud capture, intercept the data, and finally generate a 3D model of the ship according to the user's own needs.

[0012] In the above scheme, in step S1, the Alibaba Cloud data processing server performs preliminary processing on the received data, including:

[0013] (1) Performing preliminary processing on the video image data: deleting the image area with navigation marks, bridge piers and other obstacles; performing regional enhancement using the preprocessed current frame image, performing edge complementation on the video image, and obtaining the video image data of the ship;

[0014] (2) Preliminary processing of the point cloud data: Simplify the point cloud data, delete the point cloud data that does not belong to the target range, and obtain the point cloud data of the ship.

[0015] In the above scheme, step S2 specifically includes the following steps:

[0016] S2.1, the original point cloud data is processed by filtering to obtain basic point cloud data, and the point cloud graphics are distinguished by interpolating DEM;

[0017] S2.2, based on the DEM obtained by software, the DEM is registered with the video image data, and the forward video image data can be obtained by integrating the two-way data, thus completing the digital differential correction process of the image;

[0018] S2.3. After the original point cloud data undergoes image preprocessing by edge enhancement and threshold processing, the gross errors in the lidar point cloud data are eliminated and the data fusion and registration processing is completed. The point cloud data and the forward video image data can be bidirectionally matched according to the time difference to obtain a complete matching data set at the same time node.

[0019] In the above scheme, step S3 specifically includes the following steps:

[0020] S3.1. After establishing the matching data set, the image is linearly compounded and the image result is highlighted from the amplitude, thereby achieving the effect of image enhancement;

[0021] S3.2. After the enhancement is completed, the weighted fusion method is gray value weighted fusion, and the weighted fusion formula is: F(i,j)=c1A(i,j)+c2B(i,j), where i and j are the row and column expressions of pixels in the image, i=1,2,3......n, j=1,2,3......n; c1 and c2 are weighting coefficients of image information, c1+c2=1, for spectral information c1=0,c2=1, for spatial information c1=1,c2=0; A is radar point cloud data, B is video image data, A(i,j) is the pixel row and column representation of lidar point cloud data, B(i,j) is the pixel row and column representation of video image data, and F(i,j) is the final value of pixel row and column superposition;

[0022] S3.3. After the fusion is completed, the ship point cloud data with spectral characteristics is obtained.

[0023] In the above scheme, step S4 specifically includes the following steps:

[0024] S4.1. Convert the point cloud data with spectral characteristics into the "*.rcs" format or "*.rcp" format, check the integrity of the point cloud data, and complete the modeling preparation work;

[0025] S4.2. After correcting the image coordinate system, perform point cloud slicing as required and select the point cloud area; perform point cloud data analysis and preprocessing, and check whether the features are complete and meet the modeling requirements after filtering; after completion, intercept the data after the point cloud is aligned to check the processing effect;

[0026] S4.3. After judging that the point cloud data structure is complete and the features are clear, classify and fit the image features and edit the key dimensions;

[0027] S4.4. After processing, confirm whether the overall features meet the modeling requirements, use the point cloud capture function to accurately capture the three-dimensional point coordinates of the point cloud data, and complete the ship model construction.

[0028] In the above scheme, step S4.1 specifically includes:

[0029] (1) Obtaining point cloud data with spectral characteristics;

[0030] (2) Convert the data format and import the point cloud data into the Rhino software platform.

[0031] In the above scheme, step S4.2 specifically includes:

[0032] (3) Check whether the point cloud data is complete and meets the modeling requirements, and make a judgment based on the specific requirements; if the point cloud data meets the modeling requirements, proceed to the next step to extract and capture the point cloud data; otherwise, return to step (1) to obtain the point cloud data again;

[0033] (4) Correct the imported data coordinate system, unify the required coordinates, and perform point cloud cutting;

[0034] (5) Data analysis, removing noise from point cloud data and filtering coarse values;

[0035] (6) Obtain preliminary processed point cloud data, requiring that more than 70% of the point cloud data remain and the features are well preserved. If the point cloud data meets the requirements, proceed to the next step, otherwise return to step (5) and repeat the data analysis step;

[0036] (7) Align the point cloud, perform point cloud capture, and check the effect.

[0037] In the above scheme, step S4.3 specifically includes:

[0038] (8) Point cloud refinement processing;

[0039] (9) After refinement, check whether the point cloud features are intact. If the point cloud features are intact, proceed to the next step. Otherwise, return to step (7) and perform the point cloud alignment step again.

[0040] (10) Perform feature analysis and feature classification on the collected point cloud data, determine the feature morphology, and divide the feature importance level;

[0041] (11) Perform spectral feature fitting. Before matching, the image pixel spectrum and the reference spectrum must be processed by continuum removal. Then, the least square method is used to fit the curves of the pixel spectrum and the reference spectrum after the continuum removal. Error analysis is performed to reduce the error degree. If the result error is too large, return to the feature fitting step and refit the feature. If the result error is within an acceptable range, proceed to the next step.

[0042] (12) Edit and adjust key dimensions after the graphic area is segmented;

[0043] (13) Import and modify other features of the graphics;

[0044] (14) After the basic components of the graphics are completed, check the overall features to see if they meet the modeling requirements. If they do, save the graphics and score the model. If the score is not satisfactory or the modeling requirements are not met, return to the step of editing the key dimensions of the graphics.

[0045] In the above scheme, step S4.4 specifically includes:

[0046] (15) Capture point cloud and save data. If the point cloud data can be saved successfully, enter modeling, otherwise return to the key dimension editing step;

[0047] (16) Freeze the model data, output, and modeling is completed.

[0048] In the above scheme, the camera in step S1 adopts Hikvision's starlight-level ultra-low illumination high-definition network camera, which can recognize ship images during the day or at night; the camera is installed on the port bracket, and the camera is set parallel to the water surface, and the camera faces the water surface to take pictures of ships passing by on the water surface; the laser radar in step S1 adopts Velodyne VLP16-line laser radar, which has a measurement range of 100m and a scanning angle of 360°, and can provide real-time feedback on three-dimensional distance and calibrated reflectivity measurement; the laser radar is installed at a fixed position in the port, with a scanning angle of 270° and a scanning range of 100m. It regularly scans passing ships on the water surface to obtain point cloud data, and stores the point cloud data.

[0049] The beneficial effects of the present invention are:

[0050] The present invention uses a method of fusing laser radar point cloud data and video image data to assign video image spectral characteristics to ship point cloud data, and establish a 3D ship model with spectral characteristics to solve the problem of lack of actual ship data in the ship model database and enrich the ship model database. During the implementation process, the preliminary processing of point cloud data and video image data was completed through the Alibaba Cloud server, solving the problem of repeated loading and processing of previous data; next, a two-way matching data set was established based on time, solving the problem of repeated data comparison and correction, and finally enhancing the image effect based on the weighted fusion method, providing point cloud data with spectral characteristics with clear characteristics for modeling, improving the efficiency of ship modeling, and providing more data support for subsequent navigation supervision of multi-source data fusion, ensuring navigation safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments, in which:

[0052] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0053] Figure 2 It is a schematic diagram of the layout of the laser radar and camera in the method of the present invention;

[0054] Figure 3 It is a comparison diagram of ship point cloud data and video image data in the method of the present invention;

[0055] Figure 4 It is a schematic diagram of the processing flow of image data and point cloud data in the method of the present invention;

[0056] Figure 5 It is a schematic diagram of the modeling process in the embodiment of the method of the present invention. DETAILED DESCRIPTION

[0057] In order to have a clearer understanding of the technical features, purposes and effects of the present invention, specific embodiments of the present invention are now described in detail with reference to the accompanying drawings.

[0058] The present invention provides a method for establishing a ship model based on the fusion of video images and laser radar point clouds. The method requires a ship image acquisition system and a laser radar image acquisition system. The ship image acquisition system includes a camera, a memory card, and an Alibaba Cloud data processing server. The laser radar image acquisition system includes a laser radar, a memory card, and an Alibaba Cloud data processing server. The camera and the laser radar communicate with the Alibaba Cloud data processing server through a communication network. Figure 1 As shown, the method of the present invention comprises the following steps:

[0059] S1. Install a camera at a position facing the ship at the port. The camera faces the water surface and regularly takes pictures of passing ships on the water surface, and stores the video image data. At the same time, deploy a lidar at a fixed position in the port, regularly scan the passing ships on the water surface to obtain point cloud data, and store the point cloud data. After the video image data and point cloud data are stored, they are transmitted to the Alibaba Cloud data processing server through the wireless network for preliminary data processing.

[0060] For camera and lidar positions, see Figure 2 . Among them, the camera is installed on the port bracket, and the camera is set parallel to the water surface. The camera faces the water surface and continuously shoots pictures of ships passing by on the water surface. In this embodiment, the camera adopts Hikvision's starlight-level ultra-low illumination high-definition network camera, which can recognize ship images during the day or at night. At the same time, two Velodyne VLP16-line laser radars are installed at fixed positions in the port, and the measurement range is set to 100m and the scanning angle is 270° to make up for the problem of missing point cloud images caused by angle problems on the laser radar. The passing ships on the water surface are scanned regularly to obtain point cloud data, and the point cloud data is stored. Because the measurement distance of the laser radar is limited, its position can be adjusted with the distance of the ship to obtain a complete point cloud image to the greatest extent.

[0061] The laser radar point cloud image and camera image taken are as follows: Figure 3 As shown in the figure, the matching result is relatively good, ensuring the local and global features of the point cloud image. In order to ensure that the laser radar point cloud data and video image data are obtained in the same time period and the same area, the two data use the same POS (Position and Orientation System) and the same time matching segment.

[0062] The Alibaba Cloud data processing server performs preliminary processing on the received data, including: (1) preliminary processing of the video image data: deleting the image area with navigation marks, bridge piers and other obstacles; using the pre-processed current frame image to perform regional enhancement, and edge-filling the video image to obtain the video image data of the ship; (2) preliminary processing of the point cloud data: point cloud simplification, deleting the point cloud data that is not within the target range, and obtaining the point cloud data of the ship.

[0063] S2, Alibaba Cloud data processing server receives a frame of point cloud data from the laser radar and a frame of video image data taken by the camera at regular intervals. After receiving the data, it performs two-way data matching between the point cloud data and the video image data according to the time difference to obtain a matching data set at the same time node. The specific method of Alibaba Cloud data processing server for two-way data matching is:

[0064] S2.1, the original point cloud data is processed by filtering to obtain basic point cloud data, and the point cloud graphics are distinguished by interpolating DEM (Digital Elevation Model);

[0065] S2.2, based on the DEM obtained by software, the DEM is registered with the video image data, and the forward video image data can be obtained by integrating the two-way data, thus completing the digital differential correction process of the image;

[0066] S2.3, after the image preprocessing process of edge enhancement and threshold processing, the original point cloud data is eliminated from the gross errors in the LiDAR point cloud data, and the data fusion registration processing is completed. The point cloud data and the forward video image data can be matched in both directions according to the time difference to obtain a complete matching data set at the same time node. Since the two data use the same coordinate system and the same matching time period, the time matching method is used for data registration. In the Alibaba Cloud data processing server, the point cloud data angle is adjusted according to the same time period to complete the data matching work in the forward video image, and the data registration effect can be checked at the same time to decide whether to proceed to the next step of data output.

[0067] S3. After the Alibaba Cloud data processing server creates the matching data set, it has obtained the forward video image data and point cloud data. It uses the weighted fusion method to fuse the lidar point cloud data and the forward video image data, and highlights the image results in amplitude to achieve the effect of image enhancement and obtain point cloud data with spectral characteristics. The specific method for obtaining point cloud data with spectral characteristics is:

[0068] S3.1. After establishing the matching data set, the image is linearly compounded and the image result is highlighted from the amplitude, thereby achieving the effect of image enhancement;

[0069] S3.2. After the enhancement is completed, the weighted fusion method is gray value weighted fusion, and the weighted fusion formula is: F(i,j)=c1A(i,j)+c2B(i,j), where i and j are the row and column expressions of pixels in the image, i=1,2,3......n, j=1,2,3......n; c1 and c2 are weighting coefficients of image information, c1+c2=1, for spectral information c1=0,c2=1, for spatial information c1=1,c2=0; A is radar point cloud data, B is video image data, A(i,j) is the pixel row and column representation of lidar point cloud data, B(i,j) is the pixel row and column representation of video image data, and F(i,j) is the final value of pixel row and column superposition;

[0070] S3.3. After the fusion is completed, the ship point cloud data with spectral characteristics is obtained.

[0071] S4. Convert the point cloud data with spectral characteristics into a format and input it into the modeling software to capture the point cloud data. Enable point cloud capture and capture the data. Finally, generate a 3D model of the ship according to the user's needs. Figure 5 As shown:

[0072] S4.1. Convert the point cloud data with spectral characteristics to "*.rcs" or "*.rcp" format, check the integrity of the point cloud data, and complete the modeling preparation. This step specifically includes the following operation procedures:

[0073] (1) Obtaining point cloud data with spectral characteristics;

[0074] (2) Convert the data format and import the point cloud data into the Rhino software platform.

[0075] S4.2. After correcting the image coordinate system, perform point cloud slicing as required and select the point cloud area; perform point cloud data analysis and preprocessing, and filter to see whether the features are complete and meet the modeling requirements; after completion, intercept the data after the point cloud is aligned to check the processing effect. This step specifically includes the following operation flow:

[0076] (3) Check whether the point cloud data is complete and meets the modeling requirements, and make a judgment based on the specific requirements; if the point cloud data meets the modeling requirements, proceed to the next step to extract and capture the point cloud data; otherwise, return to step (1) to obtain the point cloud data again;

[0077] (4) Correct the imported data coordinate system, unify the required coordinates, and perform point cloud cutting;

[0078] (5) Data analysis, removing noise from point cloud data and filtering coarse values;

[0079] (6) Obtain preliminary processed point cloud data, requiring that more than 70% of the point cloud data remain and the features are well preserved. If the point cloud data meets the requirements, proceed to the next step, otherwise return to step (5) and repeat the data analysis step;

[0080] (7) Align the point cloud, perform point cloud capture, and observe the effect.

[0081] S4.3. After judging that the point cloud data structure is complete and the features are clear, classify and fit the image features and edit the key dimensions. This step specifically includes the following operation procedures:

[0082] (8) Point cloud refinement processing;

[0083] (9) After refinement, check whether the point cloud features are intact. If the point cloud features are intact, proceed to the next step. Otherwise, return to step (7) and perform the point cloud alignment step again.

[0084] (10) Perform feature analysis and feature classification on the collected point cloud data, determine the feature morphology, and divide the feature importance level;

[0085] (11) Perform spectral feature fitting. Before matching, the image pixel spectrum and the reference spectrum must be processed by continuum removal. Then, the least square method is used to fit the curves of the pixel spectrum and the reference spectrum after the continuum removal. Error analysis is performed to reduce the error degree. If the result error is too large, return to the feature fitting step and refit the feature. If the result error is within an acceptable range, proceed to the next step.

[0086] (12) Edit and adjust key dimensions after the graphic area is segmented;

[0087] (13) Import and modify other features of the graphics;

[0088] (14) After the basic components of the graphics are completed, check the overall features to see if they meet the modeling requirements. If they do, save the graphics and score the model. If the score is not satisfactory or the modeling requirements are not met, return to the step of editing the key dimensions of the graphics.

[0089] S4.4. After the processing is completed, confirm whether the overall features meet the modeling requirements, use the point cloud capture function to accurately capture the three-dimensional point coordinates of the point cloud data, and complete the ship model construction. This step specifically includes the following operation procedures:

[0090] (15) Capture point cloud and save data. If the point cloud data can be saved successfully, enter modeling, otherwise return to the key dimension editing step;

[0091] (16) Freeze the model data, output, and modeling is completed.

[0092] The method of establishing a ship model by fusing video images with lidar point cloud data solves the problem of lack of 3D ship models in a single navigation system, improves the modeling accuracy of the ship model, reduces the workload of manual modeling, and enhances the perception level of ship types in key areas, providing a stable and reliable data source for subsequent navigation supervision.

[0093] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the present invention and the claims, all of which are within the protection of the present invention.

Claims

1. A method for establishing a ship model based on the fusion of video images and laser radar point clouds, characterized in that: The following steps are involved: S1. Install a camera at a position facing the ship at the port. The camera faces the water surface and regularly takes pictures of ships passing by on the water surface, and stores the video image data. At the same time, deploy a laser radar at a fixed position in the port, regularly scan the ships passing by on the water surface to obtain point cloud data, and store the point cloud data. After the video image data and point cloud data are stored, they are transmitted to the Alibaba Cloud data processing server through the wireless network for preliminary data processing. S2: Alibaba Cloud data processing server receives a frame of point cloud data from the LiDAR and a frame of video image data from the camera at regular intervals. After receiving the data, it performs bidirectional data matching between the point cloud data and the video image data based on the time difference to obtain a matching data set at the same time node. S3. After the Alibaba Cloud data processing server creates a matching data set, it has obtained the forward video image data and point cloud data. It uses the weighted fusion method to fuse the lidar point cloud data and the forward video image data, and highlights the image results in terms of amplitude to achieve the effect of image enhancement, and obtain point cloud data with spectral characteristics. The specific steps include: S3.

1. After establishing the matching data set, the image is linearly compounded and the image result is highlighted from the amplitude, thereby achieving the effect of image enhancement; S3.

2. After the enhancement is completed, the weighted fusion method is gray value weighted fusion, and the weighted fusion formula is: F(i,j)=c1A(i,j)+c2B(i,j), where i and j are the row and column expressions of pixels in the image, i=1,2,3......n, j=1,2,3......n; c1 and c2 are weighting coefficients of image information, c1+c2=1, for spectral information c1=0,c2=1, for spatial information c1=1,c2=0; A is radar point cloud data, B is video image data, A(i,j) is the pixel row and column representation of lidar point cloud data, B(i,j) is the pixel row and column representation of video image data, and F(i,j) is the final value of pixel row and column superposition; S3.

3. After the fusion is completed, the ship point cloud data with spectral characteristics is obtained; S4. After the format conversion of the point cloud data with spectral characteristics, input it into the modeling software to intercept the point cloud data, start point cloud capture, intercept the data, and finally generate a 3D model of the ship according to the user's own needs.

2. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 1 is characterized in that: In step S1, the Alibaba Cloud data processing server performs preliminary processing on the received data, including: (1) Performing preliminary processing on the video image data: deleting the image area with navigation marks, bridge piers and other obstacles; performing regional enhancement using the preprocessed current frame image, performing edge complementation on the video image, and obtaining the video image data of the ship; (2) Preliminary processing of the point cloud data: Simplify the point cloud data, delete the point cloud data that does not belong to the target range, and obtain the point cloud data of the ship.

3. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 1 is characterized in that: Step S2 specifically includes the following steps: S2.1, the original point cloud data is processed by filtering to obtain basic point cloud data, and the point cloud graphics are distinguished by interpolating DEM; S2.2, based on the DEM obtained by software, the DEM is registered with the video image data, and the forward video image data can be obtained by integrating the two-way data, thus completing the digital differential correction process of the image; S2.

3. After the original point cloud data undergoes image preprocessing by edge enhancement and threshold processing, the gross errors in the lidar point cloud data are eliminated and the data fusion and registration processing is completed. The point cloud data and the forward video image data can be bidirectionally matched according to the time difference to obtain a complete matching data set at the same time node.

4. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 1 is characterized in that: Step S4 specifically includes the following steps: S4.

1. Convert the point cloud data with spectral characteristics into "*.rcs" or "*.rcp" format, check the integrity of the point cloud data, and complete the modeling preparation; S4.

2. After correcting the image coordinate system, perform point cloud slicing as required and select the point cloud area; perform point cloud data analysis and preprocessing, and check whether the features are complete and meet the modeling requirements after filtering; after completion, intercept the data after the point cloud is aligned to check the processing effect; S4.

3. After judging that the point cloud data structure is complete and the features are clear, classify and fit the image features and edit the key dimensions; S4.

4. After processing, confirm whether the overall features meet the modeling requirements, use the point cloud capture function to accurately capture the three-dimensional point coordinates of the point cloud data, and complete the ship model construction.

5. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 4 is characterized in that: The step S4.1 specifically includes: (1) Obtaining point cloud data with spectral characteristics; (2) Convert the data format and import the point cloud data into the Rhino software platform.

6. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 5 is characterized in that: The step S4.2 specifically includes: (3) Check whether the point cloud data is complete and meets the modeling requirements, and make a judgment based on the specific requirements; if the point cloud data meets the modeling requirements, proceed to the next step to extract and capture the point cloud data; otherwise, return to step (1) to obtain the point cloud data again; (4) Correct the imported data coordinate system, unify the required coordinates, and perform point cloud cutting; (5) Data analysis, removing noise from point cloud data and filtering coarse values; (6) Obtain preliminary processed point cloud data, requiring that more than 70% of the point cloud data remain and the features are well preserved. If the point cloud data meets the requirements, proceed to the next step, otherwise return to step (5) and repeat the data analysis step; (7) Align the point cloud, perform point cloud capture, and check the effect.

7. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 6 is characterized in that: The step S4.3 specifically includes: (8) Point cloud refinement processing; (9) After refinement, check whether the point cloud features are intact. If the point cloud features are intact, proceed to the next step. Otherwise, return to step (7) and perform the point cloud alignment step again. (10) Perform feature analysis and feature classification on the collected point cloud data, determine the feature morphology, and divide the feature importance level; (11) Perform spectral feature fitting. Before matching, the image pixel spectrum and the reference spectrum must be processed by continuum removal. Then, the least square method is used to fit the curves of the pixel spectrum and the reference spectrum after the continuum removal. Error analysis is performed to reduce the error degree. If the result error is too large, return to the feature fitting step and refit the feature. If the result error is within an acceptable range, proceed to the next step. (12) Edit and adjust key dimensions after the graphic area is segmented; (13) Import and modify other features of the graphics; (14) After the basic components of the graphics are completed, check the overall features to see if they meet the modeling requirements. If they do, save the graphics and score the model. If the score is not satisfactory or the modeling requirements are not met, return to the step of editing the key dimensions of the graphics.

8. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 7 is characterized in that: The step S4.4 specifically includes: (15) Capture point cloud and save data. If the point cloud data can be saved successfully, enter modeling, otherwise return to the key dimension editing step; (16) Freeze the model data, output, and modeling is completed.

9. The method for establishing a ship model based on the fusion of video images and laser radar point clouds according to claim 1 is characterized in that: The camera in step S1 is a Hikvision starlight-level ultra-low illumination high-definition network camera, which can identify ship images during the day or at night; the camera is installed on a port bracket, and the camera is set parallel to the water surface, and the camera faces the water surface to shoot pictures of passing ships on the water surface; The laser radar in step S1 uses a Velodyne VLP16-line laser radar, which has a measurement range of 100m and a scanning angle of 360°, and can provide real-time feedback of three-dimensional distance and calibration reflectivity measurement; the laser radar is installed at a fixed position in the port, with a set scanning angle of 270° and a scanning range of 100m. It regularly scans passing ships on the water surface to obtain point cloud data and stores the point cloud data.

Images