Shipping fleet size detection method based on lidar and image fusion

Abstract

The present invention relates to the technical field of shipping fleet size detection, specifically to a shipping fleet size detection method based on laser radar and image fusion, comprising the following steps: collecting three-dimensional point cloud data and two-dimensional image data, aligning the data by timestamp and spatial coordinates to obtain an aligned data set; based on the aligned data set, identifying the initial boundary of the overlapping area, marking and separating the area, and obtaining an initial boundary mark. The present invention adopts laser radar and image fusion technology to collect multi-dimensional data, and constructs a complete aligned data set by aligning timestamps and spatial coordinates to ensure comprehensive coverage of hull size information. The marking and separation of the initial boundaries of the overlapping areas reduces the ambiguity and errors of boundary identification, providing a more reliable basis for subsequent contour refinement. By combining point-to-point matching, a high-precision three-dimensional model of the hull is established, solving the problems of fracture and information loss during the modeling process.

Description

Technical Field

[0001] The present invention relates to the technical field of shipping fleet size detection, and in particular to a shipping fleet size detection method based on laser radar and image fusion. Background Art

[0002] The shipping fleet size detection method is a technical solution for accurate measurement and data optimization of the size of a single or multiple ships in a shipping fleet. Its main uses include providing basic size parameters for the design and manufacture of ships, supporting dynamic monitoring and maintenance of ships during operation, and can also be used for cargo capacity assessment and ship seaworthiness analysis.

[0003] However, the existing technology has limitations in boundary recognition and modeling accuracy in complex dynamic environments. During the data acquisition process, the processing of overlapping areas lacks separation, resulting in blurred boundaries, affecting the integrity of contour markings and the quality of subsequent modeling. In addition, the existing technology lacks a real-time response mechanism to attitude changes under dynamic conditions, and fails to effectively correct the size deviation caused by the tilt or movement of the hull. For example, when the wind speed or water flow changes and causes the hull attitude to change, the size calculation results will be biased. Therefore, improvements are needed. Summary of the invention

[0004] The purpose of the present invention is to solve the shortcomings of the prior art and propose a shipping fleet size detection method based on laser radar and image fusion.

[0005] In order to achieve the above object, the present invention adopts the following technical solution, a shipping fleet size detection method based on laser radar and image fusion, comprising the following steps:

[0006] Collecting three-dimensional point cloud data and two-dimensional image data, aligning the data according to timestamps and spatial coordinates to obtain an aligned data set; based on the aligned data set, identifying initial boundaries of overlapping areas, marking and separating areas, and obtaining initial boundary markings;

[0007] Based on the initial boundary marking, edge detection is applied to refine the edge contour of each ship to obtain refined edge information; the contour of each ship is individually marked according to the refined edge information, a separate hull mark is created, and a fine contour mark is generated;

[0008] Based on the fine contour identification, point-to-point matching is performed to reconstruct the three-dimensional model of each ship to obtain a three-dimensional construction model, and the size and volume of the hull are calculated using the three-dimensional construction model to generate optimized size data;

[0009] The optimized dimension data is applied to monitor the attitude change of the ship under dynamic conditions, perform error correction of dimension measurement, and generate corrected dimension data.

[0010] Preferably, the steps of acquiring the aligned data set are:

[0011] Collect 3D point cloud and 2D image data from LiDAR and image sensors, label each data point with a timestamp and spatial coordinates, and generate a preliminary aligned dataset;

[0012] Based on the preliminary aligned data set, the spatial deviation between data points is calculated using the formula:

[0013] ;

[0014] Get the average spatial deviation ,in, Respectively The coordinates of the data points, is the average coordinate of all data points, is the total number of data points;

[0015] According to the average spatial deviation, the positions of the data points are adjusted, the data alignment process is optimized, and an aligned data set is formed.

[0016] Preferably, the steps of obtaining the initial boundary mark are:

[0017] Through the aligned data set, spatial feature analysis is performed on each data point, and by comparing the positional relationship between points, the overlapping area is identified, the initial boundary of the overlapping area is marked, and an initial boundary data set is generated;

[0018] Based on the initial boundary data set, the geometric center of each boundary area is calculated using the following formula:

[0019] ;

[0020] in, For the The geometric center of the boundary region, In the region The three-dimensional coordinates of the points, For the The total number of points within the boundary area;

[0021] The geometric center is used to refine the boundary of each region to form an initial boundary mark.

[0022] Preferably, the step of obtaining the refined edge information is:

[0023] Using the initial boundary markers, the outer contour of each ship is scanned to capture contour changes, and the contour of each ship is identified to form a preliminary edge contour data set;

[0024] Based on the preliminary edge contour data set, the edge of each ship is refined by image enhancement to improve image contrast and connect broken contour lines, thereby obtaining a refined edge contour data set;

[0025] Based on the refined edge contour data set, the refined edge information of the vessel is defined and completed to form refined edge information.

[0026] Preferably, the steps of obtaining the fine contour mark are:

[0027] Based on the refined edge information, the contour of each ship is marked, each hull is distinguished, and a separate hull contour mark is obtained;

[0028] Based on the individual hull contour annotations, the edge feature score of each ship is calculated using the following formula:

[0029] ;

[0030] in, Score the edge features of the hull, For the The edge strength of the edge point, is the distance to the neighboring point, is the number of edge points;

[0031] According to the edge feature score, the edge contour of each ship is optimized, the edge of each ship is determined, and a fine contour mark is generated.

[0032] Preferably, the steps of acquiring the three-dimensional construction model are:

[0033] Using the fine contour identification, point-to-point matching is performed to correspond the two-dimensional contour data of each ship with the three-dimensional point cloud data to generate primary three-dimensional model data;

[0034] Based on the primary three-dimensional model data, the data points are adjusted in position and the facets are configured to obtain refined three-dimensional model data;

[0035] Based on the refined three-dimensional model data, smoothing processing is performed to obtain a three-dimensional construction model.

[0036] Preferably, the steps for obtaining the optimized size data are:

[0037] Based on the three-dimensional constructed model, the dimensions of each hull in the model are measured one by one to obtain the dimensional data of the hull and obtain basic dimensional data;

[0038] Based on the basic size data, each hull surface is meshed, and the volume of each small mesh is calculated, and the total volume data of each ship is accumulated to obtain the volume data obtained by detailed calculation;

[0039] Based on the volume data obtained by the detailed calculation, the volume data is error checked and the accuracy is optimized, and the data format and unit are adjusted to generate optimized dimension data.

[0040] Preferably, the steps of obtaining the corrected size data are:

[0041] Using the optimized size data, tracking the posture changes of the vessel in dynamic water conditions, collecting the tilt angle, acceleration and position change data of the vessel through sensors, and generating dynamic posture data;

[0042] Based on the dynamic posture data, the measurement error of the ship size is calculated using the following formula:

[0043] ;

[0044] in, Represents the total dimensional measurement error, It is The size value obtained by the measurement is is the expected size value based on the optimized size data, is the number of measurements;

[0045] According to the dimension measurement error, error correction is performed and the dimension data is adjusted to generate corrected dimension data.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are:

[0047] The present invention uses laser radar and image fusion technology to collect multi-dimensional data, and constructs a complete aligned data set through timestamp and spatial coordinate alignment to ensure comprehensive coverage of hull size information. The marking and separation of the initial boundaries of the overlapping areas reduces the ambiguity and errors of boundary identification, providing a more reliable basis for subsequent contour refinement. By combining point-to-point matching, a high-precision three-dimensional model of the hull is established, solving the problems of fracture and information loss during the modeling process. Under dynamic conditions, through real-time posture change monitoring, the measurement data is error corrected, the influence of the external environment on size detection is offset, and the corrected size data is generated, ensuring the dynamic adaptability and high precision of the hull size, and the practicality is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a schematic diagram of the steps of the present invention. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0050] See also Figure 1 The present invention provides a technical solution, a shipping fleet size detection method based on laser radar and image fusion, comprising the following steps:

[0051] Collect 3D point cloud data and 2D image data, align the data by timestamp and spatial coordinates to obtain an aligned data set; based on the aligned data set, identify the initial boundaries of the overlapping areas, mark and separate the areas, and obtain initial boundary markings;

[0052] Based on the initial boundary marking, edge detection is applied to refine the edge contour of each ship to obtain refined edge information; the contour of each ship is individually annotated according to the refined edge information, a separated hull annotation is created, and a fine contour mark is generated;

[0053] Based on the fine contour identification, point-to-point matching is performed to reconstruct the 3D model of each ship to obtain a 3D construction model. The size and volume of the hull are calculated using the 3D construction model to generate optimized size data;

[0054] Apply the optimized dimensional data, monitor the attitude changes of the ship under dynamic conditions, perform error correction of dimensional measurement, and generate corrected dimensional data.

[0055] The steps to obtain the aligned data set are:

[0056] Collect 3D point cloud and 2D image data from LiDAR and image sensors, label each data point with a timestamp and spatial coordinates, and generate a preliminary aligned dataset;

[0057] Based on the preliminary aligned data set, the spatial deviation between data points is calculated using the formula:

[0058] ;

[0059] Get the average spatial deviation ,in, Respectively The coordinates of the data points, is the average coordinate of all data points, is the total number of data points;

[0060] According to the average spatial deviation, the positions of data points are adjusted to optimize the data alignment process and form an aligned data set.

[0061] Specifically, three-dimensional point clouds and two-dimensional image data are collected from lidar and image sensors, and each data point is labeled with a timestamp and spatial coordinates. The three-dimensional spatial coordinates and capture time of each data point are recorded in real time by data acquisition equipment to ensure that the time label and spatial coordinates of each data point are accurately matched, providing a basis for data integration. In addition, time synchronization is used to ensure the consistency of timestamps, and coordinate transformation is used to ensure the spatial alignment of data. Through these processes, effective integration of data from different sources can be achieved, and the consistency of time and space data can be used to further process the data to generate a preliminary aligned data set, which is a key prerequisite for subsequent data processing and analysis.

[0062] The formula is beneficial in that it can quantify the distribution deviation of data points in three-dimensional space, helping to identify the alignment quality in the data set, thereby guiding the adjustment of alignment parameters to optimize the accuracy of data integration; Respectively The coordinates of the data points are collected in real time by 3D scanning equipment. is the coordinate average of all data points, and is calculated by taking the arithmetic average of the corresponding coordinate values ​​of all data points. is the total number of data points, obtained by counting the number of points in the data set;

[0063] Calculation process: The coordinates of the data points are and , then the average coordinate is , , put it into the formula to calculate have to:

[0064] ;

[0065] ;

[0066] ;

[0067] This result indicates that the average spatial deviation is 15, which indicates the average deviation of the data points in three-dimensional space relative to their mean position. The larger the value of , the more likely it is that the spatial alignment between data points needs to be further optimized.

[0068] According to the average spatial deviation, the position of the data points is adjusted to optimize the data alignment process. The adjustment method includes applying a deviation correction value to each data point to ensure that the position of each point is closest to the average position of other points; the overall deviation is gradually reduced through iteration until the preset accuracy requirement is reached, thereby forming an aligned data set, and the efficiency and effect of deviation reduction are monitored through a real-time feedback mechanism.

[0069] The steps to obtain the initial boundary marker are:

[0070] Through the aligned data set, the spatial feature analysis is performed on each data point, and the overlapping areas are identified by comparing the positional relationship between points, and the initial boundaries of the overlapping areas are marked to generate the initial boundary data set;

[0071] Based on the initial boundary data set, the geometric center of each boundary area is calculated using the following formula:

[0072] ;

[0073] in, For the The geometric center of the boundary region, In the region The three-dimensional coordinates of the points, For the The total number of points within the boundary area;

[0074] Using the geometric center, refine the boundaries of each region to form an initial boundary marker.

[0075] Specifically, we perform spatial feature analysis on each data point, and by comparing the relative position relationship between points, we mark the areas with high overlap. First, we extract the coordinates of each point in the data set, and then identify clusters of points with similar three-dimensional characteristics through a specific spatial comparison procedure. Next, we use the Canny edge detection algorithm to draw the initial boundaries of these areas based on the spatial proximity between points.

[0076] The benefit of the formula is that it can calculate the geometric center of each boundary region, which helps the subsequent boundary refinement and region separation process; It is The geometric center of a boundary area is obtained by calculating the root mean square of the three-dimensional coordinates of all points in the area. In the region The three-dimensional coordinates of the points, is the total number of points in the area;

[0077] Calculation process: First, obtain the three-dimensional coordinates of each point , then perform square root operations on each coordinate, add them up and divide by the total number of points For example, there are three points in an area with coordinates (1,2,3), (2,3,4), and (3,4,5). The calculation process is:

[0078] ;

[0079] This result indicates that the calculated geometric center position reflects the average position of all points in the area.

[0080] Using the geometric center, the process of refining the boundaries of each region is performed. During this process, each initial boundary is adjusted to optimize the shape and position of the boundary to better distinguish adjacent regions. It includes iterative adjustment of each boundary point to ensure the accuracy of the boundary and the uniform distribution of the point set in the region. By dynamically adjusting the boundary points, the overlap between regions is reduced and the independence of the regions is increased. All boundary adjustments are based on minimizing the spatial discreteness of the internal point set to achieve region segmentation.

[0081] The steps to obtain refined edge information are:

[0082] Using the initial boundary markers, the outer contour of each ship is scanned to capture the contour changes and identify the contour of each ship, thus forming a preliminary edge contour dataset;

[0083] Based on the preliminary edge contour dataset, the edge of each ship is refined by image enhancement to improve image contrast and connect broken contour lines to obtain a refined edge contour dataset.

[0084] Based on the refined edge contour data set, the refined edge information of the vessel is defined and completed to form refined edge information.

[0085] Specifically, the outline of each ship is initially identified. This process relies on a detailed analysis of light reflection and shadow changes. Clear outlines are extracted through edge enhancement and threshold adjustment. Continuity and breakpoints are the main detection focus at this stage. The gradient-based method is used to refine the edges. The grayscale gradient intensity is calculated for each pixel of the image. High-intensity areas are marked as potential edge points. After preliminary processing, these data are stored as a contour dataset. This dataset captures the basic shape of the ship and provides the necessary baseline for subsequent refinement steps.

[0086] Refine the edge contour of the identified ship and use image sharpening processing to enhance the local contrast of the image. In addition, breakpoint connection is a key step. The broken edges are reconnected through edge tracking, and noise is removed through morphological operations such as dilation and erosion. Smoothing reduces the jagged effect and ensures the integrity and continuity of the contour.

[0087] The final processing of each ship's refined edge information is completed by fine-tuning the image's contrast and brightness to enhance edge visibility, edge smoothing to reduce noise in the image and improve the smoothness of the outline, and further shape adjustment to correct the edges using basic geometric transformations such as rotation and scaling to ensure that each edge matches the actual shape of the ship.

[0088] The steps to obtain the fine contour mark are:

[0089] Based on the refined edge information, the outline of each ship is marked, each hull is distinguished, and a separate hull outline is obtained;

[0090] Based on the individual hull contour annotations, the edge feature score of each ship is calculated using the following formula:

[0091] ;

[0092] in, Score the edge features of the hull, For the The edge strength of the edge point, is the distance to the neighboring point, is the number of edge points;

[0093] According to the edge feature score, the edge contour of each ship is optimized, the edge of each ship is determined, and a fine contour identification is generated.

[0094] Specifically, from the refined edge information, each image pixel is first analyzed point by point to determine the start and end of the edge. Each detected contour line is verified for continuity to ensure that there are no breaks. For each hull, a contour recognition algorithm is applied separately to ensure that the edge of each ship is not confused with other objects or hull contours. In this process, the contrast and brightness of each edge point are adjusted, and the visibility of the edge is enhanced through dynamic threshold processing. In addition, the shape and size of each ship are quantitatively analyzed to ensure the accuracy and consistency of the hull annotation. Through these steps, accurate and clear individual hull contour annotations are generated.

[0095] The usefulness of the formula is that it can quantify the level of detail of each ship outline. By calculating the product of the edge strength of an edge point and the distance to its neighboring points, a score reflecting edge clarity and continuity can be obtained, which helps us further adjust the image processing algorithm to improve image quality. The edge feature score of the hull is obtained by calculating the edge strength of each edge point. and the distance from this point to its neighboring points The product sum is obtained by It can be obtained through edge detection algorithms such as Sobel operator. It is obtained by calculating the Euclidean distance between the point and the adjacent points. is the total number of edge points identified in the image;

[0096] Calculation process: Setting the number of edge points , the edge strength of each point The average is 0.8, and the distance between each point and its neighbors The average is 2 pixels, so the calculation process is:

[0097] ;

[0098] The results show that the edge clarity of the evaluated images is high, indicating that the edge detection and hull annotation are performed well and the edges of each hull can be clearly distinguished and identified.

[0099] After obtaining the edge feature score for each ship, the outline of each ship is fine-tuned and optimized. Based on the score results, the image processing parameters, such as the degree of edge enhancement and the range of outline smoothing, are adjusted to ensure that the edges of each hull are clear and continuous. A nonlinear filter is used to remove noise from the image, while a high-pass filter is applied to improve the definition of the edges. The outline of each hull is adjusted through multiple iterations, and each iteration fine-tunes the position and shape of the edge based on the previous processing results. In this way, not only the accuracy of the hull edge is improved, but also the high quality and practicality of the final generated hull outline identification is ensured.

[0100] The steps to obtain the 3D construction model are:

[0101] Using fine contour marking, point-to-point matching is performed to match the 2D contour data of each ship with the 3D point cloud data to generate primary 3D model data;

[0102] Based on the primary 3D model data, the data points are adjusted and the facets are configured to obtain refined 3D model data;

[0103] Based on the refined three-dimensional model data, smoothing processing is performed to obtain a three-dimensional construction model.

[0104] Specifically, when starting point-to-point matching, first align the 2D contour data with the 3D point cloud data, adjust each data point to match the corresponding 3D position, and adjust the coordinates of each point to ensure its accurate projection in 3D space, especially for the complex geometric shape of the hull, to ensure that each contour point of the 2D image is mapped to the corresponding point of the 3D model. In this process, the coordinate data of each point is compared and adjusted in detail to construct a preliminary 3D model that conforms to the actual physical structure;

[0105] Starting from the primary 3D model, we further refine every detail of the model, paying special attention to every facet of the hull structure, simulating the specific curved surfaces and edges of the hull through mesh adjustment, and performing separate geometric definition and mesh optimization for each part of the model, such as the deck, hull side and stern, to ensure that these areas accurately reflect the design intent;

[0106] Based on the refined 3D model, an optimization process is performed, which involves smoothing the model and enhancing the details, such as correcting small surface defects and smoothing transition areas to improve the visual quality and practicality of the model. Detailed processing includes optimization of joints and enhancement of surface texture to ensure that every part of the model can meet high standards in technology and visuals.

[0107] The steps to obtain the optimized size data are:

[0108] Based on the three-dimensional construction model, the dimensions of each hull in the model are measured one by one to obtain the hull dimension data and obtain the basic dimension data;

[0109] Based on the basic size data, each hull surface is meshed, and the volume of each small grid is calculated, and the total volume data of each ship is accumulated to obtain the volume data obtained by detailed calculation;

[0110] Based on the volume data obtained through detailed calculation, the volume data is error checked and accuracy optimized, and the data format and unit are adjusted to generate optimized dimension data.

[0111] Specifically, during the measurement process, the location and size of key points are recorded and converted into numerical data. Next, the data is used to perform basic dimensional analysis of the hull, including accurate calculation of length, width and height. This process ensures high accuracy of the data, and all measurement results are systematically organized into basic dimensional data.

[0112] Each hull surface is meshed in detail, subdividing it into thousands of small cells, each representing a specific surface area. The volume of each cell is calculated one by one. These volume cells are added up to form the total volume of the entire ship.

[0113] Error checking and accuracy optimization of volume data. First, the volume data is reviewed item by item to identify calculation errors or outliers, and necessary adjustments and corrections are made. Then, according to actual application requirements, the format and unit of the data are adjusted to optimize the usability and practicality of the data. These optimization measures ensure high accuracy and adaptability of the data.

[0114] The steps for obtaining the corrected dimension data are:

[0115] Using the optimized size data, the ship's attitude changes in dynamic water conditions are tracked, and the ship's tilt angle, acceleration and position change data are collected through sensors to generate dynamic attitude data;

[0116] Based on the dynamic posture data, the measurement error of the ship size is calculated using the following formula:

[0117] ;

[0118] in, Represents the total dimensional measurement error, It is The size value obtained by the measurement is is the expected size value based on the optimized size data, is the number of measurements;

[0119] According to the dimensional measurement error, error correction is performed and the dimensional data is adjusted to generate corrected dimensional data.

[0120] Specifically, an inclination sensor, an accelerometer and a GPS locator are set up to directly measure the attitude changes of the ship in dynamic water conditions, and to capture the inclination angle, acceleration and specific spatial position changes of the ship in real time.

[0121] The usefulness of the formula is that it provides a method to quantify the accuracy of vessel dimension measurements by calculating the difference between the actual measured dimensions and the expected dimensions, helping to identify possible systematic errors in the measurement process and make necessary corrections. It represents the total dimensional measurement error and indicates the accuracy of the measurement; It is The dimensional values ​​obtained by the laser rangefinder under controlled conditions; It is the expected dimensional value based on optimized dimensional data, usually a design parameter or a previously calibrated dimension; is the number of measurements.

[0122] Calculation process:

[0123] In one scene, the sizes of three points were measured. The expected size was 10 meters. The actual measurement results were 9.8 meters, 10.1 meters, and 9.9 meters. The error calculated by substituting these data into the formula is as follows:

[0124] First calculate the square of the error at each point:

[0125] ;

[0126] ;

[0127] ;

[0128] Then sum: , and finally, calculate the square root of the error to get the total error: ;

[0129] This result shows that the current measurement error is around 0.245 meters, based on the difference between the measured data and the expected value, indicating that the measurement accuracy is good but there is still room for improvement. Further analysis of this result can help determine the measurement equipment settings that need to be adjusted or the data processing algorithms that need to be improved to reduce the error and improve the accuracy of the hull dimension data.

[0130] Using the dynamic attitude data set and the calculation results of the dimensional measurement error, manually adjust the measurement equipment to ensure the accuracy of the measurement under different dynamic conditions. Analyze each measurement data, especially the data under complex sea conditions, identify anomalies or deviations in the data, and make appropriate adjustments. Adjust the measurement deviation, refine each data record, and generate corrected dimensional data.

[0131] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A shipping fleet size detection method based on laser radar and image fusion, characterized in that: The following steps are involved: Collect 3D point cloud data and 2D image data, align the data by timestamp and spatial coordinates to obtain an aligned data set; Based on the aligned data set, identifying initial boundaries of overlapping regions, marking and separating the regions, and obtaining initial boundary markings; Based on the initial boundary marking, edge detection is applied to refine the edge contour of each ship to obtain refined edge information; the contour of each ship is individually marked according to the refined edge information, a separate hull mark is created, and a fine contour mark is generated; Based on the fine contour identification, point-to-point matching is performed to reconstruct the three-dimensional model of each ship to obtain a three-dimensional construction model, and the size and volume of the hull are calculated using the three-dimensional construction model to generate optimized size data; Using the optimized dimension data, monitoring the attitude change of the vessel under dynamic conditions, performing error correction of dimension measurement, and generating corrected dimension data; The steps for obtaining the initial boundary mark are: Through the aligned data set, spatial feature analysis is performed on each data point, and by comparing the positional relationship between points, the overlapping area is identified, the initial boundary of the overlapping area is marked, and an initial boundary data set is generated; Based on the initial boundary data set, the geometric center of each boundary area is calculated using the following formula: ； in, For the The geometric center of the boundary region, In the region The three-dimensional coordinates of the points, For the The total number of points within the boundary area; Using the geometric center, refine the boundary of each region to form an initial boundary mark; The step of obtaining the refined edge information is: Using the initial boundary markers, the outer contour of each ship is scanned to capture contour changes, and the contour of each ship is identified to form a preliminary edge contour data set; Based on the preliminary edge contour data set, the edge of each ship is refined by image enhancement to improve image contrast and connect broken contour lines, thereby obtaining a refined edge contour data set; Based on the refined edge contour data set, defining and completing refined edge information of the vessel to form refined edge information; The steps for obtaining the fine contour mark are: Based on the refined edge information, the contour of each ship is marked, each hull is distinguished, and a separate hull contour mark is obtained; Based on the individual hull contour annotations, the edge feature score of each ship is calculated using the following formula: ； in, Score the edge features of the hull, For the The edge strength of the edge point, is the distance to the neighboring point, is the number of edge points; According to the edge feature score, the edge contour of each ship is optimized, the edge of each ship is determined, and a fine contour mark is generated; The steps of obtaining the three-dimensional construction model are: Using the fine contour identification, point-to-point matching is performed to correspond the two-dimensional contour data of each ship with the three-dimensional point cloud data to generate primary three-dimensional model data; Based on the primary three-dimensional model data, the data points are adjusted in position and the facets are configured to obtain refined three-dimensional model data; Based on the refined three-dimensional model data, smoothing processing is performed to obtain a three-dimensional construction model.

2. The method for detecting the size of a shipping fleet based on laser radar and image fusion according to claim 1 is characterized in that: The steps for obtaining the aligned data set are: Collect 3D point cloud and 2D image data from LiDAR and image sensors, label each data point with a timestamp and spatial coordinates, and generate a preliminary aligned dataset; Based on the preliminary aligned data set, the spatial deviation between data points is calculated using the formula: ； Get the average spatial deviation ,in, Respectively The coordinates of the data points, is the average coordinate of all data points, is the total number of data points; According to the average spatial deviation, the positions of the data points are adjusted, the data alignment process is optimized, and an aligned data set is formed.

3. The method for detecting the size of a shipping fleet based on laser radar and image fusion according to claim 1, characterized in that: The steps for obtaining the optimized size data are as follows: Based on the three-dimensional constructed model, the dimensions of each hull in the model are measured one by one to obtain the dimensional data of the hull and obtain basic dimensional data; Based on the basic size data, each hull surface is meshed, and the volume of each small mesh is calculated, and the total volume data of each ship is accumulated to obtain the volume data obtained by detailed calculation; Based on the volume data obtained by the detailed calculation, the volume data is error checked and the accuracy is optimized, and the data format and unit are adjusted to generate optimized dimension data.

4. The method for detecting the size of a shipping fleet based on laser radar and image fusion according to claim 1, characterized in that: The steps for obtaining the corrected dimension data are as follows: Using the optimized size data, tracking the posture changes of the vessel in dynamic water conditions, collecting the tilt angle, acceleration and position change data of the vessel through sensors, and generating dynamic posture data; Based on the dynamic posture data, the measurement error of the ship size is calculated using the following formula: ； in, Represents the total dimensional measurement error, It is The size value obtained by the measurement is is the expected size value based on the optimized size data, is the number of measurements; According to the dimension measurement error, error correction is performed and the dimension data is adjusted to generate corrected dimension data.

Images