A method for measuring the size of a ship based on a laser radar

Abstract

The application discloses a ship size measurement method based on a laser radar, and main steps of the method include: 1) using a laser radar sensor to scan and collect point cloud data in a channel in real time; 2) using a spatial coordinate noise filtering method to extract point cloud data in a region in the channel region; 3) using the characteristic difference that different objects have different light reflection intensities to filter water surface noise points; 4) using a DBSCAN algorithm to cluster the point cloud in the channel, and performing target detection on a ship; 5) judging a ship sailing direction by extracting and analyzing ship edge features; 6) through multi-frame fusion and splicing of ship point clouds, compensating for missing point clouds of a ship head and a ship tail, enhancing the point cloud data, and obtaining more complete ship point clouds; and 7) using a ship point cloud set to construct an AABB bounding box model, and completing size calculation on length features, width features and height features of the ship. The ship size measurement method can accurately detect the length, width and height size features of the ship.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of laser radar, in particular to a ship size measurement method based on laser radar. BACKGROUND

[0002] China has vast territory and numerous inland waterways, and has rich water transportation resources. With the increasing attention of the Ministry of Transport of the People's Republic of China to water transportation, the measurement of the size of the ship in the inland waterway is an important research content in water transportation. With the increasing density of the national waterway network, the digitization of the inland waterway is an important task at this stage. Through the monitoring, statistics and analysis of the ship, it is of great significance to the development of intelligent waterway, and through the accurate measurement of the main size of the ship, it is beneficial to prevent traffic congestion, strengthen the bridge anti-collision warning system, prevent ship overloading and overrunning, and prevent ship overspeed. At present, a set of scheme is needed to realize the monitoring, statistics and analysis of the ship.

[0003] The laser radar has the advantages of strong spatial perception ability, strong environmental adaptability, and accurate detection of high-resolution micro-ship data. The application of laser radar in inland waterway can accurately collect ship point cloud in the waterway, and through informationization, digitization and intelligentization, the ship flow, direction, size, tonnage and other information can be monitored, counted and analyzed, which not only saves manpower and material resources, but also greatly reduces the errors of manual observation, and further improves the efficiency of statistical results. In summary, the current digital and intelligent ship size measurement method is an urgent research direction. SUMMARY

[0004] In order to solve the shortcomings of the prior art, the purpose of the present application is to provide a ship size measurement method based on laser radar. Through the digital, information and intelligent way of ship size measurement, the current ship flow information, direction information and size information of the waterway are accurately grasped, which provides a strong data basis for evaluating and analyzing the economic situation of water transportation.

[0005] In order to achieve the above purpose, the present application provides a ship size measurement method based on laser radar, which specifically includes the following steps:

[0006] S1: Ship point cloud data acquisition based on laser radar. The laser radar acquisition system is deployed at the central position of the bridge, a tripod is welded at the middle position of the bridge and protrudes from the bridge, and the laser radar is placed on the tripod. In this way, the laser radar can be deployed to maximize the use of laser radar beam, reduce the shielding of bridge fence and the shielding between ships, and collect more complete ship point cloud data.

[0007] S2: Filtering noise by spatial coordinate noise filtering method and selecting channel region. Since the laser radar is fixedly deployed on the bridge central protruding tripod, the collected original point cloud data contains not only ship point cloud data but also other noise point cloud (including noise point cloud on the shore side, noise point cloud on the bridge and fluctuating water surface noise point cloud), and the channel region is a fixed region relative to the laser radar. Therefore, by calculating the position of the laser radar deployment position and the position of the shore side, the coordinates of the current channel region in the laser radar point cloud data coordinate system are determined, the noise points of the bridge and the shore side are filtered by setting the coordinate parameters, and the ship point cloud data in the channel region is reserved.

[0008] S3: Filtering water surface noise points by using the feature that the reflection intensity of the water surface is different from that of the ship surface. After the channel region is selected, the ship point cloud data is retained, and at the same time a part of the water surface noise point cloud (dynamic water surface noise points formed after the ship passes through the water surface or after the wind passes through the water surface) is also retained. The reflection intensity values of the water surface noise points and the ship point cloud data are obviously different. According to the set reflection intensity threshold, the water surface noise point cloud is filtered, and the form of the ship point cloud data is retained.

[0009] S4: Ship target detection based on DBSCAN clustering method. Laser point cloud data is essentially a kind of low-resolution discrete sampling data. Clustering is to divide data objects with high similarity into the same cluster and divide data objects with high dissimilarity into different clusters. DBSCAN clustering algorithm is a density-based spatial clustering algorithm. This algorithm clusters clusters with sufficient density and discovers clusters of arbitrary shape in a spatial database with noise. The disordered ship point cloud is clustered into clusters using the DBSCAN clustering method, and the ship clusters are labeled with id numbers in order.

[0010] S5: Ship flow statistics and navigation direction judgment based on edge feature. A section is used as a detection means to detect and extract the edge features of the ship bow and the edge features of the ship stern. The number of ships passing through the section is counted through the extraction of edge features, and the navigation direction of the ship is judged by comparing the time difference of the appearance of the edge features of the two sections.

[0011] S6: Method for completing missing ship point cloud based on multi-frame fusion method. Find the frame number of the ship close to the laser radar, find and select a frame with relatively complete ship point cloud as the reference frame, save the data of the reference frame in the matrix. The ship point cloud in the frame number with high concentration of ship point cloud near the reference frame is also extracted and saved in the matrix, and the ship point cloud is fused and registered with the ship point cloud of the reference frame. Take the complete bow and stern point cloud, splice them with the relatively complete ship point cloud, splice the ship point cloud into a relatively complete point cloud, realize ship point cloud data enhancement, and obtain more complete ship point cloud data, so that the calculation of ship size data is more accurate.

[0012] S5: Ship main dimension calculation based on laser radar point cloud data. The ship main dimension calculation includes calculating the length, width and height information of the ship, the ship length refers to the full length of the ship, that is, the distance from the bow to the stern; the ship width refers to the full width of the ship, that is, the distance between the two side edges of the deck of the ship; the ship height information refers to the height distance from the deck to the waterline. According to the ship point cloud obtained in the above step, an AABB bounding box model is established for the ship point cloud set. The AABB bounding box is defined as the smallest cuboid that can wrap the to-be-measured model and is always parallel to the space coordinate axis with eight edges. The area of the AABB bounding box is represented as follows:

[0013] S = {(x, y, z) | x min≤ x ≤ x max , y min≤ y ≤ y max , z min≤ z ≤ z max},

[0014] Further, the ship point cloud data acquisition system based on laser radar in step S1 includes four modules of laser radar outer machine, laser radar inner machine, switching power supply and industrial computer. Among them, the switching power supply is used to control the on-off state of the whole system, the laser radar sensor is used to scan and collect point cloud information in the channel, and the industrial computer is used to process point cloud data.

[0015] Further, the channel region division based on the space coordinate method in step S2. According to the data scanned by the laser radar, the range of the inland waterway in x, y and z axes is calculated, the coordinate range of the bridge noise points, the coordinate range of the channel and the coordinate range of the shore side noise points are found in the visualization software, and by intercepting the range of the coordinate axis, the visualization software can only display the point cloud in the inland waterway, that is, the ship point cloud in the channel can be directly obtained. This method uses the different ranges of different objects in x, y and z axes, which can effectively filter out the shore side point cloud and the bridge noise point cloud and other noise point clouds outside the channel.

[0016] Further, the water surface noise point filtering in step S3 is based on the different characteristics of the reflected light intensity. When the water surface is calm, there are almost no noise points on the water surface. But when the water surface is fluctuating, waves will form on the water surface, and water surface noise points will begin to appear. When collecting data, due to the wind blowing on the water surface, waves will be generated on the water surface, and the water surface will also generate waves when the ship passes through the water surface. Some of these wave point clouds are independent of the ship point cloud, and some are closely connected with the ship point cloud. Through observation, it is found that the reflected light intensity of the ship surface is different from that of the water surface. By screening the point cloud, the water surface noise is removed, so as to realize the filtering of the water surface noise points.

[0017] Further, the step S4 detects the ship target. After removing the wave point cloud, the remaining point cloud is still a series of unordered points, and various information of the target cannot be directly obtained. In order to obtain the position and size of the target and other information, a clustering algorithm is needed to cluster the point cloud from the same target into an independent whole. When the ship point cloud reaches the detection area D, wherein,

[0018] D = {(x, y, z) | -40≤x≤40, 0≤y≤100, -11.3≤z≤-2},

[0019] The DBSCAN clustering method is used to cluster the point cloud appearing in the detection area. DBSCAN clustering is a density-based spatial clustering algorithm. The algorithm divides the area with sufficient density into clusters and discovers clusters of arbitrary shape in a spatial database with noise. The DBSCAN algorithm does not need to determine the number of clusters in advance, but needs to determine two parameters: Eps and MinPts. Eps is the distance threshold of the point neighborhood, and Eps is too large, which will cause more points to fall into the same cluster, thereby reducing the number of clusters, and vice versa. MinPts represents the minimum number of samples in a cluster group. When Eps is fixed, smaller MinPts will produce more clusters and noise points, and larger MinPts will produce larger clusters because it has swallowed many smaller clusters. Using this clustering method, if there is a single ship passing through the detection area, the detection area will only generate one point cloud cluster, and if there are multiple ships passing through, multiple point cloud clusters will be generated. The generated clusters are labeled with id numbers in order and saved.

[0020] Further, the S5 determines the ship sailing direction based on the edge features. A cross section is used as a detection means to detect and extract the edge features of the ship bow and the edge features of the ship stern. The number of ships passing through the cross section is counted through the extraction of the edge features, and the sailing direction of the ship is determined by comparing the time difference of the appearance of the edge features of the two cross sections. When the point cloud data on the selected cross section is detected from nothing to something, that is, the edge feature of the bow is detected, and the time T head is recorded. When the point cloud data on the selected cross section is detected from something to nothing, that is, the edge feature of the stern is detected, and the time T end is recorded. Two cross sections are selected, and the time difference between T head and T end is used to determine the sailing direction of the ship.

[0021] Further, in S6, multiple frames of ship point clouds are fused. The frames of complete ship point clouds closest to the lidar are selected as reference frames, and the frames of other relatively complete ship point clouds with high density are extracted as compensation frames. The reference frames and compensation frames are then fused and registered. The bow and stern edge features of the ship are extracted and stored in a matrix. The method for detecting and extracting the ship's edge features includes: selecting a cross-section within the detection area; when the cross-section point cloud data is detected to change from zero to some, the bow edge features are detected, and the recording time is T. head When the point cloud data changes from present to absent, the stern edge feature is detected, and the recording time is T. end Extract these two frame numbers T. head T end The bow and stern features are analyzed, and the fused and registered ship point cloud is stitched together with the bow or stern point cloud to form a complete ship point cloud, resulting in an enhanced ship point cloud. This data augmentation of the original ship point cloud yields more complete ship point cloud data, preparing the ground for more accurate subsequent calculations of ship dimensions.

[0022] Furthermore, the calculation of the ship's main dimensions in S7 includes calculating the ship's length, width, and height. Based on the ship point cloud obtained in the previous step, an AABB bounding box model is established for each ship point cloud set. An AABB bounding box is defined as the smallest cuboid that can enclose the model under test and always has eight sides parallel to the spatial coordinate axes. Constructing an AABB bounding box is quite simple; it only requires finding the maximum value x of all nodes of the object on the three coordinate axes, namely x, y, and z. max y max z max and minimum value x min y min z min This yields six points of the corresponding bounding box, which define the object's AABB bounding box. The intersection test of bounding boxes determines the overlap of the projections of two AABB bounding boxes onto the three coordinate axes. Bounding boxes intersect if and only if the projections on all three axes overlap. The projection region can be derived from the maximum and minimum values ​​on each coordinate axis, requiring a maximum of six comparison operations. The region of the AABB bounding box is represented as follows:

[0023] S = {(x, y, z)|x min ≤x≤x max y min ≤y≤y max , z min ≤z≤z max},

[0024] Where, x min x max ymax , y min , z max , z min The extreme points of the projections of the object on the x, y and z axes are consistent with the definition in the bounding box.

[0025] The ship main dimension calculation method provided by the application has the following advantages:

[0026] 1. An advanced automatic data acquisition system is built, high-resolution microscopic data are acquired by using a laser radar, and intelligent data processing algorithms are combined to realize automatic and intelligent ship main dimension detection functions, improve detection accuracy, and provide solid data indicators for evaluation of waterway transportation and control.

[0027] 2. Compared with a video sensor, the laser radar sensor has the advantages of strong environmental applicability and high detection accuracy, and can realize all-weather ship size measurement work. BRIEF DESCRIPTION OF DRAWINGS

[0028] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of illustrating the preferred embodiments and are not considered to be limiting on the application. Moreover, the same reference symbols are used throughout the drawings to represent the same parts. In the drawings:

[0029] Figure 1 Flowchart for ship size measurement based on laser radar;

[0030] Figure 2 Schematic diagram of the composition of the laser radar data acquisition system;

[0031] Figure 3 Schematic diagram of the original point cloud data after coordinate rotation;

[0032] Figure 4 Effect diagram of ship point cloud data after filtering out water surface noise points;

[0033] Figure 5 Schematic diagram of the ship point cloud after DBSCAN clustering;

[0034] Figure 6 Curve graph of the number of ships passing through the detection area;

[0035] Figure 7 Schematic diagram of the fusion of a ship point cloud compensation frame and a reference frame;

[0036] Figure 8 Schematic diagram of the point cloud enhancement after splicing the edge features of a ship and the complete ship point cloud;

[0037] Figure 9 AABB bounding box model diagram of a point cloud set of a constructed ship. DETAILED DESCRIPTION

[0038] The embodiments will be described in detail with reference to the drawings, which can help researchers in the field to further understand. Obviously, the described embodiments are part of the embodiments of the present application, not all. Based on the embodiments in the present application, other embodiments obtained by researchers in the field without creativity and novelty also belong to the protection scope of the present application.

[0039] The ship size measurement method based on laser radar provided by the embodiment of the present application, as shown in Figure 1 includes the following steps:

[0040] Step S1: ship point cloud data acquisition based on laser radar sensor. The specific steps are as follows:

[0041] S101: Laser radar data acquisition system composition mode. Referring to Figure 2 , the data acquisition system includes laser radar outer machine, laser radar inner machine, switching power supply and industrial computer. The laser radar is fixed on the welded tripod protruding in the middle of the bridge, which is used for real-time acquisition of ship point cloud data. The data collected by the laser radar outer machine is transmitted to the laser radar inner machine, and then to the industrial computer for data preprocessing. The switching power supply is used to control the working time of the whole system at any time.

[0042] S102: Channel ship point cloud data acquisition deployment mode. The present application deploys laser radar and camera in the central part of the bridge to ensure that the scanning width of the laser radar can cover the entire inland waterway and obtain complete internal data of the inland waterway. Since the laser beams emitted by the laser radar are scanned downward from the central part of the bridge, this installation and deployment scheme not only avoids the point cloud shielding problem between ships, but also can obtain relatively complete ship feature point cloud close to the bridge, which provides good quality data for subsequent ship target detection and size calculation. Among them, the camera is responsible for collecting ship video data, which provides data basis for subsequent comparison with the real scene.

[0043] S103: Data preprocessing. In order to more conveniently filter out the background and construct the ship AABB bounding box model, simplify the calculation difficulty of the ship size, before carrying out the point cloud data processing work, the coordinates of all point cloud data are rotated. The point cloud data is rotated around the Z axis by a corresponding angle of coordinates, so that the channel direction is parallel to the coordinate Y axis in the new coordinate system. The rotation matrix formula is as follows:

[0044]

[0045] Wherein, (X, Y, Z) is the original coordinate of the point cloud data, (X', Y', Z') is the coordinate of the point cloud data after rotation around the Z axis, and θ is the rotation angle. Reference Figure 3 , which is the original point cloud data of the channel after coordinate rotation processing.

[0046] Step S2: Select the channel area based on the spatial coordinate method noise filtering method. The specific steps are as follows: S201-S202:

[0047] S201: Background filtering of noise point cloud data. According to the data scanned by the laser radar, the current laser radar coordinate position, the height of the laser radar from the water surface, the coordinate of the channel side, and the coordinate range of the channel are calculated and found in the visualization software.

[0048] S202: Obtain the internal area of the channel by the spatial coordinate method noise filtering method. Based on the obtained channel spatial coordinates, the point cloud background is filtered to remove the point cloud noise data of the river banks and bridges, and the ship point cloud data in the channel is retained. Thus, the X-axis value range of the channel space in the laser radar coordinate system is [-40, 40], the Y-axis value range is [0, 100], and the Z-axis value range is [-11.3, -2], all in units of m.

[0049] Step S3: Filter the water surface noise points based on the different reflection intensity of the water surface points and the ship surface point cloud. The specific steps are as follows: S301:

[0050] S301: Water surface noise point cloud removal based on reflection intensity setting. The calm water surface in the channel basically has no reflection points, but when the water surface is fluctuating, the laser beam vertically irradiated on the water surface will reflect back a part of the water surface noise point cloud, and the reflection intensity of such noise points is: Intensity = 1 and Intensity = 2; In addition, the water surface waves caused by the passing of the ship in the current water area will produce water surface noise points, and the reflection intensity of such noise points is also: Intensity = 1 and Intensity = 2. The reflection intensity range of the ship point cloud data scanned by the laser radar is [1, 60], so removing the point cloud data with reflection intensity Intensity = 1 and Intensity = 2 can filter out the water surface noise points and retain most of the ship point cloud. Reference Figure 4 , which is the ship point cloud data effect diagram after filtering out the water surface noise points.

[0051] Step S4: Detect the ship target based on the DBSCAN clustering method. The specific steps include S401-S402:

[0052] S401: Algorithm parameter setting. DBSCAN algorithm is used to cluster points with sufficient density (i.e. large density) in high-dimensional space, and mark the sparse points that cannot be divided into any category as noise. The algorithm uses the concept of local density to divide each point into core points, boundary points and noise points. DBSCAN algorithm is a classic clustering algorithm, which determines the connectivity between samples from the tightness of sample distribution represented by neighborhood radius Eps and minimum number of points in neighborhood MinPts, divides samples with connectivity into the same cluster, and then expands the cluster in the space data with noise to generate the clustering result. The selection of parameters Eps and MinPts directly affects the result of target detection. In this embodiment, based on the distribution characteristics of ship point cloud, the neighborhood radius Eps is set to 1.5 m, and the minimum number of points in neighborhood MinPts is set to 20.

[0053] S402: Ship target detection. If it is a single ship scene, only one point cloud cluster will be generated in the detection area; if it is a multi-ship scene, several clusters will be generated in the detection area. Each cluster represents a ship point cloud set, and the points of these clusters, that is, the ship point cloud set, are retained in the matrix. According to the time sequence of the ship passing through the bridge, each ship is marked with an id number. This method not only realizes the target detection of the ship, but also realizes the flow detection in the channel. Figure 5 , which is a schematic diagram of ship point cloud after DBSCAN clustering. Reference Figure 6 , which is a curve of the number of ships passing through the detection area in 1 hour and 6 hours.

[0054] Step S5: Ship flow statistics and sailing direction judgment based on edge features. The detailed content of step S5 is described as follows S501-S502:

[0055] S501: Ship edge feature detection and extraction. The edge features of the ship include the bow edge feature and the stern edge feature. A cross section is selected in the detection area to detect the edge features of the ship. When the point cloud data is detected from nothing to something, that is, the bow edge feature is detected, and the time T head is recorded. end

[0056] S502: Ship sailing direction judgment according to ship edge features. Two detection cross sections are set, which are upper cross section and lower cross section. The time when the bow edge feature of the ship passes through the upper cross section is recorded as T up , and the time when the bow edge feature of the ship passes through the lower cross section is recorded as T down . When T up T down ​When T up > T down , the direction of the ship sailing is from bottom to top.

[0057] Step S6: Point cloud data compensation based on multi-frame fusion. The details of step S6 are described as follows S601-S604:

[0058] S601: Extract a relatively complete point cloud of the ship as a reference frame. Find the point cloud of the ship close to the laser radar that displays relatively complete ship features, store the point cloud at this time in the matrix X, and find the maximum and minimum values of the point cloud on the x and y axes: x max , x min , y max , y min Store these data in matrix X, wherein,

[0059] X = {(x, y, z) | (class i , P id , x max , x min , y max , y min )}.

[0060] Wherein, class represents the ship type, P id represents the ship id number, x max , x min , y max , y min represent the maximum and minimum values of the reference frame on the x and y axes.

[0061] S602: Extract 20 frames of relatively complete ship point cloud frames near the ship reference frame as compensation frames. Store the ship point cloud data of each frame in the 20 frames in matrix Y i .

[0062] Y i = {(x, y, z) | (class i , P id , x i-max , x i-min , y i-max , y i-min )}.

[0063] i represents the frame number, x i-max , x i-min , y i-max , y i-min represent the maximum and minimum values of each frame of ship point cloud in the compensation frame on the x and y axes.

[0064] S603: Multi-frame fusion of the compensation frame and the reference frame. The minimum value y i-min of the y-axis of the reference frame min Difference: Ay1 = |y i-min- -y min | is translated by Ay units on the y-axis and fused with the reference frame to obtain multi-frame fused ship point clouds with enhanced point cloud quantity. Referring to Figure 7 , it is a schematic diagram of fusion of a ship point cloud compensation frame and a reference frame.

[0065] S604: Splicing of the ship edge feature point cloud and the fused ship point cloud. For a top-down ship, the stern point cloud is generally incomplete, so the stern point cloud is moved by Ay2 = |y i-max -y max | units on the y-axis, where y max represents the maximum value of the y-axis of the reference frame. The missing part of the ship point cloud is filled. For a bottom-up ship, the head and tail point clouds are generally incomplete, so the ship head point cloud is moved by Ay3 = |y i-min -y min | units on the y-axis. The missing part of the ship head point cloud is filled. Through the above steps, a complete feature ship point cloud can be constructed, which prepares for the next step of accurate measurement of ship size. Referring to Figure 8 , it is a schematic diagram of point cloud enhancement after splicing of a ship edge feature and a complete ship point cloud.

[0066] Step S7: Ship size extraction based on the AABB bounding box model. The specific steps include S701-S702:

[0067] S701: Construction of the AABB bounding box model based on the ship point cloud set. Based on the complete feature ship point cloud set obtained in the previous step, an AABB bounding box model is established for each ship point set to calculate the size features of the ship. The AABB bounding box is defined as the smallest cuboid that can wrap around the model to be measured and always has eight edges parallel to the spatial coordinate axes. The AABB bounding box can be represented by the maximum and minimum values: six points x min , x max , y min , y max , z min , z max are calculated to determine the six points of a bounding box. The six parameters representing the AABB bounding box can be divided into the following two groups:

[0068] P min = [x min , y min, z min ]

[0069] P max = [x max , y max , z max ]

[0070] The region of an AABB bounding box is represented as follows:

[0071] S = {(x, y, z) | x min ≤ x ≤ x max , y min ≤ y ≤ y max , z min ≤ z ≤ z max},

[0072] Referring to Figure 9 , a schematic diagram of an established AABB bounding box model of a ship point cloud set.

[0073] S702: Ship length, width, height calculation. The ship length L ship , ship width W ship and ship height H ship are extracted from the established AABB bounding box of the ship point set.

[0074] Wherein, the ship length L ship is calculated according to the following formula:

[0075] L ship = |y max -y min |

[0076] The ship width W ship is calculated according to the following formula:

[0077] W ship = |x max -x min |

[0078] The ship height H ship is calculated according to the following formula:

[0079] H ship = |z max -z min |

[0080] As shown in Table 1, the detected channel passing ship information includes ship number, ship size (length, width, height) and ship navigation direction.

[0081] Table 1 Channel ship detection information

[0082] Vessel number Vessel heading Length of vessel (m) Width of vessel (m) Height of vessel (m) 1 1 46 8.9 1.99 2 0 48 8.3 2.65 3 1 52 11.65 3.6 4 1 45 8.2 2.31 5 1 46 8.6 2.54 6 0 45 7.98 2.36 7 0 38 6.79 2.4 … … … … …

[0083] wherein the ship sailing direction, 1 represents the ship sailing from top to bottom, 0 represents the ship sailing from bottom to top.

[0084] The above merely provides the preferred but non-restrictive embodiment of the present application, and the protection scope of the present application should not be limited thereto, and any change or replacement within the technical range disclosed by the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for measuring ship dimensions based on lidar, characterized in that, The method includes the following steps: S1: Use lidar to collect real-time point cloud data of ships in the waterway space; S2: Based on the point cloud data collected by lidar, the spatial coordinate noise filtering method is used to find the coordinate range of the waterway according to the lidar coordinate system, and the corresponding x, y, z axis coordinate parameters are set to select the corresponding waterway area range and extract the ship point cloud data in the waterway. S3: By utilizing the difference in light intensity reflected from the water surface and the ship surface, noise points on the water surface are filtered out while retaining ship point cloud data; S4: When the point cloud enters the target detection area, the DBSACN clustering algorithm is used to cluster the point cloud appearing in the detection area, so that each ship point cloud forms a cluster after clustering. The ID number of each ship point cloud cluster is marked according to the order of ship passage time. S5: Detect and extract the edge features of ships, then perform ship traffic statistics based on the edge features, and determine the ship's sailing direction; S6: Perform multi-frame fusion on the ship point cloud to complete the missing point cloud at the bow and stern of the ship point cloud and enhance the point cloud data; S7: Using the ship point cloud information obtained in step S6, construct an AABB bounding box model for the ship point cloud set, and further calculate the length, width and height feature information of the ship, thereby realizing the ship size calculation; Step S6 further includes: finding a relatively complete ship point cloud close to the lidar and taking this frame as the reference frame; extracting ship point clouds with higher density from other frames near the reference frame as compensation frames; performing multi-frame fusion and registration between the compensation frames and the reference frame; extracting bow and stern edge features, and stitching the extracted bow and stern edge features with the fused ship point cloud to obtain a complete ship point cloud set without missing bow and stern. The process involves fusing and registering multiple frames of the compensation frame and the reference frame; extracting the bow and stern edge features; and stitching the extracted bow and stern edge features with the fused ship point cloud to obtain a complete ship point cloud set without missing bow and stern features. The minimum value of y on the y-axis for each frame of the compensation frame. i-min The minimum value of y on the y-axis of the reference frame min Difference: Δy1=|y i-min- y min |, translate all ship point clouds in each frame of the compensation frame by Δy1 units on the y-axis and fuse and register them with the reference frame to obtain a multi-frame fused ship point cloud with enhanced point cloud quantity; For ships viewed from top to bottom, complete the stern point cloud and move the stern point cloud along the y-axis by Δy2=|y i-max -y max | units, y max This represents the maximum value of the reference frame on the y-axis. i-max This represents the maximum value on the y-axis for each frame of the compensation frame, filling in the missing stern portion of the ship point cloud; for ships viewed from bottom to top, the bow point cloud is completed by shifting the bow point cloud along the y-axis by Δy3 = |y i-min -y min Each unit will fill in the missing portion of the ship's bow in the point cloud.

2. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, Step S1 further includes: deploying a 32-line lidar sensor in the center of the bridge in the waterway, so that the scanning width covers the ship point cloud data in the entire waterway, and collecting ship point cloud data in the current waterway in real time; rotating the coordinates of all point cloud data, rotating the point cloud data around the Z-axis by a preset angle, so that the waterway direction is parallel to the Y-axis in the new coordinate system.

3. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, Step S2 further includes: finding the coordinate range of the bridge and the coordinate range of the waterway in the lidar coordinate system, filtering out the noise points on the shore and the bridge, saving the ship point cloud in the waterway, and obtaining the target detection area.

4. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, Step S3 further includes: filtering out water surface noise points and retaining ship point cloud data by setting the intensity of reflected light.

5. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, Step S4 further includes: clustering the ship point cloud using the DBSCAN clustering method, grouping the ship point cloud into clusters, and sorting and labeling each ship point cloud cluster with an ID number according to the time when the ship clusters passed the bridge.

6. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, In step S5, the edge features of the ship are divided into bow edge features and stern edge features; a cross-section is selected within the detection area, and when the point cloud data of the cross-section changes from zero to some, the bow edge feature is detected, and the recording time is T. head When the point cloud data changes from present to absent, the stern edge feature is detected, and the recording time is T. end Two cross-sections are selected within the waterway, and the ship's course is determined by comparing the differences in the time it takes for the ship to pass through the two cross-sections.

7. A method for measuring ship dimensions based on lidar according to claim 6, characterized in that, The method of selecting two cross-sections within the waterway and determining the ship's course by comparing the time difference between the two cross-sections includes: The time it takes for the edge feature of the ship's bow to pass through the upper cross section is recorded as T. up The time record of the edge features of the ship's bow through the lower section is T. down ; When T up <T down When T indicates that the ship is sailing from top to bottom, up >T down When, it indicates that the ship is sailing in the direction of bottom to top.

8. The method for measuring ship dimensions based on lidar according to claim 1, characterized in that, Step S7 further includes: calculating the main dimensions of the ship. Using the ship point cloud set obtained in step S6, an AABB bounding box model is established. The maximum and minimum values ​​of the point cloud of each ship on the x, y, and z axes are found. The ship's length, width, and height information are calculated by the difference between the maximum and minimum values ​​of the point cloud data on the x, y, and z axes.

Images