A multi-laser-radar-based water conservancy ship lock anti-collision early warning method and system

Abstract

The application discloses a kind of water conservancy ship lock anti-collision early warning method and system based on multiple laser radars, it is related to water conservancy ship lock anti-collision early warning technical field, dynamic three-dimensional environment model S (t) is constructed by the space-time cooperation of multiple laser radars Collection and point cloud data processing of multi-source laser radar data fusion module, effectively make up the information missing problem of traditional visual monitoring system under illumination, shielding and space blind area.Except for the feature extraction and cluster analysis of environment model, abnormal obstacle identification and classification module can distinguish static obstacle, floating obstacle and potential high-risk target, and construct obstacle interference set O (t), so as to break through the limitation that the existing system is weak in "unstructured target" perception ability and reaction lag.Water disturbance field modeling module introduces disturbance field FW modeling method, and fuses water fluctuation and floating object movement trend, which can accurately predict the disturbance direction and intensity of ship heading, and provide more forward-looking anti-collision basis.

Description

Technical Field

[0001] This invention relates to the field of collision prevention and early warning technology for hydraulic ship locks, specifically to a collision prevention and early warning method and system for hydraulic ship locks based on multiple lidar sensors. Background Technology

[0002] Water transport engineering, as an important branch of water conservancy engineering, widely serves inland waterway shipping, port hubs, and river basin management systems, and is a key infrastructure supporting regional economic operation. Ship lock systems, as typical control nodes in water transport engineering, primarily function to regulate water level differences and ensure smooth passage of vessels. In the intelligent transformation of modern ship locks, more and more technologies are developing towards "environmental perception – risk assessment – ​​autonomous control." Among these, multi-laser radar systems, due to their advantages of high precision, all-weather operation, and strong structural recognition capabilities, are widely used for safety monitoring of complex spatial structures.

[0003] Current lock collision avoidance systems generally rely on single-sensor identification or visual monitoring based on fixed-point cameras. While these systems are effective in dealing with structured obstacles, they are significantly inadequate when facing dynamic environmental changes such as floating obstacles, rapid water level fluctuations, or abnormal water flow disturbances. On the one hand, existing systems lack the ability to continuously model real-time dynamic environments; on the other hand, the determination of obstacle "behavioral characteristics" (such as whether they drift or are carried by water flow) is relatively crude, failing to provide early quantitative identification of "potential collision risks."

[0004] The root cause of the above situation is that environmental factors at the lock are frequently changing and space is severely limited, making it impossible for single-point monitoring equipment or static rule-based judgments to cover the coupled interference effects of multiple dynamic factors. For example, when a sudden rise in water level causes some floating obstacles to drift into the lock gate before the traditional system has identified them, ships may enter along their original paths, which can easily lead to local collision risks such as scraping, squeezing, and jamming. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a collision avoidance and early warning method and system for hydraulic ship locks based on multiple lidar sensors, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solution: a hydraulic lock anti-collision early warning system based on multi-lidar, including a multi-source lidar data fusion module, an abnormal obstacle identification and classification module, a water surface disturbance field modeling module, a dynamic risk assessment and early warning module, a lock response control strategy module, and a ship passage behavior feedback optimization module;

[0007] The multi-source lidar data fusion module collects environmental data around the lock using lidar and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion.

[0008] The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles and fit them into an obstacle interference set O(t).

[0009] The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on environmental data collected by lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW;

[0010] The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc.

[0011] The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR.

[0012] The vessel traffic behavior feedback optimization module corrects the deviation between the vessel's actual and expected trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

[0013] Preferably, the multi-source lidar data fusion module includes a data acquisition and processing unit and a 3D data modeling unit;

[0014] The data acquisition unit collects environmental data around the lock using a multi-band 360° rotating lidar array deployed on both sides of the lock. This includes water level point cloud data Dwl(x,y), floating obstacle target reflection point data Dfa(x,y,t), water surface reflection disturbance echo point cloud Dsf(x,y,t), and fast disturbance points at the water flow edge Dfl(x,y,t). The data is then fitted to obtain the original dataset WD.

[0015] Among them, the water level point cloud data Dwl(x,y) represents the laser point set at location (x,y) that is identified as the boundary area between the water surface and the embankment;

[0016] The floating obstacle target reflection point data Dfa(x,y,t) represents the radar point identified as a floating obstacle at position (x,y) and time t;

[0017] The water surface reflection disturbance echo point cloud Dsf(x,y,t) represents the disturbance trajectory of the reflection point on the water surface at position (x,y) and time t per unit time.

[0018] The rapid disturbance point Dfl(x, y, t) at the edge of the water flow represents the reflection anomaly point caused by the local rapid water flow in the lock at location (x, y) and time t;

[0019] The 3D data modeling unit performs fusion processing on the collected raw dataset WD, eliminates redundant point clouds and filters drift noise, and reconstructs a 3D dynamic scene containing water surface, water flow direction, obstacle position and floating behavior, and obtains the fused unified 3D point cloud model Pf(x,y,z,t).

[0020] The unified 3D point cloud model Pf(x,y,z,t) after fusion is obtained through the following formula:

[0021] ;

[0022] In the formula, N represents the total number of radars, ωk represents the weight coefficient of the k-th radar, and WDk(x,y,z,t) represents the original dataset of the k-th radar.

[0023] The fused unified 3D point cloud model Pf(x,y,z,t) is filtered and denoised to obtain the structure dataset D, and a dynamic 3D environment model S(t) is constructed using the structure dataset D.

[0024] The dynamic three-dimensional environment model S(t) is obtained through the following formula:

[0025] ;

[0026] In the formula, minS(t) represents the minimization operation (finding the dynamic scene model S(t) that minimizes the objective function), M represents the number of data segments, S(t)i represents the i-th reconstruction point in the dynamic 3D environment model, Di(x,y,z,t) represents the i-th reconstruction point in the structure dataset, λ represents the regularization coefficient, and R(S(t)) represents the regularization term of the dynamic 3D environment model.

[0027] Preferably, the abnormal obstacle identification and classification module includes a candidate obstacle extraction unit and an obstacle type identification and interference set fitting unit;

[0028] The candidate obstacle extraction unit extracts the obstacle point set Co from the dynamic three-dimensional environment model S(t), including the gradient change of the point set in the height direction ∇zS(t), radar echo intensity Ir, height change threshold Hz, and echo intensity threshold Tir;

[0029] The gradient change of the point set in the height direction ∇zS(t) is obtained by calculating the gradient change of each point in the z direction in the local neighborhood using the dynamic three-dimensional environment model S(t).

[0030] The radar echo intensity Tr is obtained by simultaneously recording the distance and reflection intensity in the information returned by the lidar device at each scanning point;

[0031] The height change threshold Hz is determined by measuring the height fluctuations in multiple non-obstacle areas using a lidar device, and the maximum value is taken as the threshold.

[0032] The echo intensity threshold Tir is obtained by on-site sampling using lidar equipment, sampling typical water surfaces and typical floating objects separately, and setting a distinguishing boundary value.

[0033] The obstacle point set Co is obtained using the following formula:

[0034] ;

[0035] In the formula, p represents a point in the dynamic three-dimensional environment model S(t), ∇zS(t)(p) represents the gradient change of point p in the height direction, and Ir(p) represents the radar echo intensity of point p;

[0036] The obstacle type identification and interference set fitting unit performs clustering and motion trajectory analysis on the obstacle point set Co, identifies three types of obstacles, including static obstacles JD, floating obstacles JF, and potential dangerous objects JQ, and classifies and fits them into an obstacle interference set O(t).

[0037] The formula for identification and classification is:

[0038] ;

[0039] In the formula, Type(p) represents the classification result, and vp(t) represents the obstacle velocity at point p, obtained through a lidar device. Tvc represents the spatial standard deviation of the point set, Txc represents the preset velocity threshold, and Txc represents the morphological change threshold.

[0040] Preferably, the water surface disturbance field modeling module includes a water surface disturbance feature extraction unit and a disturbance vector field generation and impact assessment unit;

[0041] The water surface disturbance feature extraction unit extracts the dynamic features of local water surface disturbances from the environmental data collected by lidar, including the water surface disturbance degree SR(x,y), the velocity VZ of floating obstacles, and the obstacle position RZ;

[0042] The degree of water surface disturbance SR(x,y) is obtained by the following formula:

[0043] ;

[0044] In the formula, Zt(t-Δt)(x,y) represents the height of position (x,y) at time (t-Δt), Δt represents the time interval, and Zt(x,y) represents the height of position (x,y) at time t.

[0045] The obstacle position RZ is obtained through the floating obstacle target reflection point data Dfa(x, y, t):

[0046] The obstacle position RZ is obtained using the following formula:

[0047] ;

[0048] In the formula, Na represents the number of points, xa represents the x-coordinate of the a-th point, and ya represents the y-coordinate of the a-th point;

[0049] The velocity VZ of the floating obstacle is obtained using the following formula:

[0050] ;

[0051] In the formula, RZ(t) represents the position of the obstacle at time t, and RZ(t-Δt) represents the position of the obstacle at time t-Δt.

[0052] Preferably, the disturbance vector field generation and impact assessment unit constructs a water surface disturbance vector field model based on the obtained water surface disturbance degree SR(x,y), floating obstacle velocity VZ and obstacle position RZ, and calculates the impact on the ship's trajectory to generate a disturbance field FW;

[0053] The perturbation field FW is obtained using the following formula:

[0054] ;

[0055] In the formula, ∇SR(x,y) represents the gradient of the water surface disturbance, and Nj represents the number of floating objects. VZj represents the influence coefficient of the floating object on the disturbance field, VZj represents the velocity of the j-th floating obstacle, and RZj represents the position of the j-th obstacle.

[0056] Preferably, the dynamic risk assessment and early warning module includes a ship yaw response analysis unit and a risk value comprehensive assessment and early warning determination unit;

[0057] The ship yaw response analysis unit analyzes the yaw trend of the ship under the influence of the disturbance field FW. By comparing the ship's actual heading with the ideal track, the ship's yaw angle δθ is calculated.

[0058] The ship's yaw angle δθ is obtained using the following formula:

[0059] ;

[0060] In the formula, arccos represents the inverse cosine function, Diea represents the planned ideal course, and Dact represents the actual direction vector of the ship's motion.

[0061] Preferably, the risk value comprehensive assessment and early warning judgment unit combines the ship's yaw angle δθ, the obstacle interference set O(t) and the disturbance field FW to calculate and obtain the dynamic collision risk value Rc, and determine whether to trigger the early warning mechanism;

[0062] The dynamic collision risk value Rc is obtained using the following formula:

[0063] ;

[0064] In the formula, to represents the initial time, δθ(t) represents the ship's yaw angle at time t, FW(t) represents the disturbance field at time t, and dt represents the integral function;

[0065] The early warning mechanism obtains information through the following methods:

[0066] When 0 < dynamic collision risk value Rc < 0.3, it indicates safety and no intervention is needed;

[0067] When 0.3 ≤ dynamic collision risk value Rc ≤ 0.6, it indicates low risk, speed limit, and restricted access;

[0068] When 0.6 < dynamic collision risk value Rc < 1.0, it indicates a high risk, and entry will be stopped and an audible and visual alarm will be issued.

[0069] Preferably, the lock response control strategy module performs weighted processing and dynamic response amplification on the acquired dynamic collision risk value Rc within a time window T to obtain the intervention intensity value GR;

[0070] The intervention intensity value GR is obtained using the following formula:

[0071] ;

[0072] In the formula, Rc(t) represents the dynamic collision risk value at time t, e represents a constant, and kp represents the adjustment parameter;

[0073] Based on the obtained intervention intensity value GR, the control of the ship is classified and the control behavior of the lock system is dynamically adjusted, including the opening and closing status of the lock gates, the arrangement of the order of ship entry and exit, and the configuration of the passageway within the lock.

[0074] Ship control is classified and obtained through matching in the following ways:

[0075] When 0 < intervention intensity value GR < 0.3, it indicates the first level, normal passage, locks are fully open, and ships are released in the predetermined order;

[0076] When 0.3 ≤ intervention intensity value GR ≤ 0.6, it indicates the second level, early warning control, slowing down the pace of ship entry and restricting the passage distance;

[0077] When 0.6 < intervention intensity value GR < 1.0, it indicates the third level, which requires mandatory intervention, including suspending passage, closing the lock gates, and switching to one-way channels.

[0078] Preferably, the vessel traffic behavior feedback optimization module analyzes the intervention intensity value GR. When the intervention intensity value GR is greater than 0.6 continuously within a fixed period Tfc, the module provides feedback correction for the deviation between the actual and expected vessel traffic trajectories, and calculates the trajectory deviation value. And compare it with the preset trajectory threshold TKP to adjust the dynamic three-dimensional environment model S(t) and vector field model;

[0079] Trajectory deviation value Obtain it using the following formula:

[0080] ;

[0081] In the formula, Xac(t) represents the actual trajectory point of the ship, which is obtained through positioning system and radar trajectory tracking, and Xpr(t) represents the predicted trajectory point of the ship, which is obtained through dynamic three-dimensional environment model S(t) and vector field model prediction.

[0082] The adjustment method for the dynamic three-dimensional environment model S(t) and the vector field model is as follows: within a fixed period Tfc, the continuous trajectory deviation value If the trajectory threshold TKP is reached, the dynamic 3D environment model S(t) is updated, including adjusting the obstacle area and flow direction; adjusting the disturbance field FW and adjusting the coefficients.

[0083] When continuous trajectory deviation value >When the trajectory threshold TKP is reached, the dynamic collision risk value Rc is replanned to obtain a new dynamic collision risk value nRc;

[0084] The new dynamic collision risk value nRc is obtained using the following formula:

[0085] ;

[0086] In the formula, g represents the deviation response coefficient.

[0087] A collision avoidance and early warning method for hydraulic ship locks based on multiple lidar sensors includes the following steps:

[0088] Step 1: The multi-source lidar data fusion module collects environmental data around the lock using lidar, and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion.

[0089] Step 2: The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles, and fits them into an obstacle interference set O(t).

[0090] Step 3: The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on the environmental data collected by the lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW;

[0091] Step 4: The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc.

[0092] Step 5: The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR.

[0093] Step Six: The vessel traffic behavior feedback optimization module corrects the deviation between the actual and expected vessel traffic trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

[0094] This invention provides a collision avoidance and early warning method and system for hydraulic ship locks based on multiple lidar sensors, which has the following beneficial effects:

[0095] (1) During system operation, the multi-source lidar data fusion module constructs a dynamic three-dimensional environment model S(t) through spatiotemporal collaborative acquisition and point cloud data processing of multiple lidars. This effectively compensates for the information loss problem of traditional visual monitoring systems under illumination, occlusion and spatial blind spots, and significantly improves the system's comprehensive perception capability of environmental elements such as water level fluctuations, water surface structure and channel obstacles. The abnormal obstacle identification and classification module can distinguish between static obstacles, floating obstacles and potential high-risk targets through feature extraction and cluster analysis of the environmental model, and constructs an obstacle interference set O(t), thereby breaking the limitations of the existing system's weak perception capability and delayed response to "unstructured targets".

[0096] The surface disturbance field modeling module introduces the FW modeling method for disturbance fields, integrating water surface fluctuations and floating object movement trends. This enables accurate prediction of the direction and intensity of disturbances affecting the ship's course, providing a more forward-looking basis for collision avoidance and overcoming the shortcoming of traditional systems that fail to respond to dynamic changes in water flow. The dynamic risk assessment and early warning module integrates yaw angle δθ(t), disturbance intensity, and obstacle distribution to form a time-continuous risk assessment value Rc(t). This allows for real-time perception of risk trend changes, enabling quantitative judgment of potential collision risks and improving the timeliness and reliability of early warning responses.

[0097] (2) By deploying multi-band 360° rotating lidar arrays on both sides of the lock, the system can simultaneously collect dynamic point cloud information of multiple dimensions such as water level, floating obstacles, water surface disturbance, and abnormal water flow, forming a raw dataset WD. Compared with the traditional single sensor structure, this architecture has stronger environmental perception coverage and can effectively adapt to complex passage environments under different weather, water flow and line of sight conditions.

[0098] This system enables collaborative processing of different lidar data sources, eliminating redundancy, merging overlapping viewpoints, and generating a highly consistent, high-resolution fused point cloud model Pf(x,y,z,t). This fusion mechanism effectively solves the problems of data duplication, large stitching errors, and poor spatial consistency in traditional lidar systems under multiple viewpoints, ensuring that subsequent analysis has a higher geometric accuracy foundation.

[0099] (3) By using the surface disturbance feature extraction unit, local disturbance change features are extracted from the environmental point cloud acquired by the lidar, and the degree of surface disturbance is quantitatively calculated. This process can accurately reflect the real-time disturbance behavior caused by factors such as wind waves, water flow, and ship wake, providing a stable basis for subsequent judgment on whether the water surface has safe navigation conditions, and significantly improving the system's judgment accuracy in dynamic hydrological environments.

[0100] (4) By integrating multi-dimensional information such as ship yaw angle, disturbance field intensity and obstacle interference set, a continuous and calculable collision risk index is constructed, which can more comprehensively reflect the dynamic interaction between the ship and the environment. The system can not only identify the risk level, but also implement early warning decisions based on the risk development trend, thereby making up for the shortcomings of traditional systems in responding to dynamic interference factors in a timely manner and providing rough early warnings.

[0101] By applying time-weighted processing to the dynamic collision risk value Rc, the system outputs an intervention intensity value GR, and based on this, divides the control strategy into three levels, thereby achieving automated adjustment of the lock gate status, passage sequence, and channel configuration. This mechanism shifts the lock operation from a "fixed rhythm" to a "risk-driven" approach, effectively avoiding the risk superposition and local traffic congestion problems during dense ship passage, and improving the system's safety and stability under high-pressure operation. Attached Figure Description

[0102] Figure 1 This is a schematic diagram of the block flow of a hydraulic lock anti-collision early warning system based on multiple lidar radars according to the present invention.

[0103] Figure 2 This is a schematic diagram illustrating the steps of a collision avoidance and early warning method for hydraulic ship locks based on multiple lidar sensors according to the present invention.

[0104] Figure 3This is a schematic flowchart of the system block diagram for obtaining dynamic collision risk values ​​according to the present invention;

[0105] Figure 4 This is a line graph showing the dynamic collision risk value of the present invention. Detailed Implementation

[0106] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0107] Example 1

[0108] This invention provides a collision avoidance and early warning system for hydraulic ship locks based on multiple lidar sensors. Please refer to [link / reference]. Figures 1 to 4 It includes a multi-source lidar data fusion module, an abnormal obstacle identification and classification module, a water surface disturbance field modeling module, a dynamic risk assessment and early warning module, a lock response control strategy module, and a ship passage behavior feedback optimization module;

[0109] The multi-source lidar data fusion module collects environmental data around the lock using lidar and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion.

[0110] The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles and fit them into an obstacle interference set O(t).

[0111] The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on environmental data collected by lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW;

[0112] The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc.

[0113] The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR.

[0114] The vessel traffic behavior feedback optimization module corrects the deviation between the vessel's actual and expected trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

[0115] In this embodiment, the multi-source lidar data fusion module constructs a dynamic three-dimensional environment model S(t) through spatiotemporal collaborative acquisition and point cloud data processing of multiple lidars. This effectively compensates for the information loss problems of traditional visual monitoring systems under illumination, occlusion, and spatial blind spots, significantly improving the system's comprehensive perception capability of environmental elements such as water level fluctuations, water surface structure, and channel obstacles. The abnormal obstacle identification and classification module, through feature extraction and cluster analysis of the environmental model, can distinguish between static obstacles, floating obstacles, and potential high-risk targets, constructing an obstacle interference set O(t). This overcomes the limitations of existing systems in terms of weak perception capability and delayed response to "unstructured targets."

[0116] The surface disturbance field modeling module introduces the FW modeling method for disturbance fields, integrating water surface fluctuations and floating object movement trends. This enables accurate prediction of the direction and intensity of disturbances affecting the ship's course, providing a more forward-looking basis for collision avoidance and overcoming the shortcoming of traditional systems that fail to respond to dynamic changes in water flow. The dynamic risk assessment and early warning module integrates yaw angle δθ(t), disturbance intensity, and obstacle distribution to form a time-continuous risk assessment value Rc(t). This allows for real-time perception of risk trend changes, enabling quantitative judgment of potential collision risks and improving the timeliness and reliability of early warning responses.

[0117] The lock response control strategy module introduces a dynamic calculation mechanism based on the intervention intensity value GR. This mechanism automatically adjusts the vessel entry and exit rhythm, lock gate status, and passage mode according to the risk situation, effectively avoiding concentrated congestion and localized collisions caused by environmental changes, thus improving the safety and smoothness of lock passage. The vessel passage behavior feedback optimization module can optimize the flow of vessels based on the deviation between their actual and expected paths. The collision risk value is corrected in real time, thereby dynamically optimizing the risk model and path judgment logic, forming a closed-loop structure of perception-decision-execution-feedback, and improving the system's adaptability to the long-term operating environment of complex waters.

[0118] Example 2

[0119] This embodiment is an explanation based on Embodiment 1. Please refer to it. Figure 1 Specifically: the multi-source lidar data fusion module includes a data acquisition and processing unit and a 3D data modeling unit;

[0120] The data acquisition unit collects environmental data around the lock using a multi-band 360° rotating lidar array deployed on both sides of the lock. This includes water level point cloud data Dwl(x,y), floating obstacle target reflection point data Dfa(x,y,t), water surface reflection disturbance echo point cloud Dsf(x,y,t), and fast disturbance points at the water flow edge Dfl(x,y,t). The data is then fitted to obtain the original dataset WD.

[0121] Among them, the water level point cloud data Dwl(x,y) represents the laser point set at location (x,y) that is identified as the boundary area between the water surface and the embankment;

[0122] The floating obstacle target reflection point data Dfa(x,y,t) represents the radar point identified as a floating obstacle at position (x,y) and time t;

[0123] The water surface reflection disturbance echo point cloud Dsf(x,y,t) represents the disturbance trajectory of the reflection point on the water surface at position (x,y) and time t per unit time.

[0124] The rapid disturbance point Dfl(x, y, t) at the edge of the water flow represents the reflection anomaly point caused by the local rapid water flow in the lock at location (x, y) and time t;

[0125] The 3D data modeling unit performs fusion processing on the collected raw dataset WD, eliminates redundant point clouds and filters drift noise, and reconstructs a 3D dynamic scene containing water surface, water flow direction, obstacle position and floating behavior, and obtains the fused unified 3D point cloud model Pf(x,y,z,t).

[0126] The unified 3D point cloud model Pf(x,y,z,t) after fusion is obtained through the following formula:

[0127] ;

[0128] In the formula, N represents the total number of radars, ωk represents the weight coefficient of the k-th radar, and WDk(x,y,z,t) represents the original dataset of the k-th radar.

[0129] The fused unified 3D point cloud model Pf(x,y,z,t) is filtered and denoised to obtain the structure dataset D, and a dynamic 3D environment model S(t) is constructed using the structure dataset D.

[0130] The dynamic three-dimensional environment model S(t) is obtained through the following formula:

[0131] ;

[0132] In the formula, minS(t) represents the minimization operation, M represents the number of data segments, S(t)i represents the i-th reconstruction point in the dynamic 3D environment model, Di(x,y,z,t) represents the i-th reconstruction point in the structure dataset, λ represents the regularization coefficient, and R(S(t)) represents the regularization term of the dynamic 3D environment model.

[0133] In this embodiment, by deploying multi-band 360° rotating lidar arrays on both sides of the lock, the system can simultaneously collect dynamic point cloud information from multiple dimensions, such as water level, floating obstacles, water surface disturbances, and abnormal water flow, forming a raw dataset WD. Compared with traditional single-sensor structures, this architecture has stronger environmental perception coverage capabilities and can effectively adapt to complex passage environments under different weather, water flow, and line-of-sight conditions.

[0134] This system enables collaborative processing of different lidar data sources, eliminating redundancy, merging overlapping viewpoints, and generating a highly consistent, high-resolution fused point cloud model Pf(x,y,z,t). This fusion mechanism effectively solves the problems of data duplication, large stitching errors, and poor spatial consistency in traditional lidar systems under multiple viewpoints, ensuring that subsequent analysis has a higher geometric accuracy foundation.

[0135] The modeling process in this embodiment fully considers the dynamic drift noise and temporary reflection errors in radar data. Through point cloud redundancy removal, reflection intensity filtering, and regularization smoothing control, it ensures that the modeling results have stable and continuous geometric boundaries, improving the system's fault tolerance to sudden environmental changes. This has significant engineering value for dealing with unstable environmental factors such as heavy rain, instantaneous water flow fluctuations, or obstacle drift.

[0136] Example 3

[0137] This embodiment is an explanation based on Embodiment 2. Please refer to it. Figure 1 Specifically: the abnormal obstacle identification and classification module includes a candidate obstacle extraction unit and an obstacle type identification and interference set fitting unit;

[0138] The candidate obstacle extraction unit extracts the obstacle point set Co from the dynamic three-dimensional environment model S(t), including the gradient change of the point set in the height direction ∇zS(t), radar echo intensity Ir, height change threshold Hz, and echo intensity threshold Tir;

[0139] The gradient change of the point set in the height direction ∇zS(t) is obtained by calculating the gradient change of each point in the z direction in the local neighborhood using the dynamic three-dimensional environment model S(t).

[0140] The radar echo intensity Tr is obtained by simultaneously recording the distance and reflection intensity in the information returned by the lidar device at each scanning point;

[0141] The height change threshold Hz is determined by measuring the height fluctuations in multiple non-obstacle areas using a lidar device, and the maximum value is taken as the threshold.

[0142] The echo intensity threshold Tir is obtained by on-site sampling using lidar equipment, sampling typical water surfaces and typical floating objects separately, and setting a distinguishing boundary value.

[0143] The obstacle point set Co is obtained using the following formula:

[0144] ;

[0145] In the formula, p represents a point in the dynamic three-dimensional environment model S(t), ∇zS(t)(p) represents the gradient change of point p in the height direction, and Ir(p) represents the radar echo intensity of point p;

[0146] The obstacle type identification and interference set fitting unit performs clustering and motion trajectory analysis on the obstacle point set Co, identifies three types of obstacles, including static obstacles JD, floating obstacles JF, and potential dangerous objects JQ, and classifies and fits them into an obstacle interference set O(t).

[0147] The formula for identification and classification is:

[0148] ;

[0149] In the formula, Type(p) represents the classification result, and vp(t) represents the obstacle velocity at point p. Tvc represents the spatial standard deviation of the point set, Txc represents the preset velocity threshold, and Txc represents the morphological change threshold.

[0150] In this embodiment, the candidate obstacle extraction unit can quickly locate a set of suspected obstacles by extracting regions of abrupt height gradient changes and regions of abnormal echo intensity from a dynamic 3D environment model. Utilizing both height abrupt changes and reflection features for judgment not only improves the accuracy of obstacle identification but also reduces the risk of misjudgment caused by natural water surface disturbances or radar noise. Compared to traditional methods relying on image comparison or static masks, this approach is more flexible and environmentally adaptable.

[0151] The obstacle type identification and interference set fitting unit not only performs spatial clustering of obstacle point sets, but also introduces dynamic behavioral features such as motion trajectory and morphological compactness for multi-dimensional judgment, enabling it to distinguish between static structures, slowly floating objects, and high-risk moving targets. This behavior-based identification method breaks through the limitations of the traditional approach of "distinguishing obstacles solely based on static differences in location," providing a more refined risk modeling foundation for the system.

[0152] All identified obstacles are categorized into targets of different risk levels and organized into interference sets, effectively forming a well-structured and dynamically updated set of environmental impact objects. This provides a stable and logically connected input format for downstream risk calculation, path planning, and control intervention modules. This structured expression enhances the system's versatility and module synergy.

[0153] By monitoring the speed and diffusion characteristics of obstacles, this module can quickly identify potentially hazardous objects with high risk, such as large floating objects or irregular structural debris that are rapidly drifting with the current. The system can identify and issue warnings before these targets approach the waterway, buying critical time for collision avoidance intervention and significantly improving the system's decision-making and response speed in the face of sudden environmental changes.

[0154] Example 4

[0155] This embodiment is an explanation based on Embodiment 3. Please refer to it. Figure 1 and Figure 3 Specifically: the water surface disturbance field modeling module includes a water surface disturbance feature extraction unit and a disturbance vector field generation and impact assessment unit;

[0156] The water surface disturbance feature extraction unit extracts the dynamic features of local water surface disturbances from the environmental data collected by lidar, including the water surface disturbance degree SR(x,y), the velocity VZ of floating obstacles, and the obstacle position RZ;

[0157] The degree of water surface disturbance SR(x,y) is obtained by the following formula:

[0158] ;

[0159] In the formula, Zt(t-Δt)(x,y) represents the height of position (x,y) at time (t-Δt), Δt represents the time interval, and Zt(x,y) represents the height of position (x,y) at time t.

[0160] The obstacle position RZ is obtained through the floating obstacle target reflection point data Dfa(x, y, t):

[0161] The obstacle position RZ is obtained using the following formula:

[0162] ;

[0163] In the formula, Na represents the number of points, xa represents the x-coordinate of the a-th point, and ya represents the y-coordinate of the a-th point;

[0164] The velocity VZ of the floating obstacle is obtained using the following formula:

[0165] ;

[0166] In the formula, RZ(t) represents the position of the obstacle at time t, and RZ(t-Δt) represents the position of the obstacle at time t-Δt.

[0167] The disturbance vector field generation and impact assessment unit constructs a water surface disturbance vector field model based on the obtained water surface disturbance degree SR(x,y), floating obstacle velocity VZ and obstacle position RZ, and calculates the impact on the ship's trajectory, generating the disturbance field FW;

[0168] The perturbation field FW is obtained using the following formula:

[0169] ;

[0170] In the formula, ∇SR(x,y) represents the gradient of the water surface disturbance, and Nj represents the number of floating objects. VZj represents the influence coefficient of the floating object on the disturbance field, VZj represents the velocity of the j-th floating obstacle, and RZj represents the position of the j-th obstacle.

[0171] The dynamic risk assessment and early warning module includes a ship yaw response analysis unit and a risk value comprehensive assessment and early warning determination unit;

[0172] The ship yaw response analysis unit analyzes the yaw trend of the ship under the influence of the disturbance field FW. By comparing the ship's actual heading with the ideal track, the ship's yaw angle δθ is calculated.

[0173] The ship's yaw angle δθ is obtained using the following formula:

[0174] ;

[0175] In the formula, arccos represents the inverse cosine function, Diea represents the planned ideal course, and Dact represents the actual direction vector of the ship's motion.

[0176] In this embodiment, a surface disturbance feature extraction unit extracts local disturbance change features from the environmental point cloud acquired by lidar, and quantifies the degree of surface disturbance. This process can accurately reflect real-time disturbance behavior caused by factors such as wind, waves, currents, and ship wakes, providing a stable basis for subsequent judgment on whether the water surface meets safe navigation conditions, and significantly improving the system's judgment accuracy in dynamic hydrological environments.

[0177] This embodiment not only extracts the spatial location of obstacles but also calculates their velocity over continuous time, constructing the dynamic motion characteristics of the obstacles. Through the linkage analysis of velocity and position information, the system can grasp the drift trend, disturbance direction, and influence radius of floating obstacles in real time, providing a prerequisite for the system to actively avoid obstacles and solving the problem that traditional systems cannot dynamically judge obstacle behavior. The system integrates the degree of water surface disturbance, the velocity and distribution information of floating obstacles to generate a complete disturbance vector field, which is then mapped to the actual navigation channel area of ​​the ship, thereby assessing the potential interference and deviation of the disturbance on the ship's passage path. This mechanism breaks through the limitation of traditional systems that "only identify static obstacles and cannot consider the dynamic pressure of water flow," realizing the upgrade of the collision avoidance strategy from "point recognition" to "area influence."

[0178] This embodiment utilizes a ship yaw response analysis unit to compare the deviation angle between the actual trajectory and the ideal course, calculating the ship's yaw trend under the influence of a disturbance field. Based on real-time changes in the yaw angle, the system can determine whether the ship has entered an unstable navigation state, thereby enabling early identification of potential collision risks and early triggering of intervention mechanisms, enhancing the initiative and timeliness of the overall response.

[0179] Example 5

[0180] This embodiment is an explanation based on Embodiment 4. Please refer to it. Figure 1 and Figure 4 Specifically: The risk value comprehensive assessment and early warning judgment unit combines the ship's yaw angle δθ, the obstacle interference set O(t) and the disturbance field FW to calculate and obtain the dynamic collision risk value Rc, as shown in Table 1, and determines whether the early warning mechanism is triggered.

[0181] The dynamic collision risk value Rc is obtained using the following formula:

[0182] ;

[0183] In the formula, to represents the initial time, δθ(t) represents the ship's yaw angle at time t, FW(t) represents the disturbance field at time t, and dt represents the integral function;

[0184] The early warning mechanism obtains information through the following methods:

[0185] When 0 < dynamic collision risk value Rc < 0.3, it indicates safety and no intervention is needed;

[0186] When 0.3 ≤ dynamic collision risk value Rc ≤ 0.6, it indicates low risk, speed limit, and restricted access;

[0187] When 0.6 < dynamic collision risk value Rc < 1.0, it indicates a high risk, and entry will be stopped and an audible and visual alarm will be issued.

[0188] Specific examples:

[0189] Table 1: Dynamic Collision Risk Value Assessment Data Table

[0190]

[0191] The lock response control strategy module performs weighted processing and dynamic response amplification on the acquired dynamic collision risk value Rc within a time window T to obtain the intervention intensity value GR.

[0192] The intervention intensity value GR is obtained using the following formula:

[0193] ;

[0194] In the formula, Rc(t) represents the dynamic collision risk value at time t, e represents a constant, and kp represents the adjustment parameter;

[0195] Based on the obtained intervention intensity value GR, the control of the ship is classified and the control behavior of the lock system is dynamically adjusted, including the opening and closing status of the lock gates, the arrangement of the order of ship entry and exit, and the configuration of the passageway within the lock.

[0196] Ship control is classified and obtained through matching in the following ways:

[0197] When 0 < intervention intensity value GR < 0.3, it indicates the first level, normal passage, locks are fully open, and ships are released in the predetermined order;

[0198] When 0.3 ≤ intervention intensity value GR ≤ 0.6, it indicates the second level, early warning control, slowing down the pace of ship entry and restricting the passage distance;

[0199] When 0.6 < intervention intensity value GR < 1.0, it indicates the third level, which requires mandatory intervention, including suspending passage, closing the lock gates, and switching to one-way channels.

[0200] The vessel traffic behavior feedback optimization module analyzes the intervention intensity value GR. When the intervention intensity value GR is greater than 0.6 continuously within a fixed period Tfc, the module provides feedback correction for the deviation between the actual and expected vessel traffic trajectories, and calculates the trajectory deviation value. And compare it with the preset trajectory threshold TKP to adjust the dynamic three-dimensional environment model S(t) and vector field model;

[0201] Trajectory deviation value Obtain it using the following formula:

[0202] ;

[0203] In the formula, Xac(t) represents the actual trajectory point of the ship, and Xpr(t) represents the predicted trajectory point of the ship;

[0204] The adjustment method for the dynamic three-dimensional environment model S(t) and the vector field model is as follows: within a fixed period Tfc, the continuous trajectory deviation value If the trajectory threshold TKP is reached, the dynamic 3D environment model S(t) is updated, including adjusting the obstacle area and flow direction; adjusting the disturbance field FW and adjusting the coefficients.

[0205] When continuous trajectory deviation value >When the trajectory threshold TKP is reached, the dynamic collision risk value Rc is replanned to obtain a new dynamic collision risk value nRc;

[0206] The new dynamic collision risk value nRc is obtained using the following formula:

[0207] ;

[0208] In the formula, g represents the deviation response coefficient.

[0209] In this embodiment, by fusing multi-dimensional information such as ship yaw angle, disturbance field intensity, and obstacle interference set, a continuous and calculable collision risk index is constructed, which can more comprehensively reflect the dynamic interaction between the ship and the environment. The system can not only identify the risk level, but also implement early warning decisions based on the risk development trend, thereby making up for the shortcomings of traditional systems in responding to dynamic interference factors in a timely manner and providing rudimentary early warnings.

[0210] By applying time-weighted processing to the dynamic collision risk value Rc, the system outputs an intervention intensity value GR, and based on this, divides the control strategy into three levels, thereby achieving automated adjustment of the lock gate status, passage sequence, and channel configuration. This mechanism shifts the lock operation from a "fixed rhythm" to a "risk-driven" approach, effectively avoiding the risk superposition and local traffic congestion problems during dense ship passage, and improving the system's safety and stability under high-pressure operation.

[0211] By setting a cumulative risk intensity threshold judgment mechanism within a fixed period, the system can perceive the "persistence" rather than the "instantaneousness" of the risk state, thereby activating the trajectory feedback mechanism. This long-term trend recognition capability solves the problem that traditional collision avoidance systems rely solely on real-time data and struggle to judge "trend risks," improving the depth of perception and proactive response to potential safety hazards.

[0212] The system calculates the trajectory deviation value by continuously comparing the actual trajectory of the ship with the predicted trajectory. The system compares the trajectory with a set trajectory threshold (TKP). When the trajectory deviates from the threshold for an extended period, the system automatically updates the environmental model and the disturbance field model, adjusting the obstacle area and flow direction to improve the model's adaptability to changes in the actual water area. This mechanism represents an evolution of the collision avoidance system from "passive tracking" to "active correction," significantly enhancing the long-term accuracy of environmental modeling and the system's stability.

[0213] When the system detects continuous trajectory deviation value If the trajectory threshold TKP is exceeded and the intervention intensity remains at a high-risk level, the system will reassess and plan a new dynamic collision risk value Rc, dynamically adjusting subsequent decision-making logic. This mechanism ensures that under conditions of continuous accumulation of high risk, the system no longer uses outdated models or static rules, but instead rapidly iterates decision-making criteria, improving the agility of risk perception and the reliability of control at critical moments.

[0214] Example 6

[0215] A collision avoidance and early warning method for hydraulic ship locks based on multiple lidar sensors is described in the following reference: Figure 2 Specifically, it includes the following steps:

[0216] Step 1: The multi-source lidar data fusion module collects environmental data around the lock using lidar, and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion.

[0217] Step 2: The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles, and fits them into an obstacle interference set O(t).

[0218] Step 3: The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on the environmental data collected by the lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW;

[0219] Step 4: The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc.

[0220] Step 5: The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR.

[0221] Step Six: The vessel traffic behavior feedback optimization module corrects the deviation between the actual and expected vessel traffic trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

[0222] In this embodiment, a dynamic three-dimensional environment model was successfully constructed through synchronous acquisition and fusion processing of data from multi-source lidar. This model effectively perceives multiple key factors such as water level boundaries, water surface structure, and flow changes, solving the problems of traditional single-sensor systems, such as small spatial coverage, numerous environmental blind spots, and insufficient structural information acquisition. This provides a precise data foundation for subsequent collision avoidance strategies. By extracting multi-dimensional feature parameters such as height changes, reflection characteristics, and movement trajectories, the system can distinguish between static obstacles, floating debris, and potential high-risk targets. It organizes these into a unified obstacle interference set, effectively improving the system's recognition accuracy and response time when facing complex and dynamic obstacle scenarios. This fills the technical gap of traditional systems with single recognition dimensions and poor dynamic judgment capabilities.

[0223] This method establishes a vector disturbance field model based on the degree of disturbance and the behavior of floating bodies. It not only accurately reflects phenomena such as changes in water flow and wave disturbances, but also assesses their specific impact on the ship's direction of travel and stability. This extends "environmental perception" to "disturbance impact prediction," effectively supporting the system's early response and intelligent intervention under sudden disturbances. The method integrates multiple factors such as ship yaw, obstacle distribution, and disturbance impacts to generate a quantified collision risk value. Based on this value, it determines the risk level and whether to trigger a warning, replacing the traditional method that relies on experience thresholds or fixed rules. This provides a more scientific basis for decision-making and greater response flexibility in collision avoidance response.

[0224] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A collision avoidance and early warning system for hydraulic ship locks based on multiple lidar sensors, characterized in that: It includes a multi-source lidar data fusion module, an abnormal obstacle identification and classification module, a water surface disturbance field modeling module, a dynamic risk assessment and early warning module, a lock response control strategy module, and a ship passage behavior feedback optimization module; The multi-source lidar data fusion module collects environmental data around the lock using lidar and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion. The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles and fit them into an obstacle interference set O(t). The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on environmental data collected by lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW; The water surface disturbance field modeling module includes a water surface disturbance feature extraction unit and a disturbance vector field generation and impact assessment unit; The water surface disturbance feature extraction unit extracts the dynamic features of local water surface disturbances from the environmental data collected by lidar, including the water surface disturbance degree SR(x,y), the velocity VZ of floating obstacles, and the obstacle position RZ; The degree of water surface disturbance SR(x,y) is obtained by the following formula: ； In the formula, Zt(t-Δt)(x,y) represents the height of position (x,y) at time (t-Δt), Δt represents the time interval, and Zt(x,y) represents the height of position (x,y) at time t. The obstacle position RZ is obtained through the floating obstacle target reflection point data Dfa(x, y, t): The obstacle position RZ is obtained using the following formula: ； In the formula, Na represents the number of points, xa represents the x-coordinate of the a-th point, and ya represents the y-coordinate of the a-th point; The velocity VZ of the floating obstacle is obtained using the following formula: ； In the formula, RZ(t) represents the position of the obstacle at time t, and RZ(t-Δt) represents the position of the obstacle at time t-Δt. The disturbance vector field generation and impact assessment unit constructs a water surface disturbance vector field model based on the obtained water surface disturbance degree SR(x,y), floating obstacle velocity VZ and obstacle position RZ, and calculates the impact on the ship's trajectory, generating the disturbance field FW; The perturbation field FW is obtained using the following formula: ； In the formula, ∇SR(x,y) represents the gradient of the water surface disturbance, and Nj represents the number of floating objects. VZj represents the influence coefficient of the floating object on the disturbance field, VZj represents the velocity of the j-th floating obstacle, and RZj represents the position of the j-th obstacle. The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc. The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR. The vessel traffic behavior feedback optimization module corrects the deviation between the vessel's actual and expected trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

2. The anti-collision early warning system for hydraulic ship locks based on multiple lidar as described in claim 1, characterized in that: The multi-source lidar data fusion module includes a data acquisition and processing unit and a 3D data modeling unit; The data acquisition unit collects environmental data around the lock using a multi-band 360° rotating lidar array deployed on both sides of the lock. This includes water level point cloud data Dwl(x,y), floating obstacle target reflection point data Dfa(x,y,t), water surface reflection disturbance echo point cloud Dsf(x,y,t), and fast disturbance points at the water flow edge Dfl(x,y,t). The data is then fitted to obtain the original dataset WD. Among them, the water level point cloud data Dwl(x,y) represents the laser point set at location (x,y) that is identified as the boundary area between the water surface and the embankment; The floating obstacle target reflection point data Dfa(x,y,t) represents the radar point identified as a floating obstacle at position (x,y) and time t; The water surface reflection disturbance echo point cloud Dsf(x,y,t) represents the disturbance trajectory of the reflection point on the water surface at position (x,y) and time t per unit time. The rapid disturbance point Dfl(x, y, t) at the edge of the water flow represents the reflection anomaly point caused by the local rapid water flow in the lock at location (x, y) and time t; The 3D data modeling unit performs fusion processing on the collected raw dataset WD, eliminates redundant point clouds and filters drift noise, and reconstructs a 3D dynamic scene containing water surface, water flow direction, obstacle position and floating behavior, and obtains the fused unified 3D point cloud model Pf(x,y,z,t). The unified 3D point cloud model Pf(x,y,z,t) after fusion is obtained through the following formula: ； In the formula, N represents the total number of radars, ωk represents the weight coefficient of the k-th radar, and WDk(x,y,z,t) represents the original dataset of the k-th radar. The fused unified 3D point cloud model Pf(x,y,z,t) is filtered and denoised to obtain the structure dataset D, and a dynamic 3D environment model S(t) is constructed using the structure dataset D. The dynamic three-dimensional environment model S(t) is obtained through the following formula: ； In the formula, minS(t) represents the minimization operation, M represents the number of data segments, S(t)i represents the i-th reconstruction point in the dynamic 3D environment model, Di(x,y,z,t) represents the i-th reconstruction point in the structure dataset, λ represents the regularization coefficient, and R(S(t)) represents the regularization term of the dynamic 3D environment model.

3. A collision avoidance and early warning system for hydraulic locks based on multiple lidar as described in claim 2, characterized in that: The abnormal obstacle identification and classification module includes a candidate obstacle extraction unit and an obstacle type identification and interference set fitting unit; The candidate obstacle extraction unit extracts the obstacle point set Co from the dynamic three-dimensional environment model S(t), including the gradient change of the point set in the height direction ∇zS(t), radar echo intensity Ir, height change threshold Hz, and echo intensity threshold Tir; The gradient change of the point set in the height direction ∇zS(t) is obtained by calculating the gradient change of each point in the z direction in the local neighborhood using the dynamic three-dimensional environment model S(t). The radar echo intensity Tr is obtained by simultaneously recording the distance and reflection intensity in the information returned by the lidar device at each scanning point; The height change threshold Hz is determined by measuring the height fluctuations in multiple non-obstacle areas using a lidar device, and the maximum value is taken as the threshold. The echo intensity threshold Tir is obtained by on-site sampling using lidar equipment, sampling typical water surfaces and typical floating objects separately, and setting a distinguishing boundary value. The obstacle point set Co is obtained using the following formula: ； In the formula, p represents a point in the dynamic three-dimensional environment model S(t), ∇zS(t)(p) represents the gradient change of point p in the height direction, and Ir(p) represents the radar echo intensity of point p; The obstacle type identification and interference set fitting unit performs clustering and motion trajectory analysis on the obstacle point set Co, identifies three types of obstacles, including static obstacles JD, floating obstacles JF, and potential dangerous objects JQ, and classifies and fits them into an obstacle interference set O(t). The formula for identification and classification is: ； In the formula, Type(p) represents the classification result, and vp(t) represents the obstacle velocity at point p. Tvc represents the spatial standard deviation of the point set, Txc represents the preset velocity threshold, and Txc represents the morphological change threshold.

4. A collision avoidance and early warning system for hydraulic locks based on multiple lidar as described in claim 3, characterized in that: The dynamic risk assessment and early warning module includes a ship yaw response analysis unit and a risk value comprehensive assessment and early warning determination unit; The ship yaw response analysis unit analyzes the yaw trend of the ship under the influence of the disturbance field FW. By comparing the ship's actual heading with the ideal track, the ship's yaw angle δθ is calculated. The ship's yaw angle δθ is obtained using the following formula: ； In the formula, arccos represents the inverse cosine function, Diea represents the planned ideal course, and Dact represents the actual direction vector of the ship's motion.

5. A collision avoidance and early warning system for hydraulic locks based on multiple lidar as described in claim 4, characterized in that: The risk value comprehensive assessment and early warning judgment unit combines the ship's yaw angle δθ, obstacle interference set O(t) and disturbance field FW to calculate the dynamic collision risk value Rc and determine whether the early warning mechanism is triggered. The dynamic collision risk value Rc is obtained using the following formula: ； In the formula, to represents the initial time, δθ(t) represents the ship's yaw angle at time t, FW(t) represents the disturbance field at time t, and dt represents the integral function; The early warning mechanism obtains information through the following methods: When 0 < dynamic collision risk value Rc < 0.3, it indicates safety and no intervention is needed; When 0.3 ≤ dynamic collision risk value Rc ≤ 0.6, it indicates low risk, speed limit, and restricted access; When 0.6 < dynamic collision risk value Rc < 1.0, it indicates a high risk, and entry will be stopped and an audible and visual alarm will be issued.

6. A collision avoidance and early warning system for hydraulic ship locks based on multiple lidar as described in claim 5, characterized in that: The lock response control strategy module performs weighted processing and dynamic response amplification on the acquired dynamic collision risk value Rc within a time window T to obtain the intervention intensity value GR. The intervention intensity value GR is obtained using the following formula: ； In the formula, Rc(t) represents the dynamic collision risk value at time t, e represents a constant, and kp represents the adjustment parameter; Based on the obtained intervention intensity value GR, the control of the ship is classified and the control behavior of the lock system is dynamically adjusted, including the opening and closing status of the lock gates, the arrangement of the order of ship entry and exit, and the configuration of the passageway within the lock. Ship control is classified and obtained through matching in the following ways: When 0 < intervention intensity value GR < 0.3, it indicates the first level, normal passage, locks are fully open, and ships are released in the predetermined order; When 0.3 ≤ intervention intensity value GR ≤ 0.6, it indicates the second level, early warning control, and restriction of passage spacing; When 0.6 < intervention intensity value GR < 1.0, it indicates the third level, which requires mandatory intervention, including suspending passage, closing the lock gates, and switching to one-way channels.

7. A collision avoidance and early warning system for hydraulic locks based on multiple lidar as described in claim 6, characterized in that: The vessel traffic behavior feedback optimization module analyzes the intervention intensity value GR. When the intervention intensity value GR is greater than 0.6 continuously within a fixed period Tfc, the module provides feedback correction for the deviation between the actual and expected vessel traffic trajectories, and calculates the trajectory deviation value. And compare it with the preset trajectory threshold TKP to adjust the dynamic three-dimensional environment model S(t) and vector field model; Trajectory deviation value Obtain it using the following formula: ； In the formula, Xac(t) represents the actual trajectory point of the ship, and Xpr(t) represents the predicted trajectory point of the ship; The adjustment method for the dynamic three-dimensional environment model S(t) and the vector field model is as follows: within a fixed period Tfc, the continuous trajectory deviation value If the trajectory threshold TKP is reached, the dynamic 3D environment model S(t) is updated, including adjusting the obstacle area and flow direction; adjusting the disturbance field FW and adjusting the coefficients. When continuous trajectory deviation value >When the trajectory threshold TKP is reached, the dynamic collision risk value Rc is replanned to obtain a new dynamic collision risk value nRc; The new dynamic collision risk value nRc is obtained using the following formula: ； In the formula, g represents the deviation response coefficient.

8. A collision avoidance and early warning method for hydraulic locks based on multiple lidar, applied to the collision avoidance and early warning system for hydraulic locks based on multiple lidar as described in any one of claims 1 to 7, characterized in that: Includes the following steps: Step 1: The multi-source lidar data fusion module collects environmental data around the lock using lidar, and establishes a dynamic three-dimensional environment model S(t) through lidar point cloud analysis and sensor data fusion. Step 2: The abnormal obstacle identification and classification module uses a dynamic three-dimensional environment model S(t) to identify and classify different types of obstacles, and fits them into an obstacle interference set O(t). Step 3: The water surface disturbance field modeling module establishes a vector field model of water surface disturbance based on the environmental data collected by the lidar, calculates and obtains the disturbance flow on the water surface, and generates the disturbance field FW; Step 4: The dynamic risk assessment and early warning module assesses the interaction between the ship and the environment based on the acquired obstacle interference set O(t) and disturbance field FW, and calculates the dynamic collision risk value Rc. Step 5: The lock response control strategy module performs a comprehensive analysis of the dynamic collision risk value Rc and calculates the intervention intensity value GR. Step Six: The vessel traffic behavior feedback optimization module corrects the deviation between the actual and expected vessel traffic trajectories based on the obtained intervention intensity value GR, and obtains the trajectory deviation value. The dynamic collision risk value Rc is recalculated to obtain a new dynamic collision risk value nRc.

Images