A management method and system for a collision warning device for fishing vessels traveling at night

Abstract

The present invention relates to a management method and system for a collision warning device for fishing vessels traveling at night, belonging to the technical field of fishery traffic warning, obtaining the judgment result of the warning trigger abnormality, constraining the allowed planning area based on the judgment result of the warning trigger abnormality, finding the optimal solution for several radiation intervals in the allowed planning area based on the particle swarm optimization algorithm, obtaining the best warning area optimization plan, generating the optimal preset parameter adjustment plan based on the judgment result of the warning trigger abnormality, predicting the motion trajectory of the category information of the current obstacle through a convolutional neural network, calculating the trajectory overlap between the motion trajectory of the obstacle and the predetermined trajectory information, judging whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining an obstacle avoidance plan. The present invention can readjust the preset warning area and the preset parameters of the device according to the warning trigger delay, thereby eliminating the warning delay.

Description

Technical Field

[0001] The invention relates to the technical field of fishery traffic warning, and in particular to a management method and system of a collision warning device for fishing vessels traveling at night. Background Art

[0002] The collision warning device for fishing vessels at night is a device used for fishing, transportation and other vessels to issue alarms, reminders or warnings. It can be used to remind people to pay attention to potential collision dangers or emergencies, prevent collision accidents when ships are driving at night, reduce the incidence of danger, and improve the safety factor of fishing navigation. The visibility in the night environment is low. The collision warning devices currently installed on some fishing vessels cannot accurately determine whether the obstacles are in the preset warning area when the visibility is low, which causes the triggering of the collision warning device to be delayed. There are many situations that cause the delay of the collision warning device. Among them, the warning range and the errors in the parameter settings of the warning device itself are both important. They will cause the collision warning device to be unable to issue a warning signal in advance, which greatly increases the danger of sailing at night. At the same time, the existing collision warning device is difficult to judge the danger level of sailing in advance according to obstacles, has poor danger prediction ability, and has low collision warning quality. Therefore, a management method that can solve the warning delay when sailing at night is needed. Summary of the invention

[0003] The present invention overcomes the shortcomings of the prior art and provides a management method and system for a collision warning device for fishing vessels traveling at night.

[0004] To achieve the above object, the technical solution adopted by the present invention is:

[0005] A first aspect of the present invention provides a management method for a fishing vessel nighttime collision warning device, comprising the following steps:

[0006] Based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality;

[0007] If the result of the judgment of the early warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, and the maximum coverage range is obtained, and the optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area;

[0008] According to the planning interval boundaries of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution for the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best optimization plan for the warning area;

[0009] If the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a number of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result;

[0010] Obtain the category information of the current obstacle, predict the motion trajectory of the category information of the current obstacle through a convolutional neural network, obtain the motion trajectory of the obstacle, calculate the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information, obtain the trajectory overlap, determine whether the trajectory overlap is greater than the preset trajectory overlap, and obtain an obstacle avoidance plan.

[0011] Furthermore, in a preferred embodiment of the present invention, the method of obtaining a plurality of obstacle feature points based on the LDA algorithm, constructing a point cloud coordinate system to process the plurality of obstacle feature points, obtaining a three-dimensional distribution model of the obstacle, simulating and judging the warning trigger delay value when the three-dimensional distribution model enters a preset warning area, and obtaining a judgment result of the warning trigger abnormality specifically includes the following steps:

[0012] By using a laser radar to detect and scan surrounding obstacles when the ship is traveling at night, the surrounding obstacle information is obtained, and features of the surrounding obstacle information are extracted based on the LDA algorithm. The projection direction is determined by the preset eigenvalues ​​of the inter-class scatter matrix, and the distribution information is mapped to the projection direction to obtain a number of obstacle feature points;

[0013] Constructing a point cloud coordinate system, importing the plurality of obstacle feature points into the point cloud coordinate system, obtaining coordinate point coefficients of the plurality of obstacle feature points, integrating the coordinate point coefficients to obtain point cloud data of the obstacle, removing key elements whose actual standards are less than preset standards from the point cloud data, performing replacement compensation on the removed key elements and performing connection processing on the replaced point cloud data to obtain a three-dimensional distribution model of the obstacle;

[0014] The preset warning area and standard trigger distance of the ship collision warning device are obtained. When the three-dimensional distribution model of the obstacle is within the warning area, the warning device is triggered and the ship is defined as the center point of the warning area. The preset warning area is imported into the three-dimensional distribution model, and the Euclidean distance between the obstacle and the center point is calculated to obtain the actual trigger distance of the warning device.

[0015] Acquire the current speed of the ship, calculate the deviation between the actual trigger distance and the standard trigger distance to obtain a trigger distance deviation value, and calculate the time according to the speed and the trigger distance deviation value to obtain an early warning trigger delay value;

[0016] It is determined whether the trigger delay value is greater than a preset delay value. If so, the warning trigger delay value is marked as a first-class abnormality. If so, it is marked as a second-class abnormality to obtain a judgment result of a warning trigger abnormality.

[0017] Furthermore, in a preferred embodiment of the present invention, if the determination result based on the warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, and the maximum coverage range is obtained. The optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area, which specifically includes the following steps:

[0018] Analyze the determination result of the warning triggering abnormality, and if the determination result is a type-one abnormality, replan and adjust the sensing range of the preset warning area;

[0019] Based on the big data network, safety regulations information for ship driving at night is obtained, and at the same time, the ship volume information is obtained, and a safety standard system for ship driving at night is constructed according to the safety regulations information, and the sensing range is defined as an objective function, and the objective function is constrained based on the ship volume information until the constrained objective function meets the maximum constraint standard in the safety standard system for ship driving at night, thereby obtaining a maximum coverage range;

[0020] Construct a geometric area map, import the maximum coverage area into the geometric area map, obtain a maximum coverage range geometric area map, mark the preset warning area as an overlapping blind area, import the overlapping blind area into the maximum coverage range geometric area map, eliminate the coverage area of ​​the overlapping blind area in the maximum coverage range geometric area map, and obtain the allowed planning area.

[0021] Furthermore, in a preferred embodiment of the present invention, according to the planning interval limit of the allowed planning area, a plurality of radiation intervals are obtained, and the optimal solution of the plurality of radiation intervals is found by a particle swarm optimization algorithm to obtain the optimal boundary points of the plurality of warning areas, and the plurality of optimal boundary points are drawn based on a geometric area map to obtain the best optimization scheme of the warning area, which specifically includes the following steps:

[0022] Based on the geometric area map, the regional geometric information of the overlapping blind area is obtained, and a plurality of boundary intersection points of the preset warning area are extracted from the regional geometric information of the overlapping blind area, and each boundary intersection point is used as a radiation origin, and a plurality of radiation connection lines are emitted to the allowed planning area based on the radiation origin until each radiation connection line reaches the boundary of the allowed planning area and each radiation connection line emission end between two points intersects, and the boundary and the intersection point are used as the interval boundary, so as to obtain a plurality of radiation intervals;

[0023] Replanning the boundary intersections in the plurality of radiation intervals by using a particle swarm optimization algorithm, generating a plurality of particles in each of the radiation intervals and initializing the speed and position of the particles, updating the speed of the particles by the attraction between the global optimal position and the individual optimal position, and calculating the fitness of the particles according to the objective function to obtain the merits and demerits of the particle positions;

[0024] Based on the good and bad values ​​of the particle positions, the best position of each particle is updated and iterated in real time, and it is determined whether the iterative convergence value reaches a preset iterative convergence value. If so, the updating and iteration is stopped and the current position of each particle is determined to obtain the optimal boundary points of several warning areas;

[0025] In the geometric area map, the plurality of optimal warning area boundary points are connected and drawn to obtain the best warning area optimization plan and upload it to the early warning device control terminal.

[0026] Furthermore, in a preferred embodiment of the present invention, if the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a plurality of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment scheme is generated based on the performance evaluation result, specifically comprising the following steps:

[0027] If the result of the determination of the warning trigger abnormality is a Class II abnormality, the preset parameters of the ship collision warning device are reset; wherein the preset parameters include communication protocol parameters, network channel parameters, signal spectrum parameters, signal receiving frequency parameters, etc.;

[0028] Acquire the variation ranges of multiple actual parameters of the early warning device when the determination result is a second-class abnormality, establish a parameter space grid, import the variation ranges of the multiple actual parameters into the parameter space grid to obtain the parameter space grid of the actual parameters, and divide the parameter space grid of the actual parameters into a plurality of parameter combinations to obtain discrete parameter combinations;

[0029] Based on the big data network, evaluation indicators of different discrete parameter combinations are obtained, multiple evaluation scores are obtained by performing performance evaluation on each discrete parameter combination, and a weight value is assigned to each discrete parameter according to the evaluation indicator and the multiple evaluation scores to obtain performance weight values ​​of multiple discrete parameters;

[0030] Determine whether each of the performance weight values ​​is greater than a preset weight value, and if so, extract and integrate the performance weight values ​​to obtain a performance evaluation result;

[0031] Sort the performance weight values ​​of the performance evaluation results, construct a sorting table, import the performance evaluation results into the sorting table to obtain a performance evaluation sorting table, and obtain the parameter combination corresponding to the top-ranked weight value in the performance evaluation sorting table to obtain the optimal parameter combination, generate the optimal preset parameter adjustment plan based on the optimal parameter combination and upload it to the early warning device control terminal.

[0032] Furthermore, in a preferred embodiment of the present invention, the step of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by using a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps:

[0033] According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle;

[0034] Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model;

[0035] Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle;

[0036] Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap;

[0037] It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

[0038] A second aspect of the present invention provides a management system for a fishing vessel nighttime collision warning device, characterized in that the management system for a fishing vessel nighttime collision warning device comprises a memory and a processor, wherein the memory stores a management method program for a fishing vessel nighttime collision warning device, and when the management method program for a fishing vessel nighttime collision warning device is executed by the processor, the following steps are implemented:

[0039] Based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality;

[0040] If the result of the judgment of the early warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, and the maximum coverage range is obtained, and the optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area;

[0041] According to the planning interval boundaries of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution for the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best optimization plan for the warning area;

[0042] If the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a number of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result;

[0043] Obtain the category information of the current obstacle, predict the motion trajectory of the category information of the current obstacle through a convolutional neural network, obtain the motion trajectory of the obstacle, calculate the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information, obtain the trajectory overlap, determine whether the trajectory overlap is greater than the preset trajectory overlap, and obtain an obstacle avoidance plan.

[0044] Furthermore, in a preferred embodiment of the present invention, the step of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by using a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps:

[0045] According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle;

[0046] Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model;

[0047] Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle;

[0048] Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap;

[0049] It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

[0050] The present invention solves the technical defects existing in the background technology, and the beneficial technical effects of the present invention are:

[0051] Obtain the judgment result of the abnormal warning trigger, constrain the allowed planning area based on the judgment result of the abnormal warning trigger, find the optimal solution for several radiation intervals in the allowed planning area based on the particle swarm optimization algorithm, obtain the optimal boundary points of several warning areas, connect and draw the several optimal boundary points according to the geometric area diagram, obtain the best warning area optimization plan, generate the optimal preset parameter adjustment plan based on the judgment result of the abnormal warning trigger, predict the motion trajectory of the category information of the current obstacle through the convolutional neural network, calculate the trajectory overlap between the motion trajectory of the obstacle and the predetermined trajectory information, judge whether the trajectory overlap is greater than the preset trajectory overlap, and obtain the obstacle avoidance plan. The present invention can judge the situation that causes the delay according to the trigger delay of the collision warning device, and readjust the setting of the preset warning area and the parameters of the collision warning device itself based on this situation, thereby improving the obstacle perception ability of the collision warning device when driving at night and greatly reducing the occurrence of trigger delay. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, drawings of other embodiments can be obtained based on these drawings without paying creative work.

[0053] Figure 1 A first method flow chart of a method for managing a collision warning device for fishing vessels traveling at night is shown;

[0054] Figure 2 A second method flow chart of a method for managing a collision warning device for fishing vessels traveling at night is shown;

[0055] Figure 3 A third method flow chart of a method for managing a collision warning device for fishing vessels traveling at night is shown;

[0056] Figure 4 The present invention shows a system framework diagram of a management system of a collision warning device for fishing vessels traveling at night. DETAILED DESCRIPTION

[0057] In order to more clearly understand the above-mentioned purpose, features and advantages of the present invention, the present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0058] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present invention is not limited to the specific embodiments disclosed below.

[0059] The first aspect of the present invention provides a management method for a collision warning device for a fishing vessel traveling at night, such as Figure 1 As shown, the following steps are included:

[0060] S102: acquiring a number of obstacle feature points based on the LDA algorithm, constructing a point cloud coordinate system to process the number of obstacle feature points, obtaining a three-dimensional distribution model of the obstacle, simulating and determining a warning trigger delay value when the three-dimensional distribution model enters a preset warning area, and obtaining a determination result of a warning trigger abnormality;

[0061] S104: based on the determination result of the early warning triggering abnormality being a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, a maximum coverage range is obtained, and an optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain an allowed planning area;

[0062] S106: According to the planning interval boundaries of the allowed planning area, a plurality of radiation intervals are obtained, and the optimal solution for the plurality of radiation intervals is found by a particle swarm optimization algorithm to obtain the optimal boundary points of the plurality of warning areas, and the plurality of optimal boundary points are drawn based on a geometric area graph to obtain the best optimization solution for the warning area;

[0063] S108: If the determination result is a second-class abnormality, reset the preset parameters of the early warning device, establish a parameter space grid, divide the parameter space grid of the actual parameters into a plurality of parameter combinations and evaluate them to obtain a performance evaluation result, and generate an optimal preset parameter adjustment plan based on the performance evaluation result;

[0064] S110: Obtain category information of the current obstacle, predict the motion trajectory of the category information of the current obstacle through a convolutional neural network to obtain the motion trajectory of the obstacle, calculate the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determine whether the trajectory overlap is greater than the preset trajectory overlap, and obtain an obstacle avoidance plan.

[0065] Furthermore, in a preferred embodiment of the present invention, the LDA algorithm is used to obtain a plurality of obstacle feature points, a point cloud coordinate system is constructed to process the plurality of obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality, such as Figure 2 As shown, the specific steps include:

[0066] S202: Detect and scan surrounding obstacles by laser radar when the ship is traveling at night, obtain surrounding obstacle information, extract features of the surrounding obstacle information based on the LDA algorithm, determine the projection direction by preset eigenvalues ​​of the inter-class scatter matrix, and map the distribution information to the projection direction to obtain a number of obstacle feature points;

[0067] S204: constructing a point cloud coordinate system, importing the plurality of obstacle feature points into the point cloud coordinate system, obtaining coordinate point coefficients of the plurality of obstacle feature points, integrating the coordinate point coefficients to obtain point cloud data of the obstacle, removing key elements whose actual standards are less than preset standards from the point cloud data, performing replacement compensation on the removed key elements and performing connection processing on the replaced point cloud data to obtain a three-dimensional distribution model of the obstacle;

[0068] S206: obtaining a preset warning area and a standard trigger distance of a ship collision warning device; when the three-dimensional distribution model of the obstacle is within the warning area, the warning device is triggered and defines the ship as the center point of the warning area; the preset warning area is imported into the three-dimensional distribution model; the Euclidean distance between the obstacle and the center point is calculated; and the actual trigger distance of the warning device is obtained;

[0069] S208: acquiring the current speed of the ship, calculating the deviation between the actual trigger distance and the standard trigger distance to obtain a trigger distance deviation value, and calculating the time according to the speed and the trigger distance deviation value to obtain a warning trigger delay value;

[0070] S210: Determine whether the trigger delay value is greater than a preset delay value. If so, mark the warning trigger delay value as a first-class abnormality. If less than, mark it as a second-class abnormality to obtain a judgment result of a warning trigger abnormality.

[0071] It should be noted that the surrounding obstacle information includes the shape information, motion information, distance information, etc. of the obstacle. The obstacle information is extracted by the LDA algorithm. The LDA algorithm is a linear judgment analysis algorithm. The original data is mapped to a low-dimensional space through linear transformation, so that samples of different categories are separated in the new space to obtain specific features. A point cloud coordinate system is constructed to assign coordinate points to the extracted feature points, and the point cloud data of the obstacle is obtained according to the coordinate points. The point cloud data is further cleaned and the elements are replaced to obtain a three-dimensional distribution model of the obstacle. The three-dimensional distribution model of the obstacle is a continuous dynamic model. The model can reflect the changes in the surrounding environment of the ship when it is traveling at night in real time. The preset warning area of ​​the ship collision warning device is the maximum warning range for triggering the warning. The standard trigger distance is the distance between the hull and the preset warning area. The three-dimensional distribution model of the obstacle is used to simulate the perception of obstacles in the preset warning area. The delay value under the actual trigger distance is calculated and judged whether it is greater than the delay value under the standard trigger distance. If it is greater, it means that the warning trigger time of the warning device is longer, which is not conducive to danger perception. The warning trigger delay value is marked as a first-class abnormality. If it is less than, it is marked as a second-class abnormality to obtain the judgment result of the warning trigger abnormality. This method simulates and judges the trigger delay value of the current warning device by constructing a three-dimensional distribution model of the surrounding obstacles at night, and obtains the abnormal situation judgment result of the warning device according to this trigger delay value, accurately and efficiently detects the problems existing in the current collision warning device, provides a basis for solutions, and effectively improves the night warning performance of the warning device.

[0072] Furthermore, in a preferred embodiment of the present invention, if the determination result based on the abnormality of the early warning trigger is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, and the maximum coverage range is obtained. The optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area, which specifically includes the following steps:

[0073] Analyze the determination result of the warning triggering abnormality, and if the determination result is a type-one abnormality, replan and adjust the sensing range of the preset warning area;

[0074] Based on the big data network, safety regulations information for ship driving at night is obtained, and at the same time, the ship volume information is obtained, and a safety standard system for ship driving at night is constructed according to the safety regulations information, and the sensing range is defined as an objective function, and the objective function is constrained based on the ship volume information until the constrained objective function meets the maximum constraint standard in the safety standard system for ship driving at night, thereby obtaining a maximum coverage range;

[0075] Construct a geometric area map, import the maximum coverage area into the geometric area map, obtain a maximum coverage range geometric area map, mark the preset warning area as an overlapping blind area, import the overlapping blind area into the maximum coverage range geometric area map, eliminate the coverage area of ​​the overlapping blind area in the maximum coverage range geometric area map, and obtain the allowed planning area.

[0076] It should be noted that since there are two types of abnormal situations in the judgment results of the warning trigger abnormality, it is necessary to conduct in-depth analysis of each situation and formulate different solutions. When the judgment result shows a type of abnormality, it means that the warning trigger delay of the warning device is large. The preset warning area can be re-planned to expand the warning sensing range, thereby shortening the delay time and improving the warning perception ability. However, when re-planning the sensing range of the preset warning area, the coverage range cannot be infinitely expanded. Therefore, it is necessary to impose maximum restrictions on the area according to the safety regulations for ships traveling at night obtained from the big data network. The constraints on the coverage range are closely related to the hull volume. The sensing range can be defined as a constrained objective function. The maximum coverage range can be accurately determined by constraining the objective function. At the same time, a safety standard system for ships traveling at night is constructed according to the safety regulations for ships traveling at night. Finally, the objective function of the hull volume constraint meets the maximum constraint standard in the safety standard system for ships traveling at night. The constrained maximum coverage range may be an irregular geometric figure. A maximum coverage range geometric area map is constructed and the original preset warning range is eliminated in the map to finally obtain the allowed planning range. This method can re-formulate a new allowable planning range for the preset warning area according to a type of abnormal results, provide a selection area for the re-planning of the preset warning area, and improve the accuracy of the warning area planning.

[0077] Furthermore, in a preferred embodiment of the present invention, the planning interval boundaries of the allowed planning area are obtained to obtain a plurality of radiation intervals, and the optimal solution of the plurality of radiation intervals is found by a particle swarm optimization algorithm to obtain the optimal boundary points of a plurality of warning areas, and the plurality of optimal boundary points are drawn based on a geometric area graph to obtain the best optimization solution for the warning area, such as Figure 3 As shown, the specific steps include:

[0078] S302: Based on the geometric area map, obtain the regional geometric information of the overlapping blind area, extract a number of boundary intersections of the preset warning area from the regional geometric information of the overlapping blind area, take each boundary intersection as a radiation origin, and emit a number of radiation connection lines to the allowed planning area based on the radiation origin until each radiation connection line reaches the boundary of the allowed planning area and each radiation connection line transmitting end between two points intersects, and take the boundary and the intersection as the interval boundary, so as to obtain a number of radiation intervals;

[0079] S304: replanning the boundary intersections in the plurality of radiation intervals by using a particle swarm optimization algorithm, generating a plurality of particles in each of the radiation intervals and initializing the speed and position of the particles, updating the speed of the particles by the attraction between the global optimal position and the individual optimal position, and calculating the fitness of the particles according to the objective function to obtain the merit value of the particle position;

[0080] S306: based on the quality value of the particle position, the optimal position of each particle is updated and iterated in real time, and it is determined whether the iterative convergence value reaches a preset iterative convergence value. If so, the updating and iteration is stopped and the current position of each particle is determined to obtain the optimal boundary points of several warning areas.

[0081] S308: Connect and draw the plurality of optimal warning area boundary points in the geometric area map to obtain the best warning area optimization plan and upload it to the early warning device control terminal.

[0082] It should be noted that the preset warning area may be an irregular set of polygons, so there are multiple boundary intersections in the geometric area map. The purpose of re-planning the warning area can be achieved by arranging and selecting points at these boundary intersections. Therefore, it is necessary to specify the area of ​​optional points for each boundary intersection, and regard each boundary intersection as the emission origin of the radiation connection line. Based on the emission origin and the allowed planning range as the emission space, densely connected radiation lines are emitted to the surrounding area. Since there will be contact when two adjacent points emit radiation connection lines, the emission action is stopped when the connection line touches the boundary of the allowed planning range and the connection line between adjacent points intersects, and the outer boundary and the intersection point are used as the boundary to divide it into several radiation intervals. Based on the particle swarm optimization algorithm, points are re-selected in each radiation interval. The particle swarm optimization algorithm is an algorithm that gradually searches for the optimal solution by continuously adjusting the position and speed of particles, which can greatly improve the accuracy of the optimal solution of the intersection. Finally, the new optimal boundary points found are connected and drawn in the geometric area map to obtain the best warning area optimization plan. This method can reselect points within the specified radiation range through the particle swarm optimization algorithm to form the best warning area, effectively solve the problem of early warning trigger delay caused by a type of abnormality, ensure the early perception capability of the early warning device, improve the safety of ships driving at night, and reduce the occurrence rate of collision accidents. The method is reliable.

[0083] Further, in a preferred embodiment of the present invention, if the determination result is a second-class abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into several parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result, which specifically includes the following steps:

[0084] If the result of the determination of the warning trigger abnormality is a Class II abnormality, the preset parameters of the ship collision warning device are reset; wherein the preset parameters include communication protocol parameters, network channel parameters, signal spectrum parameters, signal receiving frequency parameters, etc.;

[0085] Acquire the variation ranges of multiple actual parameters of the early warning device when the determination result is a second-class abnormality, establish a parameter space grid, import the variation ranges of the multiple actual parameters into the parameter space grid to obtain the parameter space grid of the actual parameters, and divide the parameter space grid of the actual parameters into a plurality of parameter combinations to obtain discrete parameter combinations;

[0086] Based on the big data network, evaluation indicators of different discrete parameter combinations are obtained, multiple evaluation scores are obtained by performing performance evaluation on each discrete parameter combination, and a weight value is assigned to each discrete parameter according to the evaluation indicator and the multiple evaluation scores to obtain performance weight values ​​of multiple discrete parameters;

[0087] Determine whether each of the performance weight values ​​is greater than a preset weight value, and if so, extract and integrate the performance weight values ​​to obtain a performance evaluation result;

[0088] Sort the performance weight values ​​of the performance evaluation results, construct a sorting table, import the performance evaluation results into the sorting table to obtain a performance evaluation sorting table, and obtain the parameter combination corresponding to the top-ranked weight value in the performance evaluation sorting table to obtain the optimal parameter combination, generate the optimal preset parameter adjustment plan based on the optimal parameter combination and upload it to the early warning device control terminal.

[0089] It should be noted that when the judgment result shows a second type of abnormal situation, it means that the warning trigger delay of the collision warning device is not serious, and the warning impact on night driving is relatively light. The delay correction can be achieved by adjusting the parameters of the collision warning device itself, thereby restoring or improving the warning perception sensitivity of the collision warning device. The actual parameters include the actual communication protocol parameters of the current ship, the actual network channel parameters, and the actual signal receiving frequency parameters. By combining the variation range of the actual parameters in the parameter space grid, multiple discrete parameter combinations are obtained, and the weight value of each discrete parameter combination is evaluated for performance, and each performance weight value is judged to be greater than the preset weight value. If greater than, the performance evaluation result is integrated to obtain the performance evaluation result, so that the discrete parameter combination with the best performance can be selected as the best adjustment scheme for the preset parameters. This method combines the variation range of the parameters in the parameter space grid, selects the parameter combination with the best performance to output and adjust the preset parameters, and the adjustment optimization of the preset parameters can achieve the best performance through performance evaluation screening, so that the adjusted parameters eliminate the second type of abnormal situation in the warning area and improve the triggering accuracy of the collision warning device.

[0090] Furthermore, in a preferred embodiment of the present invention, the method of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by a convolutional neural network, obtaining the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information, obtaining the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps:

[0091] According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle;

[0092] Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model;

[0093] Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle;

[0094] Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap;

[0095] It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

[0096] It should be noted that, in order to test the collision perception prediction capability of the device after adjusting the early warning device based on the warning area optimization scheme and the preset parameter adjustment scheme, it is necessary to first obtain the type of obstacle according to the three-dimensional distribution model of the obstacle. The movement trajectories of different types of obstacles are different at night, and the movement trajectories of obstacles under different meteorological environmental conditions are also different. Therefore, the movement trajectories of the obstacles under different combinations of meteorological environmental conditions can be obtained in the big data network, and the movement trajectories of the obstacles under different combinations of meteorological environmental conditions can be imported into the prediction model based on the convolutional neural network to obtain the movement trajectory prediction model. The current meteorological environmental information includes Temperature, wind direction, visibility and other information are imported into the motion trajectory prediction model according to the current meteorological environment information of the ship traveling at night, and the motion trajectory of the obstacle can be predicted in the end. Since the ship should avoid obstacles when traveling at night, it is necessary to determine whether the motion trajectory of the obstacle is highly overlapped with the preset route trajectory of the ship. The trajectory overlap can be generated by calculating the Pearson correlation coefficient between the two, and it is determined whether the trajectory overlap is greater than the preset trajectory overlap. If it is greater, the adjusted early warning device sends out an early warning signal in advance and generates an obstacle avoidance plan. According to the obstacle avoidance plan, the waterway is replanned or navigation is stopped to avoid the obstacle. This method determines whether the ship collides with the obstacle by predicting the direction of the current obstacle's motion trajectory, predicts the collision risk in advance and makes a corresponding avoidance plan, improves the safety and danger perception ability of the ship's night driving, ensures the life and health safety of fishermen, and improves the early warning quality and efficiency of the collision warning device.

[0097] In addition, the management method of the fishing vessel nighttime collision warning device also includes the following steps:

[0098] Acquire initial parameters of the ship design drawing, import the initial parameters of the ship design drawing and the three-dimensional distribution model of the obstacle into the simulation software for simulation modeling and simulation based on the motion trajectory of the obstacle, and obtain multiple sets of collision test data;

[0099] Extracting the maximum test parameter from each group of collision test data, determining the maximum test parameter as a search keyword, importing the search keyword into a big data network for retrieval, and obtaining a corresponding collision type;

[0100] Obtaining the risk factor corresponding to each collision type, and determining whether the risk factor corresponding to each collision type is greater than a preset risk factor, if so, marking the risk factor as high risk, and if so, marking the risk factor as low risk;

[0101] Constructing a hazard level classification diagram of collision types, obtaining coefficient standards for each classification item in the hazard level classification diagram, importing the hazard coefficients of high risk and low risk into the hazard level line diagram and classifying them according to the coefficient standards to obtain a hazard level classification diagram;

[0102] The hazard level classification diagram is uploaded to the control terminal of the early warning device, and the early warning device performs intelligent linkage control on the equipment in the ship based on the hazard level classification diagram.

[0103] It should be noted that the risk factor is an indicator for assessing the degree of risk. Different obstacles correspond to different types of collision accidents, and different types of collision accidents cause different degrees of damage to the hull, fishermen and cargo. If the early warning device cannot assess the risk level according to the type of collision that may be caused by the obstacle, and link other equipment on the hull to take corresponding protective actions, it will cause a serious collision accident on the ship, and even threaten the personal safety of fishermen, and the safety factor will drop significantly. This method can simulate the type of collision accident of the current obstacle to the ship and obtain the corresponding risk factor, and formulate a risk level classification diagram based on the risk factor to assess the risk of the collision accident, further improve the risk prevention ability of the collision warning device, and thus link and control the internal equipment of the ship to take protective measures, so as to achieve the effect of quickly preventing dangerous accidents, with high reliability.

[0104] A second aspect of the present invention provides a management system for a fishing vessel nighttime collision warning device, characterized in that the management system for a fishing vessel nighttime collision warning device comprises a memory 41 and a processor 42, wherein the memory 41 stores a management method program for a fishing vessel nighttime collision warning device, and when the management method program for a fishing vessel nighttime collision warning device is executed by the processor 42, the following steps are implemented:

[0105] Based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality;

[0106] If the result of the judgment of the early warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum standard, and the maximum coverage range is obtained, and the optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area;

[0107] According to the planning interval boundaries of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution for the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best optimization plan for the warning area;

[0108] If the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a number of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result;

[0109] Obtain the category information of the current obstacle, predict the motion trajectory of the category information of the current obstacle through a convolutional neural network, obtain the motion trajectory of the obstacle, calculate the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information, obtain the trajectory overlap, determine whether the trajectory overlap is greater than the preset trajectory overlap, and obtain an obstacle avoidance plan.

[0110] Furthermore, in a preferred embodiment of the present invention, the step of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by using a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps:

[0111] According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle;

[0112] Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model;

[0113] Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle;

[0114] Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap;

[0115] It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

[0116] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art who is familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A management method for a collision warning device for a fishing vessel at night, characterized in that: The following steps are involved: Based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality; If the result of the early warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum constraint standard to obtain the maximum coverage range, and the optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area; According to the planning interval boundaries of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution for the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best optimization plan for the warning area; If the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a number of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result; Obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle through a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining an obstacle avoidance plan; Among them, based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the several obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and the warning trigger delay value when the three-dimensional distribution model enters the preset warning area is simulated and judged to obtain the judgment result of the warning trigger abnormality, which specifically includes the following steps: By using a laser radar to detect and scan surrounding obstacles when the ship is traveling at night, the surrounding obstacle information is obtained, and features of the surrounding obstacle information are extracted based on the LDA algorithm. The projection direction is determined by the preset eigenvalues ​​of the inter-class scatter matrix, and the obstacle information is mapped to the projection direction to obtain a number of obstacle feature points; Constructing a point cloud coordinate system, importing the plurality of obstacle feature points into the point cloud coordinate system, obtaining coordinate point coefficients of the plurality of obstacle feature points, integrating the coordinate point coefficients to obtain point cloud data of the obstacle, removing key elements whose actual standards are less than preset standards from the point cloud data, performing replacement compensation on the removed key elements and performing connection processing on the replaced point cloud data to obtain a three-dimensional distribution model of the obstacle; The preset warning area and standard trigger distance of the ship collision warning device are obtained. When the three-dimensional distribution model of the obstacle is within the preset warning area, the warning device is triggered and the ship is defined as the center point of the preset warning area. The preset warning area is imported into the three-dimensional distribution model, and the Euclidean distance between the obstacle and the center point is calculated to obtain the actual trigger distance of the warning device. Acquire the current speed of the ship, calculate the deviation between the actual trigger distance and the standard trigger distance to obtain a trigger distance deviation value, and calculate the time according to the speed and the trigger distance deviation value to obtain an early warning trigger delay value; Determine whether the trigger delay value is greater than a preset delay value, if so, mark the warning trigger delay value as a first-class abnormality, if less than, mark it as a second-class abnormality, and obtain a determination result of the warning trigger abnormality; Wherein, based on the determination result of the abnormality triggered by the early warning being a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum constraint standard, and the maximum coverage range is obtained. The optimizable area within the maximum coverage range is determined by constructing a geometric area map, and the allowed planning area is obtained, which specifically includes the following steps: Analyze the determination result of the warning triggering abnormality, and if the determination result is a type-one abnormality, replan and adjust the sensing range of the preset warning area; Based on the big data network, safety regulations information for ship driving at night is obtained, and at the same time, the ship volume information is obtained, and a safety standard system for ship driving at night is constructed according to the safety regulations information, and the sensing range is defined as an objective function, and the objective function is constrained based on the ship volume information until the constrained objective function meets the maximum constraint standard in the safety standard system for ship driving at night, thereby obtaining a maximum coverage range; Constructing a geometric area map, importing the maximum coverage range into the geometric area map to obtain a maximum coverage range geometric area map, marking the preset warning area as an overlapping blind area, importing the overlapping blind area into the maximum coverage range geometric area map, eliminating the coverage area of ​​the overlapping blind area in the maximum coverage range geometric area map, and obtaining an allowed planning area; Among them, according to the planning interval limit of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution of the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best warning area optimization plan, which specifically includes the following steps: Based on the geometric area map, regional geometric information of the overlapping blind area is obtained, and a plurality of boundary intersection points of the preset warning area are extracted from the regional geometric information of the overlapping blind area, and each boundary intersection point is used as a radiation origin, and a plurality of radiation connection lines are emitted to the allowed planning area based on the radiation origin until each radiation connection line reaches the boundary of the allowed planning area and each radiation connection line emission end between two points intersects, and the boundary and the intersection point are used as the interval boundary, so as to obtain a plurality of radiation intervals; Replanning the boundary intersections in the plurality of radiation intervals by using a particle swarm optimization algorithm, generating a plurality of particles in each of the radiation intervals and initializing the speed and position of the particles, updating the speed of the particles by the attraction between the global optimal position and the individual optimal position, and calculating the fitness of the particles according to the objective function to obtain the merits and demerits of the particle positions; Based on the good and bad values ​​of the particle positions, the best position of each particle is updated and iterated in real time, and it is determined whether the iterative convergence value reaches a preset iterative convergence value. If so, the updating and iteration is stopped and the current position of each particle is determined to obtain the optimal boundary points of several warning areas; In the geometric area map, the plurality of optimal warning area boundary points are connected and drawn to obtain the best warning area optimization plan and upload it to the early warning device control terminal.

2. A management method for a fishing vessel nighttime collision warning device according to claim 1, characterized in that: If the determination result is a second-class abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a plurality of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment scheme is generated based on the performance evaluation result, specifically including the following steps: If the determination result of the warning triggering abnormality is a Class II abnormality, the preset parameters of the ship collision warning device are reset; wherein the preset parameters include communication protocol parameters, network channel parameters, signal spectrum parameters and signal receiving frequency parameters; Acquire the variation ranges of multiple actual parameters of the early warning device when the determination result is a second-class abnormality, establish a parameter space grid, import the variation ranges of the multiple actual parameters into the parameter space grid to obtain the parameter space grid of the actual parameters, and divide the parameter space grid of the actual parameters into a plurality of parameter combinations to obtain discrete parameter combinations; Based on the big data network, evaluation indicators of different discrete parameter combinations are obtained, multiple evaluation scores are obtained by performing performance evaluation on each discrete parameter combination, and a weight value is assigned to each discrete parameter combination according to the evaluation indicator and the multiple evaluation scores to obtain performance weight values ​​of multiple discrete parameter combinations; Determine whether each of the performance weight values ​​is greater than a preset weight value, and if so, extract and integrate the performance weight values ​​to obtain a performance evaluation result; Sort the performance weight values ​​of the performance evaluation results, construct a sorting table, import the performance evaluation results into the sorting table to obtain a performance evaluation sorting table, and obtain the parameter combination corresponding to the top-ranked weight value in the performance evaluation sorting table to obtain the optimal parameter combination, generate the optimal preset parameter adjustment plan based on the optimal parameter combination and upload it to the early warning device control terminal.

3. The management method of a fishing vessel nighttime collision warning device according to claim 1, characterized in that: The method of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by using a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps: According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle; Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model; Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle; Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap; It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

4. A management system for a collision warning device for fishing vessels at night, characterized in that: The management system of the nighttime collision warning device for fishing vessels comprises a memory and a processor, wherein the memory stores a management method program of the nighttime collision warning device for fishing vessels, and when the management method program of the nighttime collision warning device for fishing vessels is executed by the processor, the following steps are implemented: Based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and a warning trigger delay value when the three-dimensional distribution model enters a preset warning area is simulated and determined to obtain a judgment result of a warning trigger abnormality; If the result of the early warning triggering abnormality is a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum constraint standard to obtain the maximum coverage range, and the optimizable area within the maximum coverage range is determined by constructing a geometric area map to obtain the allowed planning area; According to the planning interval boundaries of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution for the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best optimization plan for the warning area; If the determination result is a Class II abnormality, the preset parameters of the early warning device are reset, a parameter space grid is established, and the parameter space grid of the actual parameters is divided into a number of parameter combinations and evaluated to obtain a performance evaluation result, and an optimal preset parameter adjustment plan is generated based on the performance evaluation result; Obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle through a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining an obstacle avoidance plan; Among them, based on the LDA algorithm, a number of obstacle feature points are obtained, a point cloud coordinate system is constructed to process the several obstacle feature points, a three-dimensional distribution model of the obstacle is obtained, and the warning trigger delay value when the three-dimensional distribution model enters the preset warning area is simulated and judged to obtain the judgment result of the warning trigger abnormality, which specifically includes the following steps: By using a laser radar to detect and scan surrounding obstacles when the ship is traveling at night, the surrounding obstacle information is obtained, and features of the surrounding obstacle information are extracted based on the LDA algorithm. The projection direction is determined by the preset eigenvalues ​​of the inter-class scatter matrix, and the obstacle information is mapped to the projection direction to obtain a number of obstacle feature points; Constructing a point cloud coordinate system, importing the plurality of obstacle feature points into the point cloud coordinate system, obtaining coordinate point coefficients of the plurality of obstacle feature points, integrating the coordinate point coefficients to obtain point cloud data of the obstacle, removing key elements whose actual standards are less than preset standards from the point cloud data, performing replacement compensation on the removed key elements and performing connection processing on the replaced point cloud data to obtain a three-dimensional distribution model of the obstacle; The preset warning area and standard trigger distance of the ship collision warning device are obtained. When the three-dimensional distribution model of the obstacle is within the preset warning area, the warning device is triggered and the ship is defined as the center point of the preset warning area. The preset warning area is imported into the three-dimensional distribution model, and the Euclidean distance between the obstacle and the center point is calculated to obtain the actual trigger distance of the warning device. Acquire the current speed of the ship, calculate the deviation between the actual trigger distance and the standard trigger distance to obtain a trigger distance deviation value, and calculate the time according to the speed and the trigger distance deviation value to obtain an early warning trigger delay value; Determine whether the trigger delay value is greater than a preset delay value, if so, mark the warning trigger delay value as a first-class abnormality, if less than, mark it as a second-class abnormality, and obtain a determination result of the warning trigger abnormality; Wherein, based on the determination result of the abnormality triggered by the early warning being a type of abnormality, the sensing range of the preset warning area is replanned, the sensing range is constrained until it meets the maximum constraint standard, and the maximum coverage range is obtained. The optimizable area within the maximum coverage range is determined by constructing a geometric area map, and the allowed planning area is obtained, which specifically includes the following steps: Analyze the determination result of the warning triggering abnormality, and if the determination result is a type-one abnormality, replan and adjust the sensing range of the preset warning area; Based on the big data network, safety regulations information for ship driving at night is obtained, and at the same time, the ship volume information is obtained, and a safety standard system for ship driving at night is constructed according to the safety regulations information, and the sensing range is defined as an objective function, and the objective function is constrained based on the ship volume information until the constrained objective function meets the maximum constraint standard in the safety standard system for ship driving at night, thereby obtaining a maximum coverage range; Constructing a geometric area map, importing the maximum coverage range into the geometric area map to obtain a maximum coverage range geometric area map, marking the preset warning area as an overlapping blind area, importing the overlapping blind area into the maximum coverage range geometric area map, eliminating the coverage area of ​​the overlapping blind area in the maximum coverage range geometric area map, and obtaining an allowed planning area; Among them, according to the planning interval limit of the allowed planning area, a number of radiation intervals are obtained, and the optimal solution of the several radiation intervals is found through the particle swarm optimization algorithm to obtain the optimal boundary points of several warning areas, and the several optimal boundary points are drawn based on the geometric area map to obtain the best warning area optimization plan, which specifically includes the following steps: Based on the geometric area map, regional geometric information of the overlapping blind area is obtained, and a plurality of boundary intersection points of the preset warning area are extracted from the regional geometric information of the overlapping blind area, and each boundary intersection point is used as a radiation origin, and a plurality of radiation connection lines are emitted to the allowed planning area based on the radiation origin until each radiation connection line reaches the boundary of the allowed planning area and each radiation connection line emission end between two points intersects, and the boundary and the intersection point are used as the interval boundary, so as to obtain a plurality of radiation intervals; Replanning the boundary intersections in the plurality of radiation intervals by using a particle swarm optimization algorithm, generating a plurality of particles in each of the radiation intervals and initializing the speed and position of the particles, updating the speed of the particles by the attraction between the global optimal position and the individual optimal position, and calculating the fitness of the particles according to the objective function to obtain the merits and demerits of the particle positions; Based on the good and bad values ​​of the particle positions, the best position of each particle is updated and iterated in real time, and it is determined whether the iterative convergence value reaches a preset iterative convergence value. If so, the updating and iteration is stopped and the current position of each particle is determined to obtain the optimal boundary points of several warning areas; In the geometric area map, the plurality of optimal warning area boundary points are connected and drawn to obtain the best warning area optimization plan and upload it to the early warning device control terminal.

5. A management system for a fishing vessel nighttime collision warning device according to claim 4, characterized in that: The method of obtaining the category information of the current obstacle, predicting the motion trajectory of the category information of the current obstacle by using a convolutional neural network to obtain the motion trajectory of the obstacle, calculating the Pearson correlation coefficient between the motion trajectory of the obstacle and the predetermined trajectory information to obtain the trajectory overlap, determining whether the trajectory overlap is greater than the preset trajectory overlap, and obtaining the obstacle avoidance plan specifically includes the following steps: According to the optimal warning area optimization scheme and the optimal preset parameter adjustment scheme, the early warning device is adjusted and optimized to obtain an adjusted early warning device, a fishery knowledge graph is constructed, and the three-dimensional distribution model of the obstacle is imported into the fishery knowledge graph for identification to obtain category information of the current obstacle; Based on the big data network, the motion trajectory of the obstacle under different combinations of meteorological and environmental conditions is obtained, a motion trajectory prediction model is constructed through a convolutional neural network, and the motion rate of the obstacle under the different combinations of meteorological and environmental conditions is introduced into the motion trajectory prediction model to obtain a trained motion trajectory prediction model; Acquire current meteorological environment information of the ship traveling at night, import the current meteorological environment information into the trained motion trajectory prediction model for prediction, and obtain the motion trajectory of the obstacle; Acquire the predetermined trajectory information of the ship, set the movement trajectory of the obstacle and the predetermined trajectory information as linear variables, calculate the Pearson correlation coefficient of the linear variable, and determine the trend of the Pearson correlation coefficient in the value range between -1 and 1 to obtain the trajectory overlap; It is determined whether the trajectory overlap is greater than a preset trajectory overlap. If so, the adjusted early warning device sends out an early warning signal, generates an obstacle avoidance plan based on the trajectory overlap, and replans the route or stops navigation to avoid obstacles according to the obstacle avoidance plan.

Images