Port internal and external fleet positioning system and method

Abstract

The invention discloses a port internal and external fleet positioning system and method, and relates to the technical field of intelligent positioning. The method is used for solving the problems of frequent occurrence of positioning blind areas, sensitive dynamic interference, uncontrollable error accumulation and low anchor point resource scheduling efficiency in a complex scene of a port. Anchor point coverage is dynamically calculated through metal environment signal attenuation characteristics, blind area information is generated in combination with vehicle trajectory clustering, and signal resource allocation is optimized. Ground texture gray scale entropy segmentation and cross-frame Bayesian correlation are carried out in a blind area, interference is eliminated, synchronous pose data are generated, and visual positioning robustness is improved. An error propagation chain model is constructed based on communication topology among vehicles, accumulated errors are corrected by fusing confidence coefficients of multiple vehicles, and the motorcade positioning consistency is enhanced. And finally, analyzing a port scheduling plan and a vehicle motion state, constructing a space-time prediction model, dynamically generating an anchor point instruction, and realizing on-demand scheduling of anchor point resources. The port vehicle positioning precision and the resource utilization efficiency are remarkably improved, and reliable support is provided for port automatic operation.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent positioning, and specifically to an internal and external fleet positioning system and method for ports. Background Art

[0002] With the continuous improvement of the global port automation and intelligence level, container terminals are gradually transforming to an operation mode of "paperless scheduling, unmanned handling, and information-based collaboration". The container trucks, stackers, automated guided vehicles (AGVs) and other operating vehicles inside the port, as well as the logistics trucks outside the port, form a high-frequency collaborative operation network, which poses higher requirements for the real-time positioning accuracy of vehicles, the robustness of the system, and the scheduling response ability. In a typical port operation environment, due to the intensive metal equipment and serious signal occlusion, the traditional positioning method that solely relies on satellite navigation has poor stability and large error fluctuations, and it is difficult to support complex scheduling tasks and automated operation processes. In order to ensure operation safety and improve throughput efficiency, there is an urgent need to establish a high-precision multi-vehicle collaborative positioning system for port scenarios with dynamic adaptive capabilities.

[0003] The existing port vehicle positioning methods mainly include three categories: one is the absolute positioning system based on the Global Navigation Satellite System (GNSS), the second is the assisted positioning system that relies on the positioning of ultra-wideband (UWB) signals, radio frequency identification (RFID) tag identification, or magnetic nails deployed on the ground, and the third is the feature perception positioning system based on vision and lidar. However, these methods have the following substantial technical defects in practical applications: First, most of the anchor point positioning schemes adopt a fixed deployment method, lacking a mechanism for real-time collaboration with the vehicle operation path, resulting in low utilization rate of anchor point signals and easy formation of signal blind spots in dynamic operation areas. Second, the vision positioning method generally uses single-frame image feature extraction, which is easily affected by interference factors such as dynamic occlusions and light changes, and cannot stably output continuous pose estimation results. Third, most multi-vehicle collaborative positioning systems adopt a centralized processing architecture, without considering the differences in communication relationships and positioning uncertainties between nodes within the fleet, and errors are prone to accumulate and spread in the system. Fourth, most of the anchor point activation strategies are triggered in a fixed-period polling manner, not linked in real time with the operation plan in the port scheduling system, and cannot achieve anchor point coverage prediction based on the future movement trend of vehicles, resulting in a lag in the scheduling control response and affecting the overall collaborative performance of the system. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides an internal and external fleet positioning system and method for ports, which solves the problems in the above background art.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A port internal and external vehicle positioning system, including the following modules: a dynamic anchor collaborative construction module, an anti-interference environment feature extraction module, a vehicle fleet collaborative error suppression module, and an anchor spatio-temporal prediction module; the dynamic anchor collaborative construction module is used to obtain the dynamic position data of port mobile devices in real time, dynamically calculate the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment, and generate anchor activation instructions and anchor coverage blind area information by combining vehicle historical trajectory clustering analysis; the anti-interference environment feature extraction module is used to receive the anchor coverage blind area information, perform local gray entropy segmentation on the ground images collected by on-vehicle cameras, screen high-discrimination texture regions, and exclude dynamic interference through cross-frame Bayesian probability association to generate pose estimation data that is spatio-temporally aligned with the anchor signal; the vehicle fleet collaborative error suppression module is used to receive the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model in combination with the vehicle-to-vehicle communication topology relationship, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy; the anchor spatio-temporal prediction module is used to analyze the operation plans of mobile devices in the port scheduling system, construct a spatio-temporal reachability prediction model in combination with the motion state of the calibrated vehicle pose, predict the spatio-temporal distribution matrix of reachable anchors in the future, and generate control instructions according to the anchor activation priority in the matrix and feedback them to the dynamic anchor collaborative construction module.

[0006] Further, the specific process of obtaining the dynamic position data of port mobile devices in real time and dynamically calculating the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment is as follows: Real-time analyze the dynamic position data of port mobile devices to obtain the real-time coordinates and motion directions of gantry crane booms and temporary traffic devices; According to the metal obstacle distribution density in the container stacking area, segmentally adjust the attenuation coefficient of the signal propagation model to generate a non-uniform attenuation gradient. According to the corrected attenuation gradient, calculate the effective coverage boundary of the signal for each anchor point, generate a vector map described by the vertex coordinates of a polygon, match the real-time position of the vehicle with the vector map, mark the area where the signal strength is lower than the positioning threshold as the coverage blind area, and output the list of blind area vertex coordinates.

[0007] Further, the specific process of generating anchor activation instructions and anchor coverage blind area information by combining vehicle historical trajectory clustering analysis is as follows: Collect vehicle historical trajectory data, and identify high-frequency traffic areas, including container transfer areas and gate channels, through density clustering algorithms; Generate anchor activation priority instructions according to the clustering results, and preferentially activate the anchor points of mobile devices in high-frequency areas; Match the coordinates of the coverage blind area with the real-time position of the vehicle. When the vehicle enters the blind area boundary, activate the adjacent mobile device anchor points according to the blind area vertex coordinates, and extend their working hours. Update the anchor activation timing table according to the real-time position of the vehicle and adjust the signal coverage range.

[0008] Further, the specific process of performing local gray entropy segmentation on the ground images collected by the vehicle-mounted camera to screen out high-discrimination texture regions is as follows: within the coverage blind area, divide the ground images collected by the vehicle-mounted camera into multiple square local blocks, calculate the local gray entropy values of each block, screen out the blocks with gray entropy values higher than the environmental background as candidate texture regions, compare the entropy value changes of the same blocks in adjacent frame images, and eliminate the abnormally high-entropy blocks caused by instantaneous reflection or projection; perform morphological closing operations on the candidate texture regions to fill the broken regions and generate a continuous positioning reference mask.

[0009] Further, the specific process of generating pose estimation data that is spatio-temporally aligned with the anchor point signal by excluding dynamic interference through cross-frame Bayesian probability association is as follows: perform spatio-temporal calibration on the stable texture features in multiple consecutive frames of images, record the motion trajectories of the feature points, construct a cross-frame feature association probability matrix, calculate the association probability between the features in the current frame and the historical frames according to the Bayesian inference model, and eliminate the features if the probability is lower than the dynamic interference threshold; align the screened features with the time stamps of the anchor point signals, solve the vehicle pose through the perspective transformation matrix, and fuse the pose estimation results of multiple frames to generate smoothed vehicle trajectory data.

[0010] Further, the construction logic of constructing an error propagation chain model by receiving the pose estimation data and the vehicle inertial navigation data and combining the inter-vehicle communication topology relationship is as follows:. Dynamically establish a mesh communication topology according to the wireless communication signal strength and distance threshold between vehicles, maintain the vehicle node connection status table in real time, map the cumulative error of the single-vehicle inertial navigation to the directed graph nodes, and construct the error propagation weight through the pose difference between adjacent vehicles; calculate the transmission attenuation coefficient of the error along the communication link according to the vehicle motion direction consistency to generate an error propagation chain matrix, and recalculate the error propagation weight and attenuation coefficient when the vehicle position or communication topology changes.

[0011] Further, the specific process of correcting the cumulative positioning error through a multi-vehicle confidence weight fusion strategy is as follows: dynamically allocate confidence weights according to the alignment degree between the vehicle pose estimation data and the anchor point signal and the historical positioning stability; if the pose difference between adjacent vehicles exceeds the set threshold, reduce the weight of the abnormal vehicle and trigger local texture matching, and fuse the pose data of multiple vehicles through a weighted average algorithm based on the error propagation chain matrix to suppress the single-vehicle error; feedback the fused pose to each vehicle node to update the initial state of the inertial navigation.

[0012] Furthermore, the construction logic of the spatio-temporal reachability prediction model by parsing the operation plan of mobile devices in the port scheduling system and combining the motion state of the corrected vehicle pose is as follows: Extract the operation schedule table and path planning data of mobile devices in the port scheduling system, including the moving trajectory of the gantry crane boom and the deployment period of temporary equipment; According to the corrected pose data, construct a vehicle kinematic model to predict the spatio-temporal area that the vehicle can reach in the future, perform an intersection operation on the predicted vehicle trajectory and the spatio-temporal distribution of mobile device anchor points, and generate a spatio-temporal reachability probability distribution map; According to the spatio-temporal density of the intersection area, generate an anchor point activation priority list and mark the anchor points with high probability of being reached.

[0013] Furthermore, the specific process of predicting the spatio-temporal distribution matrix of reachable anchor points in the future and generating control instructions according to the anchor point activation priority in the matrix and feeding them back to the dynamic anchor point collaborative construction module is as follows: Quantify the spatio-temporal reachability probability distribution map into a matrix, where the matrix dimension is time-space-anchor identifier, sort according to the probability values of spatio-temporal nodes in the matrix, generate anchor point activation priority instructions, and preferentially wake up the anchor points with high probability of being reached; Feed the priority instructions back to the dynamic anchor point collaborative construction module through the port wireless private network to trigger the anchor point wake-up or sleep operation, and monitor the anchor point activation effect in real time. If the vehicle does not reach the anchor point as predicted, dynamically adjust the priority instructions and regenerate the matrix.

[0014] The positioning method for internal and external port fleets includes the following steps: S1. Obtain the dynamic position data of port mobile devices in real time, dynamically calculate the anchor point signal coverage range according to the signal propagation attenuation characteristics in the metal environment, and generate anchor point activation instructions and anchor point coverage blind area information by combining vehicle historical trajectory clustering analysis; S2. Receive the anchor point coverage blind area information, perform local gray entropy segmentation on the ground images collected by the on-vehicle camera, screen the high-distinguishability texture areas, and exclude dynamic interference through cross-frame Bayesian probability association to generate pose estimation data that is spatio-temporally aligned with the anchor point signal; S3. Combine the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model based on the vehicle-to-vehicle communication topology relationship, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy; S4. Parse the operation plan of mobile devices in the port scheduling system, construct a spatio-temporal reachability prediction model by combining the motion state of the corrected vehicle pose, predict the spatio-temporal distribution matrix of reachable anchor points in the future, and generate control instructions according to the anchor point activation priority in the matrix and feed them back to step S1.

[0015] The present invention has the following beneficial effects: (1)Port internal and external fleet positioning system. Through the dynamic anchor collaborative construction module, it dynamically calculates the signal coverage range of the anchor according to the signal propagation attenuation characteristics in the metal environment, generates anchor activation instructions and blind area information by combining vehicle historical trajectory clustering analysis, effectively solves the coverage blind area problem caused by signal multipath reflection in the metal-intensive port scenario, and improves the utilization rate of anchor resources and positioning continuity. Through the anti-interference environment feature extraction module, local gray entropy segmentation and cross-frame Bayesian probability correlation are performed on the ground images in the coverage blind area, high-discrimination texture features are screened out and dynamic interferences (such as container projections, water accumulation reflections) are excluded, significantly improving the spatio-temporal alignment accuracy of visual positioning and anchor signals, and ensuring the robustness of positioning data in complex environments.

[0016] (2)Port internal and external fleet positioning method. Based on the dynamic communication topology relationship between vehicles, an error propagation chain model is constructed, and the cumulative positioning error is corrected by combining the multi-vehicle confidence weight fusion strategy, breaking through the bottleneck of long-term error accumulation of single-vehicle inertial navigation, and improving the overall positioning stability of the fleet. Analyze the operation plan of the port scheduling system and construct a spatio-temporal reachability prediction model in combination with the vehicle motion state, dynamically generate anchor activation priority instructions and close-loop feedback to the anchor collaborative construction module, realize the on-demand scheduling and predictive response of anchor resources, greatly reduce the dependence of the positioning system on fixed infrastructure, and at the same time improve the scheduling efficiency and operation safety of the port temporary fleet.

[0017] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned advantages simultaneously. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flowchart of the port internal and external fleet positioning system of the present invention.

[0019] Figure 2 It is a flowchart of the port internal and external fleet positioning method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0020] In the embodiments of the present application, through the port internal and external fleet positioning system and method, the problems of frequent vehicle positioning blind areas, sensitivity to dynamic interference, uncontrollable error accumulation, and low anchor resource scheduling efficiency in the complex port environment are solved.

[0021] The general idea of the solution in the embodiments of the present application is as follows: Dynamic anchor collaborative construction: By real-time obtaining the dynamic position data of port mobile devices, combining the metal environment signal attenuation correction model and vehicle historical trajectory clustering analysis, dynamically generate anchor activation instructions and coverage blind area information, and solve the problems of high cost of fixed anchor deployment and randomness of signal coverage blind areas.

[0022] Anti-interference environment feature extraction: In the coverage blind area, based on local gray entropy segmentation, high-distinguishability ground textures are screened, and instantaneous interferences (such as container projections and water accumulation reflections) are excluded through cross-frame Bayesian probability correlation, generating pose data that is spatio-temporally aligned with the anchor signal, breaking through the dependence on artificial markers for visual positioning.

[0023] Fleet collaborative error suppression: Based on the dynamic communication topology relationship between vehicles, an error propagation chain model is constructed, and the cumulative error of inertial navigation is corrected through a multi-vehicle confidence weight fusion strategy to solve the problem of long-term positioning drift of a single vehicle and improve the overall positioning consistency of the fleet.

[0024] Anchor spatio-temporal prediction and closed-loop control: Analyze the operation plan of the port scheduling system, construct a spatio-temporal reachability prediction model in combination with the vehicle motion state, dynamically generate anchor activation priority instructions and feedback them to the anchor collaboration module, realizing on-demand scheduling and predictive response of anchor resources and reducing the hardware dependence cost.

[0025] Please refer to Figure 1 , the embodiment of the present invention provides a technical solution: an internal and external fleet positioning system in a port, including the following modules: a dynamic anchor collaboration construction module, an anti-interference environment feature extraction module, a fleet collaborative error suppression module, and an anchor spatio-temporal prediction module; the dynamic anchor collaboration construction module is used to obtain the dynamic position data of port mobile devices in real time, dynamically calculate the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment, and generate anchor activation instructions and anchor coverage blind area information through clustering analysis of vehicle historical trajectories; the anti-interference environment feature extraction module is used to receive the anchor coverage blind area information, perform local gray entropy segmentation on the ground images collected by the vehicle-mounted camera, screen high-distinguishability texture areas, and exclude dynamic interferences through cross-frame Bayesian probability correlation, generating pose estimation data that is spatio-temporally aligned with the anchor signal; the fleet collaborative error suppression module is used to receive the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model in combination with the communication topology relationship between vehicles, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy; the anchor spatio-temporal prediction module is used to analyze the operation plan of mobile devices in the port scheduling system, construct a spatio-temporal reachability prediction model in combination with the motion state of the corrected vehicle pose, predict the spatio-temporal distribution matrix of reachable anchors in the future, and generate control instructions according to the anchor activation priority in the matrix and feedback them to the dynamic anchor collaboration construction module.

[0026] In this implementation plan, the dynamic anchor point collaborative construction module: This module is used to dynamically establish an anchor point network according to the real-time position changes of mobile devices (such as AGVs, container trucks, etc.) within the port. Its goal is to improve the rationality of the spatial distribution of positioning signals and their dynamic adaptability, and avoid signal redundancy and blind spots. Dynamic position data: Refers to the position information uploaded in real time by mobile devices within the port (which can be obtained through GNSS, inertial navigation, or integrated navigation methods). Signal propagation attenuation characteristics: In an environment with a high density of metals, the propagation of electromagnetic waves will exhibit obvious attenuation and multipath effects. This characteristic is considered in the present invention for dynamically estimating the effective signal coverage range of anchor points. Anchor point signal coverage range: Refers to the area where a single anchor point device can stably provide positioning services under the current environmental conditions. Vehicle historical trajectory clustering analysis: By performing trajectory clustering on historical positioning trajectories, regular paths and high-frequency operation areas are extracted, providing a statistical basis for anchor point layout and activation decisions. Anchor point activation instructions and blind spot information: The former is used to indicate which anchor points should be activated to provide positioning services in the current or future time periods; the latter is used to mark the areas where the current anchor point signals are unreachable. Anti-interference environment feature extraction module: This module is used to assist in positioning by means of an on-vehicle vision perception system within the blind area of the anchor point signal coverage, ensuring that the system can still stably estimate the vehicle pose in scenarios where the anchor points are unavailable. Local gray entropy segmentation: Based on the local entropy value of the gray distribution in the image, texture regions with rich information and high discrimination are extracted, which helps to improve the stability and uniqueness of image features. Gray entropy: Reflects the degree of chaos (information entropy) of the gray distribution in the image and is used to judge the complexity of the region. High-discrimination texture region: The part of the image with obvious boundaries, clear structures, and low repeatability, which is suitable for feature matching and pose estimation. Cross-frame Bayesian probability association: Using Bayesian inference methods, probability matching of texture features is performed between time-series images to eliminate feature points affected by interference factors such as occlusion and dynamic objects. Bayesian probability association: Updates the matching confidence based on the prior distribution and observed data, and is a dynamic feature matching method with relatively high robustness. Pose estimation data: Represents the position and pose of the vehicle at a certain moment (including position coordinates and orientation angles) and is used for navigation and path tracking. Fleet collaborative error suppression module: This module is for multi-vehicle collaborative positioning applications. By introducing communication topology modeling and confidence fusion strategies, it realizes the sharing and complementarity of positioning data among vehicles, and suppresses the accumulation and diffusion of single-vehicle positioning errors. Inertial navigation data: Short-term position change information calculated based on inertial components such as accelerometers and gyroscopes, but there are long-term drift errors. Communication topology relationship: Describes the information transmission structure among vehicles in the fleet, such as star-shaped, ring-shaped, or mesh structures, and determines the paths through which error information can propagate. Error propagation chain model: Models how error information propagates in the fleet and is used to estimate the potential error transmission paths and superposition effects of each vehicle.Multi-vehicle confidence weight fusion strategy: Dynamically allocate confidence weights according to the stability of each vehicle's positioning data, sensor quality, and communication stability, and perform distributed filtering fusion to correct the differential errors between vehicles. Confidence weight: A numerical index used to reflect the credibility of the data source. Anchor point spatio-temporal prediction module: This module is used to combine the operation plan in the port scheduling system to predict the spatio-temporal position distribution of reachable anchor points in the future, so as to optimize the anchor point activation strategy in advance and achieve the forward-looking deployment of positioning services and the dynamic scheduling of system resources. Operation plan parsing: Read the data on the future operation tasks of equipment in the scheduling system, such as path planning, loading and unloading arrangements, etc. Motion state modeling: Build a future motion trajectory model based on parameters such as the current pose, speed, and acceleration of the vehicle. Spatio-temporal reachability prediction model: Integrate the operation plan and the motion model to predict the areas that the vehicle may pass through and the spatial range that can be covered by anchor points within a certain period in the future. Anchor point activation priority: According to the prediction model, calculate the importance ranking of each candidate anchor point for future positioning tasks, which is used to indicate the anchor point scheduling priority. Control instruction feedback: Transmit the priority decision result back to the dynamic anchor point module to form a task-driven closed-loop anchor point optimization system.

[0027] Specifically, the specific process of dynamically calculating the signal coverage range of anchor points according to the dynamic position data of port mobile devices in real time and the signal propagation attenuation characteristics in the metal environment is as follows: Parse the dynamic position data of port mobile devices in real time to obtain the real-time coordinates and motion directions of gantry crane booms and temporary traffic equipment; According to the distribution density of metal obstacles in the container stacking area, segmentally adjust the attenuation coefficient of the signal propagation model to generate a non-uniform attenuation gradient. According to the corrected attenuation gradient, calculate the effective signal coverage boundary for each anchor point, generate a vector map described by the vertex coordinates of a polygon, match the real-time position of the vehicle with the vector map, mark the area where the signal strength is lower than the positioning threshold as the coverage blind area, and output the list of blind area vertex coordinates.

[0028] In this implementation plan, the description of the step of parsing the dynamic position data of port mobile devices in real time: Real-time obtain the dynamic position information of port mobile devices through the on-vehicle positioning terminal (GNSS / IMU) and the positioning gateway deployed in the port scheduling system. The parsed data contains the following information: Vehicle number and type (AGV, tractor, empty container vehicle, etc.); Current coordinate point; Motion direction (in the form of a unit vector and heading angle ); Locations and movement trajectories of other dynamic facilities (such as gantry crane booms, guiding vehicles, temporary obstacles). Steps for analyzing the distribution density of metal obstacles and adjusting the signal propagation model: Use the port 3D modeling system or lidar point cloud reconstruction model to obtain the spatial distribution characteristics of metal objects such as container stacking areas and crane structures. Correct the attenuation degree of the signal propagation path according to the density of metal objects and the occlusion structure. Non-uniform attenuation modeling: Establish a set of piecewise corrected signal attenuation functions according to the density distribution of metal obstacles in the path. Let: : The propagation power of the j-th anchor point signal in the i-th path segment; : The distance of this path segment; : The environmental attenuation factor related to the metal occlusion density of this segment. Then the signal power of this segment can be expressed as: ; where: : The initial transmission power of the j-th anchor point; : Adjusted according to the on-site occlusion situation. For example, it is set to 3.0 - 4.5 in high-density container areas and 2.0 - 2.5 in open areas. The overall signal attenuation result from anchor point j to any coordinate point M can be obtained by summing multiple paths . Steps for calculating the signal coverage boundary of anchor points: Based on the port 2D geographic coordinate grid, perform ray scans from each anchor point in all directions. According to the above non-uniform model, determine the farthest point on the ray path where the signal intensity drops to the threshold as the boundary point in this direction. Polygon boundary construction method: Connect the boundary points in each direction to form an irregular polygon boundary representing the effective area of the anchor point signal. This polygon boundary is represented in the form of a vertex coordinate list: ; where, is the signal coverage area of the -th anchor point, is the number of boundary vertices. Steps for constructing a vector map and matching vehicle positions: Based on the anchor point polygon boundary, construct a real-time vector map of the port and project the current position of each vehicle onto the map. Use the point inpolygon algorithm (such as the ray method, winding number method, etc.) to determine whether the vehicle is within the effective coverage area of an anchor point: If , then the vehicle is covered by the signal of the -th anchor point; If the vehicle is not within any , or the combined signal power at this point , then this point is judged as an "anchor point blind area". Steps for marking blind areas and outputting blind area polygon coordinates: Cluster the spaces of all areas not effectively covered by any anchor point signal, and extract the continuous blind area boundaries. Adopt, for example, DBSCAN density clustering , and output its vertex coordinate list: ; where, Indicates the polygonal boundary of the th blind spot area. This coordinate information will serve as the input basis for subsequent modules (such as the anti-interference image perception module) to guide its visual-aided positioning within the blind spot area.

[0029] Specifically, the specific process of generating the anchor activation instruction and the blind spot information covered by the anchor in combination with the vehicle historical trajectory clustering analysis is as follows: Collect vehicle historical trajectory data, and identify high-frequency passing areas through the density clustering algorithm, including container transfer areas and gate channels; Generate an anchor activation priority instruction according to the clustering result, and preferentially activate the mobile device anchors within the high-frequency area; Match the blind spot coordinates with the vehicle's real-time position. When the vehicle enters the blind spot boundary, activate the adjacent mobile device anchors according to the blind spot vertex coordinates and extend their working duration, update the anchor activation timing table according to the vehicle's real-time position, and adjust the signal coverage range.

[0030] In this implementation plan, the steps of collecting and preprocessing vehicle historical trajectory data are described as follows: Retrieve the vehicle trajectory logs of multiple days from the port scheduling system. The format of the trajectory data is as follows: ; where: : The historical trajectory data of the th vehicle; : The position at the th time point; : The corresponding timestamp; : The total number of trajectory points recorded by this vehicle. After aligning the time and normalizing the space of all vehicle trajectory sets, they are sent into the clustering model. Second, the steps of identifying high-frequency passing areas based on the density clustering algorithm are described as follows: Use the density-based spatial clustering algorithm (such as DBSCAN) to process the trajectory point set to identify the high-frequency activity areas of vehicles. Let the trajectory set be: ; The output result after clustering is: ; where: : The th high-density passing cluster; : The number of trajectory points contained in this cluster; The clustering clusters will correspond to specific port area scenarios, such as the main channels of the yard and the gate entrances and exits. The steps of generating the anchor activation priority instruction are described as follows: Quantitatively analyze the clustering clusters according to the vehicle passing frequency, time period distribution, and traffic density, and calculate the priority score λ_q of each cluster. The scoring model is: ; where: : The vehicle passing frequency within the cluster; : The average vehicle density per unit time; : The time coverage rate of the passing time period within the cluster; : The weight coefficient (set according to the port operation weight). According to sort from high to low to generate the anchor priority activation list , guide the dynamic anchor collaborative construction module to preferentially enable the mobile anchors corresponding to the high-traffic areas. Instructions for dynamically activating adjacent anchors in combination with blind spot information: According to the list of blind spot polygon vertices identified in the previous steps: ; Take the current position of the vehicle and perform an Euclidean distance judgment with the blind spot boundary. Let the closest point on the blind spot boundary be , then there is: ; When (set the distance threshold), it is considered that the vehicle is about to enter the blind spot, then: Send an early activation command to the mobile device anchor near the blind spot boundary; According to the vehicle's driving speed , estimate the residence time in the blind spot , and set the anchor delay shutdown time to the current time plus . Instructions for dynamically updating the anchor activation timing table and coverage area: Establish an anchor control timing table , indicating the activation status of each anchor at time t: ; Among them: : The s-th anchor number; : The estimated shutdown time, which is dynamically updated in combination with the vehicle's current position, path prediction, and residence time in the blind spot; The activation status "on / off" is jointly determined by whether it is at the blind spot boundary and the clustering high-priority area. According to the change of the vehicle's movement trajectory, continuously update the anchor coverage priority and the working time of each anchor to ensure that the blind spot is not covered and missed for a long time, and also avoid energy consumption waste caused by redundant activation of anchor resources.

[0031] Specifically, the specific process of local gray entropy segmentation of the ground image collected by the vehicle-mounted camera to screen the high-distinguishability texture area is as follows: Within the coverage of the blind spot, divide the ground image collected by the vehicle-mounted camera into multiple square local blocks, calculate the local gray entropy value of each block, and screen the blocks with gray entropy values higher than the environmental background as candidate texture areas. Compare the entropy value changes of the same block in adjacent frame images, and eliminate the abnormally high-entropy blocks caused by instantaneous reflection or projection; Perform morphological closing operations on the candidate texture areas to fill the broken areas and generate a continuous positioning reference mask.

[0032] In this implementation plan, the instructions for image area division and local gray entropy calculation: Divide the grayscale image collected by the vehicle-mounted camera into multiple non-overlapping local square blocks: ; Among them: : The ground grayscale image of the current frame; : The -th row and -th column of the image local block; : The dimension of the divided block grid, which is determined according to the image size and the block granularity. Local gray entropy calculation: For each block Calculate its grayscale entropy value : ; where: : Grayscale level (usually 256); : The grayscale value is The normalized frequency of pixels in the block ; : To prevent the appearance of logarithmic terms A small constant (such as ). Description of the candidate texture block screening and high-entropy anomaly rejection steps: 1. Background entropy threshold setting: Calculate the entropy mean value and standard deviation of all blocks in the entire image, and set the background entropy threshold: ; where: : Adjustment coefficient, controlling the screening intensity (usually taking 1 - 2); Regions exceeding this threshold are regarded as candidate blocks with texture discrimination. 2. Inter-frame high-entropy anomaly rejection: For two consecutive frames of images and , compare the entropy difference of blocks at the same position: ; If , then judge that this change is a non-structural perturbation (such as reflection, projection), and reject this block. : Set the threshold, reflecting the maximum entropy change of stable texture (empirical value such as 0.5 - 1.0); After rejection, retain stable and repeatable texture blocks. Three. Description of the morphological closing operation repair and mask generation steps: Perform a morphological closing operation (dilate first and then erode) on the binary mask image of the candidate texture block to eliminate small holes and discontinuous connection regions: ; where Dilate and Erode respectively represent the image dilation and erosion operations; : Structure element (such as square kernel); The operation result is a closed and continuous positioning reference mask image. Output the final result of the continuous texture positioning mask: Output as the texture region applicable to positioning and matching in this frame of image; After this mask is spatio-temporally aligned with the anchor point signal coverage information, it is sent to the subsequent cross-frame Bayesian pose estimation module. Ensure that the vehicle can still rely on stable ground textures for auxiliary positioning in the blind area, improving the positioning robustness.

[0033] Specifically, by excluding dynamic interference through cross-frame Bayesian probability association, the specific process of generating pose estimation data that is spatio-temporally aligned with the anchor signal is as follows: Spatio-temporal calibration is performed on stable texture features in consecutive multiple frames of images, the motion trajectories of feature points are recorded, a cross-frame feature association probability matrix is constructed, and the association probability between the features in the current frame and historical frames is calculated according to the Bayesian inference model. If the probability is lower than the dynamic interference threshold, the feature is excluded; the filtered features are aligned with the timestamp of the anchor signal, the vehicle pose is solved through the perspective transformation matrix, and the pose estimation results of multiple frames are fused to generate smoothed vehicle trajectory data.

[0034] In this implementation, the spatio-temporal calibration of stable texture features and the construction steps of feature trajectories are described as follows: Extract the key feature point set of the stable texture region from the mask image that has completed gray entropy segmentation and morphological processing: ; where: : Image frame time index; : The -th image coordinates of the feature point; : The texture feature point set extracted at time . Use the optical flow method or feature matching algorithm to track the same points in consecutive multiple frames, and construct the motion trajectory sequence of each feature point: . The cross-frame feature association probability modeling and Bayesian inference steps are described as follows: For each pair of feature point trajectories, based on factors such as the similarity of their image space positions, the consistency of motion directions, and the time interval decay factor, construct the cross-frame feature association probability: ; where: : The Bayesian association probability of the -th feature at time and the -th feature at time ; : Likelihood function, measuring the position and motion differences; : Prior probability, set according to the regional texture stability; the denominator is the normalization factor to ensure that the sum of the association probabilities of all candidate feature points is 1. Note: The likelihood function can be defined in the form of a two-dimensional Gaussian distribution, and the specific formula is omitted and can be found in the literature of the image matching field. If the maximum association probability of a certain feature point is lower than the dynamic interference threshold , then this point is determined as the source of dynamic interference (such as pedestrians, robotic arms) and excluded: . The spatio-temporal alignment with the anchor signal and the vehicle pose solution steps are described as follows: Time synchronization: For the remaining stable feature points, align the image acquisition timestamp with the anchor signal timestamp , and perform time fusion using linear interpolation or nearest neighbor matching. Solving the single-frame perspective transformation matrix: Using the filtered planar feature points and their corresponding physical coordinates on the real port ground, through the homography matrix Establish the mapping relationship between the image and the world coordinates: ; where: : Homogeneous coordinates of the feature points in the image coordinate system; : Corresponding position in the world coordinates; : Perspective transformation matrix estimated based on the matching point pairs; through Inverse the two-dimensional pose of the vehicle body . Description of the steps for generating the trajectory by fusing multi-frame poses: For the vehicle pose estimated for each frame , adopt a weighted filtering strategy within a sliding window for trajectory smoothing: ; where: : Smoothing weight, satisfying higher weighting for the central frame (such as Gaussian weight or triangular kernel); : Radius of the sliding window; : Smoothed pose after filtering; The continuous are combined to form a trajectory for subsequent processing by the error suppression module.

[0035] Specifically, receive the pose estimation data and the vehicle inertial navigation data, and the construction logic for building the error propagation chain model by combining the vehicle - to - vehicle communication topology relationship is as follows:. According to the vehicle - to - vehicle wireless communication signal strength and the distance threshold, dynamically establish a mesh communication topology, and maintain the vehicle node connection status table in real - time. Map the cumulative error of the single - vehicle inertial navigation to the nodes of the directed graph, and construct the error propagation weight through the pose difference between adjacent vehicles; According to the vehicle motion direction consistency, calculate the transmission attenuation coefficient of the error along the communication link, generate the error propagation chain matrix, and recalculate the error propagation weight and attenuation coefficient when the vehicle position or communication topology changes.

[0036] In this implementation scheme, for the construction of the communication topology and the maintenance of the connection status, the dynamic communication topology graph is constructed according to the wireless communication signal strength (RSSI) between each vehicle within the vehicle fleet and the maximum communication distance threshold , and a mesh communication topology graph is dynamically established , where: : Represents all vehicle nodes in the vehicle fleet; : Represents the set of vehicle pairs that meet the communication conditions; The communication edge If and only if: ; where: : Vehicle and 's Euclidean distance; : Lower threshold of the communication signal strength. The node connection status table updates each vehicle node to broadcast a status heartbeat packet regularly; The central controller or edge node aggregates the communication status to generate the topology connection status table , where: . The single-vehicle error node represents the calculation of the adjacent error weight. The single-vehicle error node is defined to accumulate the current inertial navigation error of each vehicle It is represented as the node error value in the figure, where: , which represents the position residual between the inertial solution and the visual pose; : Vehicle The current pose calculated by inertial navigation; : Vehicle The pose calculated by visual / anchor alignment. The adjacent error propagation weight is defined for any pair of adjacent vehicles ( ), and the error propagation weight represents the credibility of the error transmitted from vehicle to vehicle : ; where: : The error amplification factor (set according to system experience); : Represents the Euclidean distance between the visually estimated poses; the larger the weight value, the closer the poses of the two vehicles are, and the error can be effectively propagated and compensated. Modeling of the motion direction consistency and error transmission attenuation, the direction consistency factor is set for vehicles , The motion direction vectors are respectively: ; then the direction consistency factor between the two is defined as: ; if the directions of the two vehicles are the same (collinear and in the same direction), ; if the directions are opposite, . The error attenuation coefficient is calculated by comprehensively considering the direction consistency and the communication strength, and the error attenuation coefficient is defined as: ; if the directions of the vehicles are inconsistent ( ), the attenuation coefficient is 0; the attenuation coefficient of the effective link determines the efficiency of error transmission within the vehicle fleet. Construction and dynamic update mechanism of the error propagation chain matrix, the error propagation chain matrix defines the error propagation chain matrix: ; its elements are: ; each row represents the attenuation intensity of the error of vehicle that may be transmitted to other vehicles; error tracing, weighted feedback and fusion can be realized on the matrix graph. Dynamic recalculation mechanism When the position change of any vehicle exceeds the threshold , or the communication topology changes (new / removed edges), the following two steps are re-executed: Reconstruct the communication graph and the connection status table; update , , , and reconstruct .

[0037] Specifically, the specific process of correcting the cumulative positioning error through the multi-vehicle confidence weight fusion strategy is as follows: Dynamically allocate confidence weights according to the alignment degree between vehicle pose estimation data and anchor signal, and historical positioning stability; if the pose difference between adjacent vehicles exceeds the set threshold, reduce the weight of the abnormal vehicle and trigger local texture matching. Based on the error propagation chain matrix, fuse the multi-vehicle pose data through the weighted average algorithm to suppress the single-vehicle error; feedback the fused pose to each vehicle node to update the initial state of inertial navigation.

[0038] In this implementation scheme, for any vehicle node in the vehicle fleet in the confidence weight allocation logic and dynamic confidence index calculation , let its current visual pose estimate be , the inertial navigation pose be , the anchor signal matching degree be , the historical stability score be , then the vehicle confidence weight is defined in the following way: ; where: : represents the alignment degree between the current pose of vehicle and the anchor signal; : represents the positioning stability of the vehicle in the recent period (such as the inverse of the position fluctuation variance); : represents the set of neighbor vehicles of vehicle under the communication topology; the denominator normalization ensures . Note: This formula reflects the weight evaluation mechanism based on the anchor matching quality and historical performance, avoiding the pollution of the overall positioning caused by a single abnormal node. Abnormal detection and local texture compensation trigger, pose difference detection for any adjacent vehicles , , calculate their current visual pose difference as: ; if there exists: ; then it is determined that or is abnormal, and the following operations are triggered: temporarily reduce the confidence of the abnormal vehicle node (for example, multiply by the adjustment factor ); initiate the local texture relocalization process of the abnormal node to perform texture matching verification on the recent several frames of images; if the relocalization is effective, restore the confidence, otherwise continuously reduce the influence weight of this node in the fusion model. Weighted fusion correction based on the propagation chain matrix, for the target vehicle in the multi-vehicle pose fusion model, its neighborhood vehicle set is , the error propagation chain matrix is , and the fused pose estimate is calculated as follows: ; where: : the attenuation coefficient in the error propagation chain (see the previous section for details); : Confidence based on anchor point matching and historical stability assessment; : The current visual pose of the adjacent vehicle ; After multi-vehicle weighted fusion, the drift error or local failure of a certain vehicle can be significantly suppressed. The equalization normalization strategy performs normalization processing before fusion to avoid the imbalance of the total weight: . The pose feedback and inertial navigation state re-initialization use the fused smooth pose as a high-confidence reference result; and feedback it to the vehicle 's local navigation system; which is used to update the initial state of inertial navigation. The specific update formula is as follows: ; where: : The updated initial position of inertial navigation; : The estimated inertial navigation velocity vector; : The time interval between two navigation state updates.

[0039] Specifically, the construction logic of the spatio-temporal reachability prediction model by parsing the operation plan of mobile devices in the port scheduling system and combining the motion state of the corrected vehicle pose is as follows: Extract the operation schedule and path planning data of mobile devices in the port scheduling system, including the moving trajectory of the gantry crane boom and the deployment period of temporary equipment; According to the corrected pose data, construct a vehicle kinematic model to predict the spatio-temporal area that the vehicle can reach in the future period, perform an intersection operation on the vehicle prediction trajectory and the spatio-temporal distribution of mobile device anchor points to generate a spatio-temporal reachability probability distribution map; According to the spatio-temporal density of the intersection area, generate a list of anchor point activation priorities and mark the high-probability reachable anchor points.

[0040] In this implementation plan, for the parsing of operation plan and path planning data, the operation schedule extraction obtains the operation schedule of mobile devices (such as gantry crane booms, temporary traffic equipment) from the port scheduling system, and defines the time interval set as: ; where represents the start and end time nodes of the operation. The path planning data acquisition includes the set of moving trajectory points of the equipment: ; where represents the position coordinates in three-dimensional space. Construction of the vehicle kinematic model 1. Input the corrected pose data of the vehicle , starting from the corrected pose data , establish a state vector: ; where: : The two-dimensional position coordinates of the vehicle; : The heading angle of the vehicle; : The magnitude of the vehicle speed. The kinematic prediction formula uses the vehicle kinematic model to predict the position at the future moment : ; where: : The angular velocity of the vehicle (estimated from historical data); : Prediction time step. The generation of the spatio-temporal reachability probability distribution map and the construction of the prediction trajectory are carried out through the kinematic model to generate a set of prediction trajectories of the vehicle in the future time period [t, t+T]: ; The spatio-temporal anchor distribution sets the spatio-temporal position set of the mobile device anchor as: ; where is the spatial coordinate of the anchor, is the activation time interval of the anchor. The spatio-temporal intersection operation calculates the intersection of the vehicle prediction trajectory and the anchor spatio-temporal distribution, and calculates the probability that the vehicle reaches the anchor at each moment , combining the spatial distance and time overlap relationship: ; where: represents the Euclidean distance; : The spatial position tolerance radius, which reflects the allowable error range; is the indicator function, which is 1 when and 0 otherwise. The spatio-temporal reachability probability distribution map integrates all anchors and the prediction time period to generate a probability matrix: . The generation of the anchor activation priority list, the spatio-temporal density calculation is based on the spatio-temporal probability matrix , and calculates the total probability of the anchor being reached during the prediction time period: ; The priority sorting is in descending order according to to generate the anchor activation priority list: . High-probability anchor marking For the anchor that meets the threshold condition , it is marked as a "high-priority activation" anchor, which is used as the control input of the dynamic anchor collaborative construction module.

[0041] Specifically, the specific process of predicting the spatio-temporal distribution matrix of the future reachable anchors and generating control instructions according to the anchor activation priority in the matrix and feeding them back to the dynamic anchor collaborative construction module is as follows: Quantify the spatio-temporal reachability probability distribution map into a matrix, the matrix dimension is time-space-anchor identifier, sort according to the probability values of the spatio-temporal nodes in the matrix, generate the anchor activation priority instruction, and preferentially wake up the high-probability reached anchors; Feed the priority instruction back to the dynamic anchor collaborative construction module through the port wireless private network, trigger the anchor wake-up or sleep operation, and monitor the anchor activation effect in real time. If the vehicle does not reach the anchor as predicted, dynamically adjust the priority instruction and regenerate the matrix.

[0042] In this implementation plan, the spatio-temporal reachability probability distribution map is quantified into a matrix, and the probability distribution of the future reachable anchors of the port vehicle in the spatio-temporal range is converted into a three-dimensional matrix representation. The matrix dimension includes the time series, the spatial region, and the anchor identifier. Let this matrix be , where: represents the set of discrete time nodes, and the number is ; represents a set of spatial segmentation regions, with the quantity being ; represents a set of anchor points, with the quantity being ; matrix element represents at the time point , spatial region the probability that a vehicle reaches an anchor point inside. , calculate the activation priority of the anchor points, based on the matrix calculate the activation priority index of the anchor point in it: : where: represents the weight coefficient of the spatio-temporal node , reflecting the priority or importance of this spatio-temporal node (for example, the weight is larger for the time node close to the current time, and the weight is larger for the key operation area); is the comprehensive activation priority score of the anchor point . The anchor point activation priority list is sorted in descending order according to , and the anchor point with a higher score is activated first. Generate and send the anchor point control instruction according to the activation priority threshold , and generate an activation control instruction for each anchor point : where: represents the activation state of the anchor point, 1 means activated, 0 means dormant; the threshold can be dynamically adjusted according to the system load and resource conditions. Send the control instruction to the port wireless private network and feedback it to the dynamic anchor point collaborative construction module to perform the corresponding anchor point wake-up or dormancy operation. The real-time monitoring and dynamic adjustment system continuously monitors the activation effect of the anchor points, records the actual situation of the vehicle reaching the anchor points. If the vehicle does not reach the activated anchor point as predicted, then adjust the activation priority weight or threshold according to the actual data, and regenerate the spatio-temporal distribution matrix and update the activation instruction. This process can be represented by the following correction mechanism: where: is the predicted probability of the previous cycle; is the probability estimate updated based on the actual monitoring data; is the smoothing coefficient, weighing the historical prediction and real-time adjustment.

[0043] Please refer to Figure 2, A method for positioning internal and external vehicle fleets in a port, comprising the following steps: S1. Obtain the dynamic position data of port mobile devices in real time, dynamically calculate the signal coverage range of anchor points according to the signal propagation attenuation characteristics in the metal environment, and generate anchor point activation instructions and anchor point coverage blind area information through clustering analysis of vehicle historical trajectories; S2. Receive the anchor point coverage blind area information, perform local gray entropy segmentation on the ground images collected by on-vehicle cameras, screen high-discrimination texture regions, and exclude dynamic interference through cross-frame Bayesian probability association to generate pose estimation data that is spatio-temporally aligned with the anchor point signals; S3. Combine the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model based on the vehicle-to-vehicle communication topology relationship, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy; S4. Analyze the operation plan of mobile devices in the port scheduling system, construct a spatio-temporal reachability prediction model in combination with the motion state of the corrected vehicle pose, predict the spatio-temporal distribution matrix of reachable anchor points in the future, and generate control instructions according to the anchor point activation priority in the matrix and feedback them to step S1.

[0044] In this implementation, Step S1: The real-time adjustment of the coverage range of dynamic anchor signals breaks through the traditional fixed anchor coverage assumption. It dynamically models the signal attenuation characteristics in a metal-intensive environment, accurately depicts the non-uniform propagation of signals, realizes the real-time dynamic calculation of the coverage range, and significantly improves the spatial effectiveness of positioning signals. The clustering analysis integrating the historical trajectories of vehicles innovatively uses density clustering to identify high-frequency passing areas, realizes the spatio-temporal priority scheduling of anchor activation, effectively optimizes resource allocation, avoids ineffective wake-up, and improves the energy efficiency and response speed of the system. The blind area information feedback mechanism realizes the precise positioning of the coverage blind area, dynamically activates adjacent anchors through the vertex coordinates of the blind area, compensates for the signal blind area, and enhances the continuity and stability of the positioning system. Step S2: The local gray entropy segmentation is used to locate the reference area. The local gray entropy method is adopted to screen high-discrimination textures, improve the stability and recognition rate of visual features, and effectively cope with the illumination and texture changes in the complex port environment. The cross-frame Bayesian probability association model introduces the Bayesian inference mechanism to verify the spatio-temporal consistency of texture features in consecutive frames, and innovatively eliminates dynamic interference factors such as reflection and projection, ensuring the accuracy and robustness of pose estimation data. The spatio-temporal alignment fusion mechanism synchronizes the visual features with the anchor signal timestamps, fuses multi-modal information, and improves the overall positioning accuracy. Step S3: The dynamic communication topology based on the wireless communication signal strength reflects the real-time connection status of the vehicle network, dynamically establishes an error propagation chain among multiple vehicles, breaks through the limitation of single-vehicle positioning, and realizes the cross-vehicle propagation and control of errors. The error propagation chain model innovatively maps the inertial navigation cumulative error to the nodes of a directed graph, and uses the communication topology and motion consistency to calculate the error transfer weight and attenuation coefficient, scientifically revealing the error propagation path and influence degree. The multi-vehicle confidence weight fusion strategy dynamically assigns the vehicle positioning confidence, suppresses the abnormal vehicle error through weighted fusion, improves the stability and reliability of the overall fleet positioning, and realizes collaborative error correction. Step S4: The deep fusion of the scheduling operation plan and the motion state incorporates the port scheduling system operation plan into the positioning prediction process for the first time. Combining with the vehicle kinematic model with corrected pose, it accurately predicts the spatio-temporal area where the future anchor will be reached, and realizes the forward-looking positioning resource scheduling. The spatio-temporal reachability prediction model quantifies the reach probability of the anchor based on the intersection operation of the trajectory prediction and the spatio-temporal distribution of the anchor, systematically constructs the spatio-temporal distribution matrix, and scientifically guides the priority sorting of anchor activation. The closed-loop control feedback mechanism adjusts the dynamic anchor collaborative construction module in real time through the generated control instructions, realizes the adaptive adjustment of the positioning system, and improves the real-time response ability and accuracy of the positioning system.

[0045] In summary, this application has at least the following effects: Port internal and external vehicle positioning system and method, by combining dynamic anchor point cooperation construction with anti-interference environment feature extraction, effectively suppresses the influence of signal propagation attenuation and dynamic interference in the metal environment, and realizes more accurate vehicle pose estimation. Based on cross-frame Bayesian probability association to eliminate dynamic interference features, combined with the vehicle-to-vehicle communication topology to construct an error propagation chain model, the multi-vehicle confidence weight fusion strategy effectively suppresses the cumulative error of single-vehicle inertial navigation, and improves the stability and robustness of the overall positioning system. Through vehicle historical trajectory clustering and spatio-temporal reachability prediction, the dynamic adjustment of anchor point activation is realized, avoiding the ineffective dissipation of anchor point resources, optimizing the signal coverage range, reducing the system energy consumption and operation and maintenance costs. Based on the closed-loop feedback mechanism of the spatio-temporal distribution matrix and dynamic priority instructions, the real-time adjustment and optimization of the anchor point activation strategy are realized, ensuring the system's rapid response ability to changes in the port operation environment. Using a mesh communication topology and an error propagation chain model, fusing multi-vehicle pose data, meeting the multi-node collaborative positioning requirements of large-scale vehicle fleets in the port, and improving the overall operation efficiency and safety management level of the vehicle fleet.

[0046] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0047] The present invention is described with reference to the flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram can be implemented by computer program instructions, and the combination of the processes and / or blocks in the flowchart and / or block diagram can also be implemented. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.

[0048] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory generate a manufactured article including an instruction device, and the instruction device implements in the process Figure 1 one process or multiple processes and / or blocks Figure 1 a device for the functions specified in one block or multiple blocks.

[0049] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus, causing a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one process or a plurality of processes and / or blocks Figure 1 one process or a plurality of processes and / or blocks Figure 1 or steps of the functions specified in a block or a plurality of blocks.

[0050] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic creative concept. Therefore, the appended claims are intended to be construed to cover the preferred embodiments as well as all changes and modifications falling within the scope of the present invention.

[0051] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. Port internal and external vehicle positioning system, characterized in that, It includes the following modules: Dynamic Anchor Cooperative Construction Module, Anti-interference Environment Feature Extraction Module, Fleet Cooperative Error Suppression Module, and Anchor Spatiotemporal Prediction Module; The Dynamic Anchor Cooperative Construction Module is used to obtain the dynamic position data of port mobile devices in real time, dynamically calculate the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment, and generate anchor activation instructions and anchor coverage blind area information by combining vehicle historical trajectory clustering analysis; The Anti-interference Environment Feature Extraction Module is used to receive the anchor coverage blind area information, perform local gray entropy segmentation on the ground images collected by the on-vehicle camera, screen high-distinguishability texture regions, and exclude dynamic interference through cross-frame Bayesian probability association to generate pose estimation data that is spatiotemporally aligned with the anchor signal; The Fleet Cooperative Error Suppression Module is used to receive the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model in combination with the inter-vehicle communication topology relationship, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy; The Anchor Spatiotemporal Prediction Module is used to analyze the operation plan of mobile devices in the port scheduling system, construct a spatiotemporal reachability prediction model in combination with the motion state of the corrected vehicle pose, predict the spatiotemporal distribution matrix of reachable anchors in the future, and generate control instructions according to the anchor activation priority in the matrix and feedback them to the Dynamic Anchor Cooperative Construction Module.

2. The port internal and external vehicle fleet positioning system according to claim 1, wherein: The specific process of obtaining the dynamic position data of port mobile devices in real time and dynamically calculating the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment is as follows: Parse the dynamic position data of port mobile devices in real time to obtain the real-time coordinates and motion directions of the gantry crane boom and temporary traffic equipment; According to the metal obstacle distribution density in the container stacking area, segmentally adjust the attenuation coefficient of the signal propagation model to generate a non-uniform attenuation gradient. According to the corrected attenuation gradient, calculate the effective coverage boundary of the signal for each anchor point, generate a vector map described by the polygon vertex coordinates, match the real-time vehicle position with the vector map, mark the area where the signal strength is lower than the positioning threshold as the coverage blind area, and output the list of blind area vertex coordinates.

3. The port internal and external vehicle fleet positioning system according to claim 2, characterized in that: The specific process of generating anchor activation instructions and anchor coverage blind area information by combining vehicle historical trajectory clustering analysis is as follows: Collect vehicle historical trajectory data and identify high-frequency passing areas, including container transfer areas and gate channels, through density clustering algorithms; Generate anchor activation priority instructions according to the clustering results, and preferentially activate the anchor points of mobile devices in high-frequency areas; Match the coordinates of the coverage blind area with the real-time vehicle position. When the vehicle enters the blind area boundary, activate the adjacent mobile device anchor points according to the blind area vertex coordinates and extend their working hours. Update the anchor activation timing table according to the real-time vehicle position and adjust the signal coverage range.

4. The port internal and external vehicle positioning system according to claim 3, characterized in that: The specific process of performing local gray entropy segmentation on the ground images collected by the on-vehicle camera and screening high-distinguishability texture regions is as follows: Within the coverage blind area, the ground images collected by the on-vehicle camera are divided into multiple square local blocks, the local gray entropy values of each block are calculated, and the blocks with gray entropy values higher than the environmental background are selected as candidate texture regions. The entropy value changes of the same blocks in adjacent frame images are compared, and the abnormally high entropy blocks caused by instantaneous reflection or projection are excluded; Perform morphological closing operation on the candidate texture regions to fill the broken regions and generate a continuous positioning reference mask.

5. The port internal and external vehicle fleet positioning system according to claim 4, characterized in that: And the specific process of excluding dynamic interference through cross-frame Bayesian probability association and generating pose estimation data that is spatio-temporally aligned with the anchor signal is as follows: Perform spatio-temporal calibration on the stable texture features in multiple consecutive frames of images, record the motion trajectories of feature points, construct a cross-frame feature association probability matrix, calculate the association probability between the features in the current frame and the historical frames according to the Bayesian inference model, and exclude the features if the probability is lower than the dynamic interference threshold; Align the selected features with the time stamps of the anchor signals, solve the vehicle pose through the perspective transformation matrix, and fuse the pose estimation results of multiple frames to generate smoothed vehicle trajectory data.

6. The port internal and external vehicle fleet positioning system according to claim 5, wherein: The construction logic of receiving the pose estimation data and the vehicle inertial navigation data and constructing an error propagation chain model in combination with the vehicle-to-vehicle communication topology relationship is as follows:. According to the vehicle-to-vehicle wireless communication signal strength and distance threshold, dynamically establish a mesh communication topology, maintain the vehicle node connection status table in real time, map the cumulative error of the single-vehicle inertial navigation to the nodes of the directed graph, and construct the error propagation weight through the pose difference between adjacent vehicles; According to the consistency of the vehicle movement direction, calculate the transmission attenuation coefficient of the error along the communication link, generate the error propagation chain matrix, and recalculate the error propagation weight and attenuation coefficient when the vehicle position or communication topology changes.

7. The port internal and external vehicle fleet positioning system according to claim 6, characterized in that: The specific process of correcting the cumulative positioning error through the multi-vehicle confidence weight fusion strategy is as follows: Dynamically allocate confidence weights according to the alignment degree between the vehicle pose estimation data and the anchor signal and the historical positioning stability; If the pose difference between adjacent vehicles exceeds the set threshold, reduce the weight of the abnormal vehicle and trigger local texture matching. Based on the error propagation chain matrix, fuse the pose data of multiple vehicles through the weighted average algorithm to suppress the single-vehicle error; Feed back the fused pose to each vehicle node to update the initial state of the inertial navigation.

8. The port internal and external vehicle fleet positioning system according to claim 7, characterized in that: The construction logic of parsing the operation plan of the mobile device in the port scheduling system and constructing a spatio-temporal reachability prediction model in combination with the motion state of the corrected vehicle pose is as follows: Extract the operation schedule and path planning data of the mobile device in the port scheduling system, including the moving trajectory of the gantry crane boom and the deployment period of temporary equipment; According to the corrected pose data, construct a vehicle kinematic model, predict the spatio-temporal regions that the vehicle can reach in the future period, perform an intersection operation on the predicted vehicle trajectory and the spatio-temporal distribution of the mobile device anchor points, and generate a spatio-temporal reachability probability distribution map; Generate a list of anchor activation priorities according to the spatio-temporal density of the intersection area, and mark the high-probability reachable anchor points.

9. The port internal and external vehicle fleet positioning system according to claim 8, characterized in that: The specific process of predicting the spatio-temporal distribution matrix of the anchor points that can be reached in the future and generating control instructions according to the anchor activation priorities in the matrix and feeding them back to the dynamic anchor point collaborative construction module is as follows: Quantify the spatio-temporal reachability probability distribution map into a matrix with dimensions of time - space - anchor identifier. Sort according to the probability values of spatio-temporal nodes in the matrix to generate an anchor activation priority instruction, and preferentially wake up the anchors with high probability of being reached. Feed the priority instruction back to the dynamic anchor collaborative construction module through the port wireless private network, trigger the anchor wake-up or sleep operation, and monitor the anchor activation effect in real time. If the vehicle does not reach the anchor as predicted, dynamically adjust the priority instruction and regenerate the matrix.

10. A method for positioning internal and external vehicle fleets in a port, applied to the port internal and external vehicle fleet positioning system according to any one of claims 1-9, characterized in that, It includes the following steps: S1. Obtain the dynamic position data of port mobile devices in real time, dynamically calculate the anchor signal coverage range according to the signal propagation attenuation characteristics in the metal environment, and generate an anchor activation instruction and anchor coverage blind area information by combining vehicle historical trajectory clustering analysis. S2. Receive the anchor coverage blind area information, perform local gray entropy segmentation on the ground images collected by the on-vehicle camera, screen the high-discrimination texture regions, and exclude dynamic interference through cross-frame Bayesian probability association to generate pose estimation data that is spatio-temporally aligned with the anchor signal. S3. Combine the pose estimation data and vehicle inertial navigation data, construct an error propagation chain model based on the vehicle - to - vehicle communication topology relationship, and correct the cumulative positioning error through a multi-vehicle confidence weight fusion strategy. S4. Analyze the operation plan of mobile devices in the port scheduling system, construct a spatio-temporal reachability prediction model by combining the motion state of the calibrated vehicle pose, predict the spatio-temporal distribution matrix of the future reachable anchors, and generate a control instruction according to the anchor activation priority in the matrix and feedback it to step S1.

Images