Parking self-service payment and car finding method based on mobile phone application

Abstract

The invention discloses a parking self-service payment and vehicle searching method based on a mobile phone application, and belongs to the technical field of parking lot management, and the method comprises the steps: collecting a parking lot environment image through a mobile phone camera after the parking payment is carried out through the mobile phone application, generating a first image sequence, carrying out the denoising and feature extraction of the first image sequence, and obtaining a second image sequence; obtaining a second image sequence; determining boundary coordinates of the crowded area, generating a crowded area distribution map, calculating an optimal vehicle searching path from the current position to the target vehicle in combination with the parking lot three-dimensional model, and generating a first path sequence avoiding the boundary coordinates of the crowded area; superposing the first path sequence in the second image sequence to obtain a third image sequence; and presenting a navigation interface corresponding to the third image sequence on the mobile phone application by adopting a real-time rendering technology. The parking self-service payment and vehicle searching method based on the mobile phone application solves the problems that based on an existing parking lot management technology, the vehicle searching efficiency is low, and local congestion is likely to be caused.

Description

Technical Field

[0001] The present invention relates to the technical field of parking lot management, and particularly to a method for self-service parking payment and vehicle searching based on a mobile application. Background Art

[0002] Parking lot management, as an important part of modern urban transportation, is crucial for improving travel efficiency and user experience. With the rapid increase in the number of private cars, the need for quick payment and accurate vehicle searching in parking lots has become increasingly urgent. Especially in complex parking environments such as large shopping malls and airports, efficient parking services have become a key indicator for measuring the intelligent level of a city. Currently, traditional self-service parking payment methods are like the Chinese invention patent with the application number 202111044807.1, "A Method for Self-Service Parking Payment and Vehicle Searching Based on a Mobile Terminal". However, most existing parking management methods rely on simple navigation guidance and are difficult to meet the actual needs of users to quickly find their vehicles in complex parking lots.

[0003] Traditional navigation systems usually only provide two-dimensional map guidance, lacking intuitiveness and real-time nature. Users are prone to getting lost in the parking lot, and existing systems cannot adjust routes according to real-time pedestrian dynamics, resulting in low vehicle searching efficiency and even causing local congestion. Summary of the Invention

[0004] In order to overcome the defects existing in the prior art, the present invention provides a method for self-service parking payment and vehicle searching based on a mobile application to solve the above problems.

[0005] The technical solution adopted by the present invention to solve its technical problems is: A method for self-service parking payment and vehicle searching based on a mobile application, comprising the following steps: S1: After parking payment is made through a mobile application, the parking lot environment image is collected by a mobile phone camera to generate a first image sequence, and an image processing algorithm is used to denoise and extract features from the first image sequence to obtain a second image sequence; S2: Determine the boundary coordinates of the crowded area to generate a crowded area distribution map; for the crowded area distribution map, in combination with the three-dimensional model of the parking lot, calculate the best vehicle searching path from the current position to the target vehicle, and generate a first path sequence that avoids the boundary coordinates of the crowded area; S3: Superimpose the first path sequence on the second image sequence to obtain a third image sequence; S4: Use real-time rendering technology to present the navigation interface corresponding to the third image sequence on the mobile application.

[0006] It should be noted that in the step S1, the Gaussian filtering algorithm is used to denoise the first image sequence to obtain a denoised image. For each pixel in the denoised image, if the difference between its pixel value and a preset threshold exceeds a preset value, the neighborhood mean of the pixel is calculated, and the pixel value is replaced with the neighborhood mean to obtain a second image sequence.

[0007] Optionally, in the step S2, real-time pedestrian flow distribution data is obtained from a parking lot sensor. The pedestrian flow distribution data includes multiple position point coordinates and corresponding pedestrian flow densities. The DBSCAN algorithm is used to perform clustering analysis on the pedestrian flow distribution data, and the boundary coordinates of the crowded area are obtained by determining the crowded area through a preset distance threshold and density threshold. According to the boundary coordinates of the crowded area, a crowded area distribution map is generated, and the position and range of each crowded area are marked on the crowded area distribution map.

[0008] Specifically, in the step S2, a parking lot grid map is constructed for the three-dimensional model of the parking lot. According to the channels and target vehicle position coordinates in the parking lot grid map, the A* algorithm is used to calculate the car-finding path from the current position to the target vehicle position. In the A* algorithm, the car-finding path from the current position to the target vehicle position is optimized by assigning a higher cost value to the crowded area than to the non-crowded area to avoid the boundary coordinates of the crowded area, and a first path sequence is obtained. The first path sequence includes multiple path point coordinates.

[0009] Preferably, in the step S2, the steps of constructing a parking lot grid map model for the three-dimensional model of the parking lot include: Point cloud data is extracted from the three-dimensional model of the parking lot to obtain a three-dimensional coordinate set of the wall and column. The point cloud data is discretized by using a regular cubic grid to divide the point cloud data into multiple grid cells. Each grid cell is recorded in the grid vertex coordinate set with the grid fixed-point coordinates. The grid cells corresponding to the wall or the main body are marked as occupied occupancy states, and the grid cells corresponding to the vacant area are marked as free occupancy states, generating a grid vertex coordinate set with occupancy state marks. A parking lot grid map is constructed according to the grid vertex coordinate set with occupancy state marks, and the area formed by multiple continuously occupied state-marked free grid cells is identified as a channel.

[0010] Specifically, in the step S3, according to the relative distance from the current path point to the next path point in the arrangement of the path point coordinates in the first path sequence, the relative distance is superimposed on the second image sequence by using augmented reality technology to generate a third image sequence.

[0011] Specifically, in the step S3, extract the coordinates of the key nodes that need to be turned in the navigation. If the user is approaching a key node, calculate the deflection angle when the user approaches the key node, add the deflection angle to the second image sequence, and generate a third image sequence in combination with the relative distance.

[0012] Optionally, in the step S3, the steps for generating the deflection angle include: Extract the center lines of the channels from the parking lot grid map, and use the intersection points of multiple center lines and the extreme points of the curvature when the center line is a curve as key nodes; Obtain the real-time positioning coordinates of the user and the user's movement direction vector; calculate the Euclidean distance from the real-time positioning coordinates of the user to each key node. If the minimum distance is less than a preset threshold, it is determined as an adjacent node; According to the user's movement direction vector and the direction vector from the adjacent node to the next key node, calculate the deflection angle through the cross product of three-dimensional space vectors; wherein, in the first path sequence, use the path point coordinates closest to the adjacent node as the first path point coordinates, use the next path point coordinates of the adjacent path point coordinates in the first path sequence as the second path point coordinates, and use the key node closest to the second path point coordinates as the next key node corresponding to the adjacent node.

[0013] The beneficial effects of the present invention are as follows: In the method and system for parking self-service payment and car finding based on a mobile application, the parking lot environment image is collected through the mobile phone camera, and image processing algorithms are used for denoising and feature extraction. After generating the crowded area distribution map, in combination with the three-dimensional model of the parking lot, calculate the best car-finding path to avoid the crowded area, and use augmented reality technology to dynamically display the first path sequence in the parking lot environment image collected by the mobile phone camera. Finally, use real-time rendering technology to present the navigation interface on the mobile application, providing accurate car-finding path guidance for users. The present invention effectively solves the problem of difficult car finding in large parking lots by integrating image processing and path planning, improves the user's car-finding efficiency, and improves the parking experience. Description of the Drawings

[0014] Figure 1 It is a flowchart of the method for parking self-service payment and car finding based on a mobile application in an embodiment of the present invention; Figure 2 It is a sub-step flowchart of step S2 in an embodiment of the present invention; Figure 3 It is a flowchart of the steps for generating the deflection angle in an embodiment of the present invention. Detailed Embodiment

[0015] The following further describes the specific embodiments of the present invention in conjunction with the accompanying drawings. It should be noted here that the description of these embodiments is used to help understand the present invention, but does not constitute a limitation to the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0016] As Figures 1-3 shown, a method for self-service parking payment and vehicle finding based on a mobile application includes the following steps: S1: After parking payment is made through the mobile application, the mobile phone camera is used to collect the parking lot environment image to generate a first image sequence. An image processing algorithm is used to denoise and extract features from the first image sequence to obtain a second image sequence; in this solution, parking payment through the mobile application is implemented using the traditional self-service parking payment method; S2: Determine the boundary coordinates of the crowded area to generate a crowded area distribution map; for the crowded area distribution map, in combination with the three-dimensional model of the parking lot, calculate the best vehicle finding path from the current position to the target vehicle, and generate a first path sequence that avoids the boundary coordinates of the crowded area; S3: Superimpose the first path sequence on the second image sequence to obtain a third image sequence; S4: Use real-time rendering technology (Real-time Rendering) to present the navigation interface corresponding to the third image sequence on the mobile application to complete the vehicle finding path guidance. According to the third image sequence, each frame of the image is rendered through a real-time generation mechanism, and image data is output on the mobile application.

[0017] In the system of the method for self-service parking payment and vehicle finding based on a mobile application, the mobile phone camera is used to collect the parking lot environment image, and image processing algorithms are used for denoising and feature extraction. After generating the crowded area distribution map, in combination with the three-dimensional model of the parking lot, the best vehicle finding path that avoids the crowded area is calculated, and the augmented reality technology is used to dynamically display the first path sequence in the parking lot environment image collected by the mobile phone camera. Finally, the real-time rendering technology is used to present the navigation interface on the mobile application to provide accurate vehicle finding path guidance for users. The present invention effectively solves the problem of difficult vehicle finding in large parking lots by integrating image processing and path planning, improves the vehicle finding efficiency of users, and improves the parking experience.

[0018] Preferably, in the step S1, a Gaussian filtering algorithm is used to denoise the first image sequence to obtain a denoised image; for each pixel in the denoised image, if the difference between its pixel value and the preset threshold exceeds the preset value, the neighborhood mean value of the pixel is calculated, and the pixel value is replaced with the neighborhood mean value to obtain a second image sequence.

[0019] For example, when a mobile phone camera captures images of the parking lot environment to generate a first image sequence, assume that the light is dim in an underground parking lot. The camera captures color images with a resolution of 1280x720 at a rate of 10 frames per second, generating an image sequence containing walls, parking lines, and vehicles. To ensure image quality, the noise problem in low-light environments needs to be considered. During the acquisition process, the automatic exposure function of the mobile phone can increase the image brightness, but it may introduce Gaussian noise, affecting subsequent processing. In one possible implementation, a Gaussian filtering algorithm is used to denoise the first image sequence, generating a denoised image. Gaussian filtering smooths the noise while retaining edge details by weighted averaging the neighborhood of each pixel. Assume a 5x5 Gaussian kernel with a standard deviation of 1.5 is used. After processing, the noise in the image is significantly reduced, and the edges of the parking lines are clearer. This step can improve the accuracy of subsequent feature extraction and avoid noise interference in key point detection.

[0020] Specifically, the pixel values of the denoised image need to be compared with a preset threshold. Pixels whose difference exceeds the preset value are replaced with the neighborhood mean value, generating a second image sequence. Assume the pixel value range is 0 - 255 and the threshold is 200. If a pixel value reaches 230, indicating possible specular interference, then calculate the 3x3 neighborhood mean value, such as 180, and use the neighborhood mean value to replace this pixel value, replacing 230 with 180. This process effectively removes local outliers, such as reflections or stains, ensuring image consistency and providing a reliable input for feature extraction.

[0021] Optionally, in step S2, real-time pedestrian flow distribution data is obtained from parking lot sensors. The pedestrian flow distribution data includes multiple location point coordinates and corresponding pedestrian flow densities. The DBSCAN algorithm is used to perform clustering analysis on the pedestrian flow distribution data. The boundaries of the crowded areas are determined by preset distance thresholds and density thresholds, obtaining the boundary coordinates of the crowded areas. According to the boundary coordinates of the crowded areas, a crowded area distribution map is generated. The crowded area distribution map marks the locations and ranges of each crowded area.

[0022] Parking lot sensors are deployed in key areas of the parking lot, such as near the entrance, passageways, and parking spaces, to collect location point coordinates and corresponding pedestrian flow densities. Assume the parking lot area is 5000 square meters and 100 sensors are deployed, generating data points containing the collected location point coordinates and corresponding pedestrian flow densities per second. Each data point includes three-dimensional coordinates, such as (10, 20, 1), and a density value, such as 50 people per square meter. This method can reflect the pedestrian flow distribution in real time and provide a data basis for subsequent analysis.

[0023] In a possible implementation, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm is used to perform clustering analysis on the pedestrian flow distribution data. The DBSCAN algorithm identifies dense regions by setting a distance threshold and a density threshold. Suppose the distance threshold is 2 meters and the density threshold is 30 people per square meter. After obtaining a data point with a density value exceeding the density threshold, data points within the distance threshold from this data point and with density values exceeding the density threshold are obtained with this data point as the center. Then, with the newly obtained data points as the center, other data points meeting the distance threshold and density threshold are obtained, and so on. Through this clustering method, relevant data points can be grouped into a dense region and boundary coordinates are generated. For example, the boundary coordinates of region A are [(8, 18, 1), (12, 22, 1)]. Specifically, the boundary coordinates can be mapped onto the parking lot floor plan to generate a crowded area distribution map. The crowded areas of the pedestrian flow are marked for their positions and ranges through the boundary coordinates. For example, region A is marked for its position and range through the boundary coordinates (8, 18, 1) and (12, 22, 1).

[0024] Specifically, in step S2, a parking lot grid map is constructed for the three-dimensional model of the parking lot. According to the channels and target vehicle position coordinates in the parking lot grid map, the A* algorithm is used to calculate the car-finding path from the current position to the target vehicle position; In the A* algorithm, the car-finding path from the current position to the target vehicle position is optimized by assigning a higher cost value to the crowded areas of the pedestrian flow than to the non-crowded areas of the pedestrian flow, so as to avoid the boundary coordinates of the crowded areas of the pedestrian flow, and a first path sequence is obtained. The first path sequence includes multiple path point coordinates.

[0025] In a possible implementation, the A* algorithm is used to calculate the path for finding the vehicle from the current position to the target vehicle. Assume the current position is (10, 10, 1). A grid map model of the parking lot is constructed for the three-dimensional model of the parking lot. The path cost matrix is calculated based on the grid map. A high cost value is assigned to the crowded area of people flow according to the boundary coordinates of the crowded area of people flow, and a low cost value is assigned to the unobstructed area, forming the basic data for path selection. The g value and h value of the path node are calculated. The g value represents the actual cost from the starting point to the current path node. The h value estimates the expected cost from the current path node to the target using the Manhattan distance. The f value is the sum of the g value and the h value. The core search process of the A* algorithm is executed. Starting from the starting point, if the current path node is not the target node (i.e., not the position of the target vehicle), the adjacent passable nodes are expanded, the f values of each node are calculated, and the node with the smallest f value is selected as the next visited node. Since the cost value of the crowded area of people flow is higher than that of the unobstructed area, the f value of the node with the unobstructed area as the node will be lower than the f value of the node with the crowded area of people flow as the node, so as to avoid the crowded area of people flow. The generated path sequence is like points [(10, 10, 1), (15, 15, 1), (50, 30, 1)], and the path length is about 50 meters.

[0026] It should be noted that in the step S2, the steps of constructing a grid map model of the parking lot for the three-dimensional model of the parking lot include: Extract the point cloud data from the three-dimensional model of the parking lot to obtain the three-dimensional coordinate set of the wall and the column; Extracting the point cloud data from the three-dimensional model of the parking lot is the basis for constructing the navigation system. For example, the parking lot generates point cloud through lidar scanning, including the three-dimensional coordinates of structures such as walls and columns. Assume the parking lot area is 5000 square meters, and the point cloud data contains 100 points per cubic meter. The wall coordinates are like (5, 10, 2), and the column coordinates are like (15, 20, 3). These coordinates reflect the spatial structure of the parking lot and provide an accurate reference for subsequent processing; The point cloud data is discretized using a regular cube grid to divide the point cloud data into multiple grid cells. Each grid cell is recorded in the grid vertex coordinate set with the grid fixed-point coordinates. The grid cells corresponding to the wall or the main body are marked as the occupied occupancy state, and the grid cells corresponding to the vacant area are marked as the idle occupancy state, generating a grid vertex coordinate set with occupancy state marks; In a possible implementation, the point cloud data is discretized by a regular cube grid. Assume the grid side length is 0.5 meters, and the point cloud is divided into 10 million grid cells. Each grid vertex records coordinates such as (5.5, 10.5, 2.5) and the occupancy state. The wall grid is marked as "occupied", and the vacant area is marked as "idle". This discretization method simplifies the complex point cloud into a regular grid, facilitating map construction; Construct a parking lot grid map based on the set of grid vertex coordinates with occupancy status markers, and identify the area formed by multiple continuously unoccupied grid cells as a passage. For example, generate a grid map based on the occupancy status. The map resolution is 0.5 meters. The grid cells with the occupancy status marker of occupied are shown in black, and the grid cells with the occupancy status marker of unoccupied are white. Arrange the grid cells according to the grid vertex coordinates to obtain the grid map. The continuously unoccupied grid cells are identified as passages, such as a passage area with a width of 2 meters. This grid map intuitively reflects the passable area and provides a basis for path planning.

[0027] Preferably, in the step S3, according to the relative distance from the current path point to the next path point arranged according to the path point coordinates in the first path sequence, use augmented reality technology to superimpose the relative distance on the second image sequence to generate a third image sequence.

[0028] Specifically, in the step S3, extract the coordinates of the key nodes that need to turn in the navigation. If the user is approaching a key node, calculate the deflection angle when the user approaches the key node, and superimpose the deflection angle on the second image sequence, and combine with the relative distance to generate a third image sequence.

[0029] After obtaining each frame of image in the second image sequence, generate a basic image data stream. For the basic image data stream, determine the relative distance data and deflection angle data corresponding to each frame of image according to the positioning of the mobile phone, and use information fusion technology to superimpose the relative distance and deflection angle data on the basic image data stream to generate a fused image stream. For the fused image stream, use augmented reality technology to perform visual rendering processing on each frame of image to generate a third image sequence.

[0030] It should be noted that in the step S3, the generation step of the deflection angle includes: Extract the center line of the passage from the parking lot grid map, and use the intersection points of multiple center lines and the extreme value points of the curvature when the center line is a curve as key nodes; specifically, the center line of the passage is extracted through image processing. Assume that passage A extends from (10, 10, 1) to (50, 10, 1), and the center line coordinates of passage A are [(10, 10, 1), (30, 10, 1), (50, 10, 1)]. Passage B extends from (30, 0, 1) to (30, 30, 1), and the center line coordinates of passage B are [(30, 0, 1), (30, 10, 1), (30, 30, 1)]. The intersection point (30, 10, 1) of the center lines of passage A and passage B is marked as a key node. In addition, for the center line that is a curve, mark the extreme value point of the curvature of the curve as a key node. These nodes simplify the path calculation; Obtain the user's real-time positioning coordinates and the user's motion direction vector; calculate the Euclidean distance from the user's real-time positioning coordinates to each key node. If the minimum distance is less than the preset threshold, it is determined as a nearby node. For example, the user's real-time positioning coordinates are obtained through UWB (Ultra Wide Band) technology. Suppose the user's real-time positioning coordinates are (12, 10, 1) and the user's motion direction vector is (1, 0, 0). Calculate the Euclidean distance from the user's real-time positioning coordinates to the key node A(10, 10, 1) is 2 meters, which is less than the threshold of 5 meters, and determine the key node A as a nearby node. This positioning method ensures accurate navigation; Calculate the deflection angle through the cross product of three-dimensional space vectors according to the user's motion direction vector and the direction vector from the nearby node to the next key node; among them, in the first path sequence, take the path point coordinate closest to the nearby node as the first path point coordinate, take the next path point coordinate of the nearby path point coordinate in the first path sequence as the second path point coordinate, and take the key node closest to the second path point coordinate as the next key node corresponding to the nearby node. Specifically, calculate the deflection angle through the cross product of vectors. Suppose the direction vector from the key node A as the nearby node to the next key node B is (0, 1, 0), and the cross product result with the user's motion direction vector (1, 0, 0) indicates that a left turn of 90 degrees is required. This angle calculation supports accurate turning prompts.

[0031] The above has described the embodiments of the present invention in detail in conjunction with the accompanying drawings, but the present invention is not limited to the described embodiments. For those skilled in the art, without departing from the principle and spirit of the present invention, various changes, modifications, substitutions, and variations made to these embodiments still fall within the protection scope of the present invention.

Claims

1. A method for self-service payment and car finding of parking based on a mobile phone application, characterized in that, Including the following steps: S1: After paying for parking through the mobile application, use the mobile phone camera to collect the parking lot environment images, generate the first image sequence, and use the image processing algorithm to denoise and extract features from the first image sequence to obtain the second image sequence; S2: Determine the boundary coordinates of the crowded area and generate a crowded area distribution map; for the crowded area distribution map, combine it with the 3D model of the parking lot, calculate the best car-finding path from the current position to the target vehicle, and generate the first path sequence that avoids the boundary coordinates of the crowded area; S3: Superimpose the first path sequence on the second image sequence to obtain the third image sequence; S4: Use real-time rendering technology to present the navigation interface corresponding to the third image sequence on the mobile application.

2. A method for self-service parking payment and car finding based on a mobile application according to claim 1, characterized in that: In the step S1, use the Gaussian filtering algorithm to denoise the first image sequence to obtain the denoised image; for each pixel in the denoised image, if the difference between its pixel value and the preset threshold exceeds the preset value, calculate the neighborhood mean value of the pixel and replace the pixel value with the neighborhood mean value to obtain the second image sequence.

3. A method for self-service parking payment and car finding based on a mobile application according to claim 1, characterized in that: In the step S2, obtain the real-time pedestrian flow distribution data from the parking lot sensors, and the pedestrian flow distribution data includes multiple position point coordinates and the corresponding pedestrian flow density; Use the DBSCAN algorithm to perform clustering analysis on the pedestrian flow distribution data, determine the crowded areas through the preset distance threshold and density threshold, and obtain the boundary coordinates of the crowded areas; According to the boundary coordinates of the crowded areas, generate a crowded area distribution map, and the crowded area distribution map marks the positions and ranges of each crowded area.

4. A method for self-service parking payment and car finding based on a mobile application according to claim 3, characterized in that: In the step S2, construct a parking lot grid map for the 3D model of the parking lot, and use the A* algorithm to calculate the car-finding path from the current position to the target vehicle position according to the channels and target vehicle position coordinates in the parking lot grid map; In the A* algorithm, optimize the car-finding path from the current position to the target vehicle position by assigning a higher cost value to the crowded areas than to the non-crowded areas to avoid the boundary coordinates of the crowded areas, and obtain the first path sequence, and the first path sequence includes multiple path point coordinates.

5. A method for self-service parking payment and car finding based on a mobile application according to claim 4, characterized in that: In the step S2, the steps of constructing a parking lot grid map model for the 3D model of the parking lot include: Extract the point cloud data from the 3D model of the parking lot to obtain the 3D coordinate set of the walls and columns; Use regular cubic grids to discretize the point cloud data to divide the point cloud data into multiple grid cells, and each grid cell is recorded in the grid vertex coordinate set with the grid fixed-point coordinates. Mark the grid cells corresponding to the walls or the main body as the occupied occupancy state, and mark the grid cells corresponding to the vacant areas as the idle occupancy state to generate a grid vertex coordinate set with occupancy state marks; Construct a parking lot grid map according to the grid vertex coordinate set with occupancy state marks, and identify the area formed by multiple consecutive grid cells with the occupancy state marked as idle as the channel.

6. The method for self-service parking payment and car finding based on a mobile application according to claim 5, characterized in that: In the step S3, according to the relative distance from the current path point to the next path point based on the arrangement of the path point coordinates in the first path sequence, the relative distance is superimposed on the second image sequence by using the augmented reality technology to generate a third image sequence.

7. A method for self-service parking payment and car finding based on a mobile application according to claim 6, characterized in that: In the step S3, the coordinates of the key nodes that need to be turned in the navigation are extracted. If the user is approaching a key node, the deflection angle when the user approaches the key node is calculated, and the deflection angle is superimposed on the second image sequence to generate a third image sequence in combination with the relative distance.

8. A method for self-service parking payment and car finding based on a mobile application according to claim 7, characterized in that: In the step S3, the generating step of the deflection angle includes: Extract the center line of the channel from the parking lot grid map, and use the intersection points of multiple center lines and the extreme value points of the curvature when the center line is a curve as key nodes; Obtain the real-time positioning coordinates of the user and the user's movement direction vector; calculate the Euclidean distance from the real-time positioning coordinates of the user to each key node. If the minimum distance is less than a preset threshold, it is determined as an adjacent node; According to the user's movement direction vector and the direction vector from the adjacent node to the next key node, calculate the deflection angle through the cross product of three-dimensional space vectors; among them, in the first path sequence, the path point coordinate closest to the adjacent node is used as the first path point coordinate, the next path point coordinate adjacent to the adjacent path point coordinate in the first path sequence is used as the second path point coordinate, and the key node closest to the second path point coordinate is used as the next key node corresponding to the adjacent node.

Images