Unmanned aerial vehicle autonomous path planning method based on visual perception module

Abstract

The invention discloses an unmanned aerial vehicle autonomous path planning method based on a visual perception module, and relates to the technical field of unmanned aerial vehicle inspection. Comprising the following steps: 1, collecting coordinates and image features of an inspection target and an obstacle in an inspection area, and building an environment model of the unmanned aerial vehicle inspection area; 2, based on the environment model, according to the maneuvering characteristics of the unmanned aerial vehicle, the collision risk level of the obstacle area is evaluated; step 3, based on the risk level, establishing a fitness function of the unmanned aerial vehicle safety planning path; and step 4, based on an improved particle swarm algorithm, optimizing the fitness function, and obtaining an optimal planning path of the current inspection area. According to the invention, the safety and reliability of the unmanned aerial vehicle during inspection task execution are improved.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle inspection technology, and in particular to an unmanned aerial vehicle autonomous path planning method based on a visual perception module. Background Art

[0002] At the forefront of modern technology, unmanned aerial vehicles (UAVs) have become a key tool for cross-industry applications, from agricultural monitoring to emergency rescue, to urban planning and environmental science, and their versatility and efficiency are widely recognized. However, autonomous navigation of UAVs in complex environments faces severe challenges, especially in signal-limited areas such as urban canyons and dense forests, where traditional GPS and preset path planning methods often fail to ensure real-time, accurate positioning and dynamic obstacle avoidance.

[0003] In recent years, advances in visual perception technology have opened up new avenues for autonomous flight of drones. Equipped with high-resolution cameras and advanced image processing algorithms, drones are able to capture and analyze environmental data in real time, including terrain features, obstacle locations, and moving target trajectories. This capability greatly improves the adaptability of drones in unknown or rapidly changing environments, enabling them to instantly plan safe and effective flight routes, thereby maintaining stable operation under various conditions. However, to achieve truly autonomous flight, a series of technical challenges still need to be overcome, including but not limited to real-time processing and analysis of visual data, precise positioning and navigation in complex environments, real-time avoidance of dynamic obstacles, and efficient communication with the cloud or ground stations.

[0004] In the prior art, such as the patent with application number 202410185850.7 and name “A UAV path planning algorithm for detecting power tower insulators”, path planning is performed for insulators on pole towers, and the scenario is single; and the particle swarm algorithm used for solving the problem is a basic particle swarm algorithm, which has the limitations of slow convergence speed and easy to fall into local optimality.

[0005] For example, the application number is 201811583287.X, and the name is “A global planning method for multi-objective particle swarm of three-dimensional trajectory of unmanned aerial vehicle”. The grid method is used for subsequent calculations with discrete map information. However, the grid method brings the problem of high computational complexity. The computing power required for instant obstacle avoidance of unmanned aerial vehicles is too high. There is a lack of research on the impact of the flight state of unmanned aerial vehicles on the accuracy of path planning, and there is a lack of research on path planning under dynamic obstacles, resulting in low reliability. Summary of the invention

[0006] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0007] The technical solution of the present invention is: a method for autonomous path planning of a UAV based on a visual perception module, comprising the following steps:

[0008] Step 1: Collect the coordinates and image features of the inspection targets and obstacles in the inspection area, and build an environmental model of the drone inspection area;

[0009] Step 2: Based on the environmental model and the maneuverability characteristics of the drone, assess the collision risk level in the obstacle area;

[0010] Step 3: Based on the risk level, establish the fitness function of the drone’s safe path planning;

[0011] Step 4: Based on the improved particle swarm algorithm, optimize the fitness function to obtain the optimal planning path for the current inspection area.

[0012] In step 1, a drone equipped with a high-resolution binocular camera is used to perform a full-scale scan of the inspection area to obtain high-definition image and video data. At the same time, edge detection image processing technology is used to extract feature information of inspection targets and obstacles from the image, including shape and size.

[0013] Use GPS positioning technology to record the flight position coordinates of the drone, the inspection target coordinates, and the geographic location coordinates of obstacles;

[0014] Using the collected images and coordinate data, positioning and mapping technology are used to combine the coordinates with image features to form an environmental model.

[0015] Step 1 also includes: normalizing and cross-validating the collected image and coordinate data, establishing a data index and retrieval mechanism, and storing the relevant data in the inspection database.

[0016] In step 2, the maneuverability of the UAV is evaluated based on its maneuverability characteristics, and then combined with the environmental model, the collision risk level in the obstacle area is evaluated.

[0017] Collect the basic performance parameters of the inspection drone, including maximum flight speed, maximum acceleration, minimum turning radius, maximum climb rate and descent rate.

[0018] in,

[0019]

[0020] In the formula, is the flight speed of the UAV during period t, v max is the maximum speed of the drone, The flight speed of the drone during the period t+Δt, is the acceleration of the UAV during the period of Δt;

[0021] r t UAVis the turning radius of the UAV at time t, r min is the minimum turning radius of the drone, are the climbing rate and descending rate of the UAV in the period of Δt, is the flight altitude of the UAV during the period t+Δt, is the flight altitude of the UAV during period t;

[0022] Analyze the maneuverability of the drone in different flight modes, such as hovering, straight flight, and turning, and evaluate its maneuverability based on the maneuverability range:

[0023]

[0024] Among them, 1, 2, and 3 are the levels of mobility.

[0025] The environmental features in the inspection area are modeled into similar mathematical models according to the edge features, and the minimum safe distance d between the drone and any point in the three-dimensional space is calculated. t , expressed as:

[0026]

[0027] in, is the coordinate of the center point of the drone; (x, y, z) is any point on the plane; R is the radius of the circle when the drone is mathematically modeled;

[0028] The risk level of the flight area is assessed by combining the speed, turning radius, and minimum safe distance between the drone and obstacles during the t period.

[0029] in,

[0030]

[0031] Among them, Level is the risk level, L low , L mid , L high They are low risk, medium risk, and high risk, respectively. min is the maximum centripetal acceleration, d min is the minimum safe distance between the drone and obstacles.

[0032] The fitness function is:

[0033]

[0034] Among them, F fitness is the fitness function, and P is the flight power of the UAV.

[0035] In step 4, the constructed environment model is used as the inspection area simulation for the drone path planning. Random dynamic obstacles are added to the environment model. The visual perception module carried by the drone is used to detect new obstacles in the environment, and distance measurement, edge detection and model update are performed online.

[0036] Based on the particle swarm optimization algorithm and the maneuverability of the drone, it helps the drone avoid obstacles and choose a suitable flight route. According to the value of the fitness function and the algorithm iteration time, it evaluates the quality of the path planned by the algorithm and optimizes the flight strategy.

[0037] According to the complexity of the inspection area, the initial population size and particle position are set, and each particle represents a possible flight path;

[0038] Based on the local optimal solution and the global optimal solution, guide the search direction of the particle swarm;

[0039] According to the number of iterations and the current optimization situation, the inertia weight and cognitive factor are adaptively adjusted to balance the capabilities of global search and local search, and the speed and position of each particle are updated.

[0040] Step 4 includes:

[0041] S41: Initialize population size, maximum number of iterations, inertia weight and cognitive factor, as well as particle speed and position;

[0042] S42: The particle calculates the fitness value according to the constructed fitness function;

[0043] S43: updating the inertia weight and cognitive factor according to the improved inertia weight and cognitive factor formula; at the same time, introducing the disturbance factor; updating the speed and position of the particle according to the speed and improved position formula;

[0044] S44: Update individual optimal and global optimal: compare the current fitness value of the particle with the individual historical optimal value. If the current value is better, update the individual optimal position to the current position. Among the individual optimal positions of all particles, determine the global optimal position, and find the individual optimal position of the particle with the best fitness value as the global optimal position.

[0045] S45: Determine the termination condition. If the maximum number of iterations is reached, the algorithm terminates and outputs the global optimal position and fitness value; otherwise, return to S42 to continue iterating.

[0046] The present invention has the following beneficial effects:

[0047] 1) Environmental modeling is achieved through visual perception and GPS positioning, which not only improves the height simulation of the inspection area, but also enhances the drone's perception of environmental changes during flight, that is, the identification and positioning of dynamic obstacles, dynamically improving the flight safety of drones in complex environments;

[0048] 2) Based on the minimum safe distance between the drone and obstacles and the drone's maneuverability, the risk level of the drone inspection area is evaluated, and safety warnings and drone maneuverability coefficients are provided in path planning;

[0049] 3) Construct a fitness function based on flight safety and energy efficiency, take the maneuverability characteristics of the UAV and the risk level of the flight area as constraints, better cope with environmental changes, solve path planning based on the improved particle swarm algorithm, and achieve the robustness and stability of the planned path.

[0050] The present invention improves the safety and reliability of the unmanned aerial vehicle when performing inspection tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the specific implementation of the present invention or the technical solution in the prior art, the following briefly introduces the drawings required for the specific implementation or the prior art description. In the drawings, each part is not necessarily drawn according to the actual scale.

[0052] Figure 1 It is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0053] The present invention Figure 1 As shown, the autonomous path planning method for a UAV based on a visual perception module includes the following steps:

[0054] Step 1: Collect the coordinates and image features of the inspection targets and obstacles in the inspection area, and build an environmental model of the drone inspection area;

[0055] Step 2: Based on the environmental model and the maneuverability characteristics of the drone, assess the collision risk level in the obstacle area;

[0056] Step 3: Based on the risk level, establish the fitness function of the drone’s safe path planning;

[0057] Step 4: Based on the improved particle swarm algorithm, optimize the fitness function to obtain the optimal planning path for the current inspection area.

[0058] The present invention realizes environmental modeling through visual perception and GPS positioning, which not only improves the height simulation of the inspection area, but also enhances the UAV's perception of environmental changes during flight, that is, the recognition and positioning of dynamic obstacles, dynamically improving the flight safety of the UAV in complex environments;

[0059] Based on the minimum safe distance between the drone and obstacles and the drone's maneuverability, the risk level of the drone inspection area is evaluated, and safety warnings and drone maneuverability coefficients are provided in path planning;

[0060] A fitness function based on flight safety and energy efficiency is constructed, with the maneuverability characteristics of the UAV and the risk level of the flight area as constraints to better cope with environmental changes. The path planning is solved based on the improved particle swarm algorithm to achieve the robustness and stability of the planned path.

[0061] The specific steps are as follows:

[0062] Step 1: Collect and process the coordinates and images of inspection targets and obstacles in the inspection area, build an environmental model of the inspection area, and provide accurate environmental information for path planning:

[0063] Use the high-resolution binocular camera on the drone to scan the inspection area in all directions and obtain high-definition images and video data. At the same time, use edge detection image processing technology to extract the shape, size and other feature information of the inspection target and obstacles from the image.

[0064] Use GPS positioning technology to record the flight position coordinates of the drone, the inspection target coordinates, and the geographic location coordinates of obstacles.

[0065] In order to quickly access and query specific inspection targets or obstacle information, the above data are normalized and cross-validated, and then a data index and retrieval mechanism is established. The relevant data is stored in the inspection database for subsequent updating and analysis.

[0066] Using the collected images and coordinate data, the Simultaneous Localization and Mapping (SLAM) technology is used to combine the coordinates with image features to build an environmental model with complete information including terrain, inspection targets, obstacles, etc.

[0067] Specifically, step 2 comprehensively considers the maneuverability characteristics and environmental information of the UAV to assess the collision risk in the obstacle area:

[0068] Collect the basic performance parameters (i.e. general basic parameters) of the drones used for inspection, including maximum flight speed, maximum acceleration, minimum turning radius, maximum climb rate and descent rate, etc.:

[0069]

[0070] in, is the flight speed of the UAV during period t, v max is the maximum speed of the drone, is the acceleration of the UAV during the period of Δt, is the turning radius of the UAV at time t, r min is the minimum turning radius of the drone, are the climbing rate and descending rate of the UAV in the period of Δt, is the flight altitude of the UAV during the period t+Δt;

[0071] Analyze the maneuverability of the drone in different flight modes, such as hovering, straight flight, and turning, and evaluate its maneuverability based on the maneuverability range:

[0072]

[0073] The maneuverability of the drone is defined based on whether the drone is in a flying and turning state. Among them, 1, 2, and 3 are the levels of maneuverability, with level 1 being the highest and level 3 being the lowest. Different maneuverability levels can be quantified into values ​​in different intervals during subsequent calculations.

[0074] The purpose of evaluating the maneuverability of a drone is to assess the risk level of the drone's ability to safely avoid obstacles. When a drone adjusts its path in the face of an obstacle, it needs to switch flight parameters. The time cost required to adjust flight parameters is inconsistent in the three normal states of flying, turning, and hovering.

[0075] Considering the time required to switch flight parameters, which is not instant, it is necessary to evaluate the risks brought by the adjustment time when avoiding obstacles.

[0076] Combined with the mathematical model of obstacles and the maneuverability of the UAV, the risk level of the UAV obstacle avoidance in this state is evaluated (when the maneuverability level is low, it is necessary to increase the UAV maneuverability coefficient and increase the shortest distance between the UAV and the obstacle in order to plan a safe flight path for the UAV).

[0077] The inspection environment mainly includes inspection targets and obstacles (i.e., non-inspection targets), and the environmental characteristics are mathematically modeled according to edge features.

[0078] Common obstacle shapes, such as buildings modeled as cuboids, trees modeled as cones, transmission poles modeled as cylinders, etc.; the actual objects collected are compared with mathematical models through mathematical modeling to facilitate subsequent path planning.

[0079] The environmental features in the inspection area are modeled into similar mathematical models based on edge features, and the minimum distance between the drone and the obstacle is calculated. The collision probability is determined based on the maneuverability of the drone, and low-risk, medium-risk and high-risk areas are divided accordingly:

[0080] According to the morphological characteristics of the drone, for example, the minimum safe distance d between a general quadcopter drone and any point in three-dimensional space t It is expressed as:

[0081]

[0082] in, is the coordinate of the center point of the drone; (x, y, z) is any point on the plane. When the point falls on an obstacle, it satisfies the mathematical model of the obstacle itself; R is the radius of the circle when the drone is mathematically modeled;

[0083] The risk level of the flight area is assessed by combining the drone’s speed, turning radius, and minimum safe distance from obstacles during period t.

[0084] in,

[0085]

[0086] Among them, Level is the risk level (i.e., the maneuverability coefficient of the drone), d t (maneuverability) is the mathematical relationship between the minimum safe distance (between the drone and the obstacle) and the maneuverability of the drone; L low , L mid , L high They are low risk, medium risk, and high risk respectively. The maneuverability coefficient of the drone depends on the inspection requirements. min is the maximum centripetal acceleration, d min The minimum safe distance between the drone and the obstacle. When the distance between the drone and the obstacle is close, or the speed of the drone is high and the turning radius is small, the risk level value will increase accordingly.

[0087] Based on the collision risk assessment results, a risk heat map is generated to intuitively display the collision risk levels of different areas in the inspection area. In areas with higher risks, clear visual prompts are provided to provide safety warnings for path planning and operators.

[0088] Specifically, step 3 constructs a fitness function to evaluate and plan the safe flight path of the UAV, thereby enhancing the flight efficiency and safety of the UAV in a complex environment.

[0089] Define the closest safe distance to obstacles as a safety indicator to measure the collision risk of the path, define the energy consumption of the drone as the path efficiency indicator, and combine the above indicators to construct a multi-objective fitness function to evaluate the pros and cons of the planned path:

[0090]

[0091] Among them, P is the flight power of the UAV, and the fitness function F fitness It consists of minimizing the risk level and minimizing the energy consumption.

[0092] Specifically, step 4 solves the optimal path based on the improved particle swarm optimization algorithm, and adjusts the flight path according to the simulation results to ensure that all constraints and fitness criteria are met;

[0093] The path planning in the present invention includes the constraints formed by the UAV and the environment and the minimum fitness value function pursued.

[0094] The constructed environmental model is used as a simulation of the inspection area for UAV path planning. Random dynamic obstacles are added to the environmental model. The visual perception module (RGB binocular camera) carried by the UAV is used to detect new obstacles in the environment, and online ranging, edge detection and model updating are performed.

[0095] Based on the particle swarm optimization algorithm and the maneuverability of the drone, it helps the drone avoid obstacles and choose a suitable flight route. According to the value of the fitness function and the algorithm iteration time, it evaluates the quality of the path planned by the algorithm and optimizes the flight strategy.

[0096] According to the complexity of the inspection area, the initial population size and particle position are set, and each particle represents a possible flight path. Based on the local optimal solution (individual optimal position) and the global optimal solution (group optimal position), the search direction of the particle swarm is guided. According to the number of iterations and the current optimization situation, the inertia weight and cognitive factor are adaptively adjusted to balance the capabilities of global search and local search, and the speed and position of each particle are updated.

[0097] The improved particle swarm optimization algorithm specifically includes:

[0098] S41: Initialize parameters such as population size, maximum number of iterations, inertia weight and cognitive factor, as well as particle speed and position;

[0099] S42: The particle calculates the fitness value according to the constructed fitness function; the fitness value reflects the quality of the particle in the planning model, and the goal is to find a solution that minimizes the fitness value.

[0100] S43: Update the inertia weight and cognitive factor according to the improved inertia weight and cognitive factor formula; at the same time, introduce a disturbance factor to enhance the reliability of the particle. Update the particle's speed and position according to the speed and improved position formula;

[0101] S44: Update individual optimal and global optimal: compare the current fitness value of the particle with the individual historical optimal value. If the current value is better, update the individual optimal position to the current position. Among the individual optimal positions of all particles, determine the global optimal position, and find the individual optimal position of the particle with the best fitness value as the global optimal position.

[0102] S45: Determine the termination condition. If the maximum number of iterations is reached, the algorithm terminates and outputs the global optimal position and fitness value; otherwise, return to S42 to continue iterating.

[0103] In the overall iteration process of the algorithm, in order to maximize the computing power and center of gravity of the algorithm, the inertia weight is increased in the early stage of iteration to perform global extreme value search, while in the later stage of iteration, in order to search for more precise extreme values, the inertia weight should be reduced to enhance the local search capability.

[0104] The inertia weights of individual extreme values ​​and global extreme values ​​in the comprehensive iterative progress have strong extreme value search capabilities and convergence:

[0105]

[0106] in, is the inertia weight, are the maximum and minimum inertia weights set respectively, k is the kth iteration, is the individual optimal value of the kth iteration, G best is the optimal value for the group, is the average of the individual optimal values.

[0107] The cognitive factors in the algorithm include individual cognition and social cognition. The value of individual cognition affects the individual optimal value found by the particle in the past. The importance of social cognition reflects the importance of particles to group experience. Therefore, in order to accelerate the particle diversity and global exploration ability in the early stage of the algorithm, and enhance the convergence ability of the particle's own position and the adjacent area in the later stage, the following improved cognitive factors are used in combination with the number of iterations:

[0108] c1=c1 max -e kK 1+e kK

[0109] c2=c2 min +e kK 1+e kK

[0110] Among them, c1 and c2 are individual cognitive factors and social cognitive factors respectively, k is the current iteration number, K is the total iteration number of the algorithm, and the natural constant e is a fixed constant, which is approximately 2.71828.

[0111] The algorithm continuously updates the position of particles through multiple iterations, searches for the position of the minimum value of the fitness function, and uses random perturbation methods to improve the global optimization ability of particles:

[0112]

[0113] Among them, V i k+1 , are the velocity and position of particle i at the k+1th iteration, r1 and r2 are random numbers uniformly distributed in the interval, and Δx is the random perturbation factor.

[0114] The present invention improves the convergence speed and local problems existing in the particle swarm itself, and further adjusts the weights and cognitive factors that affect the iteration speed and direction of the algorithm, and introduces a disturbance factor to enhance the robustness of the algorithm.

[0115] Output the optimal flight path obtained through optimization, including detailed parameter information such as flight route, flight duration and flight speed, and send the optimal path to the drone control system to guide the drone to perform inspections according to the planned path. During the drone's mission, the flight status is continuously monitored and the flight path is adjusted according to actual conditions.

[0116] Regarding the contents disclosed in this case, there are a few points that need to be explained:

[0117] (1) The drawings of the embodiments disclosed in this case only involve the structures involved in the embodiments disclosed in this case, and other structures can refer to the general design;

[0118] (2) In the absence of conflict, the embodiments and features of the embodiments disclosed in this case can be combined with each other to obtain new embodiments;

[0119] The above are only specific implementation methods disclosed in this case, but the protection scope of the present disclosure is not limited thereto. The protection scope disclosed in this case should be based on the protection scope of the claims.

Claims

1. The autonomous path planning method for UAV based on visual perception module is characterized by: The following steps are involved: Step 1: Collect the coordinates and image features of the inspection targets and obstacles in the inspection area, and build an environmental model of the drone inspection area; Step 2: Based on the environmental model and the maneuverability characteristics of the drone, assess the collision risk level in the obstacle area; Step 3: Based on the risk level, establish the fitness function of the drone’s safe path planning; Step 4: Based on the improved particle swarm algorithm, optimize the fitness function to obtain the optimal planning path for the current inspection area.

2. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 1 is characterized in that: In step 1, a drone equipped with a high-resolution binocular camera is used to perform a full-scale scan of the inspection area to obtain high-definition image and video data. At the same time, edge detection image processing technology is used to extract feature information of inspection targets and obstacles from the image, including shape and size. Use GPS positioning technology to record the flight position coordinates of the drone, the inspection target coordinates, and the geographic location coordinates of obstacles; Using the collected images and coordinate data, positioning and mapping technology are used to combine the coordinates with image features to form an environmental model.

3. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 1 is characterized in that: Step 1 also includes: normalizing and cross-validating the collected image and coordinate data, establishing a data index and retrieval mechanism, and storing the relevant data in the inspection database.

4. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 1 is characterized in that: In step 2, the maneuverability of the UAV is evaluated based on its maneuverability characteristics, and then combined with the environmental model, the collision risk level in the obstacle area is evaluated.

5. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 4 is characterized in that: Collect the basic performance parameters of the inspection drone, including maximum flight speed, maximum acceleration, minimum turning radius, maximum climb rate and descent rate. in, In the formula, is the flight speed of the UAV during period t, v max is the maximum speed of the drone, The flight speed of the drone during the period t+Δt, is the acceleration of the UAV during the period of Δt; r t UAV is the turning radius of the UAV at time t, r min is the minimum turning radius of the drone, are the climbing rate and descending rate of the UAV in the period of Δt, is the flight altitude of the UAV during the period t+Δt, is the flight altitude of the UAV during period t; Analyze the maneuverability of the drone in different flight modes, such as hovering, straight flight, and turning, and evaluate its maneuverability based on the maneuverability range: Among them, 1, 2, and 3 are the levels of mobility.

6. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 5 is characterized in that: The environmental features in the inspection area are modeled into similar mathematical models according to the edge features, and the minimum safe distance d between the drone and any point in the three-dimensional space is calculated. t , expressed as: in, is the coordinate of the center point of the drone; (x, y, z) is any point on the plane; R is the radius of the circle when the drone is mathematically modeled; The risk level of the flight area is assessed by combining the speed, turning radius, and minimum safe distance between the drone and obstacles during the t period. in, Among them, Level is the risk level, L low , L mid , L high They are low risk, medium risk, and high risk, respectively. min is the maximum centripetal acceleration, d min is the minimum safe distance between the drone and obstacles.

7. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 1 is characterized in that: The fitness function is: Among them, F fitness is the fitness function, and P is the flight power of the UAV.

8. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 1 is characterized in that: In step 4, the constructed environment model is used as the inspection area simulation for the drone path planning. Random dynamic obstacles are added to the environment model. The visual perception module carried by the drone is used to detect new obstacles in the environment, and distance measurement, edge detection and model update are performed online. Based on the particle swarm optimization algorithm and the maneuverability of the drone, it helps the drone avoid obstacles and choose a suitable flight route. According to the value of the fitness function and the algorithm iteration time, it evaluates the quality of the path planned by the algorithm and optimizes the flight strategy.

9. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 8 is characterized in that: According to the complexity of the inspection area, the initial population size and particle position are set, and each particle represents a possible flight path; Based on the local optimal solution and the global optimal solution, guide the search direction of the particle swarm; According to the number of iterations and the current optimization situation, the inertia weight and cognitive factor are adaptively adjusted to balance the capabilities of global search and local search, and the speed and position of each particle are updated.

10. The autonomous path planning method for unmanned aerial vehicle based on visual perception module according to claim 9 is characterized in that: Step 4 includes: S41: Initialize population size, maximum number of iterations, inertia weight and cognitive factor, as well as particle speed and position; S42: The particle calculates the fitness value according to the constructed fitness function; S43: updating the inertia weight and cognitive factor according to the improved inertia weight and cognitive factor formula; at the same time, introducing the disturbance factor; updating the speed and position of the particle according to the speed and improved position formula; S44: Update individual optimal and global optimal: compare the current fitness value of the particle with the individual historical optimal value. If the current value is better, update the individual optimal position to the current position. Among the individual optimal positions of all particles, determine the global optimal position, and find the individual optimal position of the particle with the best fitness value as the global optimal position. S45: Determine the termination condition. If the maximum number of iterations is reached, the algorithm terminates and outputs the global optimal position and fitness value; otherwise, return to S42 to continue iterating.

Images