An autonomous drone transportation system based on solar power generation

Abstract

The present invention discloses an autonomous driving UAV transportation system based on solar power generation, including a UAV, a solar panel group, an environmental perception module, a communication module, a camera, a graphics processor and a delivery device. The solar panel generates electricity through a DC / DC converter, and the battery stores energy; the environmental perception unit collects the position of the UAV and the position of obstacles; the facial photo of the recipient is collected by a high-definition camera, transmitted to the graphics processor for comparison and judged by a central controller, and the corresponding instructions are sent. In this process, the present invention uses the improved crown porcupine optimization algorithm to optimize the UAV automatic driving path to ensure that it can transport with the highest efficiency while avoiding obstacles safely. After arriving at the recipient's location, the graphics processor performs image optimization and comparison based on the OPB algorithm improved by RANSAC to improve the accuracy of the transportation operation.

Description

Technical Field

[0001] The present invention belongs to the field of drone logistics, and in particular relates to an autonomous driving drone transportation system based on solar power generation. Background Art

[0002] With the rapid development of science and technology, cities have an increasing demand for the logistics industry, which is also proved by the couriers running around in the city. However, in the process of manual transportation, it is inevitable that problems will arise in some links of the transportation logistics due to insufficient knowledge of road conditions, weather factors, personal trivial matters and other factors.

[0003] Therefore, the logistics industry has begun to use drones to carry out express delivery tasks that are difficult to deliver by manpower. Compared with manual transportation, drones are more efficient and more convenient for unified logistics management, and have a very optimistic development trend.

[0004] However, at present, although the logistics industry has made preliminary use of drone transportation, most of them rely on manual control by professionals, which is difficult for ordinary people to operate, making the drone transportation industry have a high threshold. In the face of more complex urban environments, if the route is not planned in advance to obtain the optimal transportation path, the transportation efficiency will also be reduced. Most existing drones rely on battery power, have limited endurance, and are inconvenient to charge. Drone transportation generally lacks feasible solutions that are mainly driven by new energy and overcome the above shortcomings. Summary of the invention

[0005] Purpose of the invention: To solve the above problems, the present invention discloses an autonomous driving UAV transportation system based on solar power generation; using solar energy as the driving energy of the UAV, deploying transportation tasks to the UAV through a ground command system; using the improved crown porcupine algorithm, optimizing the optimal transportation path of the UAV by comprehensively considering time, terrain, and energy consumption factors, achieving the purpose of autonomous driving through a central controller, improving the efficiency of UAV logistics transportation while maximizing economy; using the OPB algorithm improved based on the RANSAC algorithm to optimize the graphics processor, accurately identify the true identity of the recipient, and further improve the safety and stability of the transportation system.

[0006] Technical solution: The present invention discloses an autonomous driving drone transportation system based on solar power generation, and the operation of the system includes the following steps:

[0007] The S1 ground command system receives the transport mission and sends the recipient’s image, destination, and GPS information of the passing area to the UAV central controller through the communication module;

[0008] After receiving the transport mission, the S2 drone uses the crested porcupine algorithm to optimize the path;

[0009] S2.1 optimizes and initializes the crested porcupine population, brings the initialized population into the objective function evaluation, and generates a set of corresponding group values ​​of the shortest path length, the least energy consumption, and the smallest threat area;

[0010] S2.2 uses the exploration mechanism of the crested porcupine algorithm and introduces an adaptive weight factor to enable the crested porcupine individual to update its position using a smaller adaptive weight to find the area where the optimal solution is located;

[0011] S2.3 uses the developed mechanism of the crested porcupine algorithm to find a near-optimal solution based on odor and physical attack strategies;

[0012] S3 determines whether the solution satisfies the constraint conditions. If not, it returns to step S2.2. If it satisfies the constraint conditions, it outputs the optimal objective function value.

[0013] The S4 drone obtains the current optimal path based on the optimal objective function value and, under the control of the central controller, automatically drives along the optimal path to deliver the goods to the recipient's location;

[0014] S5 collects the recipient's facial photo, transmits it to the graphics processor for comparison with the recipient's image, uses the OPB algorithm improved by the RANSAC algorithm to optimize the graphics processor, determines whether it is the recipient, and sends the result to the central processor. The central processor sends corresponding instructions based on the result, and transmits its own coordinates and recipient information to the ground command system through the communication unit.

[0015] Furthermore, in step S2.1, the objective function formula is as follows:

[0016]

[0017]

[0018] Among them, constraint (1) means that the power consumed by the drone during flight cannot exceed the sum of the solar power generation power and the livestock battery power; constraint (2) ensures that the transportation task must be completed within the scheduled time; constraint (3) prevents the drone from entering the threat area, C is the transportation time; st is the constraint condition; is the weight parameter, d i is the distance of the UAV from the i-1th point to the i-th point, n is the number of nodes in the path, θ a Indicates the turning angle of the drone, N a represents the power consumed by the drone each time it turns, x is the number of turns, q represents the set of all obstacles, all obstacles are considered as cylinders, D represents the diameter of the drone, b represents the number of takeoffs, and R q represents the obstacle radius, P V is the photovoltaic power generation power; PX is the battery power, F b The power consumed in each flight.

[0019] Furthermore, the process of finding the optimal solution region in step S2.2 is as follows:

[0020] First, perform local optimization on individuals in the population to improve individual fitness. Then perform global optimization on individuals in the population to increase population diversity. The corresponding formula is as follows:

[0021]

[0022]

[0023] in, is the new position of the particle after updating; is the position of the i-th individual in the current iteration t in the crested porcupine population; is the best solution for the current iteration t, is a vector generated between the current crested porcupine individual and a crested porcupine randomly selected from the population to represent the position of the predator at iteration t, τ1 is a random number based on a normal distribution, τ2 is a random value in the interval [0, 1], X is a binary vector with 0 and 1 randomly generated to cover all possible probabilities, τ3 is a random value generated between 0 and 1, are the positions of the r1-th and r2-th individuals in the current iteration t of the crested porcupine population, where r1 and r2 are two random integers between [1, N].

[0024] Furthermore, by introducing an adaptive weight factor, the crested porcupine uses a smaller adaptive weight to update the position, thereby improving the local optimization ability of the crested porcupine algorithm. The improved formula is as follows:

[0025]

[0026] In the formula, is the new position of the particle after updating after the improved algorithm; ω is the adaptive weight factor.

[0027] Furthermore, the step S2.3 finds a near-optimal solution and adjusts its search strategy according to the individual's historical performance to avoid falling into a local optimum. The formula is as follows:

[0028]

[0029] Where r3 is a random number between [1, N], is a parameter used to control the search direction. is the position of the ith individual at iteration t, γ tis the defense factor defined using the equation, τ3 is a random value in the interval [0, 1], is the odor diffusion factor, responsible for the odor diffusion rate during the optimization process.

[0030] Furthermore, the individuals in the population are randomly disturbed to maintain the vitality of the population and find the optimal solution:

[0031]

[0032]

[0033] in, is the best solution obtained. is the position of the i-th individual at iteration t, indicating the predator at that position, α is the convergence speed factor, and τ4 is a random value in the interval [0,1]; represents the objective function, m i is the mass of the i-th individual (predator) at iteration t, is the initial velocity of the ith individual at iteration t, is a vector of random values ​​generated between 0 and 1, is the final velocity of the ith individual at the next iteration t+1 and is allocated based on selecting random solutions from the current population.

[0034] Furthermore, the step S5 uses the OPB algorithm improved based on the RANSAC algorithm to optimize the graphics processor, and the implementation of the OPB algorithm improved based on the RANSAC algorithm is as follows:

[0035] Feature point matching: First, the ORB algorithm is used to detect feature points on the two images, and the descriptor distance between the feature points is calculated to find the matching feature point pairs;

[0036] Initial matching screening: You can preliminarily screen matching point pairs by setting a distance threshold. Only when the descriptor distance between two feature points is less than this threshold, they are considered to be matched;

[0037] Use the RANSAC algorithm: For the selected matching point pairs, use the RANSAC algorithm to estimate an optimal transformation model, which can describe the geometric relationship between the two images;

[0038] Model parameter estimation: Use the homography matrix to calculate the model parameters. For n matching point pairs, the homography matrix can be calculated by the following formula:

[0039]

[0040] Among them, (x i ,y i) and (x i ',y i ') are the coordinates of the i-th matching point on the original image and the predicted image respectively, and find the homography matrix H that minimizes the error between all predicted coordinates and actual coordinates;

[0041] Eliminate mismatched points: According to the model estimated by the RANSAC algorithm, point pairs that cannot be well fitted by the model are regarded as mismatched points and are removed from the matching list;

[0042] Evaluation and iteration: Evaluate the quality and stability of the model. If the performance of the model is not satisfactory, you can optimize it by increasing the number of iterations or adjusting the threshold.

[0043] Beneficial effects:

[0044] 1. The present invention configures solar panels on the autonomous driving drone, uses solar energy to generate electricity, does not need to be charged, has a high endurance, and can meet rescue needs in complex environments such as jungles.

[0045] 2. The present invention configures a central controller on the UAV, which can realize automatic driving in cooperation with the ground command system. The operation is simple and convenient, and no professional personnel are required to operate it, thereby improving the transportation efficiency.

[0046] 3. The present invention improves the crested porcupine algorithm by introducing an adaptive weight factor, thereby improving the algorithm's local optimization capability, obtaining a more accurate objective function value, and further improving the convergence speed of the drone when searching for the optimal path.

[0047] 4. The present invention improves the OPB algorithm based on the RANSAC algorithm to improve the execution efficiency and comparison accuracy of the graphics processor, ensuring that the drone can more accurately and quickly verify the recipient information after the transportation is completed, thereby improving the transportation efficiency and avoiding unexpected situations such as delivery errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a structural schematic diagram of the present invention;

[0049] Figure 2 It is a flow chart of the optimization algorithm of crested porcupine proposed by the present invention;

[0050] Figure 3 This is a comparison chart between the crowned porcupine optimization algorithm, particle swarm algorithm, and gray wolf algorithm.

[0051] Figure 4 This is a simulation diagram of the crested porcupine optimization algorithm path proposed in the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0053] like Figure 1 As shown, the present invention provides an autonomous driving UAV jungle rescue system based on solar power generation, including a UAV, a power supply module, an environmental perception module, a communication module, a graphics processor, a central processing unit and rescue equipment. The power supply module includes a solar panel, a battery, and a DC / DC control drive circuit; the environmental perception module includes a GPS, a speedometer, a gyroscope, a laser ranging sensor, an infrared sensor and a high-definition camera; the communication module includes a wireless communication terminal, a data message and an intelligent gateway module.

[0054] The solar panels in the power supply module generate electricity through a DC / DC converter, and the battery stores energy; the environmental perception unit sends the collected position of the drone and the position of the obstacle to the central controller, and the central controller issues corresponding control instructions based on the data collected by the environmental perception unit to achieve the purpose of automatic driving; the data message in the communication unit contains the surrounding environment and the operating status of the drone itself; the intelligent gateway refers to the content of the combined data message for filtering and screening to ensure data transmission between the chip and the ground command system; the graphics processor compares the photos taken by the high-definition camera with the recipient's photos, determines the receiving target and transmits it to the central processor, and the central processor issues a delivery instruction.

[0055] like Figure 2 As shown, the present invention uses the crown porcupine optimization algorithm to optimize the path, and the implementation process is as follows:

[0056] 1) Optimize and initialize the crested porcupine population. Set the number of crested porcupines to N; each crested porcupine is randomly placed in the space as a candidate solution x = (x1, ..., xi) represented by a three-dimensional vector, set the maximum number of iterations of the algorithm, and set the search space of the algorithm according to the number of iterations;

[0057] 2) The initial population is brought into the objective function for evaluation. A corresponding set of values ​​of the shortest path length, the least energy consumption, and the smallest threat area is generated in each crested porcupine population. The objective function formula is as follows:

[0058]

[0059]

[0060] Constraint (1) indicates that the power consumed by the drone during flight cannot exceed the sum of the solar power generation power and the battery power; constraint (2) ensures that the transportation task must be completed within the scheduled time; constraint (3) prevents the drone from entering the threat area.

[0061] In the formula, C is the transportation time; st is the constraint condition; is the weight parameter, d i is the distance from the drone to the i-th point, and n is the number of nodes in the path. a Indicates the turning angle of the drone, N a represents the power consumed by the drone each time it turns, x is the number of turns, q represents the set of all obstacles, all obstacles are considered as cylinders, D represents the diameter of the drone, b represents the number of takeoffs, and R q Represents the obstacle radius. V is the photovoltaic power generation power; P X is the battery power, F b The power consumed in each flight.

[0062] 3) Use the exploration mechanism of the crested porcupine algorithm to find the area where the optimal solution is located.

[0063] First, a local search is performed on the individuals in the population to improve individual fitness. Then a global search is performed on the individuals in the population to increase population diversity.

[0064]

[0065]

[0066] In the formula, is the new position of the particle after updating; is the position of the i-th individual in the current iteration t in the crested porcupine population; is the best solution for the current iteration t, is a vector generated between the current crested porcupine individual and a crested porcupine randomly selected from the population to represent the position of the predator at iteration t, τ1 is a random number based on a normal distribution, τ2 is a random value in the interval [0, 1]. X is a binary vector, randomly generated 0 and 1, covering all possible probabilities. τ3 is a random value generated between 0 and 1. are the positions of the r1-th and r2-th individuals in the current iteration t of the crested porcupine population, where r1 and r2 are two random integers between [1, N].

[0067] 4) Use the developed mechanism of the crested porcupine algorithm to find a near-optimal solution based on scent and physical attack strategies. Adjust its search strategy based on the individual's historical performance to avoid falling into a local optimum.

[0068]

[0069] Where r3 is a random number between [1, N], is a parameter used to control the search direction. is the position of the ith individual at iteration t, γ t is the defense factor defined using Eq. τ3 is a random value in the interval [0, 1], is the odor diffusion factor, responsible for the odor diffusion rate during the optimization process.

[0070] Randomly perturb the individuals in the population to maintain the vitality of the population and find the optimal solution.

[0071]

[0072]

[0073] In the formula, is the best solution obtained. is the position of the i-th individual at iteration t, indicating the predator at that position, α is the convergence speed factor, and τ4 is a random value in the interval [0,1]; represents the objective function, m i is the mass of the i-th individual (predator) at iteration t, is the initial velocity of the ith individual at iteration t, is a vector of random values ​​generated between 0 and 1. is the final velocity of the ith individual at the next iteration t+1 and is allocated based on selecting random solutions from the current population.

[0074] 5) Determine whether the termination condition is met. If the termination condition is not met, return to step 3); if the constraint condition is met, go to step 6).

[0075] 6) Output the optimal objective function value.

[0076] 7) The step of improving the crested porcupine algorithm is to introduce an adaptive weight factor in step 3) so that the crested porcupine individual uses a smaller adaptive weight to update the position, thereby improving the local optimization ability of the crested porcupine algorithm. The improved formula is as follows:

[0077]

[0078] In the formula, is the new position of the particle after updating after the improved algorithm; ω is the adaptive weight factor.

[0079] 8) The drone obtains the current optimal path based on the optimal objective function value, and under the control of the central controller, it automatically drives according to the optimal path to deliver the goods to the recipient's location;

[0080] 9) Collect the facial photo of the recipient and transmit it to the graphics processor for comparison with the recipient's image. Use the OPB algorithm improved by the RANSAC algorithm to optimize the graphics processor to determine whether it is the recipient and send the result to the central processor. The central processor sends corresponding instructions based on the result and transmits its own coordinates and recipient information to the ground command system through the communication unit.

[0081] Among them, the implementation of the OPB algorithm improved based on the RANSAC algorithm is as follows:

[0082] Feature point matching: First, the ORB algorithm is used to detect feature points on the two images, and the descriptor distance between the feature points is calculated to find the matching feature point pairs;

[0083] Initial matching screening: You can preliminarily screen matching point pairs by setting a distance threshold. Only when the descriptor distance between two feature points is less than this threshold, they are considered to be matched;

[0084] Use the RANSAC algorithm: For the selected matching point pairs, use the RANSAC algorithm to estimate an optimal transformation model, which can describe the geometric relationship between the two images;

[0085] Model parameter estimation: Use the homography matrix to calculate the model parameters. For n matching point pairs, the homography matrix can be calculated by the following formula:

[0086]

[0087] Among them, (x i ,y i ) and (x i ',y i ') are the coordinates of the i-th matching point on the original image and the predicted image respectively, and find the homography matrix H that minimizes the error between all predicted coordinates and actual coordinates;

[0088] Eliminate mismatched points: According to the model estimated by the RANSAC algorithm, point pairs that cannot be well fitted by the model are regarded as mismatched points and are removed from the matching list;

[0089] Evaluation and iteration: Evaluate the quality and stability of the model. If the performance of the model is not satisfactory, you can optimize it by increasing the number of iterations or adjusting the threshold.

[0090] The following is a comparative test of the improved crown porcupine optimization algorithm of this embodiment and the conventional path optimization algorithm, and the results are as follows:

[0091] like Figure 3 As shown in Figure 2, the CEC2005 test function is used to test the performance of the crown porcupine optimization algorithm, the classic particle swarm optimization algorithm (PSO) and the more commonly used grey wolf optimization algorithm (GWO) in path optimization.

[0092] 1) Robustness: The optimal value of the crested porcupine algorithm is much smaller than the optimal values ​​of GWO and PSO, which shows that the crested porcupine algorithm has strong robustness in finding extreme minimum values. The worst value of the crested porcupine algorithm is also the smallest, which further proves that the crested porcupine algorithm can effectively avoid falling into the local optimum when dealing with constrained problems.

[0093] 2) Global search capability: The mean and standard deviation of the crested porcupine algorithm are much smaller than those of GWO and PSO, which indicates that the crested porcupine algorithm performs better in global search and can find better solutions. The mean and standard deviation of PSO are the largest, indicating that PSO is relatively weak in global search.

[0094] 3) Performance stability: The standard deviation of the crested porcupine algorithm is small, indicating that its performance fluctuation is small and its consistency is good. The standard deviation of GWO is slightly higher than that of the crested porcupine algorithm, but still small, indicating that its performance is relatively stable. PSO has the largest standard deviation, indicating that its performance fluctuation is large and its consistency is not as good as that of the crested porcupine algorithm and GWO.

[0095] It can be concluded that the crested porcupine algorithm performs best in terms of robustness, global search capability, accuracy, and performance stability, and is an excellent choice for solving path optimization problems.

[0096] like Figure 4 As shown, MATLAB R2022b is used for simulation, the population size is set to 30, and the maximum number of iterations is 500; 12 obstacles are set with coordinates of [30, 30, 3; 40, 30, 3; 50, 30, 3; 60, 30, 3; 30, 50, 3; 40, 50, 3; 50, 50, 3; 60, 50, 3; 30, 80, 3; 40, 80, 3; 50, 80, 3; 60, 80, 3]; with the minimum transportation time (shortest path length, minimum energy consumption, minimum threat area) as the objective function, simulation diagrams of path optimization using the crested porcupine optimization algorithm are generated, including three-dimensional and two-dimensional diagrams.

[0097] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. An autonomous driving drone transportation system based on solar power generation, characterized in that: The operation process of the transportation system includes the following steps: The S1 ground command system receives the transport mission and sends the recipient’s image, destination, and GPS information of the passing area to the UAV central controller through the communication module; After receiving the transport mission, the S2 drone uses the crested porcupine algorithm to optimize the path; S2.1 optimizes and initializes the crested porcupine population, brings the initialized population into the objective function evaluation, and generates a set of corresponding group values ​​of the shortest path length, the least energy consumption, and the smallest threat area; S2.2 uses the exploration mechanism of the crested porcupine algorithm, introduces an adaptive weight factor, uses a smaller adaptive weight to update the position, and finds the area where the optimal solution is located; S2.3 uses the developed mechanism of the crested porcupine algorithm to find a near-optimal solution based on odor and physical attack strategies; S3 determines whether the solution satisfies the constraint conditions. If not, it returns to step S2.

2. If it satisfies the constraint conditions, it outputs the optimal objective function value. The S4 drone obtains the current optimal path based on the optimal objective function value and, under the control of the central controller, automatically drives along the optimal path to deliver the goods to the recipient's location; S5 collects the recipient's facial photo, transmits it to the graphics processor for comparison with the recipient's image, uses the OPB algorithm improved by the RANSAC algorithm to optimize the graphics processor, determines whether it is the recipient, and sends the result to the central processor. The central processor sends corresponding instructions based on the result, and transmits its own coordinates and recipient information to the ground command system through the communication unit.

2. The solar-powered autonomous driving drone transportation system according to claim 1, characterized in that: In step S2.1, the objective function formula is as follows: Among them, constraint (1) means that the power consumed by the drone during flight cannot exceed the sum of the solar power generation power and the livestock battery power; constraint (2) ensures that the transportation task must be completed within the scheduled time; constraint (3) prevents the drone from entering the threat area, C is the transportation time; st is the constraint condition; is the weight parameter, d i is the distance of the UAV from the i-1th point to the i-th point, n is the number of nodes in the path, θ a Indicates the turning angle of the drone, N a represents the power consumed by the drone each time it turns, x is the number of turns, q represents the set of all obstacles, all obstacles are considered as cylinders, D represents the diameter of the drone, b represents the number of takeoffs, and R q Represents the obstacle radius; P V is the photovoltaic power generation power; P X is the battery power, F b The power consumed in each flight.

3. The solar-powered autonomous driving drone transportation system according to claim 1, characterized in that: The process of finding the optimal solution area in step S2.2 is as follows: First, local optimization is performed on individuals in the population to improve individual fitness. Second, global optimization is performed on individuals in the population to increase population diversity. The corresponding formula is as follows: in, is the new position of the particle after updating; is the position of the i-th individual in the current iteration t in the crested porcupine population; is the best solution for the current iteration t, is a vector generated between the current crested porcupine individual and a crested porcupine randomly selected from the population to represent the position of the predator at iteration t, τ1 is a random number based on a normal distribution, τ2 is a random value in the interval [0, 1], X is a binary vector with 0 and 1 randomly generated to cover all possible probabilities, τ3 is a random value generated between 0 and 1, are the positions of the r1-th and r2-th individuals in the current iteration t of the crested porcupine population, where r1 and r2 are two random integers between [1, N].

4. The autonomous driving UAV transportation system based on solar power generation according to claim 1 or 3, characterized in that: By introducing an adaptive weight factor, the crested porcupine individual uses a smaller adaptive weight to update the position, thereby improving the local optimization ability of the crested porcupine algorithm. The improved formula is as follows: In the formula, is the new position of the particle after the improved algorithm is updated; ω is the adaptive weight factor.

5. The solar-powered autonomous driving drone transportation system according to claim 1, characterized in that: The step S2.3 finds a near-optimal solution and adjusts its search strategy according to the individual's historical performance to avoid falling into the local optimum. The formula is as follows: Where r3 is a random number between [1, N], is a parameter used to control the search direction. is the position of the ith individual at iteration t, γ t is the defense factor defined using the equation, τ3 is a random value in the interval [0, 1], is the odor diffusion factor, responsible for the odor diffusion rate during the optimization process.

6. The solar-powered autonomous driving drone transportation system according to claim 5, characterized in that: Randomly perturb the individuals in the population to maintain the vitality of the population and find the optimal solution: in, is the best solution obtained. is the position of the i-th individual at iteration t, indicating the predator at that position, α is the convergence speed factor, and τ4 is a random value in the interval [0,1]; represents the objective function, m i is the mass of the i-th individual (predator) at iteration t, is the initial velocity of the ith individual at iteration t, is a vector of random values ​​generated between 0 and 1, is the final velocity of the ith individual at the next iteration t+1 and is allocated based on selecting random solutions from the current population.

7. The solar-powered autonomous driving drone transportation system according to claim 1, characterized in that: The step S5 uses the OPB algorithm improved based on the RANSAC algorithm to optimize the graphics processor. The implementation of the OPB algorithm improved based on the RANSAC algorithm is as follows: Feature point matching: First, the ORB algorithm is used to detect feature points on the two images, and the descriptor distance between the feature points is calculated to find the matching feature point pairs; Initial matching screening: You can preliminarily screen matching point pairs by setting a distance threshold. Only when the descriptor distance between two feature points is less than this threshold, they are considered to be matched; Use the RANSAC algorithm: For the selected matching point pairs, use the RANSAC algorithm to estimate an optimal transformation model, which can describe the geometric relationship between the two images; Model parameter estimation: Use the homography matrix to calculate the model parameters. For n matching point pairs, the homography matrix can be calculated by the following formula: Among them, (x i ,y i ) and (x i ',y i ') are the coordinates of the i-th matching point on the original image and the predicted image respectively, and find the homography matrix H that minimizes the error between all predicted coordinates and actual coordinates; Eliminate mismatched points: According to the model estimated by the RANSAC algorithm, point pairs that cannot be well fitted by the model are regarded as mismatched points and are removed from the matching list; Evaluation and iteration: Evaluate the quality and stability of the model. If the performance of the model is not satisfactory, you can optimize it by increasing the number of iterations or adjusting the threshold.

Images