Target positioning error optimization method based on unmanned aerial vehicle

Abstract

The invention relates to a target positioning error optimization method based on an unmanned aerial vehicle, and provides an error processing method based on coordinate transformation of an airborne measurement system and an improved dung beetle optimization algorithm (CCDBO) by focusing on the requirement of the unmanned aerial vehicle for accurate space positioning of a target. The method comprises the following steps of: firstly, calculating a total standard deviation by integrating information such as an aerial carrier position, an attitude angle, a pod azimuth angle and a pitch angle and combining a given pod installation azimuth alignment error, and introducing a target position correction strategy of an improved dung beetle optimization algorithm to further improve the positioning precision; according to the improved dung beetle optimization algorithm, Kent mapping is adopted to broaden the population range, a crisscross algorithm is introduced to update and simulate the target position, and flight experiment results show that the method can significantly reduce the target positioning error of the unmanned aerial vehicle, the positioning precision is averagely improved by 20.42% compared with an original dung beetle algorithm, and the target positioning accuracy is greatly improved. And powerful support is provided for high-precision target observation and identification of the unmanned aerial vehicle in a complex environment.

Description

Technical Field

[0001] The present invention relates to the technical field of positioning methods, and in particular to a target positioning error optimization method based on an unmanned aerial vehicle (UAV). Background Art

[0002] Driven by the Fourth Scientific and Technological Revolution, the second robotics revolution, driven by intelligent technology, is reshaping the reconnaissance sector. Airborne mobile platforms, as a symbol of innovation, are central to reconnaissance due to their innovative and practical capabilities, demonstrating increasing strategic value. The development of optoelectronic payloads for drones enables them to perform multiple tasks, including target tracking and location. Continuous advances in sensor technology, computer vision, and other fields have also led to significant progress in drone target positioning technology. Precise target positioning is a key technology in the drone sector, providing information support for subsequent mission planning.

[0003] Currently, positioning methods can be divided into active positioning and passive positioning. The most common active positioning method is to obtain the target position through laser ranging combined with homogeneous coordinate transformation. It has high accuracy, but the laser range is limited and is not conducive to the concealment and safety requirements of drones. Passive positioning does not transmit signals to the detected target, but only uses the received target radiation source signal to scout and locate the target, which can improve its own concealment. It has been widely studied by scholars in recent years. Bai Guanbing et al. established a multi-target positioning model and reduced the target positioning error by improving the filtering effect. Qian et al. studied the pure bearing passive positioning problem of drones in an electromagnetic silent environment. Feng et al. established a drone positioning model based on single-target optimization to guide the flight of drones in passive positioning. With the complexity of application scenarios, how to further reduce positioning errors and improve positioning accuracy has become an urgent problem to be solved.

[0004] During the UAV target positioning process, multiple factors interact, leading to diverse error sources. Therefore, how to effectively process and distribute these errors is of vital practical significance for improving the accuracy of UAV target positioning. Tan et al. established a single-station passive positioning equation, analyzed target positioning errors, explored the error factors affecting positioning, and provided application recommendations for single-station passive positioning. Sun et al. studied the error distribution problem of the sighting angle of the optoelectronic pod on small and micro UAVs from the perspective of multi-objective optimization. However, traditional algorithms often have shortcomings such as high computational complexity, poor adaptability, and high requirements for initial conditions. With the continuous development of swarm intelligence algorithms, the problem of poor performance of traditional algorithms in optimization problems has been solved. Examples include particle swarm optimization (PSO), sparrow algorithm (SSA), and grey wolf algorithm (GWO). Although intelligent algorithms have shown significant advantages in many fields, their applications are becoming increasingly widespread, and they have achieved remarkable results in dealing with complex problems, optimizing decision-making processes, and improving computing efficiency. However, there is still room for improvement in intelligent algorithms. The Dung Beetle Algorithm (DBO) is a new swarm intelligent optimization algorithm proposed by Xue et al. that imitates the survival behaviors of dung beetles, such as rolling balls, dancing, foraging, stealing, and reproduction. It shows excellent performance in solving function optimization problems, has a strong ability to search for optimal values and a fast convergence speed, and is widely used in path planning, engineering applications, and other aspects. Summary of the Invention

[0005] To better address the error allocation problem in drone target positioning, this paper proposes an improved dung beetle algorithm (CCDBO) that integrates cross-cutting and principled approaches. This algorithm uses the error to be allocated as an optimization parameter, providing an optimal solution under certain constraints. This improves the algorithm's ability to escape local optimal states and increases population quality and diversity. Simulation and field experiments have demonstrated that the improved dung beetle optimization algorithm is more effective. To address the above technical issues, this paper adopts the following technical solutions:

[0006] A target positioning error optimization method based on an unmanned aerial vehicle comprises the following steps:

[0007] Step 1: Utilize a drone to collect target position information, and use a range-angle target positioning method to locate the target. In a coordinate system conversion step, when converting the geographic coordinate system to the world coordinate system, an improved dung beetle optimization algorithm is introduced in the world coordinate system to optimize the coordinate results in the world coordinate system. The improved dung beetle optimization algorithm uses Kent mapping to broaden the population range, and introduces a vertical and horizontal intersection algorithm to update the target position to obtain an updated position.

[0008] Step 2: Change the position or attitude of the drone and repeat step 1 for the same target.

[0009] Step 3: Repeat step 2 multiple times to obtain multiple updated positions. Use the 3σ principle to filter the multiple updated positions to obtain filtered positions. Average the filtered positions to obtain the final optimized world coordinate system coordinates.

[0010] Step 4: Convert the final optimized world coordinate system coordinates to the geodetic coordinate system coordinates and output the final positioning result.

[0011] Furthermore, the dung beetle optimization algorithm uses Kent mapping to initialize the population.

[0012] Furthermore, the improved dung beetle optimization algorithm includes the following steps:

[0013] Step 1.1: In each iteration, the target position is obtained using the conventional dung beetle optimization algorithm;

[0014] Step 1.2: Use the cross-cross algorithm to update the target position in step 1.1 to obtain a new target position. Compare the new target position with the target position in step 1.1, select the target position with a higher fitness value and retain it, and obtain the updated position after this iteration.

[0015] Furthermore, in step 1.2, the position update formula of the cross-cross algorithm is:

[0016] x i ′(t)=min{r1*x(i,d)+(1-r1)*x(j,d)+c1*(x(i,d)-x(j,d)),r*x(i,d1)

[0017] +(1-r)*x(i,d2)} |p +a×k×{r1*x(j,d)+(1-r1)*x(i,d)

[0018] +c1*(x(j,d)-x(i,d))} |p +b×Δx

[0019] Among them, t represents the current iteration number; x i (t) is the position of the i-th dung beetle after t iterations, a is the natural coefficient, when a = 1 there is no deviation, when a = -1 it deviates from the original direction, k is the deflection coefficient, b∈(0,1) is a constant, Δx simulates the change of light intensity, X w is the global worst position, x(i,d) and x(j,d) are the d-th dimension of individual particles x(i) and x(j) in the parent population, respectively, XS hc (i,d) and XS hc(j,d) represents the d-th dimension offspring generated by horizontal crossover of x(i,d) and x(j,d), r1 and r2 are random numbers in [0,1], and c1 and c2 are random numbers in [-1,1].

[0020] After adopting the above technical solution, the present invention has the following advantages compared with the prior art:

[0021] This paper focuses on the need for precise spatial positioning of targets by unmanned aerial vehicles (UAVs). It proposes an error processing method based on coordinate transformation and an improved CCDBO algorithm (CCDBO) for airborne measurement systems. First, by integrating information such as the aircraft's position, attitude angle, pod azimuth, and pitch angle, the total standard deviation is calculated based on the given pod installation azimuth alignment error. The Taylor series expansion method is then used to linearize and approximate the coordinate transformation process, simplifying the error propagation model. Finally, to further improve positioning accuracy, a target position correction strategy based on the improved CCDBO algorithm is introduced. This paper proposes an improved CCDBO algorithm (CCDBO) that integrates the vertical and horizontal intersection and 3σ principles. This algorithm uses the error to be assigned as an optimization parameter to determine the optimal solution under certain constraints. This improves the algorithm's ability to escape local optimal states and increases population quality and diversity. Simulation and flight test results demonstrate that this method can significantly reduce UAV target positioning errors, improving positioning accuracy by an average of 20.42% compared to the original CCDBO algorithm. This method provides strong support for high-precision target observation and recognition by UAVs in complex environments.

[0022] The present invention is described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 It is a schematic diagram of the spatial rectangular coordinate system;

[0024] Figure 2 It is a schematic diagram of the geographic coordinate system;

[0025] Figure 3 It is a schematic diagram of the body coordinate system;

[0026] Figure 4 It is a schematic flow chart of the method of the present invention;

[0027] Figure 5 It is the CCDBO algorithm flow chart;

[0028] Figure 6 This is a schematic diagram comparing the average simulation results of the CCDBO algorithm and other algorithms;

[0029] Figure 7 It is the target point simulation latitude and longitude scatter plot distribution map;

[0030] Figure 8 It is the target point simulation latitude and longitude histogram distribution diagram;

[0031] Figure 9 is the distance diagram between the simulation point and the true value point;

[0032] Figure 10 This is a schematic diagram of the measurement data of target point 1;

[0033] Figure 11 It is a schematic diagram of the body coordinate system;

[0034] Figure 12 is a schematic diagram of the target point positioning error; DETAILED DESCRIPTION

[0035] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0036] 1 Definition of coordinate system

[0037] The basis for UAV target positioning is to establish a suitable coordinate system and define the geodetic coordinate system, spatial rectangular coordinate system, geographic coordinate system, body coordinate system, and camera coordinate system as needed.

[0038] 1.1 Geodetic coordinate system (O-BLH)

[0039] The origin of the WGS84 coordinate system is located at the geometric center of the Earth's ellipsoid, coinciding with the Earth's center of mass. B It is defined as the angle between the ellipsoid normal and the equatorial plane at a certain point in space. This angle is measured with the equator as the reference line, following the principle that north is positive and south is negative. Its range covers the global latitude domain from -90° to +90°, fully mapping the latitude variation characteristics of the earth's surface. L The determination of is based on the angle between the meridian plane where the spatial point lies and the prime meridian plane. This angle, also relative to the prime meridian plane, follows the rule of positive values to the east and negative values to the west, ranging from -180° to +180°. The geodetic height H quantifies the vertical distance from the spatial point to the ellipsoid surface, with positive values above the ellipsoid and negative values below. The coordinates of a point can be expressed as (B, L, H).

[0040] 1.2 Space rectangular coordinate system (OX e Y e Z e )

[0041] The spatial geodetic rectangular coordinate system has the same coordinate axes as the geodetic coordinate system. The coordinates can be expressed by the projection of the point on the various coordinate axes of this coordinate system, such as Figure 1 As shown. The coordinates of a point can be expressed as (X e ,Y e ,Z e )express.

[0042] 1.3 Geographic Coordinate System (O n -X n Y n Z n )

[0043] The coordinate system uses the North-East coordinate system, with the center of mass of the drone as the origin of the coordinate system, the X axis points to the north of the earth, the Y axis points to the east of the earth, and the Z axis is perpendicular to the earth's surface and downward, forming a right-handed coordinate system, such as Figure 2 As shown. The geographical coordinates of a point are (X n ,Y n ,Z n ).

[0044] 1.4 Body coordinate system (O h -X h Y h Z h )

[0045] The origin of the aircraft coordinate system coincides with the origin of the geographic coordinate system, the X-axis points to the front of the aircraft, the Y-axis points to the right wing of the drone, and the Z-axis is parallel to the vertical axis of the drone and points downward. Figure 3 When the UAV has no attitude change, the coordinate system coincides with the spatial rectangular coordinate system, and the body coordinate system of a certain point is expressed as (X h ,Y h ,Z h ).

[0046] 1.5 Camera Coordinate System (O c -X c Y c Z c )

[0047] The origin is located at the intersection of the camera optical axis and the camera lens plane. When the camera is in the initial position, the X axis points to the right wing of the drone, the Y axis points to the bottom of the drone body, and the Z axis is the camera optical axis. When the camera attitude angle changes, the camera coordinates of a certain point can be expressed as (X c ,Y c ,Z c ).

[0048] 2 Error Propagation Theory

[0049] During drone positioning, various error factors can affect the positioning results. Since coordinate transformation is inherently nonlinear, errors undergo corresponding nonlinear transformations during propagation. To effectively analyze this process, we first need to linearize the errors using Taylor series, and then further analyze the error propagation. This section will explore the error propagation mechanisms during target positioning, starting with the principle of Taylor expansion.

[0050] For a random vector x that undergoes a nonlinear transformation f y :

[0051]

[0052] in: For a random vector x=(x1,x2,…x n ) is the mean; Δx is the mean 0 and the covariance is P x The random error variable.

[0053] The first-order Taylor expansion of the above equation at x yields:

[0054] y≈f(x)+d Δx f (2)

[0055] Where: d Λx f is the function f(·) in The total differential of Δx.

[0056] The variance of the random error variable after nonlinear coordinate transformation is:

[0057]

[0058] Among them J x For f(·) The Jacobian matrix at .

[0059] The target positioning process of UAV will be affected by multiple errors. This paper mainly selects UAV position error, UAV attitude angle error, pod heading angle and pitch angle error for analysis. The above errors are all random errors, obey the normal distribution and are assumed to be independent of each other.

[0060] 2.1 Error transfer from camera coordinate system to body coordinate system

[0061] Camera coordinate system coordinate A c =(x c ,y c ,z c ) T , there exists A c ~N(μ c ,P c ), the error between the heading angle α and the pitch angle β of the pod is Ω=(α,β,0) T ~N(μ Ω ,P Ω ),in Diagonal matrix The coordinates of the body coordinate system A obtained after the coordinate transformation of the camera coordinate system are obtained h =(x h,y h ,z h ) T .

[0062] The coordinate transformation is:

[0063]

[0064] in:

[0065]

[0066] 2.2 Error transfer from the body coordinate system to the geographic coordinate system

[0067] Body coordinate system coordinate A h =(x h ,y h ,z h ) T , there exists A h ~N(μ h ,P h ), UAV attitude angle Ψ=(φ,θ,σ) T ~N(μ Ψ ,P Ψ ),in Diagonal matrix The coordinates of the geographic coordinate system A obtained after the coordinate transformation of the body coordinate system n =(x n ,y n ,z n ) T .

[0068] The coordinate transformation is:

[0069]

[0070] in:

[0071]

[0072] 2.3 Error transfer from geographic coordinate system to spatial rectangular coordinate system

[0073] Geographic coordinate system coordinate A n =(x n ,y n ,z n ) T , UAV geodetic coordinate A gfc =(B,L,H) T ~N(μ gfc ,P gfc ),in Diagonal matrix The coordinates of the geographic coordinate system A obtained after the coordinate transformation of the body coordinate system e =(x e ,y e ,z e ) T .

[0074] The coordinate transformation is:

[0075]

[0076] in: It is the translation distance from the spatial rectangular coordinate system to the origin of the geographic coordinate system.

[0077]

[0078] 3 Error synthesis and distribution

[0079] According to equations (4) to (12), the covariance matrix of the target coordinate position in the spatial rectangular coordinate system after coordinate transformation is:

[0080]

[0081] Therefore, the standard deviation of the target positioning error in the spatial rectangular coordinate system is:

[0082]

[0083] Among them, trace(P e ) is the matrix P e The sum of the diagonal elements.

[0084] The total error in a rectangular coordinate system is determined by the combined effects of each individual error. Therefore, error allocation can be used to determine the individual errors and, in turn, to correct the results. Common error allocation methods include the equal influence principle and the probability principle. This paper proposes an improved dung beetle optimization algorithm for error allocation.

[0085] 3.1 Allocating errors according to the principle of equal influence

[0086] Assuming that all errors are random errors and uncorrelated, we have:

[0087]

[0088] The equal influence principle allocates errors based on the principle that each error has equal influence on the total error.

[0089]

[0090] Where σ total is the total error, and n is the number of errors.

[0091] Therefore, the x, y, and z axis errors are:

[0092]

[0093] The equal influence principle can achieve rapid allocation of errors, that is, it assumes that each error has the same impact on the total error. It is the preferred allocation method to meet low-precision requirements, but it will make some measurement errors easier to achieve, while other measurement error requirements are difficult to meet.

[0094] 3.2 Allocation error based on improved dung beetle optimization algorithm

[0095] 3.2.1 Target Allocation Model

[0096] In this paper, the latitude and longitude difference between the target point and the calculation point is used as an indicator to evaluate the positioning accuracy of the UAV. The smaller the difference, the closer the straight-line distance between the target point and the calculation point. At the same time, the error allocation should be as close as possible to the total error to be allocated so that the error to be allocated can be fully utilized. In summary, the objective function and constraints are shown in Equation (18).

[0097]

[0098] Among them: B true and L true Indicates the true latitude and longitude of the target point, B i and L i represents the latitude and longitude of the calculation point, σ total is the total error to be allocated, σ xe ,σ ye ,σ ze is the error assigned to the x, y, and z axes. The constraint restricts the error to be assigned to not exceed the total error.

[0099] 3.2.2 Dung Beetle Optimization Algorithm

[0100] If the error standard deviation is considered the variable to be optimized, and the objective is to minimize the objective function formed by this variable, the originally complex error allocation problem can be abstracted into a mathematical problem of parameter optimization. Therefore, this paper introduces the improved dung beetle optimization algorithm as an efficient solution strategy to achieve the optimal state of error allocation.

[0101] The dung beetle optimization algorithm is inspired by the biological behavior of dung beetles and has the advantages of strong optimization ability and fast convergence speed. The principle of the original dung beetle algorithm is briefly described as follows:

[0102] (1) Ball rolling behavior: When the dung beetle has no obstacles in moving forward, it will use the sun for navigation during the rolling process. The position update formula of the dung beetle rolling ball is shown in formula (19):

[0103]

[0104] Among them, t represents the current iteration number; x i (t) is the position of the i-th dung beetle after t iterations, a is the natural coefficient, when a = 1 there is no deviation, when a = -1 it deviates from the original direction, k is the deflection coefficient, b∈(0,1) is a constant, Δx simulates the change of light intensity, X w is the global worst position.

[0105] (2) Dancing behavior: When a dung beetle encounters an obstacle and cannot move forward, it needs to use dancing to reposition itself. The position update of the dancing behavior is shown in formula (20).

[0106] x i (t+1)=x i (t)+tan(θ)|x i (t)-x i (t+1)| (20)

[0107] Where θ is the deflection angle, which ranges from [0,π].

[0108] (3) Reproduction behavior: In order to provide a safe environment for their offspring, it is crucial for dung beetles to choose a suitable egg-laying site. The strategy for selecting the boundary of the area where dung beetles lay eggs is shown in formula (21).

[0109] Lb * =max(X * ×(1-R),Lb)

[0110] Ub * =min(X * ×(1+R),Ub) (21)

[0111] Among them, Lb * , Ub * Represent the lower and upper bounds of the spawning area, X * is the current local optimal position, T max is the maximum number of iterations set, Lb and Ub represent the lower bound and upper bound of the optimization problem.

[0112] As the number of iterations increases, the spawning area will be dynamically updated, as shown in formula (22).

[0113] T i (t+1)=X * +k1×(T i (t)-Lb * )+k2×(T i (t)-Ub *) (twenty two)

[0114] Among them, T i (t) represents the position of the i-th dung beetle egg at the t-th iteration; k1 and k2 are random vectors of size 1×D, where D represents the dimension of the optimization problem.

[0115] (4) Foraging behavior: Grown dung beetles will emerge from the ground to forage. An optimal dung beetle area needs to be established to guide the dung beetles in foraging. The boundary of the optimal dung beetle area is defined as shown in formula (23).

[0116] Lb b =max(X b ×(1-R),Lb)

[0117] Ub b =min(X b ×(1+R),Ub) (23)

[0118] Among them, X b is the global optimal foraging position, Lb b and Ub b They represent the lower and upper bounds of the optimal foraging area, respectively.

[0119] The position update of the dung beetle is shown in formula (24).

[0120] x i (t+1)=x i (t)+k3×(x i (t)-Lb b )+k4×(x i (t)-Ub b ) (twenty four)

[0121] Among them, k3 and k4 are two independent random numbers, the former obeys the normal distribution, and the latter ranges from 0 to 1.

[0122] (5) Stealing behavior: Stealing dung beetles will steal the dung balls of other dung beetles. The position of the stealing dung beetle is updated as shown in formula (25).

[0123] x i (t+1)=X b +S×g×(|x i (t)-X * |)+(|x i (t)-X b |) (25)

[0124] Where S is a constant value, g is a random vector of size 1×D, and obeys the normal distribution.

[0125] 3.2.3 Improved dung beetle optimization algorithm

[0126] As the number of iterations increases, the DBO algorithm suffers from problems such as low convergence accuracy and easy falling into local optimality. To address these shortcomings, this paper, based on the dung beetle optimization algorithm, performs a vertical and horizontal cross strategy on the target new position after each behavior position is updated.

[0127] The Kent mapping is used to initialize the population. The initialized population has a broader solution space, which helps to enhance the global search ability of the algorithm and improve the convergence speed of the algorithm. The mathematical model of the Kent mapping is:

[0128]

[0129] Where a is a random number in [0,1].

[0130] During the iteration process, the algorithm performs both horizontal and vertical crossover in each generation, so that some dimensions in the population that are trapped in the local optimum have the opportunity to jump out of the iteration. At the same time, the crossover operation can increase the diversity of the population. The horizontal and vertical crossovers are performed successively, and the two crossovers influence each other and work together to improve the algorithm's solution accuracy and accelerate the convergence speed.

[0131] Horizontal crossover is an arithmetic crossover between two different individual particles of the same dimension in the population. The formula is as follows:

[0132] XS hc (i,d)=r1*x(i,d)+(1-r1)*x(j,d)+c1*(x(i,d)-x(j,d)) (27)

[0133] XS hc (j,d)=r1*x(j,d)+(1-r1)*x(i,d)+c1*(x(j,d)-x(i,d)) (28)

[0134] Where: x(i,d) and x(j,d) are the d-th dimensions of individual particles x(i) and x(j) in the parent population, respectively. hc (i,d) and XS hc (j, d) represents the d-th dimension offspring generated by horizontal crossover of x(i, d) and x(j, d), r1 and r2 are random numbers in [0, 1], c1 and c2 are random numbers in [-1, 1]. The obtained offspring are compared with the parent generation, and the individuals with smaller objective function are retained.

[0135] Vertical crossover is an arithmetic crossover between two different dimensions of a particle in the population. Each individual performs a vertical crossover and only updates one of the dimensions, while the other dimensions remain unchanged. The formula is as follows:

[0136] XS lc (i,d1)=r*x(i,d1)+(1-r)*x(i,d2) (29)

[0137] Where: x(i,d1) and x(i,d2) represent the d1th and d2th dimensions of the individual particle x(i), XS lc (i, d1) is the d1-th dimension offspring generated by vertical crossover of the d1-th and d2-th dimensions of x(i). The obtained offspring are compared with the parent generation, and the individuals with smaller objective functions are retained.

[0138] Therefore, the position of the dung beetle becomes equation (29):

[0139]

[0140] Among them, t represents the current iteration number; x i (t) is the position of the i-th dung beetle after t iterations, a is the natural coefficient, when a = 1 there is no deviation, when a = -1 it deviates from the original direction, k is the deflection coefficient, b∈(0,1) is a constant, Δx simulates the change of light intensity, X w is the global worst position, x(i,d) and x(j,d) are the d-th dimension of individual particles x(i) and x(j) in the parent population, respectively, XS hc (i,d) and XS hc (j, d) represents the d-th dimension offspring generated by horizontal crossover of x(i, d) and x(j, d), r1 and r2 are random numbers in [0, 1], c1 and c2 are random numbers in [-1, 1]. The obtained offspring are compared with the parent generation, and the individuals with smaller objective function are retained.

[0141] At the same time, the 3σ principle is introduced. In the mathematical framework of normal distribution, the mean μ represents the average level of all values in the data, and the standard deviation σ measures the degree of dispersion of the data points relative to the mean. That is, in a normal distribution, the probability of a value being distributed within the interval (μ-3σ, μ+3σ) is 0.9973, and the probability of exceeding this range is only 0.3%. Therefore, data outside this interval are discarded. The overall flow chart of the present invention is as follows: Figure 4 As shown, the CCDBO algorithm flow chart is as follows Figure 5 shown.

[0142] 4 Simulation Experiment

[0143] In order to verify the effectiveness of the improved dung beetle algorithm proposed in this paper in improving the target positioning accuracy of UAVs, the particle swarm optimization algorithm (PSO), sparrow optimization algorithm (SSA), gray wolf optimization algorithm (GWO) and original dung beetle algorithm (DBO) are added as comparison to verify the optimization effect of the improved CCDBO algorithm strategy. The population size is set to 30 and the number of algorithm iterations is set to 200. To verify the reliability of the simulation results, the average value of 100 simulation experiment results is taken to draw the fitness change curve. Figure 6 shown.

[0144] Figure 6 The results show that although the initial fitness value of the CCDBO algorithm is large, it can converge quickly during iteration and eventually obtain the global optimal solution with the lowest fitness, which is better than other optimization algorithms.

[0145] Tables 1 and 2 list the objective function fitness value results and the adjusted optimal parameters of DBO, CCDBO, GWO, PSO, and SSA algorithms with 100 and 200 iterations, respectively.

[0146] In order to intuitively demonstrate the effect of CCDBO algorithm on improving the positioning accuracy of UAV targets, this paper uses Monte Carlo statistical method for simulation analysis. The relevant parameters of the UAV are set as shown in Table 1. A single target point is measured. 1) The target point is measured without considering the error to obtain the error-free latitude and longitude coordinates of the target point; 2) The random number function provided by MATLAB is used to generate error terms that obey the respective probability distributions; 3) The generated error is substituted into formula (14) to obtain the error standard deviation of the target point, and the target point is simulated 10,000 times to obtain the unprocessed data; 4) The results are processed according to the equal influence principle, dung beetle optimization algorithm and improved dung beetle algorithm, where the dung beetle population size is set to 30 and the number of algorithm iterations is 200. The simulation results are shown in Figure 2. Figure 6 , Figure 7 As shown, (a) is the unprocessed data, (b) is the data processed according to the equal influence principle, (c) is the data processed by the original dung beetle algorithm, and (d) is the data processed by the improved dung beetle algorithm.

[0147] Table 1 Simulation experimental data

[0148]

[0149] The root mean square error can reflect the positioning performance of the algorithm, and the calculation formula is shown in formula (29).

[0150]

[0151] Where: n is the number of target information, X is the true value of the target information, X iis the simulated value of the target information. According to the above processing method, the root mean square error of the target point latitude and longitude is calculated respectively, and the results are shown in Table 2.

[0152] Table 2 Root mean square error

[0153]

[0154] pass Figure 7 and Figure 8 It can be seen that the processed simulation points are closer to the true value points. Figure 9 The distance between the simulation point and the true value point is intuitively displayed. Table 2 shows that the algorithm proposed in this paper has the smallest root mean square error, proving that the CCDBO algorithm can effectively improve the accuracy of target positioning results. Compared with other results, the accuracy of the simulation points processed by the CCDBO algorithm is improved by 38.97%, 23.48%, and 20.42%, respectively.

[0155] 5 Flight experiment verification

[0156] The experiment uses a CW-25E UAV equipped with an MG150E optoelectronic pod. The installation parameters for the CW-25E UAV are: That is, the distance between the optoelectronic pod and the center of mass of the UAV. In the experiment, the UAV flew at an altitude of about 1100 meters and flew above the target area.

[0157] The experiment sets up 1 center point and 6 target points, such as Figure 10 As shown in the figure, the latitude and longitude of six target points were measured using differential positioning. Because the differential positioning error is at the centimeter level, the measurement results are used as the true values. Ten data collections were performed for each target point. Each set of data was processed using the equal influence principle, the DBO algorithm, and the CCDBO algorithm. The longitude and latitude of the target point were calculated using the pixel coordinate system of the target point captured by the UAV's electro-optical pod. Finally, the average of each data set was compared with the true value, as shown in Tables 3 and 4. The longitude and latitude data in these tables are all north latitude and east longitude.

[0158] Table 3 Target point latitude and longitude

[0159]

[0160] Table 4 Latitude and longitude of target points after processing

[0161]

[0162] The error distances between the three sets of data and the true value points are calculated and shown in Table 5.

[0163] Table 5 Error distance

[0164]

[0165] Taking target point 1 as an example, the positioning results of ten measurement data are as follows: Figure 11 As shown in the figure, with true north as 0°, the target point error distance results processed by different methods are displayed in polar coordinates. By comparison, it can be seen that the data error after CCDBO processing is smaller than that of other processing methods.

[0166] Figure 12 (a) The latitude and longitude of the target point and the true value after taking the average of ten measurement data. Figure 12 (b) is the straight-line distance between the average value point and the target point. It can be seen from the figure that the processed data has a significant decrease in error compared to the unprocessed data. The decrease percentage is shown in Table 6, indicating that the introduction of the CCDBO algorithm can reduce the error of target positioning.

[0167] Table 6 Error reduction percentage

[0168]

[0169] 5 Conclusion

[0170] To achieve more refined target positioning results, this paper establishes a coordinate system combined with spatial homogeneous coordinate transformations and analyzes error propagation based on Taylor series expansion. The latitude and longitude of the target point are calculated based on the drone's position and attitude information, the pod's heading and pitch angles, and the pod's installation position. To reduce positioning error, this paper introduces a total standard deviation, calculated by integrating the standard deviations of each error source. This standard deviation is then used to synthesize and distribute the errors in the data. Furthermore, a CCDBO algorithm is proposed, which incorporates a vertical and horizontal crossover strategy and the 3σ criterion to further improve data accuracy. This process significantly reduces positioning deviations caused by random errors and improves the overall accuracy of UAV target positioning. Monte Carlo simulations of the target point positioning process are conducted, and the results demonstrate the superiority of this method. Flight experiments also validate the reliability of the simulation results, demonstrating that the proposed method can effectively improve UAV target positioning accuracy in practical applications. Future research should focus on exploring more precise error allocation mechanisms, particularly by determining the specific influencing factors of each error source through experiments or data analysis to achieve personalized and refined error allocation. In short, this method can not only further improve the accuracy of UAV target positioning, but also provide more solid technical support for the autonomous navigation and mission execution of future UAVs in complex environments.

[0171] The foregoing is an example of the best mode of carrying out the present invention. Any portion not described in detail herein is common knowledge within the skill of one of ordinary skill in the art. The scope of protection of the present invention is determined by the claims. Any equivalent transformation based on the technical teachings of the present invention is also within the scope of protection of the present invention.

Claims

1. A target positioning error optimization method based on UAV, characterized in that: The following steps are included: Step 1: Utilize a drone to collect target position information, and use a range-angle target positioning method to locate the target. In a coordinate system conversion step, when converting the geographic coordinate system to the world coordinate system, an improved dung beetle optimization algorithm is introduced in the world coordinate system to optimize the coordinate results in the world coordinate system. The improved dung beetle optimization algorithm uses Kent mapping to broaden the population range, and introduces a vertical and horizontal intersection algorithm to update the target position to obtain an updated position. Step 2: Change the position or attitude of the drone and repeat step 1 for the same target. Step 3: Repeat step 2 multiple times to obtain multiple updated positions. Use the 3σ principle to filter the multiple updated positions to obtain filtered positions. Average the filtered positions to obtain the final optimized world coordinate system coordinates. Step 4: Convert the final optimized world coordinate system coordinates to the geodetic coordinate system coordinates and output the final positioning result.

2. The target positioning error optimization method based on UAV according to claim 1 is characterized in that: The dung beetle optimization algorithm uses Kent mapping to initialize the population.

3. The target positioning error optimization method based on UAV according to claim 1 is characterized in that: The improved dung beetle optimization algorithm comprises the following steps: Step 1.1: In each iteration, the target position is obtained using the conventional dung beetle optimization algorithm; Step 1.2: Use the cross-cross algorithm to update the target position in step 1.1 to obtain a new target position. Compare the new target position with the target position in step 1.1, select the target position with a higher fitness value and retain it, and obtain the updated position after this iteration.

4. The target positioning error optimization method based on UAV according to claim 3 is characterized in that: In step 1.2, the position update formula of the cross-cross algorithm is: x i ′(t)=min{r1*x(i,d)+(1-r1)*x(j,d)+c1*(x(i,d)-x(j,d)),r*x(i,d1)+(1-r)*x(i,d2)} |p +a×k×{r1*x(j,d)+(1-r1)*x(i,d)+c1*(x(j,d)-x(i,d))} |p +b×Δx Among them, t represents the current iteration number; x i (t) is the position of the i-th dung beetle after t iterations, a is the natural coefficient, when a = 1 there is no deviation, when a = -1 it deviates from the original direction, k is the deflection coefficient, b∈(0,1) is a constant, Δx simulates the change of light intensity, X w is the global worst position, x(i,d) and x(j,d) are the d-th dimension of individual particles x(i) and x(j) in the parent population, respectively, XS hc (i,d) and XS hc (j,d) represents the d-th dimension offspring generated by horizontal crossover of x(i,d) and x(j,d), r1 and r2 are random numbers in the range [0,1], and c1 and c2 are random numbers in the range [-1,1].

Images