A path planning method for unmanned systems based on image recognition

Through the unmanned system path planning method based on image recognition and improved Tianniu Tribe algorithm, the problem of high resource consumption of artificial bird-driving vehicles is solved, and the efficient path planning of unmanned autonomous bird-driving vehicles is realized, and bird strikes are improved Prevention ability.

CN114913486BActive Publication Date: 2024-11-26CIVIL AVIATION UNIV OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210669428.X
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-06-14
Publication Date
2024-11-26
Estimated Expiration
2042-06-14

AI Technical Summary

Technical Problem

In the prior art, artificially driven bird-driving vehicles require a lot of human and material resources in preventing bird strikes, and their response speed and efficiency are low. How to effectively save human resources and improve bird-driving efficiency through unmanned autonomous driving systems is a Issues that need to be solved urgently.

Method used

Using an unmanned system path planning method based on image recognition, the bird flock activity location and bird-repellent vehicle position are determined by building an environmental model, and the mobile path is planned using the improved Tianniu herd algorithm to ensure that the bird-repellent vehicle can reach the target automatically and effectively Position to drive away the birds.

Benefits of technology

The path planning of unmanned autonomous driving bird-driving vehicles has been realized, human resources have been saved, bird-driving efficiency and response speed have been improved, and bird strike prevention capabilities at the airport have been ensured.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114913486B_ABST
    Figure CN114913486B_ABST
Patent Text Reader

Abstract

The present invention provides an unmanned system path planning method based on image recognition, comprising: step S1: constructing an environmental model; step S2: determining the activity position of a flock of birds by using image recognition technology; step S3: obtaining the position of a bird-repelling vehicle located in an airport at the current moment; step S4: determining the target moving position point corresponding to each of the bird-repelling vehicles based on the activity position of the flock of birds and the position of the bird-repelling vehicle; step S5: planning a moving path based on the position of the bird-repelling vehicle as a starting point and the target moving position point as an end point. The unmanned system path planning method based on image recognition of the present invention realizes the unmanned automatic bird-repelling of the bird-repelling vehicle, saves human resources of the existing method of manually driving the bird-repelling vehicle, and can respond faster, thereby improving the efficiency of bird-repelling.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of path planning, and in particular to an unmanned system path planning method based on image recognition. Background Art

[0002] At present, with the rapid development of the aviation industry and the increasing number of air routes, the competition between aircraft and birds for airspace has become increasingly fierce. The probability of bird strike accidents around the world has shown a clear upward trend, causing serious loss of life and property. Faced with the severe situation of bird strike prevention, domestic and foreign airports have carried out extensive and in-depth bird surveys and adopted a variety of bird strike prevention measures. For example: fixed laser bird repellents and gas cannons and other bird repellent equipment; compared with fixed laser bird repellents and gas cannons and other bird repellent equipment, bird repellent vehicles have the characteristics of flexible operation and large effective radius. Staff can drive vehicles equipped with bird repellent equipment to designated locations and repel birds. However, this method requires a lot of manpower and material resources. If an unmanned automatic driving bird repellent vehicle is used, human resources can be effectively saved. Relatively speaking, how to plan the path of the unmanned driving system is a technical problem that needs to be solved urgently. Summary of the invention

[0003] One of the purposes of the present invention is to provide an unmanned system path planning method based on image recognition to solve the above-mentioned technical problems.

[0004] An unmanned system path planning method based on image recognition provided by an embodiment of the present invention includes:

[0005] Step S1: constructing an environment model;

[0006] Step S2: Determine the location of the flock of birds by using image recognition technology;

[0007] Step S3: obtaining the location of the bird-scaring vehicle in the airport at the current moment;

[0008] Step S4: based on the activity position of the flock of birds and the position of the bird-repelling vehicle, determining the target moving position point corresponding to each of the bird-repelling vehicles;

[0009] Step S5: planning a moving path based on the location of the bird-repelling vehicle as a starting point and the target moving location point as an end point.

[0010] Preferably, the step S5: planning a moving path based on the location of the bird-repelling vehicle as a starting point and the target moving location point as an end point, comprises:

[0011] Step S51: adopting the improved beetle swarm algorithm to realize the planning of the moving path.

[0012] Preferably, the step S51: adopting an improved beetle swarm algorithm to implement mobile path planning includes:

[0013] Step S501: Initialize the position and speed of the longhorn beetle group, set the maximum number of iterations N, the optimization dimension d, the number of iterations parameter t, and other parameters in the initial algorithm;

[0014] Step S502: Calculate the fitness value and path length value of each individual in the beetle group in the initial state, set the current position of the particle as the optimal position of the particle, and the initial beetle group position as the optimal position;

[0015] Step S503: Determine whether the algorithm has reached the maximum number of iterations; if yes, go to step S510, otherwise go to step S504;

[0016] Step S504: calculating the distance between the left and right whiskers of each longhorn beetle, the position coordinates of the left and right whiskers, and calculating the initial value of the learning factor;

[0017] Step S505: updating the speed and position of the longhorn beetle group, comparing the current fitness value of each longhorn beetle individual with the fitness value of the last iteration, and selecting the population position with the smaller fitness value;

[0018] Step S506: If the random number is less than the crossover probability pc, a crossover operation is performed to update the particle position, and the fitness value of each individual longhorn beetle at the current position is compared with the fitness value before the crossover operation, and the population position with the smaller fitness is selected;

[0019] Step S507: If the random number is less than the mutation probability pm, a mutation operation is performed to update the particle position, and the fitness value of each individual position of the longhorn beetle is compared with the fitness value before the Gaussian mutation operation, and the population position with the smaller fitness value is selected;

[0020] Step S508: Determine whether the current number of iterations satisfies the preset conditions at the same time; if yes, update the position of the particle, otherwise go to step S509;

[0021] Step S509: Update the step size and increase the number of iterations by one.

[0022] Step S510: Save and output the current fitness value.

[0023] Preferably, in step S501, when initializing the position and speed of the longhorn beetle group, firstly, a random solution is generated by using the improved Tent with random variables, and the formula is as follows:

[0024]

[0025] X 1=min+max-min×reshape(z, N, d)

[0026] In the formula, z i Represents a column vector of d×N numbers, z i+1 Indicates z i The next scalar, N represents the number of pre-set beetle groups, X 1 represents the initial vector of the first iteration; min is the minimum value of the domain range of each longhorn beetle individual, max is the maximum value of the domain range of each longhorn beetle individual, the reshape function converts the column vector z into a d×N dimensional matrix; rand represents a random variable.

[0027] Preferably, in step S505: updating the speed and position of the longhorn beetle group, wherein:

[0028] The speed update formula is as follows:

[0029]

[0030] In the formula, represents the velocity of the ith particle after the tth iteration in dimension d, represents the position of the ith particle in the tth iteration under dimension d, rand represents a random number between 0 and 1, and X t+1 It indicates the position of the beetle after searching the algorithm. c1, c2, c3 indicate the preset learning factors. gbest indicates the extreme value of the beetle group. pbest indicates the extreme value of the beetle individual.

[0031] The position update formula is as follows:

[0032]

[0033] represents the position of the i-th particle after the t-th iteration in dimension d

[0034] Preferably, performing a crossover operation to update the particle position in step S506 includes:

[0035]

[0036] in, is the position of the beetle group after the tth crossover of the ith particle in dimension d. The crossover rule is that the random coordinate values ​​of two particles are crossed. pc is the crossover probability, which is a constant in the range [0,1].

[0037] Preferably, performing a mutation operation to update the particle position in step S507 includes:

[0038]

[0039] Among them, pm is the mutation probability, which is a constant ranging from [0,1].

[0040] Preferably, step S508: determine whether the current number of iterations satisfies the preset conditions at the same time; if yes, update the position of the particle, otherwise go to step S509;

[0041] The preset conditions include:

[0042] tmod50=0

[0043]

[0044] The particle update formula is as follows:

[0045]

[0046] Among them, mod means the remainder operation, The number of particles is N / 10 and the particle position with higher fitness value is is the position of the particle with the lowest fitness value among the N / 10 particles.

[0047] Preferably, the step length update formula in step S509 is as follows:

[0048] δ t+1 =eta×δ t ;

[0049] In the formula, eta is a preset positive parameter; δ t represents the step size at the tth iteration; δ t+1 Indicates the step size at the t+1th iteration.

[0050] Preferably, step S4: determining the target moving position points corresponding to each of the bird-repelling vehicles based on the activity positions of the flock of birds and the positions of the bird-repelling vehicles, comprises:

[0051] Determine the area where the flock of birds is active and the first coordinates of the center point;

[0052] Determine the second coordinates of the closest point between the take-off and landing area in the airport and the activity position of the flock of birds;

[0053] Determining the third coordinate of each of the bird-repelling vehicles;

[0054] Based on the first coordinate and the second coordinate, determine a first vector, where the first vector starts from the first coordinate and points to the second coordinate;

[0055] Determine a second vector based on the first coordinate and the third coordinate, where the second vector starts from the first coordinate and points to the third coordinate;

[0056] Calculating a first angle between the first vector and each of the second vectors;

[0057] When the first angle is the smallest, extracting the third coordinate corresponding to the second vector, and using an intersection of a line connecting the extracted third coordinate and the first coordinate and a boundary of the region as a first reference point;

[0058] On the boundary of the region, on both sides of the first reference point, a second reference point is extracted at every preset angle along the boundary, and a preset number of second reference points are extracted in total;

[0059] determining a third vector based on a fourth coordinate corresponding to the second reference point and the first coordinate;

[0060] Calculating a second angle between a fourth vector and each of the second vectors;

[0061] When the second angle is the smallest, determining the target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector as the second reference point corresponding to the fourth vector;

[0062] The target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector of the first reference point is determined as the first reference point.

[0063] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings.

[0064] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0066] Figure 1 A schematic diagram of an unmanned system path planning method based on image recognition in an embodiment of the present invention;

[0067] Figure 2 This is a flow chart of an attention mechanism of an improved YOLOv5 network in an embodiment of the present invention;

[0068] Figure 3 It is an improved CSPDarkNet53 backbone network structure in an embodiment of the present invention;

[0069] Figure 4 This is a flow chart of an improved beetle swarm algorithm in an embodiment of the present invention;

[0070] Figure 5 The figure is a schematic diagram of path planning of an unmanned bird-repelling vehicle in an embodiment of the present invention. DETAILED DESCRIPTION

[0071] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0072] The embodiment of the present invention provides a method for unmanned system path planning based on image recognition, such as Figure 1 As shown, including:

[0073] Step S1: constructing an environment model;

[0074] Step S2: Determine the location of the flock of birds by using image recognition technology;

[0075] Step S3: obtaining the location of the bird-scaring vehicle in the airport at the current moment;

[0076] Step S4: based on the activity position of the flock of birds and the position of the bird-repelling vehicle, determining the target moving position point corresponding to each bird-repelling vehicle;

[0077] Step S5: planning a moving path based on the location of the bird-repelling vehicle as the starting point and the target moving location point as the end point.

[0078] The working principle and beneficial effects of the above technical solution are:

[0079] The current environment of the airport is constructed as an environmental model, and the location of the bird flock identified by image recognition technology is determined in the environmental model. Then, according to the current location of each bird-repelling vehicle, the target location point of each bird-repelling vehicle is determined. Based on the location of the bird-repelling vehicle as the starting point and the target moving location point as the end point, the moving path is planned, and the unmanned automatic bird-repelling of the bird-repelling vehicle is realized, which saves the human resources of the existing manual driving bird-repelling vehicle, and can respond faster, thereby improving the efficiency of bird-repelling. Among them, the location of the bird flock activity position through image recognition technology can adopt a target detection algorithm based on deep learning; specifically, it can be divided into two categories, one is a two-stage target detection algorithm, which mainly includes networks such as SPPNet, R-CNN, Fast R-CNN and Faster R-CNN. In the detection process, this type of network first generates a regional candidate frame, extracts features in the candidate frame, and then determines the position frame based on the extracted features and predicts the bird. The other is a one-stage target detection algorithm, which mainly includes a series of networks of SSD and YOLO. This type of algorithm predicts the target while generating the candidate frame, and is characterized by high detection speed. In order to achieve accurate and rapid detection of birds at low altitudes at airports, in addition, the improved YOLOv5 network based on the coordinate attention mechanism can be used. While maintaining the detection speed of the YOLOv5 network, the coordinate attention mechanism is used to capture the network's channel, direction perception and position perception information, which can effectively improve the network's detection accuracy for small bird targets at airports, so as to achieve accurate and efficient detection results. In the bird detection task, the target only exists in a small area in the image or video, which makes it easy for the network to lose the information of the bird during the feature extraction process, resulting in poor detection results. The coordinate attention mechanism (CA) is used to improve the CSPDarkNet53 backbone network in YOLOv5. This attention mechanism can not only capture cross-channel information, but also capture direction perception and position perception information, and is lightweight and flexible. The coordinate attention mechanism is as follows: Figure 2 As shown. The coordinate attention mechanism can be divided into two steps. The first step is to use the pooling layer to encode each channel of the given input according to the horizontal and vertical coordinates, obtain the direction-aware feature map, and embed the coordinate information. The second step is to splice the extracted direction information, transform the information using 1×1 convolution, and then decompose it into two separate tensors along the spatial dimension and use 1×1 convolution to make them have the same number of channels, generate the coordinate information mask as the attention weight, and finally perform pixel-by-pixel multiplication with the input vector as the output of the module. The improved CSPDarkNet53 backbone network structure based on the coordinate attention mechanism is shown in the figure. Figure 3 As shown. Figure 3It can be seen that a coordinate attention module is added after each convolutional layer of the CSPDarkNet53 backbone network, which can be used to capture the small target area features of birds without affecting the overall complexity of the network, helping the network to more accurately locate and identify bird targets and improve the detection accuracy.

[0080] In one embodiment, the step S5: planning a moving path based on the location of the bird-repelling vehicle as a starting point and the target moving location point as an end point, comprises:

[0081] Step S51: adopting the improved beetle swarm algorithm to realize the planning of the moving path.

[0082] Wherein, the step S51: adopting the improved beetle swarm algorithm to realize the planning of the moving path includes:

[0083] First, the improved Tent chaotic sequence is introduced to generate the initial solution to speed up the optimization of the path planning algorithm. The initial solution of the beetle swarm algorithm is randomly generated by the system, but its uneven characteristics will reduce the optimization speed of the algorithm. The Tent mapping can be used as the initial solution of the optimization algorithm. However, the Tent chaotic sequence has unstable periodic points. In order to avoid it from falling into unstable periodic points, the random variable rand2 is introduced. The improved Tent generates the initial random solution formula as follows:

[0084]

[0085] X 1 =min+(max-min)×reshape(z, N, d) (2)

[0086] In formula (1), z i Represents a column vector of d×N numbers, z i+1 Indicates z i The next scalar, N represents the number of pre-set longhorn beetle groups. 1 represents the initial vector of the first iteration; min is the minimum value of the domain range of each longhorn beetle individual, max is the maximum value of the domain range of each longhorn beetle individual, the reshape function converts the column vector z into a d×N dimensional matrix; rand represents a random variable.

[0087] Initialize the position and speed X of the beetle group i =(x i1 , x i2 ,...x in ), V i =(v i1 , v i2 ,...v in ) After that, set the beetle group to search in d-dimensional space, generate a random vector and normalize it:

[0088]

[0089]

[0090]

[0091] In formula (3), t is the normalized vector, formula (4) is used to calculate the coordinates of the left and right whiskers of the longhorn beetle at the tth iteration, I is the distance from the longhorn beetle whisker to the center of mass, and formula (5) is the position update formula of the longhorn beetle whisker search algorithm, where δ is the step size, sign is the sign function, and f is the function to be optimized.

[0092] The position update method in the beetle whisker search algorithm is used to replace the inertia weight part in the particle swarm algorithm, and the position and speed are updated using the individual extreme value pbest of the beetle and the group extreme value gbest of the beetle. In order to further improve the optimization speed and accuracy, the speed, position and step size update formulas of the improved beetle swarm algorithm are shown in equations (6), (7) and (8), respectively:

[0093]

[0094]

[0095] δ t+1 =eta×δ t (8)

[0096] In formula (6), rand is a random number between (0 and 1), t is the number of iterations, c1, c2, c3 are learning factors, and eta in formula (8) is a positive parameter. The sigmoid function is introduced to propose a learning factor update algorithm based on step size. The update formulas are shown in formulas (9), (10) and (11):

[0097]

[0098] c2=2.6-c1 (10)

[0099]

[0100] In formula (9) and formula (11), h<0.

[0101] In order to further improve the optimization accuracy in the later stage of iteration, the particle position update method is shown in formula (12):

[0102]

[0103] tmod50=0 (13)

[0104]

[0105] When the number of iterations t satisfies the conditions of formula (13) and formula (14), the position mutation operation of formula (12) is performed. The number of particles is N / 10 and the particle position with higher fitness value is is the position of the particle with the lowest fitness value among the N / 10 particles.

[0106] The Gaussian mutation strategy is introduced to make the improved beetle swarm algorithm more suitable for the path planning problem of unmanned bird-repelling vehicles. In view of the defect that the diversity of the beetle swarm is reduced with the increase of iteration times, the crossover and mutation method in the genetic algorithm is introduced to generate candidate solutions, and the Gaussian mutation strategy is introduced. After the new beetle swarm position is generated according to formula (6), the crossover operation is introduced, and the position update formula is shown in formula (15):

[0107]

[0108] In formula (15), is the position of the beetle group after crossing. The crossing rule is to cross the random coordinates of two particles. pc is the crossing probability, which is a constant in the range [0,1];

[0109] After the new position of the longhorn beetle group is generated according to formula (15), the fitness value is compared with the population before the crossover. After selecting the better value, the Gaussian mutation operation is introduced. The position update formula is shown in formula (16):

[0110]

[0111] In formula (16), pm is the mutation probability, which is a constant in the range [0,1].

[0112] The path length is the sum of the distances of multiple trajectory segments. Generally, in order to save energy and time for the unmanned bird-repelling vehicle, the path length is required to be as short as possible. The path length cost is shown in formula (17):

[0113]

[0114] In formula (17), fL is the path length cost. This paper uses the penalty function method. If the planned path passes through Figure 2 For the black obstacle area shown in the figure, fL is set to 100 times, that is, 100×fL. P is the number of track points, P-1 represents the number of track segments in a path, and the length of the i-th track segment Li is expressed in Euclidean distance:

[0115]

[0116] Algorithm flow:

[0117] Step 1: According to formula (2), the initialization position and speed of the beetle group are set, and the maximum number of iterations N, the optimization dimension d, the number of iterations t, the initial step size and other parameters needed in the algorithm are set.

[0118] Step 2: Calculate the fitness value and path length value of each individual in the initial situation of the beetle group, set the current position of the particle as the optimal position of the particle, and the initial position of the beetle group as the optimal position.

[0119] Step 3: Determine whether the algorithm has reached the maximum number of iterations. If yes, go to Step 10, otherwise go to Step 4.

[0120] Step 4: Calculate the distance between the left and right whiskers of each longicorn beetle, the position coordinates of the left and right whiskers, and then calculate the initial value of the learning factor.

[0121] Step 5: Update the speed and position of the beetle group according to formula (7), compare the current fitness value of each beetle individual with the fitness value of the previous iteration, and select the population position with the smaller fitness value.

[0122] Step 6: If the random number is less than the crossover probability pc, a crossover operation is performed according to formula (15) to update the particle position, compare the fitness value of each individual beetle at the current position with the fitness value before the crossover operation, and select the population position with the smaller fitness.

[0123] Step 7: If the random number is less than the mutation probability pm, the mutation operation is performed according to formula (16) to update the particle position, compare the fitness value of each individual position of the longhorn beetle with the fitness value before the Gaussian mutation operation, and select the population position with the smaller fitness value.

[0124] Step 8: Determine whether the current number of iterations satisfies the conditions of both equations (13) and (14). If yes, perform position update according to equation (12), otherwise go to Step 9.

[0125] Step 9: Update the step size according to formula (8) and increase the number of iterations by one.

[0126] Step 10: Save the current fitness value and output it, and the algorithm ends.

[0127] The flowchart of the improved beetle swarm algorithm is as follows: Figure 3 As shown; the path planning algorithm of the unmanned bird-repelling vehicle is as follows Figure 4 As shown; Figure 5 This is a schematic diagram of the planned route.

[0128] In order to determine the target moving position points of each bird-repelling vehicle, step S4: based on the activity position of the flock of birds and the position of the bird-repelling vehicle, determining the target moving position points corresponding to each bird-repelling vehicle includes:

[0129] Determine the area of ​​the bird flock activity position and the first coordinate of the center point; spatially locate the bird flock activity position through image recognition technology, and then map it in the environment model, thereby determining the area of ​​the bird flock activity position and the first coordinate of the center point of the area; the center point is a position point inside the area, and the sum of the distances from the center point to each position point on the boundary of the area is the smallest;

[0130] Determine the second coordinates of the closest point between the take-off and landing area in the airport and the location of the flock of birds; determine the take-off and landing area as the take-off and landing area of ​​the airport in the area division in the environment model; determine the second coordinates of the closest point on the take-off and landing area in the airport to the location of the flock of birds based on the minimum value of the distance between the boundary point of the take-off and landing area and the boundary point on the area of ​​the flock of birds;

[0131] Determine the third coordinate of each of the bird-repelling vehicles; obtain positioning data through the positioning module on the bird-repelling vehicle, and then determine the position of each bird-repelling vehicle in the environmental model according to the positioning data, and the coordinate of the position is the third coordinate; of course, the bird-repelling vehicle in the area where the bird activity area is located needs to be stopped;

[0132] Based on the first coordinate and the second coordinate, determine a first vector, where the first vector starts from the first coordinate and points to the second coordinate;

[0133] Determine a second vector based on the first coordinate and the third coordinate, where the second vector starts from the first coordinate and points to the third coordinate;

[0134] Calculate a first angle between the first vector and each of the second vectors; use a cosine vector angle formula to calculate;

[0135] When the first angle is the smallest, extracting the third coordinate corresponding to the second vector, and using an intersection of a line connecting the extracted third coordinate and the first coordinate and a boundary of the region as a first reference point;

[0136] On the boundary of the region, on both sides of the first reference point, a second reference point is extracted at every preset angle (for example, 30 degrees) along the boundary, and a preset number (for example, 4) of the second reference points are extracted in total;

[0137] determining a third vector based on a fourth coordinate corresponding to the second reference point and the first coordinate;

[0138] Calculating a second angle between a fourth vector and each of the second vectors;

[0139] When the second angle is the smallest, determining the target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector as the second reference point corresponding to the fourth vector;

[0140] The target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector of the first reference point is determined as the first reference point.

[0141] The working principle and beneficial effects of the above technical solution are:

[0142] By comprehensively considering the area and center of the bird flock activity location, the location of the airport take-off and landing area, and the current location of each bird-repelling vehicle, a bird-repelling target moving location point is allocated to each bird-repelling vehicle nearby; when the bird-repelling vehicle moves toward the target moving location point, it can be ensured that the bird-repelling vehicle drives away from the airport take-off and landing area, and drives in a wide range of 240 degrees from the bird flock activity area to the take-off and landing area, thereby ensuring the effectiveness of the drive. In addition, by determining the target location point and its corresponding bird-repelling vehicle based on the vector, it can be ensured that the bird-repelling vehicle will not pass through the area where the bird flock activity location is located during driving.

[0143] In one embodiment, the unmanned system path planning method based on image recognition further includes:

[0144] Get the planned movement path;

[0145] Sampling the movement trajectory based on a preset sampling rule to determine a plurality of sampling points;

[0146] Determine the shortest distance between each sampling point and the area where the flock of birds is active;

[0147] An evaluation vector is constructed based on the shortest distance corresponding to each of the sampling points and the position coordinates of the sampling points; each shortest distance is used as a dimension of the evaluation vector, and the shortest distances and the position coordinates of the sampling points are sorted in intervals according to the order of the sampling points on the moving trajectory to construct an evaluation vector;

[0148] Obtain a preset route evaluation library;

[0149] Matching the evaluation vector with the scoring vectors corresponding to the respective scoring values ​​in the route evaluation library; the matching may be performed by calculating the similarity between the evaluation vector and the scoring vector, and specifically, a cosine similarity calculation formula may be used, and when the similarity is greater than a preset threshold (e.g., 0.90) and is the largest in the route evaluation library, determining that the evaluation vector matches the scoring vector;

[0150] Get the score value corresponding to the matching score vector;

[0151] When the score value is less than a preset score threshold, the moving route is deleted, and a new bird-repelling vehicle is selected and / or the moving route is replanned;

[0152] Among them, the route evaluation library is constructed in advance based on a large amount of data analysis; the analysis idea is that the closer the distance difference between the two sampling points is to the bird acquisition area, the closer it is to the sampling distance interval; the corresponding score value is also higher; when the angle between the vector formed by the two sampling points closer to the bird acquisition area and the vector formed by the target moving point and the center is smaller, the score value is higher.

[0153] The principle and beneficial effects of the above technical solution are:

[0154] The moving trajectory is scored through the route evaluation database, and then the bird-repelling effect of the moving trajectory is evaluated; when the trajectory is close to the area where the birds are active, the bird-repelling vehicle is facing the center of the activity area to ensure the bird-repelling effect and direction of the bird-repelling vehicle.

[0155] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A path planning method for unmanned systems based on image recognition, It is characterized in that include: Step S1: constructing an environment model; Step S2: Determine the location of the flock of birds by using image recognition technology; Step S3: obtaining the location of the bird-scaring vehicle in the airport at the current moment; Step S4: based on the activity position of the flock of birds and the position of the bird-repelling vehicle, determining the target moving position point corresponding to each of the bird-repelling vehicles; Step S5: planning a moving path based on the location of the bird-repelling vehicle as a starting point and the target moving location point as an end point; Wherein, the step S5: planning a moving path based on the location of the bird-repelling vehicle as a starting point and the target moving location point as an end point, comprises: Step S51: using an improved beetle swarm algorithm to implement moving path planning; Wherein, the step S51: adopting the improved beetle swarm algorithm to realize the planning of the moving path includes: Step S501: Initialize the position and speed of the longhorn beetle group, set the maximum number of iterations N, the optimization dimension d, the number of iterations parameter t, and other parameters in the initial algorithm; Step S502: Calculate the fitness value and path length value of each individual in the beetle group in the initial state, set the current position of the particle as the optimal position of the particle, and the initial beetle group position as the optimal position; Step S503: Determine whether the algorithm has reached the maximum number of iterations; if yes, go to step S510, otherwise go to step S504; Step S504: calculating the distance between the left and right whiskers of each longhorn beetle, the position coordinates of the left and right whiskers, and calculating the initial value of the learning factor; Step S505: updating the speed and position of the longhorn beetle group, comparing the current fitness value of each longhorn beetle individual with the fitness value of the last iteration, and selecting the population position with the smaller fitness value; Step S506: If the random number is less than the crossover probability pc, a crossover operation is performed to update the particle position, and the fitness value of each individual longhorn beetle at the current position is compared with the fitness value before the crossover operation, and the population position with the smaller fitness is selected; Step S507: If the random number is less than the mutation probability pm, a mutation operation is performed to update the particle position, and the fitness value of each individual position of the longhorn beetle is compared with the fitness value before the Gaussian mutation operation, and the population position with the smaller fitness value is selected; Step S508: Determine whether the current number of iterations satisfies the preset conditions at the same time; if yes, update the position of the particle, otherwise go to step S509; Step S509: Update the step size and increase the number of iterations by one; Step S510: save the current fitness value and output it; Wherein, step S4: based on the activity position of the flock of birds and the position of the bird-repelling vehicle, determining the target moving position point corresponding to each of the bird-repelling vehicles comprises: Determine the area where the flock of birds is active and the first coordinates of the center point; Determine the second coordinates of the closest point between the take-off and landing area in the airport and the activity position of the flock of birds; Determining the third coordinate of each of the bird-repelling vehicles; Based on the first coordinate and the second coordinate, determine a first vector, where the first vector starts from the first coordinate and points to the second coordinate; Determine a second vector based on the first coordinate and the third coordinate, where the second vector starts from the first coordinate and points to the third coordinate; Calculating a first angle between the first vector and each of the second vectors; When the first angle is the smallest, extracting the third coordinate corresponding to the second vector, and using an intersection of a line connecting the extracted third coordinate and the first coordinate and a boundary of the region as a first reference point; On the boundary of the region, on both sides of the first reference point, a second reference point is extracted at every preset angle along the boundary, and a preset number of second reference points are extracted in total; determining a third vector based on a fourth coordinate corresponding to the second reference point and the first coordinate; Calculating a second angle between a fourth vector and each of the second vectors; When the second angle is the smallest, determining the target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector as the second reference point corresponding to the fourth vector; The target moving position point of the bird-repelling vehicle at the third coordinate corresponding to the second vector of the first reference point is determined as the first reference point.

2. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that In step S501, when initializing the position and speed of the longhorn beetle group, the random variable improved Tent is first used to generate a random solution, and the formula is as follows: In the formula, express A column vector of numbers, express The next scalar, Indicates the preset number of longhorn beetle groups, represents the initial vector of the first iteration; min is the minimum value of the domain range of each longhorn beetle individual, max is the maximum value of the domain range of each longhorn beetle individual, and the reshape function converts the column vector z into a d×N dimensional matrix; represents a random variable.

3. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that In step S505: the speed and position of the longhorn beetle group are updated, wherein: The speed update formula is as follows: In the formula, represents the velocity of the ith particle after the tth iteration in dimension d, represents the position of the ith particle in the tth iteration under dimension d, rand represents a random number between 0 and 1, represents the position of the beetle whiskers after the search algorithm, c 1 , c 2 , c 3 represents the preset learning factor; gbest represents the extreme value of the longhorn beetle group; pbest represents the extreme value of the longhorn beetle individual; The position update formula is as follows: represents the position of the i-th particle after the t-th iteration in dimension d.

4. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that In step S506, a crossover operation is performed to update the particle position, including: in, is the position of the beetle group after the tth crossover of the ith particle in dimension d. The crossover rule is that the random coordinate values ​​of two particles are crossed. pc is the crossover probability, which is a constant in the range [0,1].

5. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that In step S507, a mutation operation is performed to update the particle position, including: Among them, pm is the mutation probability, which is a constant ranging from [0,1].

6. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that Step S508: Determine whether the current number of iterations satisfies the preset conditions at the same time; If yes, update the position of the particle, otherwise go to step S509; The preset conditions include: The particle update formula is as follows: Among them, mod means the remainder operation, The number of particles is N / 10 and the particle position with higher fitness value is is the position of the particle with the lowest fitness value among the N / 10 particles.

7. The unmanned system path planning method based on image recognition as claimed in claim 1, It is characterized in that The step length update formula in step S509 is as follows: In the formula, eta is a preset positive parameter; Indicates the step size at the tth iteration; Indicates the step size at the t+1th iteration.