High-adaptability unmanned aerial vehicle automatic charging docking device

Abstract

The invention discloses a high-adaptability unmanned aerial vehicle automatic charging docking device which comprises an unmanned aerial vehicle and a charging device. The unmanned aerial vehicle comprises an unmanned aerial vehicle main body, a camera, an electric quantity monitoring unit, an unmanned aerial vehicle positioning unit, an unmanned aerial vehicle flight height monitoring unit, an unmanned aerial vehicle speed monitoring unit, an unmanned aerial vehicle control unit and a flight driving unit; the automatic charging docking device for the unmanned aerial vehicle can cope with different environments, and the cruising ability of the unmanned aerial vehicle is improved.

Description

Technical Field

[0001] The invention relates to the field of drone charging, and in particular to a highly adaptable drone automatic charging docking device. Background Art

[0002] With the development of science and technology, many high-tech products have emerged in all walks of life. Some high-tech inventions have enriched people's entertainment life, and some inventions have improved human work efficiency. Among them, drones not only meet the needs of life, but also provide people with a new way of entertainment.

[0003] Drones generally refer to aircraft that do not require a pilot to operate the aircraft. There are two main ways to control a drone. The simplest one is to use a remote control, similar to a remote control car, to control the various postures of the aircraft by sending action commands. The other method is more complicated. It is to install a flight control device on the fuselage of the drone, mainly a single-chip microcomputer or an integrated control system, and then write the software into the controller to control the aircraft using the program.

[0004] Drones are small, flexible and convenient. They can be programmed to perform a variety of functions, perform tasks that traditional aircraft cannot perform, and reach environments that ordinary aircraft cannot reach. Especially in harsh geographical environments, using drones to perform flight missions can free up manpower and improve work efficiency.

[0005] The flight time of small drones on the market is generally 20 to 30 minutes, which limits the flight distance of the drone and is not conducive to the long-term operation of the drone.

[0006] Therefore, there is a need for an effective and reliable automatic charging and docking device to improve the endurance of the UAV. Summary of the invention

[0007] The object of the present invention is to provide a highly adaptable UAV automatic charging docking device, comprising a UAV and a charging device;

[0008] The drone includes a drone body, a camera, a power monitoring unit, a drone positioning unit, a drone flight height monitoring unit, a drone speed monitoring unit, a drone control unit and a flight drive unit;

[0009] The power monitoring unit, the UAV positioning unit, the flight control unit and the flight drive unit are integrated inside the UAV body;

[0010] The power monitoring unit monitors the remaining power P of the drone body and transmits it to the drone control unit;

[0011] The remaining power of the drone control unit is P≤50%P max+ΔP or when receiving positioning request information from the charging device, the UAV positioning unit, the UAV flight altitude monitoring unit, and the UAV speed monitoring unit are started; P max is the available power when the drone takes off; ΔP is the reserved power;

[0012] When the drone control unit receives the positioning request information from the charging device, the drone control unit transmits the remaining power of the drone body to the charging device;

[0013] When the drone control unit receives the return charging information from the charging device, it generates a first return signal Q1={Q 11 ,Q 12}, and transmit them to the flight drive unit and camera respectively; Q 11 , Q 12 is the first return sub-signal;

[0014] The drone control unit stores a charging device recognition model and a positioning pattern recognition model;

[0015] The drone control unit uses the charging device recognition model to recognize the image captured by the charging device, and determines whether there is a charging device in the image captured by the charging device. If so, a second return signal Q2 is generated and transmitted to the flight drive unit;

[0016] After receiving the hovering response signal from the flight drive unit, the UAV control unit generates a third return signal Q3={Q 31 ,Q 32}, and transmit them to the flight drive unit and camera respectively; Q 31 , Q 32 is the third return sub-signal;

[0017] After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit;

[0018] The charging device identification model includes an input layer, hidden and output layers; wherein the hidden layer includes an input gate, a forget gate and an output gate;

[0019] The forget control unit f of the hidden layer t , Input control unit i t , input element Memory control unit C t , output control unit O t and output element h t They are as follows:

[0020] f t =σ(W f h t-1 +U f x t +b f ) (1)

[0021] i t =σ(W i h t-1 +U i x t +b i ) (2)

[0022]

[0023]

[0024] o t =σ(W o h t-1 +U o x t +b o ) (5)

[0025] h t =o t ⊙tanh(C t ) (6)

[0026] Where tanh(.) is the hyperbolic tangent function; σ(·) is the loss function; W f , W i , W C , W o Represent the weight matrices of the forget gate, input gate, memory control unit and output gate respectively; b f 、b i 、b C 、b o Represents the offset vectors of the forget gate, input gate, memory control unit and output gate respectively; x t Represents the input of the hidden layer node of the current sequence, h t Represents the hidden layer node output of the current sequence, h t-1 represents the output of the hidden layer node at time t-1; C t is the node state of the hidden layer of the current sequence; ⊙ is the Hamiltonian multiplier; U f , U i , U C , U o represents weight;

[0027] The positioning pattern recognition model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer;

[0028] The input layer obtains a positioning pattern captured image and uses the positioning pattern captured image as input data;

[0029] The number of convolutional layers is F;

[0030] The convolution layer performs a convolution operation on the input data through a filter to extract features at different positions, thereby outputting a feature map;

[0031] Among them, the feature map E output by the convolution layer f f As shown below:

[0032]

[0033] Where: E f-1 represents the feature map of convolutional layer f-1, represents the convolution operation, W f , B f represents the weight and bias of the convolutional layer f; σ() represents the activation function; f = 1, 2, ..., F;

[0034] The number of pooling layers is R;

[0035] The pooling layer reduces the data dimension of the feature map output by the convolution layer through a downsampling operation, and outputs the feature map after dimensionality reduction;

[0036] Among them, the feature map G output by the pooling layer r r As shown below:

[0037] G r =δ(W' r down(G r-1 )+B' r ) (8)

[0038] Where: G r-1 Represents the feature map output by pooling layer r-1, W' r , B' r Represents the weight and bias of the pooling layer r; down() represents the pooling function;

[0039] The fully connected layer maps the extracted feature map into a positioning pattern frame;

[0040] The output of the lth layer of the fully connected layer is as follows:

[0041]

[0042] Where: is the jth neuron output of the lth layer; F i l-1 is the output of the i-th neuron in the l-1th layer; are weights and biases;

[0043] The output layer is used to output the positioning pattern frame;

[0044] The camera receives the first return sub-signal Q 11 When the charging device is in use, an image directly below the drone body is captured, and an image captured by the charging device is generated and transmitted to the drone control unit;

[0045] The camera receives the third return sub-signal Q 31 When the drone is in the position of the target, the image directly below the drone body is captured, a positioning pattern capture image is generated, and the image is transmitted to the drone control unit;

[0046] The flight driving unit receives the first return sub-signal Q 12 Then, the main body of the UAV is driven to fly horizontally at a speed V1 toward the location of the charging device;

[0047] The flight driving unit receives the second return sub-signal Q 22 After that, the UAV body is driven to descend at a speed of V2 until the flight altitude of the UAV body drops to h0; h0 is the hovering altitude; after the flight altitude of the UAV body drops to h0, the flight drive unit sends a hovering response signal to the UAV control unit;

[0048] At the hovering height h0, the flight drive unit generates a signal based on the third return sub-signal Q 32 , driving the drone body to hover over the charging device at a speed V3; the third return sub-signal Q 23 Carrying the two-dimensional coordinate information of the target position I I=(I x ,I y );I x ,I y is the horizontal and vertical coordinates of the target position I;

[0049] After receiving the fourth return sub-signal Q4, the flight drive unit drives the drone body to descend on the charging device at a speed V4; the speed V1>V2>V3>V4;

[0050] The drone positioning unit monitors the two-dimensional position information (x, y) of the drone and transmits it to the charging control unit of the charging device;

[0051] The human-machine flight altitude monitoring unit monitors the flight altitude h of the UAV and transmits it to the charging control unit of the charging device;

[0052] The drone speed monitoring unit monitors the drone flight speed v and transmits it to the charging control unit of the charging device;

[0053] The charging device comprises a housing (2), a storage battery, a charging control unit, a charging device positioning unit, two sets of charging contact units, and an electromagnetic locking unit;

[0054] The battery, charging control unit and charging device positioning unit are integrated inside the housing;

[0055] The charging device positioning unit is used to monitor the two-dimensional position information of the charging device;

[0056] The charging control unit calculates the distance d between the charging device and the main body of the drone in the horizontal direction based on the position of the charging device and the two-dimensional position information (x, y) of the drone;

[0057] The charging control unit determines the remaining power of the drone body Is it true? If so, send return charging information to the drone. If not, After a certain time, the positioning request information is sent to the UAV; P0 is the power consumption per unit time of the UAV flight; Δt is the lead time;

[0058] Two groups of charging contact units are located on the upper surface of the housing (2) and are spaced apart from the electromagnetic locking units;

[0059] The drone body contacts any one of the charging contact units;

[0060] The charging contact unit serves as the positive and negative electrodes of the battery to charge the drone body;

[0061] The electromagnetic locking unit is located on the upper surface of the housing (2) and electromagnetically locks the drone body, thereby fixing the drone on the upper surface of the charging device;

[0062] The process of the drone body hovering over the charging device is as follows:

[0063] 1) The flight drive unit reads the third return sub-signal Q 32 The target position I carried by the drone is determined, and whether the drone body is located at the target position I. If not, the drone body is driven to reach the target position I. The initial coordinates of the target position I are the two-dimensional coordinates of the drone body when the charging device is first photographed.

[0064] 2) The camera is based on the third return sub-signal Q 31 , take an image directly below the drone body, generate a positioning pattern capture image, and transmit it to the drone control unit;

[0065] 3) After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit. If not, it proceeds to step 4);

[0066] The step of determining whether the positioning pattern is located in the central area of ​​the positioning pattern captured image comprises:

[0067] 3.1) The positioning pattern capture image is gridded to obtain a grid matrix A of the positioning pattern capture image, namely:

[0068]

[0069] In the formula, a mn The grid of the mth row and nth column of the positioning pattern capture image; m = 1, 2, ..., M; n = 1, 2, ..., N; M and N are the number of rows and columns of the grid matrix A;

[0070] 3.2) Use the positioning pattern recognition model to recognize the positioning pattern captured image, mark the positioning pattern frame, and construct the grid matrix S of the positioning pattern frame, that is:

[0071]

[0072] In the formula, s m'n' The grid of the m'th row and n'th column of the positioning pattern capture image; m'=1, 2, ..., M'; n'=1, 2, ..., N'; M', N' are the number of rows and columns of the grid matrix S;

[0073] 3.3) Based on the number of rows and columns M' and N' of the grid matrix S, determine the center grid of the positioning pattern frame, and record the grid number of the center grid of the positioning pattern frame in the positioning pattern capture image;

[0074] 3.4) Calculate the grid distances from the center grid of the positioning pattern frame to the four edges of the positioning pattern capture image, denoted as C1 respectively; wherein b1 and b3 are the grid distances from the center grid of the positioning pattern frame to the edges I and III of the positioning pattern capture image; b2 and b4 are the grid distances from the center grid of the positioning pattern frame to the edges II and IV of the positioning pattern capture image; edges I and III are relative; edges II and IV are relative; wherein edges I, II, III and IV of the positioning pattern capture image correspond to the first direction, the second direction, the third direction and the fourth direction around the drone body respectively;

[0075] 3.4) Determine whether b1=b3 and b2=b4. If so, the positioning pattern is located in the central area of ​​the positioning pattern capture image. Otherwise, the positioning pattern is not located in the central area of ​​the positioning pattern capture image.

[0076] 4) Based on b1, b2, b3, and b4, update the two-dimensional coordinate information of the target position I, and update the third return sub-signal Q according to the target position I 32 , and return to step 1).

[0077] Further, the charging contact unit includes a charging contact unit I, a charging contact unit II, a charging contact unit III, and a charging contact unit IV;

[0078] The charging contact unit I, charging contact unit II, charging contact unit III, and charging contact unit IV are all covered on the upper surface of the housing;

[0079] Among them, the charging contact unit I is opposite to the charging contact unit III, serving as the first set of positive and negative electrodes of the battery;

[0080] The charging contact unit II is opposite to the charging contact unit IV and serves as a second group of positive and negative electrodes of the battery.

[0081] Further, when the UAV slide tube contacts the charging contact unit I and the charging contact unit III, the charging contact unit I and the charging contact unit III serve as the positive and negative electrodes of the battery to charge the UAV;

[0082] When the UAV slide pipe contacts the charging contact unit II and the charging contact unit IV, the charging contact unit II and the charging contact unit IV serve as the positive and negative electrodes of the battery to charge the UAV.

[0083] Further, the electromagnetic locking unit includes an electromagnetic locking unit I, an electromagnetic locking unit II, an electromagnetic locking unit III, and an electromagnetic locking unit IV;

[0084] Wherein, the two ends of the electromagnetic locking unit I are in contact with the charging contact unit I and the charging contact unit II respectively;

[0085] The two ends of the electromagnetic locking unit II are in contact with the charging contact unit II and the charging contact unit III respectively;

[0086] Two ends of the electromagnetic locking unit III are in contact with the charging contact unit III and the charging contact unit IV respectively;

[0087] Two ends of the electromagnetic locking unit IV are in contact with the charging contact unit IV and the charging contact unit I respectively;

[0088] The charging contact unit I, the electromagnetic locking unit I, the charging contact unit II, the electromagnetic locking unit II, the charging contact unit III, the electromagnetic locking unit III, the charging contact unit IV and the electromagnetic locking unit IV sequentially form a closed rectangle.

[0089] Further, when determining whether the positioning pattern is located in the center area of ​​the positioning pattern capture image, if the number of grids in the horizontal and vertical directions of the positioning pattern frame is an odd number, the center grid of the positioning pattern frame selects a grid located at the center position, which is recorded as C0; b1, b2, b3, and b4 are the grid distances from the center grid C0 to the edge I, edge II, edge III, and edge IV of the positioning pattern capture image;

[0090] If the number of grids in the horizontal direction of the positioning pattern frame is an even number, and the number of grids in the vertical direction is an odd number, the center grid of the positioning pattern frame selects two grids located at the center position and in the same vertical direction, which are recorded as C1 and C2; b1 and b3 are the grid distances from the center grid C1 or the center grid C2 to the edge I and edge III of the positioning pattern capture image; b2 is the grid distance from the center grid C1 to the edge II of the positioning pattern capture image; b4 is the grid distance from the center grid C2 to the edge IV of the positioning pattern capture image;

[0091] If the number of grids in the horizontal direction of the positioning pattern frame is an odd number, and the number of grids in the vertical direction is an even number, then the center grid of the positioning pattern frame selects two grids located in the center and in the same horizontal direction, which are recorded as C3 and C4; b2 and b4 are the grid distances from the center grid C3 or the center grid C4 to the edge II and edge IV of the positioning pattern capture image; b1 is the grid distance from the center grid C3 to the edge I of the positioning pattern capture image; b3 is the grid distance from the center grid C4 to the edge III of the positioning pattern capture image;

[0092] If the number of grids in the horizontal and vertical directions of the positioning pattern frame is an even number, the central grids of the positioning pattern frame select the four grids located at the center, which are recorded as C5, C6, C7, and C8 respectively; b1 is the grid distance from the central grid C5 to the edge I of the positioning pattern captured image; b2 is the grid distance from the central grid C6 to the edge II of the positioning pattern captured image; b3 is the grid distance from the central grid C7 to the edge III of the positioning pattern captured image; b4 is the grid distance from the central grid C8 to the edge IV of the positioning pattern captured image.

[0093] Furthermore, the step of constructing the charging device identification model includes:

[0094] a1) Obtain multiple sets of positive and negative sample images and label them; the positive sample images are images with a charging device, and the negative sample images are images without a charging device; the labels of the positive sample images are images with a charging device, and the labels of the negative sample images are images without a charging device

[0095] a2) expanding the positive and negative sample images to obtain a sample image set with sample images as input and labels as output; during the expansion, the number of positive sample images is greater than the number of negative sample images;

[0096] a3) randomly dividing the sample image set into a training set and a test set; the number of sample images in the training set is greater than the number of sample images in the test set;

[0097] a4) using the training set to train the convolutional neural network to obtain a charging device recognition model;

[0098] a5) Using the test set to test the charging device identification model, if the error is greater than the error threshold, return to step a1).

[0099] Furthermore, in step a1), the method of expanding the positive and negative sample images includes resampling, flipping, scaling, and translation.

[0100] Further, in step 4), the step of updating the two-dimensional coordinate information of the target position I includes:

[0101] If b1>b3, the updated target position moves in the first direction of the target position before updating, and the moving distance is Δx;

[0102] If b1<b3, the updated target position moves in the third direction of the target position before updating, and the moving distance is Δx;

[0103] If b2>b4, the updated target position moves in the second direction of the target position before updating, and the moving distance is Δy;

[0104] If b2<b4, the updated target position moves in the fourth direction of the target position before updating, and the moving distance is Δy.

[0105] Furthermore, the charging device is movable.

[0106] Furthermore, it also includes a wireless communication module I integrated in the drone body, and a wireless communication module II integrated in the charging device;

[0107] The wireless communication module I and the wireless communication module II are used to realize information exchange between the drone and the charging device.

[0108] The technical effect of the present invention is unquestionable. The present invention provides an automatic charging and docking device for a drone, which can cope with different environments and improve the endurance of the drone.

[0109] The present invention adopts two groups of charging contact units to power the drone in an alternative manner, without adjusting the falling posture of the drone, thereby improving the charging reliability.

[0110] The present invention adopts a four-stage landing method to maximize the return speed and landing accuracy. At the same time, the present invention uses the positioning pattern as a reference for the landing of the drone, further improving the landing accuracy of the drone. BRIEF DESCRIPTION OF THE DRAWINGS

[0111] Figure 1 This is a schematic diagram of the UAV structure;

[0112] Figure 2 is a schematic diagram of the charging device structure;

[0113] Figure 3 It is a schematic diagram of the upper surface structure of the charging device;

[0114] Figure 4 This is a schematic diagram with 1 central grid;

[0115] Figure 5 This is a schematic diagram of a center grid with 2 (horizontal);

[0116] Figure 6 This is a schematic diagram of a center grid with 2 (vertical) grids;

[0117] Figure 7 This is a schematic diagram of a center grid with 4;

[0118] In the figure, charging contact unit I11, charging contact unit II12, charging contact unit III13, charging contact unit IV14, housing 2, electromagnetic locking unit I31, electromagnetic locking unit II32, electromagnetic locking unit III33, electromagnetic locking unit IV34. DETAILED DESCRIPTION

[0119] The present invention is further described below in conjunction with the embodiments, but it should not be understood that the above subject matter of the present invention is limited to the following embodiments. Without departing from the above technical ideas of the present invention, various substitutions and changes are made according to the common technical knowledge and customary means in the art, which should all be included in the protection scope of the present invention.

[0120] Embodiment 1:

[0121] See also Figures 1 to 7 , a highly adaptable UAV automatic charging docking device, comprising a UAV and a charging device;

[0122] The drone includes a drone body, a camera, a power monitoring unit, a drone positioning unit, a drone flight height monitoring unit, a drone speed monitoring unit, a drone control unit and a flight drive unit;

[0123] The power monitoring unit, the UAV positioning unit, the flight control unit and the flight drive unit are integrated inside the UAV body;

[0124] The power monitoring unit monitors the remaining power P of the drone body and transmits it to the drone control unit;

[0125] The remaining power of the drone control unit is P≤50%P max +ΔP or when receiving positioning request information from the charging device, the UAV positioning unit, the UAV flight altitude monitoring unit, and the UAV speed monitoring unit are started; P max is the available power when the drone takes off; ΔP is the reserved power;

[0126] When the drone control unit receives the positioning request information from the charging device, the drone control unit transmits the remaining power of the drone body to the charging device;

[0127] When the drone control unit receives the return charging information from the charging device, it generates a first return signal Q1={Q 11 ,Q 12}, and transmit them to the flight drive unit and camera respectively; Q 11 , Q 12 is the first return sub-signal;

[0128] The drone control unit stores a charging device recognition model and a positioning pattern recognition model;

[0129] The drone control unit uses the charging device recognition model to recognize the image captured by the charging device, and determines whether there is a charging device in the image captured by the charging device. If so, a second return signal Q2 is generated and transmitted to the flight drive unit;

[0130] After receiving the hovering response signal from the flight drive unit, the UAV control unit generates a third return signal Q3={Q 31 ,Q 32}, and transmit them to the flight drive unit and camera respectively; Q 31 , Q 32 is the third return sub-signal;

[0131] After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit;

[0132] The charging device identification model includes an input layer, hidden and output layers; wherein the hidden layer includes an input gate, a forget gate and an output gate;

[0133] The forget control unit f of the hidden layer t , Input control unit i t , input element Memory control unit C t , output control unit O tand output element h t They are as follows:

[0134] f t =σ(W f h t-1 +U f x t +b f ) (1)

[0135] i t =σ(W i h t-1 +U i x t +b i ) (2)

[0136]

[0137]

[0138] o t =σ(W o h t-1 +U o x t +b o ) (5)

[0139] h t =o t ⊙tanh(C t ) (6)

[0140] Where tanh(·) is the hyperbolic tangent function; σ(·) is the loss function; W f , W i , W C , W o Represent the weight matrices of the forget gate, input gate, memory control unit and output gate respectively; b f 、b i 、b C 、b o Represents the offset vectors of the forget gate, input gate, memory control unit and output gate respectively; x t Represents the input of the hidden layer node of the current sequence, h t Represents the hidden layer node output of the current sequence, h t-1 represents the output of the hidden layer node at time t-1; C t is the node state of the hidden layer of the current sequence; ⊙ is the Hamiltonian multiplier; U f , U i , U C , U o represents weight;

[0141] The positioning pattern recognition model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer;

[0142] The input layer obtains a positioning pattern captured image and uses the positioning pattern captured image as input data;

[0143] The number of convolutional layers is F;

[0144] The convolution layer performs a convolution operation on the input data through a filter to extract features at different positions, thereby outputting a feature map;

[0145] Among them, the feature map E output by the convolution layer f f As shown below:

[0146]

[0147] Where: E f-1 represents the feature map of convolutional layer f-1, represents the convolution operation, W f , B f represents the weight and bias of the convolutional layer f; σ() represents the activation function; f = 1, 2, ..., F;

[0148] The number of pooling layers is R;

[0149] The pooling layer reduces the data dimension of the feature map output by the convolution layer through a downsampling operation, and outputs the feature map after dimensionality reduction;

[0150] Among them, the feature map G output by the pooling layer r r As shown below:

[0151] G r =δ(W' r down(G r-1 )+B' r ) (8)

[0152] Where: G r-1 Represents the feature map output by pooling layer r-1, W' r , B' r Represents the weight and bias of the pooling layer r; down() represents the pooling function;

[0153] The fully connected layer maps the extracted feature map into a positioning pattern frame;

[0154] The output of the lth layer of the fully connected layer is as follows:

[0155]

[0156] Where: is the jth neuron output of the lth layer; F il-1 is the output of the i-th neuron in the l-1th layer; are weights and biases;

[0157] The output layer is used to output the positioning pattern frame;

[0158] The camera receives the first return sub-signal Q 11 When the charging device is in use, an image directly below the drone body is captured, an image captured by the charging device is generated, and the image is transmitted to the drone control unit;

[0159] The camera receives the third return sub-signal Q 31 When the drone is in the position of the target, the image directly below the drone body is captured, a positioning pattern capture image is generated, and the image is transmitted to the drone control unit;

[0160] The flight driving unit receives the first return sub-signal Q 12 Then, the main body of the UAV is driven to fly horizontally at a speed V1 toward the location of the charging device;

[0161] The flight driving unit receives the second return sub-signal Q 22 After that, the UAV body is driven to descend at a speed of V2 until the flight altitude of the UAV body drops to h0; h0 is the hovering altitude; after the flight altitude of the UAV body drops to h0, the flight drive unit sends a hovering response signal to the UAV control unit;

[0162] At the hovering height h0, the flight drive unit generates a signal based on the third return sub-signal Q 32 , driving the drone body to hover over the charging device at a speed V3; the third return sub-signal Q 23 Carrying the two-dimensional coordinate information of the target position I I=(I x ,I y );I x ,I y is the horizontal and vertical coordinates of the target position I;

[0163] After receiving the fourth return sub-signal Q4, the flight drive unit drives the drone body to descend on the charging device at a speed V4; the speed V1>V2>V3>V4;

[0164] The drone positioning unit monitors the two-dimensional position information (x, y) of the drone and transmits it to the charging control unit of the charging device;

[0165] The human-machine flight altitude monitoring unit monitors the flight altitude h of the UAV and transmits it to the charging control unit of the charging device;

[0166] The drone speed monitoring unit monitors the drone flight speed v and transmits it to the charging control unit of the charging device;

[0167] The charging device comprises a housing, a storage battery, a charging control unit, a charging device positioning unit, two sets of charging contact units, and an electromagnetic locking unit;

[0168] The battery, charging control unit and charging device positioning unit are integrated inside the housing;

[0169] The charging device positioning unit is used to monitor the two-dimensional position information of the charging device;

[0170] The charging control unit calculates the distance d between the charging device and the main body of the drone in the horizontal direction based on the position of the charging device and the two-dimensional position information (x, y) of the drone;

[0171] The charging control unit determines the remaining power of the drone body Is it true? If so, send return charging information to the drone. If not, After a certain time, the positioning request information is sent to the UAV; P0 is the power consumption per unit time of the UAV flight; Δt is the lead time;

[0172] Two groups of charging contact units are located on the upper surface of the housing and are spaced apart from the electromagnetic locking units;

[0173] The drone body contacts any one of the charging contact units;

[0174] The charging contact unit serves as the positive and negative electrodes of the battery to charge the drone body;

[0175] The electromagnetic locking unit is located on the upper surface of the housing and electromagnetically locks the drone body, thereby fixing the drone on the upper surface of the charging device;

[0176] The positioning pattern is the whole or part of the area on the upper surface of the housing that does not have the charging contact unit and the electromagnetic locking unit. The positioning pattern can be a rectangular pattern.

[0177] The process of the drone body hovering over the charging device is as follows:

[0178] 1) The flight drive unit reads the third return sub-signal Q 32 The target position I carried by the drone is determined, and whether the drone body is located at the target position I. If not, the drone body is driven to reach the target position I. The initial coordinates of the target position I are the two-dimensional coordinates of the drone body when the charging device is first photographed.

[0179] 2) The camera is based on the third return sub-signal Q 31 , take an image directly below the drone body, generate a positioning pattern capture image, and transmit it to the drone control unit;

[0180] 3) After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit. If not, it proceeds to step 4);

[0181] The step of determining whether the positioning pattern is located in the central area of ​​the positioning pattern captured image comprises:

[0182] 3.1) The positioning pattern capture image is gridded to obtain a grid matrix A of the positioning pattern capture image, namely:

[0183]

[0184] In the formula, a mn The grid of the mth row and nth column of the positioning pattern capture image; m = 1, 2, ..., M; n = 1, 2, ..., N; M and N are the number of rows and columns of the grid matrix A;

[0185] 3.2) Use the positioning pattern recognition model to recognize the positioning pattern captured image, mark the positioning pattern frame, and construct the grid matrix S of the positioning pattern frame, that is:

[0186]

[0187] In the formula, s m'n' The grid of the m'th row and n'th column of the positioning pattern capture image; m'=1, 2, ..., M'; n'=1, 2, ..., N'; M', N' are the number of rows and columns of the grid matrix S;

[0188] 3.3) Based on the number of rows and columns M' and N' of the grid matrix S, determine the center grid of the positioning pattern frame, and record the grid number of the center grid of the positioning pattern frame in the positioning pattern capture image;

[0189] 3.4) Calculate the grid distances from the center grid of the positioning pattern frame to the four edges of the positioning pattern capture image, denoted as C1 respectively; wherein b1 and b3 are the grid distances from the center grid of the positioning pattern frame to the edges I and III of the positioning pattern capture image; b2 and b4 are the grid distances from the center grid of the positioning pattern frame to the edges II and IV of the positioning pattern capture image; edges I and III are relative; edges II and IV are relative; wherein edges I, II, III and IV of the positioning pattern capture image correspond to the first direction, the second direction, the third direction and the fourth direction around the drone body respectively;

[0190] 3.4) Determine whether b1=b3 and b2=b4. If so, the positioning pattern is located in the central area of ​​the positioning pattern capture image. Otherwise, the positioning pattern is not located in the central area of ​​the positioning pattern capture image.

[0191] 4) Based on b1, b2, b3, and b4, update the two-dimensional coordinate information of the target position I, and update the third return sub-signal Q according to the target position I 32 , and return to step 1).

[0192] The charging contact unit includes a charging contact unit I, a charging contact unit II, a charging contact unit III, and a charging contact unit IV;

[0193] The charging contact unit I, charging contact unit II, charging contact unit III, and charging contact unit IV are all covered on the upper surface of the housing;

[0194] Among them, the charging contact unit I is opposite to the charging contact unit III, serving as the first set of positive and negative electrodes of the battery;

[0195] The charging contact unit II is opposite to the charging contact unit IV and serves as a second group of positive and negative electrodes of the battery.

[0196] When the UAV slide tube contacts the charging contact unit I and the charging contact unit III, the charging contact unit I and the charging contact unit III act as the positive and negative electrodes of the battery to charge the UAV;

[0197] When the UAV slide pipe contacts the charging contact unit II and the charging contact unit IV, the charging contact unit II and the charging contact unit IV serve as the positive and negative electrodes of the battery to charge the UAV.

[0198] The electromagnetic locking unit includes an electromagnetic locking unit I, an electromagnetic locking unit II, an electromagnetic locking unit III, and an electromagnetic locking unit IV;

[0199] Wherein, the two ends of the electromagnetic locking unit I are in contact with the charging contact unit I and the charging contact unit II respectively;

[0200] The two ends of the electromagnetic locking unit II are in contact with the charging contact unit II and the charging contact unit III respectively;

[0201] Two ends of the electromagnetic locking unit III are in contact with the charging contact unit III and the charging contact unit IV respectively;

[0202] Two ends of the electromagnetic locking unit IV are in contact with the charging contact unit IV and the charging contact unit I respectively;

[0203] The charging contact unit I, the electromagnetic locking unit I, the charging contact unit II, the electromagnetic locking unit II, the charging contact unit III, the electromagnetic locking unit III, the charging contact unit IV and the electromagnetic locking unit IV sequentially form a closed rectangle.

[0204] When judging whether the positioning pattern is located in the center area of ​​the positioning pattern capture image, if the number of grids in the horizontal and vertical directions of the positioning pattern frame is an odd number, the center grid of the positioning pattern frame selects a grid located at the center position, which is recorded as C0; b1, b2, b3, and b4 are the grid distances from the center grid C0 to the edge I, edge II, edge III, and edge IV of the positioning pattern capture image;

[0205] If the number of grids in the horizontal direction of the positioning pattern frame is an even number, and the number of grids in the vertical direction is an odd number, the center grid of the positioning pattern frame selects two grids located at the center position and in the same vertical direction, which are recorded as C1 and C2; b1 and b3 are the grid distances from the center grid C1 or the center grid C2 to the edge I and edge III of the positioning pattern capture image; b2 is the grid distance from the center grid C1 to the edge II of the positioning pattern capture image; b4 is the grid distance from the center grid C2 to the edge IV of the positioning pattern capture image;

[0206] If the number of grids in the horizontal direction of the positioning pattern frame is an odd number, and the number of grids in the vertical direction is an even number, then the center grid of the positioning pattern frame selects two grids located in the center and in the same horizontal direction, which are recorded as C3 and C4; b2 and b4 are the grid distances from the center grid C3 or the center grid C4 to the edge II and edge IV of the positioning pattern capture image; b1 is the grid distance from the center grid C3 to the edge I of the positioning pattern capture image; b3 is the grid distance from the center grid C4 to the edge III of the positioning pattern capture image;

[0207] If the number of grids in the horizontal and vertical directions of the positioning pattern frame is an even number, the central grids of the positioning pattern frame select the four grids located at the center, which are recorded as C5, C6, C7, and C8 respectively; b1 is the grid distance from the central grid C5 to the edge I of the positioning pattern captured image; b2 is the grid distance from the central grid C6 to the edge II of the positioning pattern captured image; b3 is the grid distance from the central grid C7 to the edge III of the positioning pattern captured image; b4 is the grid distance from the central grid C8 to the edge IV of the positioning pattern captured image.

[0208] The steps of constructing the charging device identification model include:

[0209] a1) Obtain multiple sets of positive and negative sample images and label them; the positive sample images are images with a charging device, and the negative sample images are images without a charging device; the labels of the positive sample images are images with a charging device, and the labels of the negative sample images are images without a charging device

[0210] a2) expanding the positive and negative sample images to obtain a sample image set with sample images as input and labels as output; during the expansion, the number of positive sample images is greater than the number of negative sample images;

[0211] a3) randomly dividing the sample image set into a training set and a test set; the number of sample images in the training set is greater than the number of sample images in the test set;

[0212] a4) using the training set to train the convolutional neural network to obtain a charging device recognition model;

[0213] a5) Using the test set to test the charging device identification model, if the error is greater than the error threshold, return to step a1).

[0214] In step a1), the method of expanding the positive and negative sample images includes resampling, flipping, scaling, and translation.

[0215] In step 4), the step of updating the two-dimensional coordinate information of the target position I includes:

[0216] If b1>b3, the updated target position moves in the first direction of the target position before updating, and the moving distance is Δx;

[0217] If b1<b3, the updated target position moves in the third direction of the target position before updating, and the moving distance is Δx;

[0218] If b2>b4, the updated target position moves in the second direction of the target position before updating, and the moving distance is Δy;

[0219] If b2<b4, the updated target position moves in the fourth direction of the target position before updating, and the moving distance is Δy.

[0220] The charging device is movable.

[0221] It also includes a wireless communication module I integrated in the drone body, and a wireless communication module II integrated in the charging device;

[0222] The wireless communication module I and the wireless communication module II are used to realize information exchange between the drone and the charging device.

[0223] Embodiment 2:

[0224] A highly adaptable UAV automatic charging docking device, comprising a UAV and a charging device;

[0225] The drone includes a drone body, a camera, a power monitoring unit, a drone positioning unit, a drone flight height monitoring unit, a drone speed monitoring unit, a drone control unit and a flight drive unit;

[0226] The power monitoring unit, the UAV positioning unit, the flight control unit and the flight drive unit are integrated inside the UAV body;

[0227] The power monitoring unit monitors the remaining power P of the drone body and transmits it to the drone control unit;

[0228] The remaining power of the drone control unit is P≤50%P max +ΔP or when receiving positioning request information from the charging device, the UAV positioning unit, the UAV flight altitude monitoring unit, and the UAV speed monitoring unit are started; P max is the available power when the drone takes off; ΔP is the reserved power;

[0229] When the drone control unit receives the positioning request information from the charging device, the drone control unit transmits the remaining power of the drone body to the charging device;

[0230] When the drone control unit receives the return charging information from the charging device, it generates a first return signal Q1={Q 11 ,Q 12}, and transmit them to the flight drive unit and camera respectively; Q 11 , Q 12 is the first return sub-signal;

[0231] The drone control unit stores a charging device recognition model and a positioning pattern recognition model;

[0232] The drone control unit uses the charging device recognition model to recognize the image captured by the charging device, and determines whether there is a charging device in the image captured by the charging device. If so, a second return signal Q2 is generated and transmitted to the flight drive unit;

[0233] After receiving the hovering response signal from the flight drive unit, the UAV control unit generates a third return signal Q3={Q 31 ,Q 32}, and transmit them to the flight drive unit and camera respectively; Q 31 , Q 32 is the third return sub-signal;

[0234] After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit;

[0235] The charging device identification model includes an input layer, hidden and output layers; wherein the hidden layer includes an input gate, a forget gate and an output gate;

[0236] The forget control unit f of the hidden layer t , Input control unit i t , input element Memory control unit C t , output control unit O t and output element h t They are as follows:

[0237] f t =σ(W f h t-1 +U f x t +b f )(1)

[0238] i t =σ(W i h t-1 +U i x t +b i )(2)

[0239]

[0240] o t =σ(W o h t-1 +U o x t +b o )(5)

[0241] h t =o t ⊙tanh(C t )(6)

[0242] Where tanh(.) is the hyperbolic tangent function; σ(·) is the loss function; W f , W i , W C , W o Represent the weight matrices of the forget gate, input gate, memory control unit and output gate respectively; b f , b i , b C , bo Represents the offset vectors of the forget gate, input gate, memory control unit and output gate respectively; x t Represents the input of the hidden layer node of the current sequence, h t Represents the hidden layer node output of the current sequence, h t-1 represents the output of the hidden layer node at time t-1; C t is the node state of the hidden layer of the current sequence; ⊙ is the Hamiltonian multiplier; U f , U i , U C , U o represents weight;

[0243] The positioning pattern recognition model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer;

[0244] The input layer obtains a positioning pattern captured image and uses the positioning pattern captured image as input data;

[0245] The number of convolutional layers is F;

[0246] The convolution layer performs a convolution operation on the input data through a filter to extract features at different positions, thereby outputting a feature map;

[0247] Among them, the feature map E output by the convolution layer f f As shown below:

[0248]

[0249] Where: E f-1 represents the feature map of convolutional layer f-1, represents the convolution operation, W f , B f represents the weight and bias of the convolutional layer f; σ() represents the activation function; f = 1, 2, ..., F;

[0250] The number of pooling layers is R;

[0251] The pooling layer reduces the data dimension of the feature map output by the convolution layer through a downsampling operation, and outputs the feature map after dimensionality reduction;

[0252] Among them, the feature map G output by the pooling layer r r As shown below:

[0253] G r =δ(W' r down(G r-1 )+B' r ) (8)

[0254] Where: G r-1 Represents the feature map output by pooling layer r-1, W'r , B' r Represents the weight and bias of the pooling layer r; down() represents the pooling function;

[0255] The fully connected layer maps the extracted feature map into a positioning pattern frame;

[0256] The output of the lth layer of the fully connected layer is as follows:

[0257]

[0258] Where: is the jth neuron output of the lth layer; F i l-1 is the output of the i-th neuron in the l-1th layer; are weights and biases;

[0259] The output layer is used to output the positioning pattern frame;

[0260] The camera receives the first return sub-signal Q 11 When the charging device is in use, an image directly below the drone body is captured, an image captured by the charging device is generated, and the image is transmitted to the drone control unit;

[0261] The camera receives the third return sub-signal Q 31 When the drone is in the position of the target, the image directly below the drone body is captured, a positioning pattern capture image is generated, and the image is transmitted to the drone control unit;

[0262] The flight driving unit receives the first return sub-signal Q 12 Then, the main body of the UAV is driven to fly horizontally at a speed V1 toward the location of the charging device;

[0263] The flight driving unit receives the second return sub-signal Q 22 After that, the UAV body is driven to descend at a speed of V2 until the flight altitude of the UAV body drops to h0; h0 is the hovering altitude; after the flight altitude of the UAV body drops to h0, the flight drive unit sends a hovering response signal to the UAV control unit;

[0264] At the hovering height h0, the flight drive unit generates a signal based on the third return sub-signal Q 32 , driving the drone body to hover over the charging device at a speed V3; the third return sub-signal Q 23 Carrying the two-dimensional coordinate information of the target position I I=(I x ,I y );I x ,I y is the horizontal and vertical coordinates of the target position I;

[0265] After receiving the fourth return sub-signal Q4, the flight drive unit drives the drone body to descend on the charging device at a speed V4; the speed V1>V2>V3>V4;

[0266] The drone positioning unit monitors the two-dimensional position information (x, y) of the drone and transmits it to the charging control unit of the charging device;

[0267] The human-machine flight altitude monitoring unit monitors the flight altitude h of the UAV and transmits it to the charging control unit of the charging device;

[0268] The drone speed monitoring unit monitors the drone flight speed v and transmits it to the charging control unit of the charging device;

[0269] The charging device comprises a housing (2), a storage battery, a charging control unit, a charging device positioning unit, two sets of charging contact units, and an electromagnetic locking unit;

[0270] The battery, charging control unit and charging device positioning unit are integrated inside the housing;

[0271] The charging device positioning unit is used to monitor the two-dimensional position information of the charging device;

[0272] The charging control unit calculates the distance d between the charging device and the main body of the drone in the horizontal direction based on the position of the charging device and the two-dimensional position information (x, y) of the drone;

[0273] The charging control unit determines the remaining power of the drone body Is it true? If so, send return charging information to the drone. If not, After a certain time, the positioning request information is sent to the UAV; P0 is the power consumption per unit time of the UAV flight; Δt is the lead time;

[0274] Two groups of charging contact units are located on the upper surface of the housing (2) and are spaced apart from the electromagnetic locking units;

[0275] The drone body contacts any one of the charging contact units;

[0276] The charging contact unit serves as the positive and negative electrodes of the battery to charge the drone body;

[0277] The electromagnetic locking unit is located on the upper surface of the housing (2) and electromagnetically locks the drone body, thereby fixing the drone on the upper surface of the charging device;

[0278] The process of the drone body hovering over the charging device is as follows:

[0279] 1) The flight drive unit reads the third return sub-signal Q 32The target position I carried by the drone is determined, and whether the drone body is located at the target position I. If not, the drone body is driven to reach the target position I. The initial coordinates of the target position I are the two-dimensional coordinates of the drone body when the charging device is first photographed.

[0280] 2) The camera is based on the third return sub-signal Q 31 , take an image directly below the drone body, generate a positioning pattern capture image, and transmit it to the drone control unit;

[0281] 3) After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit. If not, it proceeds to step 4);

[0282] The step of determining whether the positioning pattern is located in the central area of ​​the positioning pattern captured image comprises:

[0283] 3.1) The positioning pattern capture image is gridded to obtain a grid matrix A of the positioning pattern capture image, namely:

[0284]

[0285] In the formula, a mn The grid of the mth row and nth column of the positioning pattern capture image; m = 1, 2, ..., M; n = 1, 2, ..., N; M and N are the number of rows and columns of the grid matrix A;

[0286] 3.2) Use the positioning pattern recognition model to recognize the positioning pattern captured image, mark the positioning pattern frame, and construct the grid matrix S of the positioning pattern frame, that is:

[0287]

[0288] In the formula, s m'n' The grid of the m'th row and n'th column of the positioning pattern capture image; m'=1, 2, ..., M'; n'=1, 2, ..., N'; M', N' are the number of rows and columns of the grid matrix S;

[0289] 3.3) Based on the number of rows and columns M' and N' of the grid matrix S, determine the center grid of the positioning pattern frame, and record the grid number of the center grid of the positioning pattern frame in the positioning pattern capture image;

[0290] 3.4) Calculate the grid distances from the center grid of the positioning pattern frame to the four edges of the positioning pattern capture image, denoted as C1 respectively; wherein b1 and b3 are the grid distances from the center grid of the positioning pattern frame to the edges I and III of the positioning pattern capture image; b2 and b4 are the grid distances from the center grid of the positioning pattern frame to the edges II and IV of the positioning pattern capture image; edges I and III are relative; edges II and IV are relative; wherein edges I, II, III and IV of the positioning pattern capture image correspond to the first direction, the second direction, the third direction and the fourth direction around the drone body respectively;

[0291] 3.4) Determine whether b1=b3 and b2=b4. If so, the positioning pattern is located in the central area of ​​the positioning pattern capture image. Otherwise, the positioning pattern is not located in the central area of ​​the positioning pattern capture image.

[0292] 4) Based on b1, b2, b3, and b4, update the two-dimensional coordinate information of the target position I, and update the third return sub-signal Q according to the target position I 32 , and return to step 1).

[0293] Embodiment 3:

[0294] A highly adaptable automatic charging docking device for unmanned aerial vehicles, the technical content of which is the same as that of Example 2, further, the charging contact unit includes a charging contact unit I11, a charging contact unit II12, a charging contact unit III13, and a charging contact unit IV14;

[0295] The charging contact unit I11, charging contact unit II12, charging contact unit III13, and charging contact unit IV14 are all covered on the upper surface of the shell;

[0296] Among them, the charging contact unit I11 and the charging contact unit III13 are opposite to each other, serving as the first set of positive and negative electrodes of the battery;

[0297] The charging contact unit II12 is opposite to the charging contact unit IV14 and serves as a second group of positive and negative electrodes of the battery.

[0298] Embodiment 4:

[0299] A highly adaptable automatic charging docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-3, further, when the unmanned aerial vehicle sliding tube contacts the charging contact unit I11 and the charging contact unit III13, the charging contact unit I11 and the charging contact unit III13 serve as the positive and negative electrodes of the battery to charge the unmanned aerial vehicle;

[0300] When the UAV sliding pipe contacts the charging contact unit II12 and the charging contact unit IV14, the charging contact unit II12 and the charging contact unit IV14 act as the positive and negative electrodes of the battery to charge the UAV.

[0301] Embodiment 5:

[0302] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-4, further, the electromagnetic locking unit includes an electromagnetic locking unit I31, an electromagnetic locking unit II32, an electromagnetic locking unit III33, and an electromagnetic locking unit IV34;

[0303] Wherein, the two ends of the electromagnetic locking unit I31 are in contact with the charging contact unit I11 and the charging contact unit II12 respectively;

[0304] The two ends of the electromagnetic locking unit II32 are in contact with the charging contact unit II12 and the charging contact unit III13 respectively;

[0305] The two ends of the electromagnetic locking unit III33 are in contact with the charging contact unit III13 and the charging contact unit IV14 respectively;

[0306] The two ends of the electromagnetic locking unit IV34 are in contact with the charging contact unit IV14 and the charging contact unit I11 respectively;

[0307] The charging contact unit I11, the electromagnetic locking unit I31, the charging contact unit II12, the electromagnetic locking unit II32, the charging contact unit III13, the electromagnetic locking unit III33, the charging contact unit IV14 and the electromagnetic locking unit IV34 sequentially form a closed rectangle.

[0308] Embodiment 6:

[0309] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-5, further, when judging whether the positioning pattern is located in the central area of ​​the positioning pattern capture image, if the number of grids in the horizontal direction and the vertical direction of the positioning pattern frame is an odd number, then the central grid of the positioning pattern frame selects a grid located at the center position, which is recorded as C0; b1, b2, b3, and b4 are the grid distances from the central grid C0 to the edge I, edge II, edge III, and edge IV of the positioning pattern capture image;

[0310] If the number of grids in the horizontal direction of the positioning pattern frame is an even number, and the number of grids in the vertical direction is an odd number, the center grid of the positioning pattern frame selects two grids located at the center position and in the same vertical direction, which are recorded as C1 and C2; b1 and b3 are the grid distances from the center grid C1 or the center grid C2 to the edge I and edge III of the positioning pattern capture image; b2 is the grid distance from the center grid C1 to the edge II of the positioning pattern capture image; b4 is the grid distance from the center grid C2 to the edge IV of the positioning pattern capture image;

[0311] If the number of grids in the horizontal direction of the positioning pattern frame is an odd number, and the number of grids in the vertical direction is an even number, then the center grid of the positioning pattern frame selects two grids located in the center and in the same horizontal direction, which are recorded as C3 and C4; b2 and b4 are the grid distances from the center grid C3 or the center grid C4 to the edge II and edge IV of the positioning pattern capture image; b1 is the grid distance from the center grid C3 to the edge I of the positioning pattern capture image; b3 is the grid distance from the center grid C4 to the edge III of the positioning pattern capture image;

[0312] If the number of grids in the horizontal and vertical directions of the positioning pattern frame is an even number, the central grids of the positioning pattern frame select the four grids located at the center, which are recorded as C5, C6, C7, and C8 respectively; b1 is the grid distance from the central grid C5 to the edge I of the positioning pattern captured image; b2 is the grid distance from the central grid C6 to the edge II of the positioning pattern captured image; b3 is the grid distance from the central grid C7 to the edge III of the positioning pattern captured image; b4 is the grid distance from the central grid C8 to the edge IV of the positioning pattern captured image.

[0313] Embodiment 7:

[0314] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of Embodiments 2-6, and further, the steps of constructing the charging device identification model include:

[0315] a1) Obtain multiple sets of positive and negative sample images and label them; the positive sample images are images with a charging device, and the negative sample images are images without a charging device; the labels of the positive sample images are images with a charging device, and the labels of the negative sample images are images without a charging device

[0316] a2) expanding the positive and negative sample images to obtain a sample image set with sample images as input and labels as output; during the expansion, the number of positive sample images is greater than the number of negative sample images;

[0317] a3) randomly dividing the sample image set into a training set and a test set; the number of sample images in the training set is greater than the number of sample images in the test set;

[0318] a4) using the training set to train the convolutional neural network to obtain a charging device recognition model;

[0319] a5) Using the test set to test the charging device identification model, if the error is greater than the error threshold, return to step a1).

[0320] Embodiment 8:

[0321] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-7. Further, in step a1), the method for expanding the positive and negative sample images includes resampling, flipping, scaling, and translation.

[0322] Embodiment 9:

[0323] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-8, further, in step 4), the step of updating the two-dimensional coordinate information of the target position I includes:

[0324] If b1>b3, the updated target position moves in the first direction of the target position before updating, and the moving distance is Δx;

[0325] If b1<b3, the updated target position moves in the third direction of the target position before updating, and the moving distance is Δx;

[0326] If b2>b4, the updated target position moves in the second direction of the target position before updating, and the moving distance is Δy;

[0327] If b2<b4, the updated target position moves in the fourth direction of the target position before updating, and the moving distance is Δy.

[0328] Embodiment 10:

[0329] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of Examples 2-9, and further, the charging device is movable.

[0330] Embodiment 11:

[0331] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of embodiments 2-10, and further comprising a wireless communication module I integrated in the main body of the unmanned aerial vehicle, and a wireless communication module II integrated in the charging device;

[0332] The wireless communication module I and the wireless communication module II are used to realize information exchange between the drone and the charging device.

[0333] Embodiment 12:

[0334] A highly adaptable automatic charging and docking device for unmanned aerial vehicles, the technical content of which is the same as any one of Examples 2-10. Furthermore, when constructing a charging device recognition model, the charging device recognition model is trained using a labeled charging device image, and a 0-1 recognition variable is output. If it is 0, it indicates that the charging device is not recognized, and if it is 1, it indicates that the charging device is recognized.

[0335] When training the charging device recognition model, the loss function used is as follows:

[0336]

[0337] In the formula, y represents the output value and true value of the machine learning model respectively; n is the number of samples.

[0338] When constructing a positioning pattern recognition model, images with labeled positioning patterns are used to train weights and biases in the positioning pattern recognition model.

[0339] Embodiment 13:

[0340] A method for using the highly adaptable drone automatic charging docking device described in any one of Embodiments 1-12, comprising the following steps:

[0341] 1) Record the power P of the drone when it takes off max .

[0342] 2) Use the power monitoring unit to monitor the remaining power P of the drone body in real time and transmit it to the drone control unit.

[0343] When the remaining power of the drone body P≤50%P max +ΔP, the drone positioning unit, the drone flight altitude monitoring unit, and the drone speed monitoring unit are started; ΔP is the reserved power, which is used to ensure the basic flight function of the drone body and prevent the drone body from running out of power before landing on the charging device. In this embodiment, it is set to 1% P'. P' is the maximum power that the drone body battery can store.

[0344] 3) The drone positioning unit monitors the two-dimensional position information (x, y) of the drone and transmits it to the charging control unit of the charging device;

[0345] The human-machine flight altitude monitoring unit monitors the flight altitude h of the UAV and transmits it to the charging control unit of the charging device;

[0346] The drone speed monitoring unit monitors the drone flight speed v and transmits it to the charging control unit of the charging device;

[0347] 4) The charging control unit determines the remaining power of the drone Is it true? If so, send return charging information to the drone and jump to step 6). If not, After a certain time, the positioning request information is sent to the UAV, and the process goes to step 5); P0 is the power consumption per unit time of the UAV flight; Δt is the lead time, which is equal to the sum of the positioning request information transmission time, the UAV main body response time, and the retention time to prevent the power from running out, which can be set to 30s-1min;

[0348] 5) Start the drone positioning unit, the drone flight altitude monitoring unit, and the drone speed monitoring unit, and return to step 3);

[0349] 6) When the UAV control unit receives the return charging information from the charging device, it generates a first return signal Q1={Q 11 ,Q 12}, and transmit them to the flight drive unit and camera respectively;

[0350] 7) The flight drive unit receives the first return sub-signal Q 12 Then, the main body of the UAV is driven to fly horizontally at a speed V1 toward the location of the charging device;

[0351] During the flight, the camera is based on the first return sub-signal Q 11 Taking an image directly below the drone body, generating a charging device capture image, and transmitting it to the drone control unit;

[0352] 8) The UAV control unit uses the charging device recognition model to identify the image captured by the charging device, and determines whether there is a charging device in the image captured by the charging device. If so, a second return signal Q2 is generated and transmitted to the flight drive unit, and the process proceeds to step 9). If not, the process returns to step 7);

[0353] 9) The flight drive unit receives the second return sub-signal Q 22 After that, the UAV body is driven to descend at a speed of V2 until the flight altitude of the UAV body drops to h0; h0 is the hovering altitude; after the flight altitude of the UAV body drops to h0, the flight drive unit sends a hovering response signal to the UAV control unit;

[0354] After receiving the hovering response signal from the flight drive unit, the UAV control unit generates a third return signal Q3={Q 31 ,Q 32}, and transmit them to the flight drive unit and camera respectively; Q 31 , Q 32 is the third return sub-signal;

[0355] 10) At the hovering height h0, the flight drive unit generates a signal based on the third return sub-signal Q32 , driving the drone body to hover over the charging device at a speed V3; during the hovering process, the camera is based on the third return sub-signal Q 31 When the drone body is in the center of the center area, a fourth return signal Q4 is generated and transmitted to the flight drive unit, and the drone control unit receives the positioning pattern capture image from the camera, and then uses the positioning pattern recognition model to recognize the positioning pattern capture image, extract the positioning pattern frame, and determine whether the positioning pattern is located in the center area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit, and the process goes to step 11);

[0356] The specific steps include:

[0357] 10.1) The flight drive unit reads the third return sub-signal Q 32 The target position I carried by the drone is determined, and whether the drone body is located at the target position I. If not, the drone body is driven to reach the target position I. The initial coordinates of the target position I are the two-dimensional coordinates of the drone body when the charging device is first photographed.

[0358] 10.2) The camera is based on the third return sub-signal Q 31 , take an image directly below the drone body, generate a positioning pattern capture image, and transmit it to the drone control unit;

[0359] 10.3) After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit to enter step 11); if not, enter step 10.4);

[0360] The step of determining whether the positioning pattern is located in the central area of ​​the positioning pattern captured image comprises:

[0361] 10.3.1) The positioning pattern capture image is gridded to obtain a grid matrix A of the positioning pattern capture image, namely:

[0362]

[0363] In the formula, a mn The grid of the mth row and nth column of the positioning pattern capture image; m = 1, 2, ..., M; n = 1, 2, ..., N; M and N are the number of rows and columns of the grid matrix A;

[0364] 10.3.2) Use the positioning pattern recognition model to recognize the positioning pattern captured image, mark the positioning pattern frame, and construct the grid matrix S of the positioning pattern frame, that is:

[0365]

[0366] In the formula, s m'n' The grid of the m'th row and n'th column of the positioning pattern capture image; m'=1, 2, ..., M'; n'=1, 2, ..., N'; M', N' are the number of rows and columns of the grid matrix S;

[0367] 10.3.3) Determine the center grid of the positioning pattern frame based on the number of rows and columns M' and N' of the grid matrix S, and record the grid number of the center grid of the positioning pattern frame in the positioning pattern capture image;

[0368] 10.3.4) Calculate the grid distances from the center grid of the positioning pattern frame to the four edges of the positioning pattern capture image, denoted as C1 respectively; wherein, b1 and b3 are the grid distances from the center grid of the positioning pattern frame to the edges I and III of the positioning pattern capture image; b2 and b4 are the grid distances from the center grid of the positioning pattern frame to the edges II and IV of the positioning pattern capture image; edges I and III are relative; edges II and IV are relative; wherein edges I, II, III and IV of the positioning pattern capture image correspond to the first direction, the second direction, the third direction and the fourth direction around the drone body respectively;

[0369] 10.3.4) Determine whether b1=b3 and b2=b4. If so, the positioning pattern is located in the central area of ​​the positioning pattern capture image. Otherwise, the positioning pattern is not located in the central area of ​​the positioning pattern capture image.

[0370] 10.4) Based on b1, b2, b3, and b4, update the two-dimensional coordinate information of the target position I, and update the third return sub-signal Q according to the target position I 32 , and return to step 10.1).

[0371] The steps of updating the two-dimensional coordinate information of the target position I include:

[0372] If b1>b3, the updated target position moves in the first direction of the target position before updating, and the moving distance is Δx;

[0373] If b1<b3, the updated target position moves in the third direction of the target position before updating, and the moving distance is Δx;

[0374] If b2>b4, the updated target position moves in the second direction of the target position before updating, and the moving distance is Δy;

[0375] If b2<b4, the updated target position moves in the fourth direction of the target position before updating, and the moving distance is Δy.

[0376] 11) After receiving the fourth return sub-signal Q4, the flight drive unit drives the drone body to descend onto the charging device at a speed V4.

[0377] 12) The electromagnetic locking unit electromagnetically locks the drone body, thereby fixing the drone on the upper surface of the charging device;

[0378] 13) The sliding tube of the drone body contacts the charging contact unit, and the charging contact unit serves as the positive and negative electrodes of the battery to charge the drone body.

[0379] Embodiment 14:

[0380] A method for using the highly adaptable automatic charging docking device for unmanned aerial vehicles described in any one of Embodiments 1-12, the technical content of which is the same as that of Embodiment 13, further, the charging device recognition model includes an input layer, hidden and output layers; wherein the hidden layer includes an input gate, a forget gate and an output gate;

[0381] The forget control unit f of the hidden layer t , Input control unit i t , input element Memory control unit C t , output control unit O t and output element h t They are as follows:

[0382] f t =σ(W f h t-1 +U f x t +b f ) (3)

[0383] i t =σ(W i h t-1 +U i x t +b i ) (4)

[0384]

[0385] o t =σ(W o h t-1 +U o x t +b o ) (7)

[0386] h t =o t ⊙tanh(C t ) (8)

[0387] Where tanh(.) is the hyperbolic tangent function; σ(·) is the loss function; W f , W i , W C , W o Represent the weight matrices of the forget gate, input gate, memory control unit and output gate respectively; b f , b i , b C , b o Represents the offset vectors of the forget gate, input gate, memory control unit and output gate respectively; x t Represents the input of the hidden layer node of the current sequence, h t Represents the hidden layer node output of the current sequence, h t-1 represents the output of the hidden layer node at time t-1; C t is the node state of the hidden layer of the current sequence; ⊙ is the Hamiltonian multiplier; U f , U i , U C , U o Represents weight.

[0388] Embodiment 15:

[0389] A method for using the highly adaptable automatic charging docking device for unmanned aerial vehicles described in any one of Examples 1-12, the technical content of which is the same as any one of Examples 13-14, and further, the positioning pattern recognition model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer;

[0390] The input layer obtains a positioning pattern captured image and uses the positioning pattern captured image as input data;

[0391] The number of convolutional layers is F;

[0392] The convolution layer performs a convolution operation on the input data through a filter to extract features at different positions, thereby outputting a feature map;

[0393] Among them, the feature map E output by the convolution layer f f As shown below:

[0394]

[0395] Where: E f-1 represents the feature map of convolutional layer f-1, represents the convolution operation, W f , B f represents the weight and bias of the convolutional layer f; σ() represents the activation function; f = 1, 2, ..., F;

[0396] The number of pooling layers is R;

[0397] The pooling layer reduces the data dimension of the feature map output by the convolution layer through a downsampling operation, and outputs the feature map after dimensionality reduction;

[0398] Among them, the feature map G output by the pooling layer r r As shown below:

[0399] G r =σ(W' r down(G r-1 )+B' r ) (10)

[0400] Where: G r-1 Represents the feature map output by pooling layer r-1, W' r , B' r Represents the weight and bias of the pooling layer r; down() represents the pooling function;

[0401] The fully connected layer maps the extracted feature map into a positioning pattern frame;

[0402]

[0403] Where: is the jth neuron in the lth layer;

[0404] The output layer is used to output the positioning pattern frame.

[0405] Embodiment 16:

[0406] A method for using the highly adaptable automatic charging docking device for unmanned aerial vehicles described in any one of Examples 1-12, the technical content is the same as any one of Examples 13-14, further, when judging whether the positioning pattern is located in the central area of ​​the positioning pattern capture image, if the number of grids in the horizontal and vertical directions of the positioning pattern frame is an odd number, then the central grid of the positioning pattern frame selects a grid located at the center position, which is recorded as C0; b1, b2, b3, and b4 are the grid distances from the central grid C0 to the edge I, edge II, edge III, and edge IV of the positioning pattern capture image;

[0407] If the number of grids in the horizontal direction of the positioning pattern frame is an even number, and the number of grids in the vertical direction is an odd number, the center grid of the positioning pattern frame selects two grids located at the center position and in the same vertical direction, which are recorded as C1 and C2; b1 and b3 are the grid distances from the center grid C1 or the center grid C2 to the edge I and edge III of the positioning pattern capture image; b2 is the grid distance from the center grid C1 to the edge II of the positioning pattern capture image; b4 is the grid distance from the center grid C2 to the edge IV of the positioning pattern capture image;

[0408] If the number of grids in the horizontal direction of the positioning pattern frame is an odd number, and the number of grids in the vertical direction is an even number, then the center grid of the positioning pattern frame selects two grids located in the center and in the same horizontal direction, which are recorded as C3 and C4; b2 and b4 are the grid distances from the center grid C3 or the center grid C4 to the edge II and edge IV of the positioning pattern capture image; b1 is the grid distance from the center grid C3 to the edge I of the positioning pattern capture image; b3 is the grid distance from the center grid C4 to the edge III of the positioning pattern capture image;

[0409] If the number of grids in the horizontal and vertical directions of the positioning pattern frame is an even number, the central grids of the positioning pattern frame select the four grids located at the center, which are recorded as C5, C6, C7, and C8 respectively; b1 is the grid distance from the central grid C5 to the edge I of the positioning pattern captured image; b2 is the grid distance from the central grid C6 to the edge II of the positioning pattern captured image; b3 is the grid distance from the central grid C7 to the edge III of the positioning pattern captured image; b4 is the grid distance from the central grid C8 to the edge IV of the positioning pattern captured image.

[0410] Embodiment 17:

[0411] A method for using the highly adaptable automatic charging docking device for unmanned aerial vehicles described in any one of Examples 1-12, the technical content of which is the same as any one of Examples 13-16, further, the step of constructing the charging device identification model includes:

[0412] a1) Obtain multiple sets of positive and negative sample images and label them; the positive sample images are images with a charging device, and the negative sample images are images without a charging device; the labels of the positive sample images are images with a charging device, and the labels of the negative sample images are images without a charging device

[0413] a2) expanding the positive and negative sample images to obtain a sample image set with sample images as input and labels as output; during the expansion, the number of positive sample images is greater than the number of negative sample images;

[0414] a3) randomly dividing the sample image set into a training set and a test set; the number of sample images in the training set is greater than the number of sample images in the test set;

[0415] a4) using the training set to train the convolutional neural network to obtain a charging device recognition model;

[0416] a5) Using the test set to test the charging device identification model, if the error is greater than the error threshold, return to step a1).

Claims

1. A highly adaptable drone automatic charging docking device, characterized by: Includes drone and charging device. The drone includes a drone body, a camera, a power monitoring unit, a drone positioning unit, a drone flight height monitoring unit, a drone speed monitoring unit, a drone control unit and a flight drive unit; The power monitoring unit, the UAV positioning unit, the flight control unit and the flight drive unit are integrated inside the UAV body; The power monitoring unit monitors the remaining power P of the drone body and transmits it to the drone control unit; The remaining power of the drone control unit is P≤50%P max +ΔP or when receiving positioning request information from the charging device, the UAV positioning unit, the UAV flight altitude monitoring unit, and the UAV speed monitoring unit are started; P max is the available power when the drone takes off; ΔP is the reserved power; When the drone control unit receives the positioning request information from the charging device, the drone control unit transmits the remaining power of the drone body to the charging device; When the drone control unit receives the return charging information from the charging device, it generates a first return signal Q1={Q 11 ,Q 12 }, and transmit them to the flight drive unit and camera respectively; Q 11 , Q 12 is the first return sub-signal; The drone control unit stores a charging device recognition model and a positioning pattern recognition model; The drone control unit uses the charging device recognition model to recognize the image captured by the charging device, and determines whether there is a charging device in the image captured by the charging device. If so, a second return signal Q2 is generated and transmitted to the flight drive unit; After receiving the hovering response signal from the flight drive unit, the UAV control unit generates a third return signal Q3={Q 31 ,Q 32 }, and transmit them to the flight drive unit and camera respectively; Q 31 , Q 32 is the third return sub-signal; After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit; The charging device identification model includes an input layer, hidden and output layers; wherein the hidden layer includes an input gate, a forget gate and an output gate; The forget control unit f of the hidden layer t , Input control unit i t , input element Memory control unit C t , output control unit O t and output element h t They are as follows: f t =σ(W f h t-1 +U f x t +b f ) (1) i t =σ(W i h t-1 +U i x t +b i ) (2) the t =σ(W o h t-1 +U o x t +b o ) (5) h t =o t ⊙tanh(C t ) (6) Where tanh(.) is the hyperbolic tangent function; σ(·) is the loss function; W f , W i , W C , W o Represent the weight matrices of the forget gate, input gate, memory control unit and output gate respectively; b f 、b i 、b C 、b o Represents the offset vectors of the forget gate, input gate, memory control unit and output gate respectively; x t Represents the input of the hidden layer node of the current sequence, h t Represents the hidden layer node output of the current sequence, h t-1 represents the output of the hidden layer node at time t-1; C t is the node state of the hidden layer of the current sequence; ⊙ is the Hamiltonian multiplier; U f , U i , U C , U o represents weight; The positioning pattern recognition model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; The input layer obtains a positioning pattern captured image and uses the positioning pattern captured image as input data; The number of convolutional layers is F; The convolution layer performs a convolution operation on the input data through a filter to extract features at different positions, thereby outputting a feature map; Among them, the feature map E output by the convolution layer f f As shown below: Where: E f-1 represents the feature map of convolutional layer f-1, represents the convolution operation, w f 、b f represents the weight and bias of the convolutional layer f; δ() represents the activation function; f = 1, 2, ..., F; The number of pooling layers is R; The pooling layer reduces the data dimension of the feature map output by the convolution layer through a downsampling operation, and outputs the feature map after dimensionality reduction; Among them, the feature map G output by the pooling layer r r As shown below: G r =δ(W' r down(G r-1 )+B' r ) (8) Where: G r-1 Represents the feature map output by pooling layer r-1, W' r , B' r Represents the weight and bias of the pooling layer r; down() represents the pooling function; The fully connected layer maps the extracted feature map into a positioning pattern frame; The output layer is used to output the positioning pattern frame; The camera receives the first return sub-signal Q 11 When the charging device is in use, an image directly below the drone body is captured, an image captured by the charging device is generated, and the image is transmitted to the drone control unit; The camera receives the third return sub-signal Q 31 When the drone is in the air, it takes an image directly below the drone body, generates a positioning pattern capture image, and transmits it to the drone control unit; The flight driving unit receives the first return sub-signal Q 12 Then, the main body of the UAV is driven to fly horizontally at a speed V1 toward the location of the charging device; The flight driving unit receives the second return sub-signal Q 22 After that, the UAV body is driven to descend at a speed of V2 until the flight altitude of the UAV body drops to h0; h0 is the hovering altitude; after the flight altitude of the UAV body drops to h0, the flight drive unit sends a hovering response signal to the UAV control unit; At the hovering height h0, the flight drive unit generates a signal based on the third return sub-signal Q 32 , driving the drone body to hover over the charging device at a speed V3; the third return sub-signal Q 23 Carrying the two-dimensional coordinate information of the target position I I=(I x ,I y );I x ,I y is the horizontal and vertical coordinates of the target position I; After receiving the fourth return sub-signal Q4, the flight drive unit drives the drone body to descend on the charging device at a speed V4; the speed V1>V2>V3>V4; The drone positioning unit monitors the two-dimensional position information (x, y) of the drone and transmits it to the charging control unit of the charging device; The human-machine flight altitude monitoring unit monitors the flight altitude h of the UAV and transmits it to the charging control unit of the charging device; The drone speed monitoring unit monitors the drone flight speed v and transmits it to the charging control unit of the charging device; The charging device comprises a housing (2), a storage battery, a charging control unit, a charging device positioning unit, two sets of charging contact units, and an electromagnetic locking unit; The battery, charging control unit and charging device positioning unit are integrated inside the housing; The charging device positioning unit is used to monitor the two-dimensional position information of the charging device; The charging control unit calculates the distance d between the charging device and the main body of the drone in the horizontal direction based on the position of the charging device and the two-dimensional position information (x, y) of the drone; The charging control unit determines the remaining power of the drone body Is it true? If so, send return charging information to the drone. If not, After a certain time, the positioning request information is sent to the UAV; P0 is the power consumption per unit time of the UAV flight; Δt is the lead time; Two groups of charging contact units are located on the upper surface of the housing (2) and are spaced apart from the electromagnetic locking units; The drone body contacts any one of the charging contact units; The charging contact unit serves as the positive and negative electrodes of the battery to charge the drone body; The electromagnetic locking unit is located on the upper surface of the housing (2) and electromagnetically locks the drone body, thereby fixing the drone on the upper surface of the charging device; The process of the drone body hovering over the charging device is as follows: 1) The flight drive unit reads the third return sub-signal Q 32 The target position I carried by the drone is determined, and whether the drone body is located at the target position I. If not, the drone body is driven to reach the target position I. The initial coordinates of the target position I are the two-dimensional coordinates of the drone body when the charging device is first photographed. 2) The camera is based on the third return sub-signal Q 31 , take an image directly below the drone body, generate a positioning pattern capture image, and transmit it to the drone control unit; 3) After receiving the positioning pattern capture image from the camera, the UAV control unit uses the positioning pattern recognition model to recognize the positioning pattern capture image, extracts the positioning pattern frame, and determines whether the positioning pattern is located in the central area of ​​the positioning pattern capture image. If so, a fourth return signal Q4 is generated and transmitted to the flight drive unit. If not, it proceeds to step 4); The step of determining whether the positioning pattern is located in the central area of ​​the positioning pattern captured image comprises: 3.1) The positioning pattern capture image is gridded to obtain a grid matrix A of the positioning pattern capture image, namely: In the formula, a mn The grid of the mth row and nth column of the positioning pattern capture image; m = 1, 2, ..., M; n = 1, 2, ..., N; M and N are the number of rows and columns of the grid matrix A; 3.2) Use the positioning pattern recognition model to recognize the positioning pattern captured image, mark the positioning pattern frame, and construct the grid matrix S of the positioning pattern frame, that is: In the formula, s m'n' The grid of the m'th row and n'th column of the positioning pattern capture image; m'=1, 2, ..., M'; n'=1, 2, ..., N'; M', N' are the number of rows and columns of the grid matrix S; 3.3) Based on the number of rows and columns M' and N' of the grid matrix S, determine the center grid of the positioning pattern frame, and record the grid number of the center grid of the positioning pattern frame in the positioning pattern capture image; 3.4) Calculate the grid distances from the center grid of the positioning pattern frame to the four edges of the positioning pattern capture image, denoted as C1 respectively; wherein b1 and b3 are the grid distances from the center grid of the positioning pattern frame to the edges I and III of the positioning pattern capture image; b2 and b4 are the grid distances from the center grid of the positioning pattern frame to the edges II and IV of the positioning pattern capture image; edges I and III are relative; edges II and IV are relative; wherein edges I, II, III and IV of the positioning pattern capture image correspond to the first direction, the second direction, the third direction and the fourth direction around the drone body respectively; 3.4) Determine whether b1=b3 and b2=b4. If so, the positioning pattern is located in the central area of ​​the positioning pattern capture image. Otherwise, the positioning pattern is not located in the central area of ​​the positioning pattern capture image. 4) Based on b1, b2, b3, and b4, update the two-dimensional coordinate information of the target position I, and update the third return sub-signal Q according to the target position I 32 , and return to step 1).

2. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: The charging contact unit comprises a charging contact unit I (11), a charging contact unit II (12), a charging contact unit III (13), and a charging contact unit IV (14); The charging contact unit I (11), the charging contact unit II (12), the charging contact unit III (13), and the charging contact unit IV (14) are all covered on the upper surface of the shell; Wherein, the charging contact unit I (11) is opposite to the charging contact unit III (13), serving as the first set of positive and negative electrodes of the battery; The charging contact unit II (12) is opposite to the charging contact unit IV (14) and serves as the second group of positive and negative electrodes of the battery.

3. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 2, characterized in that: When the UAV slide pipe contacts the charging contact unit I (11) and the charging contact unit III (13), the charging contact unit I (11) and the charging contact unit III (13) act as the positive and negative electrodes of the battery to charge the UAV; When the UAV sliding pipe contacts the charging contact unit II (12) and the charging contact unit IV (14), the charging contact unit II (12) and the charging contact unit IV (14) serve as the positive and negative electrodes of the battery to charge the UAV.

4. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 2, characterized in that: The electromagnetic locking unit comprises an electromagnetic locking unit I (31), an electromagnetic locking unit II (32), an electromagnetic locking unit III (33), and an electromagnetic locking unit IV (34); Wherein, two ends of the electromagnetic locking unit I (31) are in contact with the charging contact unit I (11) and the charging contact unit II (12) respectively; Two ends of the electromagnetic locking unit II (32) are in contact with the charging contact unit II (12) and the charging contact unit III (13) respectively; Two ends of the electromagnetic locking unit III (33) are in contact with the charging contact unit III (13) and the charging contact unit IV (14) respectively; Two ends of the electromagnetic locking unit IV (34) are in contact with the charging contact unit IV (14) and the charging contact unit I (11) respectively; The charging contact unit I (11), the electromagnetic locking unit I (31), the charging contact unit II (12), the electromagnetic locking unit II (32), the charging contact unit III (13), the electromagnetic locking unit III (33), the charging contact unit IV (14) and the electromagnetic locking unit IV (34) sequentially form a closed rectangle.

5. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: When judging whether the positioning pattern is located in the center area of ​​the positioning pattern capture image, if the number of grids in the horizontal and vertical directions of the positioning pattern frame is an odd number, the center grid of the positioning pattern frame selects a grid located at the center position, which is recorded as C0; b1, b2, b3, and b4 are the grid distances from the center grid C0 to the edge I, edge II, edge III, and edge IV of the positioning pattern capture image; If the number of grids in the horizontal direction of the positioning pattern frame is an even number, and the number of grids in the vertical direction is an odd number, the center grid of the positioning pattern frame selects two grids located at the center position and in the same vertical direction, which are recorded as C1 and C2; b1 and b3 are the grid distances from the center grid C1 or the center grid C2 to the edge I and edge III of the positioning pattern capture image; b2 is the grid distance from the center grid C1 to the edge II of the positioning pattern capture image; b4 is the grid distance from the center grid C2 to the edge IV of the positioning pattern capture image; If the number of grids in the horizontal direction of the positioning pattern frame is an odd number, and the number of grids in the vertical direction is an even number, then the center grid of the positioning pattern frame selects two grids located in the center and in the same horizontal direction, which are recorded as C3 and C4; b2 and b4 are the grid distances from the center grid C3 or the center grid C4 to the edge II and edge IV of the positioning pattern capture image; b1 is the grid distance from the center grid C3 to the edge I of the positioning pattern capture image; b3 is the grid distance from the center grid C4 to the edge III of the positioning pattern capture image; If the number of grids in the horizontal and vertical directions of the positioning pattern frame is an even number, the central grids of the positioning pattern frame select the four grids located at the center, which are recorded as C5, C6, C7, and C8 respectively; b1 is the grid distance from the central grid C5 to the edge I of the positioning pattern captured image; b2 is the grid distance from the central grid C6 to the edge II of the positioning pattern captured image; b3 is the grid distance from the central grid C7 to the edge III of the positioning pattern captured image; b4 is the grid distance from the central grid C8 to the edge IV of the positioning pattern captured image.

6. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: The steps of constructing the charging device identification model include: a1) Obtain multiple sets of positive and negative sample images and label them; the positive sample images are images with a charging device, and the negative sample images are images without a charging device; the labels of the positive sample images are images with a charging device, and the labels of the negative sample images are images without a charging device a2) expanding the positive and negative sample images to obtain a sample image set with sample images as input and labels as output; during the expansion, the number of positive sample images is greater than the number of negative sample images; a3) randomly dividing the sample image set into a training set and a test set; the number of sample images in the training set is greater than the number of sample images in the test set; a4) using the training set to train the convolutional neural network to obtain a charging device recognition model; a5) Using the test set to test the charging device identification model, if the error is greater than the error threshold, return to step a1).

7. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 6, characterized in that: In step a1), the method of expanding the positive and negative sample images includes resampling, flipping, scaling, and translation.

8. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: In step 4), the step of updating the two-dimensional coordinate information of the target position I includes: If b1>b3, the updated target position moves in the first direction of the target position before updating, and the moving distance is Δx; If b1<b3, the updated target position moves in the third direction of the target position before updating, and the moving distance is Δx; If b2>b4, the updated target position moves in the second direction of the target position before updating, and the moving distance is Δy; If b2<b4, the updated target position moves in the fourth direction of the target position before updating, and the moving distance is Δy.

9. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: The charging device is movable.

10. The highly adaptable automatic charging and docking device for unmanned aerial vehicles according to claim 1, characterized in that: It also includes a wireless communication module I integrated in the drone body, and a wireless communication module II integrated in the charging device; The wireless communication module I and the wireless communication module II are used to realize information exchange between the drone and the charging device.

Images