Fixed-wing unmanned aerial vehicle runway identification method, system and application

Abstract

The invention provides a fixed-wing unmanned aerial vehicle runway recognition method and system and application, and the method comprises the following steps: S1, carrying out the information collection of an unmanned aerial vehicle airport runway through a foresight camera on a fixed-wing unmanned aerial vehicle, and obtaining the historical runway information; s2, a BiSeNet network is established, the BiSeNet network is improved through a space attention mechanism, and the improved BiSeNet network is obtained; s3, labeling the historical runway information to obtain a historical runway information data set; s4, inputting the historical runway information data set into the improved BiSeNet network for training and parameter adjustment, and obtaining an optimal runway semantic segmentation model; s5, current runway information collected by the fixed-wing unmanned aerial vehicle is obtained in real time and input into the optimal runway semantic segmentation model, and a runway recognition result is output. The method is high in applicability and easy to implement, and it can be guaranteed that the unmanned aerial vehicle lands safely under the satellite navigation denial condition.

Description

Technical Field

[0001] The present invention relates to the technical field of fixed-wing UAV autonomous landing, and particularly relates to a runway recognition method, system and application for fixed-wing UAVs.

Background Art

[0002] The key technology of fixed-wing UAV vision-guided landing occupies a very important position in the field of UAVs and is currently being intensively studied by many researchers. The accurate recognition of the UAV runway determines whether the UAV can land safely. Currently, the recognition of the UAV runway is mainly divided into a detection method based on the runway boundary and an image segmentation method based on the road surface, and the latter mainly uses deep learning technology to complete.

[0003] The BiSeNet network is a lightweight end-to-end bilateral semantic segmentation network, mainly composed of a spatial path and a context path. The spatial path mainly uses wide and shallow network channels to obtain relatively rich spatial information, and the context path mainly uses a fast downsampling method to obtain a large receptive field. Through the above methods, the problem of balancing accuracy and real-time performance can be better solved. However, since the present invention is carried on an aircraft, there is a higher requirement for real-time performance, and it is necessary to improve on the basis of this network.

[0004] Therefore, it is necessary to study a runway recognition method, system and application for fixed-wing UAVs to address the deficiencies of the existing technology and solve or alleviate one or more of the above problems.

Summary of the Invention

[0005] In view of this, the present invention provides a runway recognition method, system and application for fixed-wing UAVs. For the situation where the UAV lands under satellite navigation denial conditions, a recognition process of the runway during the UAV landing based on an improved BiSeNet network is provided, which has strong applicability, is easy to implement, and can ensure the safe landing of the UAV under satellite navigation denial conditions.

[0006] On the one hand, the present invention provides a runway recognition method for fixed-wing UAVs, and the runway recognition method for fixed-wing UAVs includes the following steps:

[0007] S1: Collect information on the UAV airport runway through the forward-looking camera on the fixed-wing UAV to obtain historical runway information;

[0008] S2: Build a BiSeNet network, improve the BiSeNet network through a spatial attention mechanism, and obtain the improved BiSeNet network;

[0009] S3: Perform annotation processing on the historical runway information to obtain a historical runway information dataset;

[0010] S4: Input the historical runway information dataset into the improved BiSeNet network for training and parameter adjustment to obtain the best runway semantic segmentation model;

[0011] S5: Obtain the current runway information collected by the fixed-wing drone in real time and input it into the best runway semantic segmentation model to output the runway recognition result.

[0012] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. The content of information collection in S1 includes the drone airport runway information under different times, different angles, and different weather conditions.

[0013] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. The improved BiSeNet network in S2 includes the following processing procedures:

[0014] Upsampling processing: Use the spatial branch in downsampling to extract the spatial features of the runway;

[0015] Semantic feature extraction: Use the context branch to extract the semantic features of the runway;

[0016] Fusion processing: Fuse the spatial features and semantic features;

[0017] Upsampling processing: Use upsampling to improve the resolution of the image.

[0018] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. The improvement of the BiSeNet network by the spatial attention mechanism in S2 specifically includes:

[0019] S21: Improve the context path network into a lightweight MobileNetV2 network as the backbone network;

[0020] S22: Improve the simple channel attention mechanism in the ARM optimization module in the context path into a combination of using the channel attention mechanism and the position attention mechanism;

[0021] S23: Improve the fusion processing to first reduce the dimension of the features output by the spatial branch through 1*1 convolution, secondly further refine the features through residual processing, and finally splice the features with the output features of the context branch, and optimize the fused features through the attention mechanism.

[0022] In the aspect and any possible implementation manner as described above, a further implementation manner is provided. Before the improvement of the context path network in S21, it is the Xception network or the ResNet18 backbone network.

[0023] For the aspects and any possible implementation manners described above, a further implementation manner is provided. In the process of verifying the effect of the improved BiSeNet network in S2, the technical effect of the improved BiSeNet network model is calculated by using a loss function and the MIOU value.

[0024] For the aspects and any possible implementation manners described above, a further implementation manner is provided. In S3, the annotation process is performed by using the Labelme software.

[0025] For the aspects and any possible implementation manners described above, a further implementation manner is provided. Before the historical runway information dataset is input into the improved BiSeNet network in S4, a process of proportionally allocating the historical runway information dataset is further included, where the training set: test set is 4:1.

[0026] For the aspects and any possible implementation manners described above, a further implementation manner provides a fixed-wing UAV runway recognition system, and the fixed-wing UAV runway recognition system includes:

[0027] An information collection module, which collects information on the UAV airport runway through a forward-looking camera on the fixed-wing UAV to obtain historical runway information;

[0028] A network construction module, which constructs a BiSeNet network and improves the BiSeNet network through a spatial attention mechanism to obtain an improved BiSeNet network;

[0029] A dataset acquisition module, which performs annotation processing on the historical runway information to obtain a historical runway information dataset;

[0030] An optimal model acquisition module, which inputs the historical runway information dataset into the improved BiSeNet network for training and parameter adjustment to obtain an optimal runway semantic segmentation model;

[0031] A runway recognition result output module, which obtains the current runway information collected by the fixed-wing UAV in real time and inputs it into the optimal runway semantic segmentation model to output the runway recognition result.

[0032] For the aspects and any possible implementation manners described above, a further implementation manner provides an application of a fixed-wing UAV runway recognition method, and the application is specifically: obtaining a runway recognition result through the fixed-wing UAV runway recognition method, obtaining the slope of the runway edge line through the runway recognition result and performing UAV relative position calculation to obtain the UAV relative runway vertex position, and completing the UAV guided landing.

[0033] Compared with the prior art, the present invention can obtain the following technical effects:

[0034] (1) By replacing the backbone network with the MobileNetV2 network, the number of parameters is reduced, and the real-time performance of the network is greatly improved, making it more adaptable to the running speed of the drone;

[0035] (2) By improving the optimization module and the fusion module, the segmentation accuracy of the runway by the network can be improved, which is beneficial to accurately calculating the slope of the runway edge line subsequently and more accurately resolving the position of the drone;

[0036] (3) Through the above improvements, the network can improve the real-time performance while ensuring the segmentation accuracy, making it more suitable for application in the drone scenario.

[0037] Of course, it is not necessary for any product implementing the present invention to achieve all the above-mentioned technical effects simultaneously.

BRIEF DESCRIPTION OF THE DRAWINGS

[0038] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings based on these drawings without creative efforts.

[0039] Figure 1 It is a schematic diagram of the vision-guided landing of a fixed-wing drone based on semantic segmentation provided by an embodiment of the present invention;

[0040] Figure 2 It is a block diagram of the improved BiSeNetV1 network structure provided by an embodiment of the present invention;

[0041] Figure 3 It is a top view of the fixed-wing drone runway recognition method provided by an embodiment of the present invention;

[0042] Figure 4 It is a framework diagram of the MobilNetV2 network structure provided by an embodiment of the present invention;

[0043] Figure 5 It is a structure diagram of the attention mechanism module provided by an embodiment of the present invention;

[0044] Figure 6 It is a structure diagram of the fusion module provided by an embodiment of the present invention;

[0045] Figure 7 It is a structure diagram of the improved fusion module provided by an embodiment of the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To better understand the technical solutions of the present invention, the embodiments of the present invention will be described in detail below with reference to the drawings.

[0047] It should be clear that the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0048] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms of "a", "the" and "said" used in the embodiments of the present invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0049] The present invention provides a method for identifying a runway of a fixed-wing unmanned aerial vehicle. The method for identifying a runway of a fixed-wing unmanned aerial vehicle includes the following steps:

[0050] S1: Collect information on the airport runway of the unmanned aerial vehicle through the forward-looking camera on the fixed-wing unmanned aerial vehicle to obtain historical runway information;

[0051] S2: Build a BiSeNet network, improve the BiSeNet network through a spatial attention mechanism, and obtain the improved BiSeNet network;

[0052] S3: Perform annotation processing on the historical runway information to obtain a historical runway information data set;

[0053] S4: Input the historical runway information data set into the improved BiSeNet network for training and parameter adjustment to obtain the best runway semantic segmentation model;

[0054] S5: Real-time obtain the current runway information collected by the fixed-wing unmanned aerial vehicle and input it into the best runway semantic segmentation model to output the runway recognition result.

[0055] In a specific embodiment, the content of the information collection in S1 includes the airport runway information of the unmanned aerial vehicle under different times, different angles, and different weather conditions.

[0056] In a specific embodiment, the improved BiSeNet network in S2 includes the following processing procedures:

[0057] Upsampling processing: Extract the spatial features of the runway by using the spatial branch in downsampling;

[0058] Semantic feature extraction: Extract the semantic features of the runway by using the context branch;

[0059] Fusion processing: Fuse the spatial features and semantic features;

[0060] Upsampling processing: Use upsampling to improve the resolution of the image.

[0061] In a specific embodiment, the improvement of the BiSeNet network by the spatial attention mechanism in S2 specifically includes:

[0062] S21: Improve the context path network into a lightweight MobileNetV2 network as the backbone network;

[0063] S22: Improve the simple channel attention mechanism in the ARM optimization module in the context path into a combination of using channel attention mechanism and position attention mechanism;

[0064] S23: Improve the fusion processing by first reducing the dimension of the features output by the spatial branch through 1*1 convolution, secondly further refining the features through residual processing, and finally splicing the output features of the context branch, and optimizing the fused features through the attention mechanism.

[0065] In a specific embodiment, before the improvement of the context path network in S21, it is an Xception network or a ResNet18 backbone network.

[0066] In a specific embodiment, the effect verification process of the improved BiSeNet network in S2 is: Calculate the technical effect of the improved BiSeNet network model by using the loss function and the MIOU value.

[0067] In a specific embodiment, the annotation processing in S3 is performed by Labelme software.

[0068] In a specific embodiment, before the historical runway information dataset is input into the improved BiSeNet network in S4, it also includes a process of proportionally allocating the historical runway information dataset, where the training set: test set is 4:1.

[0069] The present invention also provides a fixed-wing UAV runway recognition system, and the fixed-wing UAV runway recognition system includes:

[0070] An information acquisition module, which collects information on the UAV airport runway through a forward-looking camera on the fixed-wing UAV to obtain historical runway information;

[0071] A network construction module, which constructs a BiSeNet network, improves the BiSeNet network through a spatial attention mechanism, and obtains the improved BiSeNet network;

[0072] A dataset acquisition module, which performs annotation processing on the historical runway information to obtain a historical runway information dataset;

[0073] The optimal model acquisition module inputs the historical runway information dataset into the improved BiSeNet network for training and parameter adjustment to obtain the optimal runway semantic segmentation model;

[0074] The runway recognition result output module obtains the current runway information collected by the fixed-wing UAV in real time and inputs it into the optimal runway semantic segmentation model to output the runway recognition result.

[0075] The present invention also provides an application of a fixed-wing UAV runway recognition method, and the application specifically is: obtaining the runway recognition result through the fixed-wing UAV runway recognition method, obtaining the slope of the runway edge line through the runway recognition result and performing UAV relative position calculation to obtain the relative position of the UAV with respect to the runway vertex, and completing the guided landing of the UAV.

[0076] Embodiment 1:

[0077] The present invention provides a fixed-wing UAV runway recognition method, which specifically includes the following technical solutions:

[0078] Step 1: Use the forward-looking camera on the fixed-wing UAV to collect the UAV airport runway at different times, different angles, and different weathers.

[0079] Step 2: Build an improved BiSeNet network, which includes a downsampling module, an upsampling module, an attention mechanism module, and a fusion module. The spatial branch in the downsampling is used to extract the spatial features of the runway, the context branch is used to extract the semantic features of the runway, then the fusion module is used for fusion, then the upsampling module is used to improve the resolution of the image, and finally the loss function and the MIOU value are used to calculate the technical effect of the improved network model.

[0080] Step 3: Make a dataset, use the Labelme software to label different airport runways at different times, different weathers, and different angles to obtain the dataset.

[0081] Step 4: Use the above-mounted model and the labeled dataset for training, adjusting parameters, and obtaining the optimal segmentation model.

[0082] In the present invention, the network model built in step 2 of the above-mentioned fixed-wing UAV runway recognition based on the improved BiSeNet includes.

[0083] Step 2.1: Improve the Xception network or the ResNet18 backbone network used in the context path to the lightweight MobileNetV2 network as the backbone network to greatly enhance the recognition speed.

[0084] Step 2.2: Improve the simple channel attention mechanism in the ARM optimization module in the context path to a combination of channel attention mechanism and position attention mechanism to optimize the output features of downsampling and improve the accuracy.

[0085] Step 2.3: Improve the fusion module by first reducing the dimension of the features output by the spatial branch through 1*1 convolution, secondly further refining the features through the residual module, and finally splicing the output features of the context branch, and optimizing the fused features through the attention mechanism.

[0086] In the present invention, the specific steps of step 4 of the above-mentioned fixed-wing UAV runway recognition based on the improved BiSeNet are as follows.

[0087] Step 4.1: Divide the labeled data pictures in step 3 into two parts, the training set and the test set, in a ratio of 4:1.

[0088] Step 4.2: Use the network built in step 2 for training.

[0089] Step 4.3: Adopt a joint loss function, that is, the output of the entire network is supervised by the main loss function, and the output of the large receptive field branch is supervised by two auxiliary loss functions. Calculate the effect of the improved BiSeNet network model. The specific expression is as follows.

[0090] All loss functions are Softmax losses.

[0091]

[0092] The above formula is the expression of the Softmax loss function, where p is the predicted output of the network.

[0093]

[0094] The above formula l p is the main loss of the network output, l i is the auxiliary loss; M i represents the output characteristics of the i-th stage of the MobileNetV2 model; W is the weight of the network; the parameter α is the weight for balancing the main loss function and the auxiliary loss function; L is the joint loss function; n is 3.

[0095] Step 4.4: Use the mean intersection over union ratio to evaluate the segmentation result. The intersection over union ratio represents the ratio of the intersection to the union of the region measured by the semantic segmentation model for a certain category and the region marked in the annotation map. The maximum value of the IOU is 1, which means that the measured region completely coincides with the marked region, and the prediction accuracy is 100%; the minimum value of the IOU is 0, and the prediction accuracy is 0. Calculate the IOU value for each category and take its average as the MIOU value. The expression is as follows.

[0096]

[0097] Among them, p ii represents identifying class i as class i, p ij represents identifying class i as class j, p ji represents identifying class j as class i.

[0098] As Figure 1 shown is the schematic diagram of fixed-wing vision-guided landing based on semantic segmentation. First, the corresponding runway semantic segmentation map is obtained through the runway segmentation module in the figure, then the slope of the runway edge line is extracted through the runway edge line extraction, and then the position of the UAV is calculated through the position calculation algorithm, so as to achieve the purpose of guiding the UAV to safely land on the runway.

[0099] As Figure 2 shown is the block diagram of the improved BiSeNetV1 network structure. The spatial path of the improved BiSeNetV1 network consists of four convolutional layers, and each convolutional layer is accompanied by batch normalization operation and ReLU activation operation. The kernel sizes of the convolutional layers are 7*7, 3*3, 3*3, and 1*1 respectively; the stride of the first three stages is 2, and the stride of the fourth stage is 1, achieving the purpose of 8-fold downsampling. In the context path, the MobilNetV2 network is used as the backbone network. First, a convolutional layer with a stride of 2 and a kernel size of 7 is used for 2-fold downsampling, and then 4 MobilNetV2 basic blocks with a stride of 2, and 3 MobilNetV2 basic blocks with a stride of are used for 4-fold, 8-fold, 16-fold, and 32-fold downsampling. Then, a convolutional layer with a kernel size of 1*1 and global average pooling are used as the output of the backbone network. The outputs of 16-fold downsampling and 32-fold downsampling are further optimized by the improved attention mechanism, and then the outputs of the spatial branch and the context path are fused through the improved multi-scale fusion module to obtain the final feature output. Finally, the 8-fold upsampling is performed through the bilinear interpolation upsampling module to convert the low-resolution feature map into a high-resolution output image.

[0100] As Figure 3 shown is the MobilNetV2 basic block. The left side is the basic block with a stride of 1, and the right side is the basic block with a stride of 2. The basic block with a stride of 2 achieves the purpose of downsampling.

[0101] As Figure 4 shown is the composition of the attention mechanism module of the original network. This attention mechanism module first captures global information by means of global average pooling, secondly performs feature transformation through a 1*1 convolution, then performs batch normalization, and finally maps the weights to the range of 0-1 through the Sigmoid activation function to generate an attention feature vector, which is used to optimize the output features after downsampling.

[0102] Figure 5 Shown is the improved attention mechanism. Compared with the original module, the improved module adds a spatial attention mechanism on this basis. It not only assigns different weights to learn in the channels, but also assigns different weights to different spatial pixels for key learning, thereby improving the accuracy.

[0103] As Figure 6 Shown is the fusion module of the original network. First, the different features output by two different branches are concatenated. Secondly, feature extraction is performed through a convolutional layer. Then, the feature is processed through an attention mechanism module similar to that mentioned in the previous section, and a multiplication operation is performed with the feature before processing to obtain a new feature. Finally, the new feature is added to the feature after the previous convolutional processing to obtain the final fusion feature.

[0104] Figure 7 Shown is the improved fusion module. Compared with the original fusion module, first, the output of the spatial branch is passed through a 1×1 convolution to reduce the number of channels. Then, similar to the residual module, the feature is further refined, and the refined feature is concatenated with the output feature of the context branch. The improvement of this module can further refine the output feature of the spatial branch and improve the accuracy.

[0105] The above has introduced in detail a fixed-wing UAV runway recognition method, system and application provided by the embodiments of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present application.

[0106] As used in the specification and claims, certain terms are used to refer to specific components. Those skilled in the art should understand that hardware manufacturers may use different terms to refer to the same component. The specification and claims do not use the difference in names as a way to distinguish components, but use the difference in functions of components as the criterion for distinction. As mentioned throughout the specification and claims, "comprising" and "including" are open-ended terms, so they should be interpreted as "including / including but not limited to". "Roughly" means within an acceptable error range. Those skilled in the art can solve the technical problem within a certain error range and basically achieve the technical effect. The subsequent description in the specification is the preferred implementation manner for implementing the present application, but the description is for the purpose of explaining the general principle of the present application and is not used to limit the scope of the present application. The protection scope of the present application shall be determined by the scope defined by the appended claims.

[0107] It should also be noted that the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a commodity or system comprising a series of elements not only includes those elements but also includes other elements not expressly listed, or elements inherent to such commodity or system. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the commodity or system comprising said element.

[0108] It should be understood that the term "and / or" used herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0109] The above description shows and describes several preferred embodiments of the present application. However, as mentioned above, it should be understood that the present application is not limited to the forms disclosed herein, should not be regarded as an exclusion of other embodiments, but can be used in various other combinations, modifications and environments, and can be changed within the scope of the application concept described herein through the above teachings or the technology or knowledge in the relevant field. And any changes and modifications made by those skilled in the art that do not depart from the spirit and scope of the present application shall fall within the protection scope of the appended claims of the present application.

Claims

1. A runway recognition method for a fixed-wing unmanned aerial vehicle, characterized in that The fixed-wing UAV runway recognition method includes the following steps: S1: Collect information on the UAV airport runway through the forward-looking camera on the fixed-wing UAV to obtain historical runway information; S2: Build a BiSeNet network, improve the BiSeNet network through a spatial attention mechanism, and obtain the improved BiSeNet network; S3: Perform annotation processing on the historical runway information to obtain a historical runway information dataset; S4: Input the historical runway information dataset into the improved BiSeNet network for training and parameter adjustment to obtain the best runway semantic segmentation model; S5: Real-time obtain the current runway information collected by the fixed-wing UAV and input it into the best runway semantic segmentation model to output the runway recognition result.

2. The fixed-wing UAV runway recognition method according to claim 1, characterized in that The content of the information collection in S1 includes the UAV airport runway information under different times, different angles, and different weather conditions.

3. The fixed-wing UAV runway recognition method according to claim 1, characterized in that The improved BiSeNet network in S2 includes the following processing procedures: Upsampling processing: Use the spatial branch in downsampling to extract the spatial features of the runway; Semantic feature extraction: Use the context branch to extract the semantic features of the runway; Fusion processing: Fuse the spatial features and semantic features; Upsampling processing: Use upsampling to improve the resolution of the image.

4. The fixed-wing UAV runway recognition method according to claim 1, wherein The improvement of the BiSeNet network by the spatial attention mechanism in S2 specifically includes: S21: Improve the context path network into a lightweight MobileNetV2 network as the backbone network; S22: Improve the simple channel attention mechanism in the ARM optimization module in the context path into a combination of a channel attention mechanism and a position attention mechanism; S23: Improve the fusion processing to first reduce the dimension of the features output by the spatial branch through 1*1 convolution, secondly further refine the features through residual processing, and finally splice them with the output features of the context branch, and optimize the fused features through the attention mechanism.

5. The fixed-wing UAV runway recognition method according to claim 4, wherein, Before the improvement in S21, the context path network was an Xception network or a ResNet18 backbone network.

6. The fixed-wing UAV runway recognition method according to claim 4, wherein The effect verification process of the improved BiSeNet network in S2 is: Calculate the technical effect of the improved BiSeNet network model using the loss function and the MIOU value.

7. The fixed-wing UAV runway recognition method according to claim 1, wherein The annotation processing in S3 is performed through the Labelme software.

8. The fixed-wing UAV runway recognition method according to claim 1, wherein Before the historical runway information dataset is input into the improved BiSeNet network in S4, it also includes the process of proportionally allocating the historical runway information dataset, where the training set: test set is 4:

1.

9. A fixed-wing UAV runway recognition system, characterized in that, The fixed-wing UAV runway recognition system includes: An information collection module that collects information on the UAV airport runway through the forward-looking camera on the fixed-wing UAV to obtain historical runway information; A network construction module that builds a BiSeNet network, improves the BiSeNet network through a spatial attention mechanism, and obtains the improved BiSeNet network; A dataset acquisition module that performs annotation processing on the historical runway information to obtain a historical runway information dataset; The optimal model acquisition module inputs the historical runway information data set into the improved BiSeNet network for training and parameter adjustment to obtain the optimal runway semantic segmentation model; The runway recognition result output module obtains the current runway information collected by the fixed-wing UAV in real time and inputs it into the optimal runway semantic segmentation model to output the runway recognition result.

10. Application of a runway recognition method for a fixed-wing unmanned aerial vehicle, characterized in that, The specific application is as follows: the runway recognition result is obtained by the fixed-wing UAV runway recognition method according to any one of claims 1-8, the slope of the runway edge line is obtained through the runway recognition result and the relative position of the UAV is calculated, and the relative position of the UAV to the runway vertex is obtained to complete the guided landing of the UAV.

Images