WO2021051278A1

WO2021051278A1 - Earth surface feature identification method and device, unmanned aerial vehicle, and computer readable storage medium

Info

Publication number: WO2021051278A1
Application number: PCT/CN2019/106228
Authority: WO
Inventors: 董双; 李鑫超; 王涛; 李思晋; 梁家斌; 田艺
Original assignee: 深圳市大疆创新科技有限公司
Priority date: 2019-09-17
Filing date: 2019-09-17
Publication date: 2021-03-25
Also published as: CN112154447A

Abstract

An earth surface feature identification method and device, an unmanned aerial vehicle, and a computer readable storage medium. The method comprises: obtaining earth surface image information (S101); processing a plurality of pieces of color channel information, and image depth information to obtain a feature map containing earth surface semantic information (S102); and determining an earth surface feature identification result according to the earth surface semantic information in the feature map (S103). The method improves the accuracy and convenience in earth surface feature identification.

Description

Surface feature recognition method, equipment, drone and computer readable storage medium

Technical field

This application relates to the field of artificial intelligence, and in particular to a method, equipment, drone, and computer-readable storage medium for recognizing ground features.

Background technique

With the rapid development of my country's UAV manufacturing industry, UAVs are growing rapidly in the fields of agriculture, aerial surveys, power line inspections, natural gas (oil) pipeline inspections, forest fire prevention, emergency rescue and disaster relief, and smart cities. In the agricultural field, the automatic spraying of pesticides on crops can be achieved through drones.

At present, various natural disasters, diseases and insect pests on the ground have a great impact on the nature. People can observe the changes on the ground regularly, such as the growth of plants, etc., and then identify the corresponding ground features to determine whether the nature has been affected by nature. Disasters, diseases and insect pests, etc., but it takes a lot of time and labor costs to identify the desired surface features, and the accuracy of the identified results cannot be guaranteed. Therefore, how to improve the accuracy and convenience of surface feature recognition is a problem that needs to be solved urgently.

Summary of the invention

Based on this, the present application provides a method, equipment, unmanned aerial vehicle, and computer-readable storage medium for recognizing ground features, aiming to improve the accuracy and convenience of recognition results of ground features.

In the first aspect, this application provides a land surface feature recognition method, including:

Acquiring ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information;

Processing the multiple color channel information and image depth information to obtain a feature map containing semantic information on the surface;

According to the semantic information of the surface in the feature map, the recognition result of the surface feature is determined.

In a second aspect, the present application also provides an unmanned aerial vehicle including a spraying device and a processor, and the processor is configured to implement the following steps:

Acquiring a flying spraying task, wherein the flying spraying task is determined according to the recognition result of the ground surface features;

Perform the flying spraying task, and control the spraying device to execute the corresponding spraying action according to the spraying parameters in the flying spraying task.

In a third aspect, the present application also provides a surface feature recognition device, the surface feature recognition device including a memory and a processor;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the following steps:

In a fourth aspect, the present application also provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor realizes the above The surface feature recognition method described.

The embodiments of the present application provide a method for identifying surface features, a drone, and a computer-readable storage medium. By processing multiple color channel information and image depth information in the surface image information, features containing surface semantic information can be obtained Figure, through the surface semantic information in the feature map, the recognition result of the surface feature can be accurately determined, and the entire recognition process does not require manual participation, which can improve the accuracy and convenience of surface feature recognition.

It should be understood that the above general description and the following detailed description are only exemplary and explanatory, and cannot limit the application.

Description of the drawings

In order to explain the technical solutions of the embodiments of the present application more clearly, the following will briefly introduce the drawings used in the description of the embodiments. Obviously, the drawings in the following description are some embodiments of the present application. Ordinary technicians can obtain other drawings based on these drawings without creative work.

FIG. 1 is a schematic flow chart of the steps of a method for recognizing ground features according to an embodiment of the present application;

Fig. 2 is a schematic flowchart of sub-steps of the land surface feature recognition method in Fig. 1;

Fig. 3 is a schematic diagram of splicing ground surface images in an embodiment of the present application;

FIG. 4 is a schematic flow chart of the steps of another land surface feature recognition method provided by an embodiment of the present application;

FIG. 5 is a schematic flowchart of the steps of another method for recognizing ground features according to an embodiment of the present application;

Figure 6 is a schematic structural diagram of a drone provided by an embodiment of the present application;

FIG. 7 is a schematic flow chart of the steps of a drone performing a spraying task provided by an embodiment of the present application;

FIG. 8 is a schematic diagram of a flying spray route in an embodiment of the present application;

FIG. 9 is a schematic diagram of a flying spraying route in an embodiment of the present application;

Figure 10 is a schematic diagram of a disaster spread boundary in an embodiment of the present application;

FIG. 11 is a schematic diagram of overlapping spraying operation areas in an embodiment of the present application;

FIG. 12 is another schematic diagram of overlapping spraying operation areas in an embodiment of the present application

FIG. 13 is a schematic block diagram of a surface feature recognition device provided by an embodiment of the present application.

detailed description

The technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of this application.

The flowchart shown in the drawings is only an example, and does not necessarily include all contents and operations/steps, nor does it have to be executed in the described order. For example, some operations/steps can also be decomposed, combined or partially combined, so the actual execution order may be changed according to actual conditions.

The surface feature recognition method provided in this application can be applied to ground control platforms, servers, and/or drones to recognize surface features. Among them, the ground control platform includes laptop computers and PC computers. The server can be a single server or a server cluster composed of multiple servers. UAVs include rotary-wing UAVs, such as quadrotor UAVs, The six-rotor UAV, the eight-rotor UAV can also be a fixed-wing UAV, or a combination of a rotary-wing type and a fixed-wing UAV, which is not limited here.

Hereinafter, some embodiments of the present application will be described in detail with reference to the accompanying drawings. In the case of no conflict, the following embodiments and features in the embodiments can be combined with each other.

Please refer to FIG. 1. FIG. 1 is a schematic flowchart of steps of a method for identifying ground features according to an embodiment of the present application.

Specifically, as shown in FIG. 1, the method for identifying land features includes step S101 to step S103.

S101. Acquire ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information.

When it is necessary to identify the surface features, obtain the required surface image information, where the surface image information is obtained by fusing the image information of multiple color channels and the image depth information, and the image information of multiple color channels includes at least R, G , B three-channel information. By fusing the image information and image depth information of multiple color channels, the surface image information can be enriched, and the accuracy of surface feature recognition can be indirectly improved.

Wherein, the ground surface image information also includes a top view and the image depth information is height information in the top view. The surface image is obtained through aerial photography. In the process of aerial photography, due to the tilt of the movable platform and other reasons, the surface image obtained by aerial photography is not a normal top view image. By converting the surface image to a top view image, it can be guaranteed The accuracy of surface image information can improve the accuracy of surface feature recognition.

Wherein, the surface image information also includes geographic location information corresponding to the surface image. The geographic location information includes positioning information obtained through a global satellite navigation and positioning system; and/or positioning information obtained through a real-time differential positioning system. In the aerial photography process, the mobile platform can obtain the geographic location information of the surface image through the global satellite navigation and positioning system or the real-time differential positioning system, which can further enrich the surface image information and facilitate subsequent query of the area to which the identified surface feature belongs.

The method for determining the image depth information may be based on a binocular ranging algorithm and image information of multiple color channels, or may be based on a monocular ranging algorithm and image information of multiple color channels. The associated frame is an overlapped image frame in the image information of multiple color channels. The overlapped image frame is regarded as the same two image frames, and the disparity of the two image frames is calculated, and then the image can be determined by the disparity In-depth information. It should be noted that the binocular ranging algorithm can be set based on actual conditions, which is not specifically limited in this application. The binocular ranging algorithm can be selected as a semi-global matching algorithm (Semi-Global Matching, SGM).

In one embodiment, as shown in FIG. 2, step S101 includes sub-step S1011 to sub-step S1012.

S1011, acquiring a ground surface image set, and generating a corresponding depth map according to each ground surface image in the ground surface image set;

Obtain a land surface image set from the local or cloud. The land surface image set includes a plurality of surface images, and each of the surface images is a top view of the surface image; according to each surface image in the surface image set, a corresponding depth map is generated. Among them, the surface image set can be obtained by aerial photography of the ground surface by a drone. Specifically, the drone acquires surface aerial photography tasks, where the surface aerial photography tasks include aerial photography flight routes and aerial photography parameters; the drone performs surface aerial photography tasks in order to The ground surface is aerial photographed to obtain a ground surface image set, where the ground surface image in the ground surface image set includes geographic location information. The drone can store the surface image collection obtained by aerial photography locally, or upload the surface image collection obtained by aerial photography to the cloud.

Among them, the method for obtaining surface aerial photography tasks is specifically as follows: the drone obtains surface aerial photography mission files, where the surface aerial photography task files include waypoint information and aerial photography parameters of each waypoint; according to the waypoint information, an aerial photography flight route is generated, where , The aerial photography flight route is marked with multiple waypoints; according to the aerial photography parameters of each waypoint, the aerial photography parameters of each waypoint on the aerial photography flight route are set to generate surface aerial photography tasks. Among them, the aerial photography parameters include aerial photography altitude and aerial photography frequency. The aerial photography frequency is used to control the camera for continuous photography, and the waypoint information includes the position and sequence of each waypoint.

Among them, the method of generating the depth map is as follows: based on the monocular ranging algorithm, according to the associated frame of each surface image in the surface image set, generate the depth map corresponding to each surface image, and perform the calculation of the depth map corresponding to each surface image. By stitching, a stitched depth map is obtained, and the stitched depth map is used as the depth map corresponding to the ground surface image set. Or based on the binocular ranging algorithm, according to every two related surface images in the surface image set, generate the depth map corresponding to each two related surface images, and stitch the depth maps corresponding to each two related surface images to get Splice the depth map, and use the spliced depth map as the depth map corresponding to the ground surface image set.

S1012. Process each ground surface image and the depth map in the ground surface image set to obtain ground surface image information.

After the depth map corresponding to the ground surface image set is generated, each ground surface image and depth map in the ground surface image set is processed to obtain ground surface image information. Specifically, each surface image in the surface image set is spliced to obtain a spliced surface image; the depth map and the spliced surface image are merged to obtain surface image information, which includes image information and image depth of multiple color channels information.

Among them, the splicing method of the surface images is specifically: determining the corresponding splicing parameters of each surface image, where the splicing parameters include splicing order and splicing relationship; according to the corresponding splicing parameters of each surface image, each surface image is spliced , Get the spliced surface image. Please refer to FIG. 3, which is a schematic view of stitching the ground surface image in the embodiment of the application. As shown in FIG. 3, the right side of the ground image A is stitched with the left side of the ground image B, and the right side of the ground image B is connected with The left side of the surface image C is spliced, the upper side of the surface image D is spliced with the lower side of the surface image A, the right side of the surface image E is spliced with the left side of the surface image D, and the left side of the surface image F is spliced to the right of the surface image E. Side stitching.

Among them, the method of determining the stitching parameters is: acquiring the respective aerial time point and aerial position of each surface image; determining the respective stitching sequence of each surface image according to the aerial time point corresponding to each surface image; The corresponding aerial position of each surface image is determined, and the corresponding splicing relationship of each surface image is determined. It should be noted that the smaller the aerial time point, the higher the splicing order of the surface image, the larger the aerial time point, the lower the splicing order of the surface image. The aerial position of the surface image can determine the difference between the surface images. The location relationship, and the location relationship between the surface images of each region is regarded as the corresponding splicing relationship of each surface image.

S102. Process the multiple color channel information and image depth information to obtain a feature map containing semantic information of the ground surface.

After the surface image information is obtained, multiple color channel information and image depth information in the surface image information are processed to obtain a feature map containing the semantic information of the surface. Wherein, the surface semantic information includes each surface feature and the recognition probability value of each surface feature.

In one embodiment, the multiple color channel information and image depth information are fused to obtain a fused image block; the fused image block is matched with an image block in a preset image block set to obtain a fused image block and each image The matching degree between the blocks; according to the matching degree between the fused image block and each image block, the feature map containing the semantic information of the surface is determined. It should be noted that the preset image block set includes a plurality of image blocks marked with surface features, and the image blocks in the preset image block set can be set based on actual conditions, which is not specifically limited in this application.

Among them, the matching method between the fused image block and the image block is specifically: split the fused image block into a preset number of fused image sub-blocks, and split the image blocks into a preset number of image sub-blocks, where, There is a one-to-one correspondence between fused image sub-blocks and image sub-blocks; calculate the similarity between each fused image sub-block and the corresponding image sub-block, and accumulate the difference between each fused image sub-block and the corresponding image sub-block Similarity, the degree of matching between the fused image block and the image block is obtained.

The method for determining the feature map is specifically: acquiring the image block whose matching degree is greater than or equal to the preset matching degree threshold as the target image block, and obtaining the difference between each image sub-block in each target image block and the corresponding fused sub-block The similarity between each target image block and the corresponding land surface feature of each image sub-block in each target image block; obtain the land surface feature corresponding to the image sub-block whose similarity is greater than or equal to the preset similarity threshold, and compare the image corresponding to the land surface feature The similarity of the sub-blocks is used as the recognition probability value; the land surface feature and the recognition probability value are marked in the corresponding fused image sub-block in the fused image block, thereby obtaining a feature map containing the semantic information of the land surface.

In one embodiment, a plurality of color channel information and image depth information are fused to obtain a fused image block; the fused image block is processed through a pre-trained neural network to obtain a feature map containing semantic information on the surface. Among them, the pre-trained neural network can extract the surface semantic information from the fused image block, thereby obtaining a feature map containing the surface semantic information. It should be noted that the neural network may be a convolutional neural network or a cyclic convolutional neural network, which is not specifically limited in this application.

Among them, the neural network training method is specifically: acquiring a large number of surface images marked with surface semantic information, and performing normalization and data enhancement processing on the surface images marked with surface semantic information, thereby obtaining sample data; inputting the sample data To the neural network, the neural network is trained until the neural network converges, so that the ground surface image information is processed through the converged neural network, and a feature map containing the semantic information of the ground surface can be obtained. Through the normalization processing and data enhancement processing of the surface image marked with the surface semantic information, the processing effect of the trained neural network on the surface image can be guaranteed, and the accurate feature map containing the surface semantic information can be obtained, and the surface feature recognition can be improved. Accuracy.

S103: Determine the recognition result of the surface feature according to the semantic information of the surface in the feature map.

After obtaining the surface semantic information contained in the feature map, the recognition result of the surface feature is determined according to the surface semantic information in the feature map. Specifically, the confidence level of each surface feature is obtained from the surface semantic information, and the confidence level of each surface feature is compared with a preset confidence threshold to obtain the surface feature whose confidence is greater than or equal to the confidence threshold , And determine the recognition result of the surface features based on the surface features whose confidence is greater than or equal to the confidence threshold. It should be noted that the above-mentioned confidence threshold may be set based on actual conditions, which is not specifically limited in this application.

Among them, the recognition results of surface features include surface disaster type, surface disaster area information, surface disaster degree information, and surface disaster area information. Surface disaster type is used to describe the type of disaster that occurs on the surface, and surface disaster area information is used to describe the disaster on the surface. The information on the degree of damage on the surface is used to describe the degree of damage in the affected area, and the information on the damage on the surface of the surface is used to describe the area of the affected area. It should be noted that the types of disasters include but are not limited to lodging, plant diseases and insect pests, floods and droughts, which are not specifically limited in this application.

The surface feature recognition method provided by the foregoing embodiment can obtain a feature map containing surface semantic information by processing multiple color channel information and image depth information in the surface image information, and by using the surface semantic information in the feature map, Accurately determine the recognition results of surface features, the entire recognition process does not require human involvement, which can improve the accuracy and convenience of surface feature recognition.

Please refer to FIG. 4, which is a schematic flowchart of the steps of another method for identifying ground features according to an embodiment of the present application.

Specifically, as shown in FIG. 4, the method for identifying features of the land surface includes steps S201 to S205.

S201. Acquire ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information.

S202. Process the multiple color channel information and image depth information to obtain a feature map containing semantic information of the ground surface.

S203: Determine the recognition result of the surface feature according to the semantic information of the surface in the feature map.

S204. Acquire at least one historical recognition result of the surface feature, where the historical recognition result is the recognition result of the surface feature determined before the current moment.

After the recognition result of the surface feature is determined, at least one historical recognition result of the surface feature is obtained. Wherein, the historical recognition result is the recognition result of the surface features determined before the current moment. The historical recognition result is stored in the local disk, or the historical recognition result is stored in the cloud.

In one embodiment, the recognition result of the surface feature can be stored in regions according to the geographic location information of the recognition result of the surface feature, and the storage can also be further divided into surface regions under the dimension of the region to facilitate subsequent queries.

S205. Determine a change trend of the surface according to the recognition result of the surface feature and the at least one historical recognition result of the surface feature.

After acquiring at least one historical recognition result of the surface feature, determine the surface change trend according to the recognition result of the surface feature and the at least one historical recognition result of the surface feature, that is, compare the recognition result of the surface feature with the historical recognition result to know The changing trend of the surface. Among them, the surface change trend includes, but is not limited to, the change trend of plant diseases and insect pests, the change trend of lodging, the change trend of floods, and the change trend of drought. The change trend of plant diseases and insect pests includes the continuous spread of plant diseases and insect pests and the weakening of the intensity of plant diseases and insect pests. The trend of change includes the continued spread of floods and the weakening of the intensity of the flood, and the trend of changes of droughts includes the continued spread of drought and the weakening of the intensity of drought.

In an embodiment, the first definite time point of the recognition result of the surface feature and the second definite time point of each historical recognition result are acquired; according to the first definite time point and each second definite time point, the recognition result and Each historical recognition result is sorted to obtain a recognition result queue; according to every two adjacent recognition results in the recognition result queue, multiple candidate surface change trends are determined; multiple candidate surface change trends are processed to obtain a surface change trend.

The method for determining the change trend of the candidate land surface is specifically: obtaining every two adjacent recognition results in the recognition result queue, and comparing the two recognition results with each other to obtain the change trend of the candidate land surface. It should be noted that the smaller the identification result's determination time point is, the higher the position of the identification result in the identification result queue is, and the larger the identification result time point is, the greater the position of the identification result in the identification result queue. lean back. Among them, the processing method of the change trend of the candidate land surface is specifically: obtain the time sequence corresponding to each candidate land surface change trend, and connect each candidate land surface change trend in turn according to the time sequence corresponding to each candidate land surface change trend, so as to obtain The changing trend of the surface.

The land surface feature recognition method provided by the above embodiment obtains the recognition result of the previously determined land surface feature after the recognition result of the land surface feature is accurately determined, through the recognition result of the currently determined land feature and the recognition result of the previously determined land feature , Can accurately determine the trend of surface changes, facilitate user decision-making, and greatly improve user experience.

Please refer to FIG. 5, which is a schematic flowchart of the steps of yet another method for identifying ground features according to an embodiment of the present application.

Specifically, as shown in FIG. 5, the land surface feature recognition method includes step S301 to step S305.

S301. Acquire ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information.

S302. Process the multiple color channel information and image depth information to obtain a feature map containing semantic information of the ground surface.

S303. Determine the recognition result of the surface feature according to the semantic information of the surface in the feature map.

S304. Obtain a three-dimensional surface map, and obtain surface disaster area information, surface disaster degree information, and surface disaster area information from the recognition result.

After determining the recognition result of the surface features, a three-dimensional surface map is obtained, and the surface disaster area information, the surface damage degree information, and the surface damage area information are obtained from the recognition results. Among them, the recognition results of surface features include surface disaster type, surface disaster area information, surface disaster degree information, and surface disaster area information. Surface disaster type is used to describe the type of disaster that occurs on the surface, and surface disaster area information is used to describe the disaster on the surface. The information on the degree of damage on the surface is used to describe the degree of damage in the affected area, and the information on the damage on the surface of the surface is used to describe the area of the affected area. It should be noted that the types of disasters include but are not limited to lodging, plant diseases and insect pests, floods and droughts, which are not specifically limited in this application.

It should be noted that the three-dimensional surface map is generated based on a three-dimensional construction algorithm, and the three-dimensional construction algorithm can be set based on actual conditions, which is not specifically limited in this application.

S305: Mark the three-dimensional surface map according to the information on the surface disaster area, the degree of damage information, and the information on the surface area affected by the disaster, to obtain a target three-dimensional map marked with the area affected by the disaster, the extent of the disaster, and the area affected by the disaster.

Mark the three-dimensional surface map according to the information of the surface disaster area, the degree of surface damage, and the information of the surface area affected to obtain a target three-dimensional map marked with the affected area, the extent of the disaster, and the area affected by the disaster. The user can quickly and simply know the disaster area, the degree of the disaster and the disaster area through the target three-dimensional map, which is convenient for users to refer to and improves the user experience. Among them, after obtaining the target three-dimensional map, the target three-dimensional map can be stored; and/or the target three-dimensional map is sent to the terminal device for the terminal device to display the target three-dimensional map; and/or the target three-dimensional map is sent to the cloud for the cloud Store a three-dimensional map of the target.

Specifically, according to the surface disaster area information, mark each disaster area in the three-dimensional surface map, that is, obtain the geographic location information of each disaster area from the surface disaster area information. According to the geographic location information of each disaster area, the three-dimensional surface Mark each disaster area on the map; mark the corresponding disaster degree of each disaster area according to the information of the disaster degree on the surface; mark the disaster area corresponding to each disaster area according to the information of the disaster area on the surface, which is obtained from the information of the disaster area on the surface Each disaster area corresponds to the disaster area, and the disaster area is marked in the disaster area on the three-dimensional surface map. It should be noted that the marking method of the affected area, the extent of the disaster, and the affected area can be set based on the actual situation, which is not specifically limited in this application. Optionally, the marking method of the degree of damage is color marking.

Among them, according to the information of the degree of damage on the surface, the corresponding color of each affected area is determined, that is, the degree of damage corresponding to each area is obtained from the information of the degree of damage on the surface, and the pre-stored mapping between the degree of damage and the color is obtained The relationship table, and then query the mapping relationship table, can determine the corresponding disaster degree color of each disaster area; according to the disaster degree color of each disaster area, mark the disaster degree of each disaster area.

The surface feature recognition method provided by the above embodiment, after determining the recognition result of the surface feature, mark the three-dimensional surface map according to the surface disaster area information, the surface damage degree information and the surface damage area information in the surface feature recognition result, and obtain The target three-dimensional map marked with the affected area, the extent of the disaster and the affected area allows the user to quickly and simply know the affected area, the extent of the disaster and the affected area through the target three-dimensional map, which is convenient for users to refer to and improve the user experience.

Please refer to FIG. 6, which is a schematic structural diagram of a drone provided by an embodiment of the present application. The UAV can be a rotary-wing UAV, such as a quadrotor UAV, a six-rotor UAV, an eight-rotor UAV, a fixed-wing UAV, or a rotary-wing UAV and a fixed-wing UAV. The combination of is not limited here.

As shown in FIG. 6, the drone 400 includes a spraying device 401 and a processor 402. The drone 400 is used for spraying agricultural products, forests, and other liquids such as pesticides and water. The unmanned aerial vehicle 400 can realize movement, rotation, turning, etc., and can drive the spraying device 401 to move to different positions or different angles to perform spraying operations in a preset area. The processor 402 is installed inside the drone 400 and is not visible in FIG. 6.

Referring to FIG. 6, in some embodiments, the spraying device 401 includes a pump assembly 4011, a liquid supply tank 4012, a spray head assembly 4013 and a catheter 4014. The liquid supply tank 4012 communicates with the pump assembly 4011. The spray head assembly 4013 is used to implement spraying operations. The liquid guide tube 4014 is connected with the pump assembly 4011 and the spray head assembly 4013 and is used to transport the liquid pumped from the pump assembly 4011 to the spray head assembly 4013. Wherein, the number of the nozzle assembly 4013 is at least one, which can be one, two, three, four or more, which is not limited in this application.

Specifically, the processor 402 may be a micro-controller unit (MCU), a central processing unit (CPU), a digital signal processor (Digital Signal Processor, DSP), or the like.

Please refer to FIG. 7, which is a schematic flow chart of the steps of a drone performing a spraying task according to an embodiment of the application. As shown in FIG. 7, the processor 402 is configured to implement step S401 to step S402.

S401. Acquire a flying spraying task, where the flying spraying task is determined according to the recognition result of the ground surface feature.

Specifically, after obtaining the recognition results of ground features, it is found that crops or fruit trees are found to be lodging, diseases and insect pests, or water shortages, and the crops or fruit trees need to be sprayed with pesticides or watering, etc., and the flying spraying tasks can be determined according to the recognition results of the ground features . The drone 400 acquires a flying spraying task, where the flying spraying task includes a flying spraying route and spraying parameters of each waypoint, and the spraying parameters include spraying time, spraying angle, spraying flow rate, and spraying box.

In one embodiment, the recognition result of the surface feature is obtained from a local disk, a ground terminal or a server, where the recognition result of the surface feature includes the information of the surface disaster area and the information of the degree of damage to the surface; according to the information of the disaster area of the surface and the information of the degree of damage of the surface, Generate the corresponding flying spray task.

Further, the method of generating the flight spraying task is specifically: according to the information of the disaster area on the surface, determine the waypoint information of the flight spraying route to be planned, and generate the corresponding flight spraying route according to the waypoint information; set according to the information on the degree of surface damage The spraying parameters of each waypoint on the flying spraying route to generate the corresponding flying spraying task.

Further, the waypoint information is determined specifically as follows: determine the shape and area of the disaster area according to the information of the disaster area on the surface; determine the type of flight spraying route to be planned according to the shape of the disaster area; determine according to the area of the disaster area The number of waypoints of the spraying route to be planned; the waypoint information of the spraying route to be planned is determined according to the route type, the information of the disaster-affected area on the surface and the number of waypoints.

Further, the method for determining the shape and area of the disaster area is specifically as follows: obtaining the contour information of the disaster area on the surface and the geographic location of each contour point from the information of the disaster area on the surface, and determining the disaster area according to the geographic location of each contour point According to the contour information, determine the contour shape of the disaster area on the surface, and calculate the similarity between the contour shape and each preset shape, and use the preset shape with the highest similarity as the shape of the disaster area.

Further, the method for determining the route type is specifically: obtaining the mapping relationship table between the pre-stored shape and the route type, and querying the mapping relationship table, obtaining the route type corresponding to the shape of the disaster-affected area, and using the obtained route type as The route type of the spraying route to be planned. Route types include strip routes and loop routes. It should be noted that the above-mentioned mapping relationship table between the shape and the route type can be set based on the actual situation, which is not specifically limited in this application.

Further, the method for determining the number of waypoints is specifically: obtaining a mapping relationship table between the pre-stored area and the number of waypoints, and querying the mapping relationship table to obtain the number of waypoints corresponding to the area of the disaster-affected area, and to obtain the obtained The number of waypoints is used as the number of waypoints of the spraying route to be planned. It should be noted that the above-mentioned mapping relationship table between the area and the number of waypoints can be set based on actual conditions, which is not specifically limited in this application.

Further, the waypoint information is determined further by: obtaining a pre-stored map, and marking the corresponding surface disaster area in the pre-stored map according to the information of the surface disaster area; calculating the area of the marked surface disaster area, and calculating the area according to the area and navigation The number of points, determine the distance between the waypoints; according to the distance and route type, mark each waypoint in the disaster-stricken area on the surface in turn, and obtain the marking order of each waypoint and the location of each waypoint in the disaster-affected area on the surface The location of each waypoint; the marking sequence of each waypoint and the geographical location of each waypoint in the disaster area on the surface are used as the waypoint sequence and position of each waypoint, so as to obtain the flight spraying route to be planned Waypoint information.

Further, the method of generating the flight spray route is specifically as follows: obtain the navigation order and position of each waypoint from the waypoint information; connect each waypoint position in turn according to the navigation sequence of each waypoint, and Generate the corresponding flight spraying route. Among them, the flight spraying route includes a circumnavigation route and/or a strip route.

FIG. 8 is a schematic diagram of the flying spraying route in the embodiment of the application. As shown in FIG. 8, the flying spraying route is a circle route, and the flying spraying route includes four waypoints, and the four waypoints are respectively waypoints. A. Waypoint B, Waypoint C and Waypoint D, and the navigation sequence is waypoint A→waypoint B→waypoint C→waypoint D. In this way, this generates a circle route waypoint A→waypoint B→waypoint C→waypoint D→waypoint A enclosed by waypoint A, waypoint B, waypoint C, and waypoint D.

Fig. 9 is a schematic diagram of the flying spraying route in the embodiment of the application. As shown in Fig. 9, the flying spraying route is a strip route, and the flying spraying route includes four waypoints, and the four waypoints are respectively waypoints. E. Waypoint F, Waypoint G and Waypoint H, where the starting point is waypoint E and the ending point is waypoint G. Connect waypoint E, waypoint F, waypoint G and waypoint H in turn to form a closed spray area, where the spray area is based on the preset start waypoint E, end waypoint G, and preset route interval Wait for automatic route planning, such as the bow-shaped route shown in Figure 9.

Further, the method of generating the flight spraying task is specifically: obtaining the pre-stored mapping relationship table between the surface damage degree and the spraying parameter; according to the surface damage degree information and the mapping relationship table, determining the spraying parameter of each waypoint on the flight spraying route ; According to the determined spraying parameters of each waypoint on the flying spraying route, set the spraying parameters of each waypoint on the flying spraying route to generate the corresponding flying spraying task. It should be noted that the above-mentioned mapping relationship table between the degree of surface damage and spraying parameters can be set based on actual conditions, which is not specifically limited in this application.

Further, the method of generating the flying spray task can also be: determining the disaster spread boundary of the disaster area on the surface according to the obtained surface disaster degree information, and determining the position relationship between each waypoint and the disaster spread boundary; according to each The positional relationship between the waypoint and the disaster spread boundary, and determine the spraying parameters of each waypoint on the flight spraying route, so that the spraying time, spraying concentration and/or spraying flow rate of the waypoints on the spread side of the disaster spreading boundary are determined. The spraying parameters are greater than the spraying time, spraying concentration and/or spraying flow rate of the waypoints on the side to be spread on the disaster spreading boundary; according to the spraying parameters of each waypoint on the determined flying spraying route, set every spraying route on the flight The spraying parameters of each waypoint to generate the corresponding flying spraying mission. Among them, the positional relationship includes the spread side and the side to be spread located on the boundary of the disaster spread. The extent of the disaster on the spread side is greater than that of the side to be spread. According to the positional relationship between the waypoint and the disaster spread boundary, the spraying parameters are determined, so that the UAV can spray the disaster-stricken area on the surface according to the spraying parameters, so as to suppress or delay the continuous spread of the disaster.

Please refer to FIG. 10, which is a schematic diagram of the disaster spread boundary in an embodiment of the present application. As shown in FIG. 10, the spread side of the disaster spread boundary is the side where the surface disaster area A is located, and the disaster spread boundary is to be spread The side is the side where the surface area B is located, and the disaster spreads from the surface disaster area A to the surface area B. The crops of the surface disaster area A and the surface area B can be the same or different.

Further, the spraying parameters also include spray box labels, which are used to identify spray boxes. The drone includes at least two spray boxes, and each spray box has different pesticide types and/or pesticide concentrations, corresponding to different levels of surface damage Different types of pesticides and/or concentrations of pesticides. The higher the degree of surface damage, the higher the corresponding pesticide concentration. The lower the degree of surface damage, the lower the corresponding pesticide concentration. For a surface disaster area, according to the location area of each waypoint Set the spray box label corresponding to the pesticide type and/or pesticide concentration of each waypoint. By spraying different types of pesticides and/or pesticide concentrations in combination, disasters can be effectively treated, and the continued spread of disasters can also be suppressed or delayed.

Furthermore, for a surface disaster area, at least two drones can be used to coordinate the flight spraying task. Each drone is responsible for a spraying area in the surface disaster area, and spraying operations between at least two drones. The areas are overlapped. The overlapped area is the area corresponding to the severe damage on the surface. In the overlapped area, the drones are located at different heights or the spraying time points of the drones are different, or avoidance can be achieved through sensors. It can avoid the collision of drones when spraying in the overlapping area. Since at least two drones are spraying the overlapping area (the severely damaged area), it can effectively improve the damage to the area. The treatment effect of severe areas can also quickly complete the spraying of disaster-stricken areas on the surface to inhibit or delay the continued spread of disasters.

Take two drones to complete the flight spraying tasks in the disaster area on the surface as an example to explain. The first drone of the two drones is responsible for a spraying area in the disaster area on the surface, and the two drones are The second drone is responsible for another spraying area in the disaster area on the surface. The spraying area of the first drone and the spraying area of the second drone are overlapped, and the overlapping area is the surface In the area corresponding to the severe damage degree, in the overlapping area, the first drone and the second drone are located at different heights, or the spraying time of the first drone and the second drone are different. Or the first UAV and the second UAV can avoid obstacles through sensors, which can prevent the first UAV and the second UAV from colliding when spraying in the overlapping area. Please refer to Figure 11. Figure 11 is a schematic diagram of the overlap of the spraying operation area in an embodiment of the present application. As shown in Figure 11, the surface disaster area includes the spraying operation area A and the spraying operation area B, and the spraying operation area A and the spraying area The overlapping area of operation area B is area C. The first drone is responsible for spraying operation area A, and the second drone is responsible for spraying operation area B. Both the first drone and the second drone are in the overlapping area C Perform spraying operations.

Please refer to FIG. 12, which is another schematic diagram of the overlap of the spraying operation area in an embodiment of the present application. As shown in FIG. 12, according to the obtained information on the degree of surface damage, the determined disaster spreading direction of the surface affected area is the surface damage The area A spreads to the surface area B. The four drones are assigned the flight spraying area, and the flight spraying route is planned in their corresponding flight spraying area. The flight spraying area of UAV 1 is a, and the flight of UAV 2 is flying. The spraying area is b, the flying spraying area of drone 3 is c, the flying spraying area of drone 4 is d, and the flying spraying area a, the flying spraying area is b, and the flying spraying area is c and the flying spraying area d exists. Overlapping, UAV 1, UAV 2 and UAV 3 are mainly responsible for spraying on the surface affected area A on the spread side, and UAV 4 is mainly responsible for spraying on the surface area B on the side to be spread, the overlapping area Contains a part of the disaster-affected area A and a part of the surface area B. After perceiving the information on the degree of damage to the surface, it is possible to allocate the spraying area to multiple drones and plan the flight spraying route, so that the multiple drones can coordinate the spraying operation, and the spraying can be planned reasonably within the range of the drone At the same time, the combined spraying of overlapping areas can inhibit or delay the spread of disasters. As a comparative example, if multiple drones are not used for cooperative operation to implement overlapping spraying, if a single drone sprays areas with more serious disasters or disasters, it needs to stay in a fixed area for a long time. Increase the amount of spraying agent in this area, and the UAV has a fixed cruising range due to energy constraints. In this case, it is difficult to spray the entire area to be sprayed.

It is understandable that the above embodiment is only an exemplary description of the spraying operation performed by multiple drones in coordination, and the number of drones can also be flexibly set according to actual needs. For example, the number of drones is two. 3 sets, 4 sets, 5 sets, etc., this application does not limit this.

S402: Execute the flying spraying task, and control the spraying device to execute a corresponding spraying action according to the spraying parameters in the flying spraying task.

The drone 400 obtains the flying spraying task, executes the flying spraying task, and controls the spraying device to perform the corresponding spraying action according to the spraying parameters in the flying spraying task, that is, obtain the flying spraying route and the flight spraying route and each waypoint from the flying spraying task. Spraying parameters and flying according to the flying spraying route, and during the flight, the spraying device 401 is controlled to perform corresponding spraying actions according to the spraying parameters of each waypoint to complete the flying spraying task.

UAVs can perform flying spraying tasks determined based on the recognition results of ground features, and can automatically spray pesticides or water crops or fruit trees, and prevent and control crops or fruit trees from lodging, diseases and insect pests, or water shortages.

The application also provides a ground feature recognition device.

Please refer to FIG. 13, which is a schematic block diagram of a surface feature recognition device provided by an embodiment of the present application. As shown in FIG. 13, the surface feature recognition device 500 includes a processor 501 and a memory 502, and the processor 501 and the memory 502 are connected by a bus 503, which is, for example, an I2C (Inter-integrated Circuit) bus. Wherein, the surface feature recognition device 500 can be a ground control platform, a server or a drone. The ground control platform includes a laptop computer and a PC computer. The server can be a single server or a server cluster composed of multiple servers. , UAVs include rotary-wing UAVs, such as four-rotor UAVs, six-rotor UAVs, and eight-rotor UAVs. It can also be a fixed-wing UAV, or a rotary-wing UAV and a fixed-wing UAV. The combination of is not limited here.

Specifically, the processor 501 may be a micro-controller unit (MCU), a central processing unit (CPU), a digital signal processor (Digital Signal Processor, DSP), or the like.

Specifically, the memory 502 may be a Flash chip, a read-only memory (ROM, Read-Only Memory) disk, an optical disk, a U disk, or a mobile hard disk.

Wherein, the processor 501 is configured to run a computer program stored in the memory 502, and implement the following steps when the computer program is executed:

Further, the surface image information includes a top view and a front view.

Further, the image depth information is height information in the top and front view.

Further, the surface image information includes geographic location information corresponding to the surface image.

Further, the geographic location information includes positioning information obtained through a global satellite navigation and positioning system;

And/or, positioning information obtained through a real-time differential positioning system.

Further, the image information of the multiple color channels includes at least R, G, and B three-channel information.

Further, the image depth information is determined based on a binocular ranging algorithm and image information of the multiple color channels.

Further, the image depth information is determined based on a monocular ranging algorithm and an associated frame of the image information of the multiple color channels.

Further, when the processor realizes the processing of the multiple color channel information and the image depth information to obtain a feature map containing the semantic information of the ground surface, it is used to realize:

Performing fusion processing on the multiple color channel information and image depth information to obtain a fused image block;

Matching the fused image block with an image block in a preset image block set to obtain a degree of matching between the fused image block and each of the image blocks;

According to the degree of matching between the fused image block and each of the image blocks, a feature map containing the semantic information of the ground surface is determined.

The fused image block is processed by a pre-trained neural network to obtain a feature map containing semantic information of the ground surface.

Further, when the processor realizes the acquisition of surface image information, it is used to realize:

Acquiring a ground surface image set, and generating a corresponding depth map according to each ground surface image in the ground surface image set;

Each surface image in the surface image set and the depth map are processed to obtain surface image information.

Further, when the processor realizes the processing of each ground surface image and the depth map in the ground surface image set to obtain ground surface image information, it is used to realize:

Splicing each ground surface image in the ground surface image set to obtain a spliced ground surface image;

The depth map and the spliced ground surface image are fused to obtain ground surface image information.

Further, when the processor implements the splicing of each ground surface image in the ground surface image set to obtain a spliced ground surface image, it is used to implement:

Determine the respective stitching parameters corresponding to each of the surface images, where the stitching parameters include a stitching order and a stitching relationship;

According to the splicing parameters corresponding to each of the surface images, splicing each of the surface images to obtain a spliced surface image.

Further, when the processor implements the determination of the stitching parameters corresponding to each of the surface images, it is used to implement:

Acquiring an aerial photographing time point and aerial photographing position corresponding to each of the ground surface images;

Determine the corresponding splicing sequence of each of the surface images according to the respective aerial time points corresponding to each of the surface images;

According to the aerial position corresponding to each of the surface images, the stitching relationship corresponding to each of the surface images is determined.

Further, after the processor realizes the recognition result of the ground surface feature is determined according to the ground surface semantic information in the feature map, it is further used to realize:

Acquiring at least one historical recognition result of a surface feature, where the historical recognition result is a recognition result of the surface feature determined before the current moment;

According to the recognition result of the land surface feature and the at least one historical recognition result of the land surface feature, a change trend of the land surface is determined.

Further, when the processor realizes the determination of the change trend of the surface according to the recognition result of the surface feature and the at least one historical recognition result of the surface feature, it is used to achieve:

Acquiring the first determined time point of the recognition result of the surface feature and the second determined time point of each of the historical recognition results;

Sort the recognition results and each of the historical recognition results according to the first determined time point and each of the second determined time points to obtain a recognition result queue;

Determine a plurality of candidate land surface change trends according to every two adjacent recognition results in the recognition result queue;

The multiple candidate land surface change trends are processed to obtain the land surface change trends.

Obtaining a three-dimensional surface map, and obtaining surface disaster area information, surface damage degree information, and surface damage area information from the recognition result;

Mark the three-dimensional surface map according to the information of the surface disaster area, the degree of surface damage, and the information of the surface area.

Further, when the processor implements marking the three-dimensional surface map according to the information on the surface disaster area, the information on the degree of the surface disaster, and the information on the surface area affected by the disaster, the processor is used to implement:

Mark each disaster-stricken area in the three-dimensional surface map according to the information of the disaster-stricken area on the surface;

According to the information on the degree of damage to the surface, mark the degree of damage corresponding to each of the disaster-stricken areas;

According to the information of the disaster area on the surface, the disaster area corresponding to each of the disaster areas is marked.

Further, when the processor realizes marking the corresponding disaster degree of each disaster-affected area according to the information of the disaster degree of the surface, it is used to realize:

Determine the color of the disaster degree corresponding to each of the disaster-affected areas according to the information on the degree of disaster on the surface;

According to the disaster degree color corresponding to each disaster area, mark the disaster degree corresponding to each disaster area.

Further, the processor implements marking the three-dimensional surface map according to the information of the surface disaster area, the information of the degree of damage of the surface and the information of the area of the surface of the disaster, to obtain the target three-dimensional map marked with the disaster area, the degree of damage, and the area of the disaster. After the map, it is used to achieve:

Store the three-dimensional map of the target; and/or

Sending the target three-dimensional map to a terminal device for the terminal device to display the target three-dimensional map; and/or

The target three-dimensional map is sent to the cloud, so that the cloud stores the target three-dimensional map.

It should be noted that those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working process of the surface feature recognition device described above can refer to the corresponding process in the foregoing flight mission generation method embodiment. This will not be repeated here.

The embodiments of the present application also provide a computer-readable storage medium, the computer-readable storage medium stores a computer program, the computer program includes program instructions, and the processor executes the program instructions to implement the foregoing implementation The steps of the surface feature recognition method provided in the example.

Wherein, the computer-readable storage medium may be an internal storage unit of the surface feature recognition device described in any of the foregoing embodiments, such as a hard disk or memory of the surface feature recognition device. The computer-readable storage medium may also be an external storage device of the surface feature recognition device, such as a plug-in hard disk equipped on the surface feature recognition device, a smart media card (SMC), a secure digital ( Secure Digital, SD card, Flash Card, etc.

It should be understood that the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in the specification of this application and the appended claims, unless the context clearly indicates other circumstances, the singular forms "a", "an" and "the" are intended to include plural forms. It should also be understood that the term "and/or" used in the specification and appended claims of this application refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

The above are only specific implementations of this application, but the protection scope of this application is not limited to this. Anyone familiar with the technical field can easily think of various equivalents within the technical scope disclosed in this application. Modifications or replacements, these modifications or replacements shall be covered within the scope of protection of this application. Therefore, the protection scope of this application should be subject to the protection scope of the claims.

Claims

A land surface feature recognition method, which is characterized in that it includes:

Acquiring ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information;

Processing the multiple color channel information and image depth information to obtain a feature map containing semantic information on the surface;

According to the semantic information of the surface in the feature map, the recognition result of the surface feature is determined.
The land surface feature recognition method according to claim 1, wherein the land surface image information includes a top view.
The method for recognizing ground features according to claim 2, wherein the image depth information is height information in the top and front view.
The land surface feature recognition method according to claim 1, wherein the land surface image information includes geographic location information corresponding to the land surface image.
The land surface feature recognition method according to claim 4, wherein the geographic location information includes positioning information obtained through a global satellite navigation and positioning system;

And/or, positioning information obtained through a real-time differential positioning system.
The land surface feature recognition method according to claim 1, wherein the image information of the multiple color channels includes at least R, G, and B three-channel information.
The land surface feature recognition method according to claim 1, wherein the image depth information is determined based on a binocular ranging algorithm and image information of the multiple color channels.
The land surface feature recognition method according to claim 1, wherein the image depth information is determined based on a monocular ranging algorithm and an associated frame of the image information of the multiple color channels.
The land surface feature recognition method according to claim 1, wherein the processing the multiple color channel information and image depth information to obtain a feature map containing surface semantic information comprises:

Performing fusion processing on the multiple color channel information and image depth information to obtain a fused image block;

Matching the fused image block with an image block in a preset image block set to obtain a degree of matching between the fused image block and each of the image blocks;

According to the degree of matching between the fused image block and each of the image blocks, a feature map containing the semantic information of the ground surface is determined.
The land surface feature recognition method according to claim 1, wherein the processing the multiple color channel information and image depth information to obtain a feature map containing surface semantic information comprises:

Performing fusion processing on the multiple color channel information and image depth information to obtain a fused image block;

The fused image block is processed by a pre-trained neural network to obtain a feature map containing semantic information of the ground surface.
The land surface feature recognition method according to any one of claims 1 to 10, wherein said acquiring land surface image information comprises:

Acquiring a ground surface image set, and generating a corresponding depth map according to each ground surface image in the ground surface image set;

Each surface image in the surface image set and the depth map are processed to obtain surface image information.
The land surface feature recognition method according to claim 11, wherein the processing each land surface image in the land surface image set and the depth map to obtain land surface image information comprises:

Splicing each ground surface image in the ground surface image set to obtain a spliced ground surface image;

The depth map and the spliced ground surface image are fused to obtain ground surface image information.
The land surface feature recognition method according to claim 12, wherein said splicing each land surface image in said land surface image set to obtain a spliced land surface image comprises:

Determine the respective stitching parameters corresponding to each of the surface images, where the stitching parameters include a stitching order and a stitching relationship;

According to the splicing parameters corresponding to each of the surface images, splicing each of the surface images to obtain a spliced surface image.
The land surface feature recognition method according to claim 13, wherein said determining the respective stitching parameters corresponding to each of the land surface images comprises:

Acquiring an aerial photographing time point and aerial photographing position corresponding to each of the ground surface images;

Determine the corresponding splicing sequence of each of the surface images according to the respective aerial time points corresponding to each of the surface images;

According to the aerial position corresponding to each of the surface images, the stitching relationship corresponding to each of the surface images is determined.
The land surface feature recognition method according to any one of claims 1 to 10, characterized in that, after determining the recognition result of the land surface feature according to the surface semantic information in the feature map, the method further comprises:

Acquiring at least one historical recognition result of a surface feature, where the historical recognition result is a recognition result of the surface feature determined before the current moment;

According to the recognition result of the land surface feature and the at least one historical recognition result of the land surface feature, a change trend of the land surface is determined.
The land surface feature recognition method according to claim 15, wherein the determining the land surface change trend based on the recognition result of the land surface feature and the at least one historical recognition result of the land surface feature comprises:

Acquiring the first determined time point of the recognition result of the surface feature and the second determined time point of each of the historical recognition results;

Sort the recognition results and each of the historical recognition results according to the first determined time point and each of the second determined time points to obtain a recognition result queue;

Determine a plurality of candidate land surface change trends according to every two adjacent recognition results in the recognition result queue;

The multiple candidate land surface change trends are processed to obtain the land surface change trends.
The land surface feature recognition method according to any one of claims 1 to 10, characterized in that, after determining the recognition result of the land surface feature according to the surface semantic information in the feature map, the method further comprises:

Obtaining a three-dimensional surface map, and obtaining surface disaster area information, surface damage degree information, and surface damage area information from the recognition result;

Mark the three-dimensional surface map according to the information of the surface disaster area, the degree of surface damage, and the information of the surface area.
The land surface feature recognition method according to claim 17, wherein the marking the three-dimensional surface map according to the information on the disaster area on the surface, the information on the degree of damage on the surface, and the information on the area on the surface by the disaster comprises:

Mark each disaster-stricken area in the three-dimensional surface map according to the information of the disaster-stricken area on the surface;

According to the information on the degree of damage to the surface, mark the degree of damage corresponding to each of the disaster-stricken areas;

According to the information of the disaster area on the surface, the disaster area corresponding to each of the disaster areas is marked.
The land surface feature recognition method according to claim 18, wherein the marking the corresponding disaster degree of each of the disaster-stricken areas according to the information of the disaster degree of the ground surface comprises:

Determine the color of the disaster degree corresponding to each of the disaster-stricken areas according to the information about the degree of disaster on the surface;

According to the disaster degree color corresponding to each disaster area, mark the disaster degree corresponding to each disaster area.
The land surface feature recognition method according to claim 17, wherein the three-dimensional surface map is marked according to the information of the surface disaster area, the information of the degree of surface damage, and the information of the area of the surface disaster, so as to obtain the marked area of the disaster. After the target three-dimensional map of the extent of the disaster and the area affected by the disaster, it also includes:

Store the three-dimensional map of the target; and/or

Sending the target three-dimensional map to a terminal device for the terminal device to display the target three-dimensional map; and/or

The target three-dimensional map is sent to the cloud, so that the cloud stores the target three-dimensional map.
An unmanned aerial vehicle, characterized in that the unmanned aerial vehicle includes a spraying device and a processor, and the processor is configured to implement the following steps:

Acquiring a flying spraying task, wherein the flying spraying task is determined according to the recognition result of the ground surface features;

Execute the flying spraying task, and control the spraying device to execute the corresponding spraying action according to the spraying parameters in the flying spraying task.
The unmanned aerial vehicle according to claim 21, wherein the processor is used to realize:

Acquiring the recognition result of the ground surface feature, wherein the recognition result of the ground surface feature includes the information of the disaster area on the ground and the information of the degree of the disaster on the ground;

According to the surface disaster area information and the surface disaster degree information, a corresponding flying spray mission is generated.
The unmanned aerial vehicle according to claim 22, characterized in that, when the processor is configured to generate a corresponding flying spray task according to the information on the disaster area on the surface and the information on the degree of damage on the surface, it is used to achieve:

Determine the waypoint information of the flight spraying route to be planned according to the information of the disaster-affected area on the surface, and generate the corresponding flight spraying route according to the waypoint information;

According to the information on the degree of damage to the surface, the spraying parameters of each waypoint on the flying spraying route are set to generate a corresponding flying spraying task.
The unmanned aerial vehicle according to claim 23, wherein the determining the waypoint information of the flight spraying route to be planned according to the information of the disaster-affected area on the surface comprises:

Determine the shape and area of the disaster area according to the information of the disaster area on the surface;

According to the shape of the disaster-affected area, determine the type of flight spraying route to be planned;

According to the area of the disaster-affected area, determine the number of waypoints of the flight spraying route to be planned;

Determine the waypoint information of the flight spraying route to be planned according to the route type, the information of the disaster-affected area on the surface and the number of waypoints.
The UAV according to claim 23, wherein said generating a corresponding flight spraying route according to said waypoint information comprises:

Obtain the navigation sequence and position of each waypoint from the waypoint information;

According to the sequence of the navigation sequence of each waypoint, the position of each waypoint is connected in turn to generate the corresponding flight spraying route.
The unmanned aerial vehicle according to claim 23, wherein the setting spraying parameters of each waypoint on the flying spraying route to generate a corresponding flying spraying task according to the information on the degree of damage to the ground surface comprises:

Obtain the mapping relationship table between the pre-stored surface damage degree and spraying parameters;

Determine the spraying parameters of each waypoint on the flying spraying route according to the information on the degree of damage to the surface and the mapping relationship table;

According to the determined spraying parameters of each waypoint on the flying spraying route, the spraying parameters of each waypoint on the flying spraying route are set to generate a corresponding flying spraying task.
A surface feature recognition device, characterized in that the surface feature recognition device includes a memory and a processor;

The memory is used to store a computer program;

The processor is configured to execute the computer program and, when executing the computer program, implement the following steps:

Acquiring ground surface image information, where the ground surface image information includes image information of multiple color channels and image depth information;

Processing the multiple color channel information and image depth information to obtain a feature map containing semantic information on the surface;

According to the semantic information of the surface in the feature map, the recognition result of the surface feature is determined.
The land surface feature recognition device according to claim 27, wherein the land surface image information includes a top view.
The surface feature recognition device according to claim 28, wherein the image depth information is height information in the top and front view.
The land surface feature recognition device according to claim 27, wherein the land surface image information includes geographic location information corresponding to the land surface image.
The land surface feature recognition device according to claim 30, wherein the geographic location information includes positioning information obtained through a global satellite navigation and positioning system;

And/or, positioning information obtained through a real-time differential positioning system.
The surface feature recognition device according to claim 27, wherein the image information of the multiple color channels includes at least R, G, and B three-channel information.
The surface feature recognition device according to claim 27, wherein the image depth information is determined based on a binocular ranging algorithm and image information of the multiple color channels.
The surface feature recognition device according to claim 27, wherein the image depth information is determined based on a monocular ranging algorithm and an associated frame of the image information of the multiple color channels.
The land surface feature recognition device according to claim 27, wherein the processor is used to implement processing of the multiple color channel information and image depth information to obtain a feature map containing surface semantic information. :

Performing fusion processing on the multiple color channel information and image depth information to obtain a fused image block;

Matching the fused image block with an image block in a preset image block set to obtain a degree of matching between the fused image block and each of the image blocks;

According to the degree of matching between the fused image block and each of the image blocks, a feature map containing the semantic information of the ground surface is determined.
The land surface feature recognition device according to claim 27, wherein the processor is used to implement processing of the multiple color channel information and image depth information to obtain a feature map containing surface semantic information. :

Performing fusion processing on the multiple color channel information and image depth information to obtain a fused image block;

The fused image block is processed by a pre-trained neural network to obtain a feature map containing semantic information of the ground surface.
The land surface feature recognition device according to any one of claims 27 to 36, wherein the processor is configured to achieve: when realizing the acquisition of ground surface image information:

Acquiring a ground surface image set, and generating a corresponding depth map according to each ground surface image in the ground surface image set;

Each surface image in the surface image set and the depth map are processed to obtain surface image information.
The land surface feature recognition device according to claim 37, wherein the processor is used to implement processing for each land surface image and the depth map in the land surface image set to obtain land surface image information. :

Splicing each ground surface image in the ground surface image set to obtain a spliced ground surface image;

The depth map and the spliced ground surface image are fused to obtain ground surface image information.
The land surface feature recognition device according to claim 38, wherein, when the processor implements splicing each land surface image in the land surface image set to obtain a spliced land surface image, it is used to implement:

Determine the respective stitching parameters corresponding to each of the surface images, where the stitching parameters include a stitching order and a stitching relationship;

According to the splicing parameters corresponding to each of the surface images, splicing each of the surface images to obtain a spliced surface image.
The land surface feature recognition device according to claim 39, wherein, when the processor realizes the determination of the stitching parameters corresponding to each of the land surface images, it is configured to realize:

Acquiring an aerial photographing time point and aerial photographing position corresponding to each of the ground surface images;

Determine the corresponding splicing sequence of each of the surface images according to the respective aerial time points corresponding to each of the surface images;

According to the aerial position corresponding to each of the surface images, the stitching relationship corresponding to each of the surface images is determined.
The land surface feature recognition device according to any one of claims 27 to 36, wherein the processor is further configured to determine the recognition result of the land surface feature based on the surface semantic information in the feature map. achieve:

Acquiring at least one historical recognition result of a surface feature, where the historical recognition result is a recognition result of the surface feature determined before the current moment;

According to the recognition result of the land surface feature and the at least one historical recognition result of the land surface feature, a change trend of the land surface is determined.
The land surface feature recognition device according to claim 41, wherein the processor is configured to determine a land surface change trend based on the recognition result of the land surface feature and at least one historical recognition result of the land surface feature. achieve:

Acquiring the first determined time point of the recognition result of the surface feature and the second determined time point of each of the historical recognition results;

Sort the recognition results and each of the historical recognition results according to the first determined time point and each of the second determined time points to obtain a recognition result queue;

Determine a plurality of candidate land surface change trends according to every two adjacent recognition results in the recognition result queue;

The multiple candidate land surface change trends are processed to obtain the land surface change trends.
The land surface feature recognition device according to any one of claims 27 to 36, wherein the processor is further configured to determine the recognition result of the land surface feature based on the surface semantic information in the feature map. achieve:

Obtaining a three-dimensional surface map, and obtaining surface disaster area information, surface damage degree information, and surface damage area information from the recognition result;

Mark the three-dimensional surface map according to the information of the surface disaster area, the degree of surface damage, and the information of the surface area.
The surface feature recognition device according to claim 43, wherein when the processor implements marking the three-dimensional surface map according to the surface disaster area information, the surface disaster degree information, and the surface disaster area information, Used to achieve:

Mark each disaster-stricken area in the three-dimensional surface map according to the information of the disaster-stricken area on the surface;

According to the information on the degree of damage to the surface, mark the degree of damage corresponding to each of the disaster-stricken areas;

According to the information of the disaster area on the surface, the disaster area corresponding to each of the disaster areas is marked.
The land surface feature recognition device according to claim 44, wherein the processor is configured to: when the processor realizes marking the corresponding disaster degree of each of the disaster-stricken areas according to the information of the disaster degree of the ground surface:

Determine the color of the disaster degree corresponding to each of the disaster-affected areas according to the information on the degree of disaster on the surface;

According to the disaster degree color corresponding to each disaster area, mark the disaster degree corresponding to each disaster area.
The surface feature recognition device according to claim 43, wherein the processor implements marking the three-dimensional surface map according to the surface disaster area information, the surface disaster degree information, and the surface disaster area information to obtain After marking the target three-dimensional map with the affected area, the extent of the disaster and the affected area, it is used to achieve:

Store the three-dimensional map of the target; and/or

Sending the target three-dimensional map to a terminal device for the terminal device to display the target three-dimensional map; and/or

The target three-dimensional map is sent to the cloud, so that the cloud stores the target three-dimensional map.
A computer-readable storage medium, characterized in that, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor realizes as described in any one of claims 1 to 20. The surface feature recognition method described.