Method and system for agricultural machinery trajectory recognition

Abstract

A method and system for agricultural machinery trajectory recognition, relating to the technical field of agricultural application. The method comprises: on the basis of data sent by a GPS device, determining trajectory point-related data of an agricultural machine, and obtaining a corresponding satellite image; on the basis of a neural network model having a special architecture, processing and analyzing the above data, and determining image features of each trajectory point corresponding to the satellite image and classification information corresponding to each trajectory point; and marking, on the satellite image, a movement trajectory formed by connecting the trajectory points and the classification information corresponding to the trajectory points, and outputting a marked map. On the basis of the solution, the movement trajectory of an agricultural machine can be obtained and identified without manual on-site surveying, efficiently and accurately marking the movement trajectory of the agricultural machine on a satellite map, and visually presenting distribution information of farmland.

Description

Methods and systems for agricultural machinery trajectory recognition

[0001] This application claims priority to Chinese Patent Application No. 202311127954.4, filed on September 1, 2023, entitled “Method and System for Identifying Agricultural Machinery Trajectory”, the entire contents of which are incorporated herein by reference. Technical Field

[0002] This application relates to the field of agricultural application technology, and more specifically, to methods and systems for agricultural machinery trajectory recognition. Background Technology

[0003] In agricultural work, obtaining information on the distribution of farmland is crucial. Typically, farmers obtain this information through on-site surveys and then mark it on a designated map for display. This method is inefficient and has poor accuracy.

[0004] Therefore, how to efficiently and accurately determine the distribution information of farmland marked on a map without the need for manual on-site surveys is an urgent technical problem to be solved.

[0005] Summary of the Invention

[0006] This application provides a method and system for agricultural machinery trajectory recognition, which can match the movement trajectory of agricultural machinery with satellite maps, analyze the farmland and road portions corresponding to the movement trajectory, and mark them on the satellite map, so as to achieve efficient and accurate distribution information of marked farmland on the map without human intervention.

[0007] In a first aspect, a method for identifying a track of an agricultural machine is provided. The method comprises: determining a first track graph based on first track point information sent by a global positioning system (GPS) device carried by the agricultural machine, the first track graph comprising a first movement track connected by a plurality of track points, the first movement track representing a movement track of the agricultural machine in a first area, the first track graph representing a relationship between the first movement track and first state information, the first state information comprising at least one of speed information, time information, and mileage information; determining an initial point sequence based on the first track point information, the initial point sequence comprising, for each track point, a longitude and a latitude of a position of the agricultural machine, an instantaneous speed of the agricultural machine, a travel mileage of the agricultural machine, and a time point of dotting; determining a first satellite map, the first satellite map representing the first area, the first satellite map and the first track graph having the same size parameter and being in the same geodetic coordinate system; and inputting the first track graph, the first satellite map, and the initial point sequence into a first neural network model to output a first marked map, the first marked map comprising an identification result for the first movement track, the first neural network model comprising an encoder and a processor, each of the encoder and the processor comprising n stages, the first neural network model being configured to:

[0008] The first trajectory map and the first satellite map are superimposed to determine a first input image; in the first stage of the encoder, feature extraction is performed on the first input image at a first down-sampling scale to determine a first feature map, and in the first stage of the processor, a full connection operation is performed on the initial point sequence to determine a first point sequence, the number of channels for representing point information in the first point sequence being the same as the number of channels for representing point information in the first feature map; and based on a multi-headed self-attention (MSA) mechanism, a self-attention operation is performed on the first feature map and the first point sequence to determine a first fused point sequence; in the mth stage of the encoder, feature extraction is performed on the (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and in the mth stage of the processor, a full connection operation is performed on the (m-1)th fused point sequence to determine an mth point sequence, the number of channels for representing point information in the mth point sequence being the same as the number of channels for representing point information in the mth feature map; and based on the MSA mechanism, a self-attention operation is performed on the mth feature map and the mth point sequence to determine an mth fused point sequence, m∈[2,n] and is an integer; in the case where m is equal to n, a full connection operation is performed on the nth fused point sequence to determine a target point sequence, the target point sequence including the latitude and longitude coordinates of each trajectory point constituting the first moving trajectory and classification information for the candidate, the classification information for the candidate including at least farmland and road; through a normalized exponential function, classification information to which each trajectory point constituting the first moving trajectory respectively belongs is determined from the classification information for the candidate in the target point sequence, and the first moving trajectory and the classification information to which each trajectory point constituting the first moving trajectory respectively belongs are mapped to the first satellite map according to the latitude and longitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, and a first labeled map is output.

[0009] For example, the first trajectory map described above can be an image based on STD. The first trajectory map described above is drawn according to the dotting time corresponding to each trajectory point, and is a gray-scale image. Each trajectory point constituting the first moving trajectory respectively corresponds to a different gray value, which can be associated with the first state information described above.

[0010] For example, since the size parameters of the first satellite map and the first trajectory map need to be the same, and both are used to represent the same region, they also need to be in the same geodetic coordinate system, so the map level and size parameters of the two maps need to be determined in advance.

[0011] For example, the n stages described above are also called n stages of the encoder, and can be specifically four stages. In each stage, the input image needs to pass through multiple transformer processing blocks to extract image features.

[0012] Based on the above technical solution, the moving track of the agricultural machine can be efficiently and accurately identified based on the track point information collected by the GPS device and the corresponding satellite map through the neural network model without manual on-site investigation, and the moving track is marked on the satellite map. Moreover, the moving track can indicate that each part thereof belongs to a farmland or a road, and thus the distribution information of the farmland can be directly determined based on the part of the moving track belonging to the farmland.

[0013] With reference to the first aspect, in some implementations of the first aspect, before the first track map and the first satellite map are input into the first neural network model, a first length and a first width in size parameters of the first track map and the first satellite map are respectively expanded to a second length and a second width.

[0014] For example, the first length and the first width can be represented based on the number of pixels, and can be specifically represented as 1024x1024. Similarly, the second length and the second width can be specifically represented as 1280x1280.

[0015] Based on the above technical solution, it can be ensured that the model can obtain sufficient image edge information.

[0016] With reference to the first aspect, in some implementations of the first aspect, before the first track map and the first satellite map are superimposed to determine the first input image, a downsampling operation is performed on the first track map and the first satellite map, the second length and the second width in the size parameters of the first track map and the first satellite map are reduced to a third length and a third width, and the number of image channels in the size parameters of the first track map and the first satellite map is correspondingly expanded.

[0017] For example, two convolution kernels with a step of 2 can be used to perform the downsampling operation on the first track map and the first satellite map respectively to change the size parameters of the images.

[0018] Based on the above technical solution, the size of the input image can be reduced, and the number of image features is also enriched, which helps the accuracy of subsequent neural network model feature extraction of the input image.

[0019] With reference to the first aspect, in some implementations of the first aspect, the trajectory points with the classification information of the farmland in the first moving track are connected by first color lines, and the trajectory points with the classification information of the road in the first moving track are connected by second color lines.

[0020] Based on the above technical solution, the classification of the moving track can be intuitively distinguished on the satellite map, and the distribution information of the farmland can be directly obtained, which helps subsequent statistical analysis of the distribution information of the farmland and determination of the corresponding farming plan.

[0021] With reference to the first aspect, in some implementations of the first aspect, the first neural network model is sequentially input with Q sets of data including the first trajectory map, the first satellite map and the initial point sequence, the first neural network model corresponds to output Q first marked maps, Q is an integer greater than 1; the Q first marked maps are spliced to determine the second marked map.

[0022] Based on the above technical solutions, while ensuring the processing efficiency of the neural network model, the multiple first marked maps are spliced to obtain a second marked map that can completely present a large-scale farmland, thereby further increasing the intuitiveness of presenting the farmland distribution information.

[0023] The second aspect provides a system for recognizing the track of agricultural machinery, which comprises a GPS server, a satellite map server and an image processing server.

[0024] The GPS server is configured to determine a first trajectory map according to first trajectory point information sent by a GPS device carried by the agricultural machinery, the first trajectory map comprising a first moving trajectory connected by multiple trajectory points, the first moving trajectory being used to represent the moving trajectory of the agricultural machinery in a first region, the first trajectory map being used to represent the relationship between the first moving trajectory and first state information, the first state information comprising at least one of the following information: speed information, time information and mileage information; and determine an initial point sequence according to the first trajectory point information, the initial point sequence comprising the longitude and latitude coordinates of the position of the agricultural machinery, the instantaneous speed of the agricultural machinery, the travel mileage of the agricultural machinery and the time of dotting for each trajectory point.

[0025] The satellite map server is configured to determine a first satellite map, the first satellite map being used to represent the first region, the first satellite map and the first trajectory map having the same size parameters and being in the same geodetic coordinate axis.

[0026] The image processing server is configured to input the first trajectory map, the first satellite map and the initial point sequence into a first neural network model, and output a first marked map, the first neural network model comprising an encoder and a processor, both the encoder and the processor comprising n stages, the first neural network model being configured to:

[0027] The first trajectory map and the first satellite map are superimposed to determine a first input image; in the first stage of the encoder, feature extraction is performed on the first input image at a first down-sampling scale to determine a first feature map, and in the first stage of the processor, a full connection operation is performed on the initial point sequence to determine a first point sequence, the number of channels for representing point information in the first point sequence being the same as the number of channels for representing point information in the first feature map; and based on the MSA mechanism, a self-attention operation is performed on the first feature map and the first point sequence to determine a first fused point sequence; in the mth stage of the encoder, feature extraction is performed on the (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and in the mth stage of the processor, a full connection operation is performed on the (m-1)th fused point sequence to determine an mth point sequence, the number of channels for representing point information in the mth point sequence being the same as the number of channels for representing point information in the mth feature map; and based on the MSA mechanism, a self-attention operation is performed on the mth feature map and the mth point sequence to determine an mth fused point sequence, m∈[2,n] and is an integer; in the case where m is equal to n, a full connection operation is performed on the nth fused point sequence to determine a target point sequence, the target point sequence including the longitude and latitude coordinates of each trajectory point constituting the first moving trajectory and classification information for the candidate, the classification information for the candidate including at least farmland and road; through a normalized exponential function, classification information to which each trajectory point constituting the first moving trajectory respectively belongs is determined from the classification information for the candidate in the target point sequence, and the first moving trajectory and the classification information to which each trajectory point constituting the first moving trajectory respectively belongs are mapped to the first satellite map according to the longitude and latitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, and a first marked map is output.

[0028] With reference to the second aspect, in some implementations of the second aspect, before the image processing server inputs the first trajectory map and the first satellite map into the first neural network model, the image processing server is further configured to: respectively expand a first length and a first width in the size parameters of the first trajectory map and the first satellite map to a second length and a second width.

[0029] With reference to the second aspect, in some implementations of the second aspect, before the image processing server superimposes the first trajectory map and the first satellite map, the image processing server is further configured to: perform a down-sampling operation on the first trajectory map and the first satellite map, reduce a second length and a second width in the size parameters of the first trajectory map and the first satellite map to a third length and a third width, and correspondingly expand the number of image channels in the size parameters of the first trajectory map and the first satellite map.

[0030] With reference to the second aspect, in some implementations of the second aspect, the image processing server is further configured to sequentially input, to the first neural network model, Q sets of data including the first trajectory map, the first satellite map and the first point sequence, the first neural network model corresponding to output Q first labeled maps, Q being an integer greater than 1; and stitch the Q first labeled maps to determine the second labeled map.

[0031] In a third aspect, a device for agricultural machine trajectory recognition is provided. The device includes a processor and a memory. The processor and the memory are connected. The memory is configured to store program code. The processor is configured to invoke the program code to execute any one of the methods in the possible implementation manners of the method design of the first aspect.

[0032] In a fourth aspect, a chip system is provided. The chip system is applied to an electronic device. The chip system includes one or more interface circuits and one or more processors. The interface circuits and the processors are connected through lines. The interface circuits are configured to receive signals from a memory of the electronic device and send signals to the processors. The signals include computer instructions stored in the memory. When the processors execute the computer instructions, the electronic device executes any one of the methods in the possible implementation manners of the method design of the first aspect.

[0033] In a fifth aspect, a computer readable storage medium is provided. The computer readable storage medium stores computer instructions. The computer instructions are executed by a processor to implement any one of the methods in the possible implementation manners of the method design of the first aspect.

[0034] In a sixth aspect, a computer program product is provided. The computer program code or instructions, when executed on a computer, cause the computer to execute any one of the methods in the possible implementation manners of the method design of the first aspect. BRIEF DESCRIPTION OF DRAWINGS

[0035] FIG. 1 is a schematic block diagram of a system 100 for agricultural machine trajectory recognition according to an embodiment of the present application;

[0036] FIG. 2 is a schematic block diagram of a method 200 for agricultural machine trajectory recognition according to an embodiment of the present application;

[0037] FIG. 3 is a schematic diagram of a trajectory map according to an embodiment of the present application;

[0038] FIG. 4 is a schematic block diagram of a neural network model architecture 400 according to an embodiment of the present application;

[0039] FIG. 5 is a schematic block diagram of another system 500 for agricultural machine trajectory recognition according to an embodiment of the present application. DETAILED DESCRIPTION

[0040] In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of or, for example, A / B can represent A or B; "and / or" herein is a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B can represent: A alone, A and B exist at the same time, and B alone. In the present application, "at least one" means one or more, and "multiple" means two or more. "At least one of the following" or the like means any combination of the items, including any combination of single item or multiple items. For example, at least one of a, b, or c can represent: a, b, c, a-b, a-c, b-c, or a-b-c, where a, b, and c can be single or multiple.

[0041] In the embodiments of the present application, the prefix words such as "first", "second" are only used to distinguish different description objects, and have no limiting effect on the position, order, priority, quantity or content of the described object. The use of ordinal words and other prefix words in the embodiments of the present application for distinguishing description objects does not constitute a limitation on the described objects, and the description of the described objects should be referred to the description of the context in the claims or embodiments, and should not constitute an unnecessary limitation because of the use of such prefix words.

[0042] The technical solutions in the embodiments of the present application will be described below with reference to the drawings.

[0043] Acquiring farmland distribution information is an important part of farming work, which affects the subsequent farming work strategy. Usually, the farming personnel adopts the method of on-site survey to acquire the farmland distribution information, and then marks the distribution information of the farmland on the specified map for display. This method is low in efficiency and poor in accuracy.

[0044] With the advancement of agricultural automation, informatization and intelligentization technology, the farming mode requiring human participation will be gradually replaced. And without human on-site survey, realizing marking the corresponding farmland distribution information on the map is also an important part of the development of agricultural technology.

[0045] In view of this, the embodiments of the present application propose a method, device and system for recognizing the track of agricultural machinery, which combines the information collected by satellite map and global positioning system (GPS) equipment through a transformer model configured with special functions, performs feature extraction and data fusion operations, accurately and efficiently recognizes the farmland area and road area in the satellite map, and marks the farmland area and road area on the satellite map.

[0046] FIG. 1 is a schematic block diagram of a system 100 for recognizing the trajectory of an agricultural machine according to an embodiment of the present application.

[0047] The system 100 includes an agricultural machine 110, a GPS device 120, a GPS server 130, a satellite map server 140, an image processing server 150, and a control center 160. These devices can be connected through a wireless network to realize data interaction between the devices.

[0048] The agricultural machine 110 is a machine used in agricultural production, such as a seeder, a drone for spraying fertilizer, or a combine harvester, and has a feature that the movement speed and trajectory of the agricultural machine in the working area are different from those in the non-working area.

[0049] The GPS device 120 is mounted on the agricultural machine 110 and is used to collect GPS data of the agricultural machine 110 during movement of the agricultural machine 110. The GPS data can be a sequence of points, each of which can be referred to as a trajectory point of the agricultural machine 110, and each of which can correspond to the following information: the longitude and latitude coordinates of the location of the agricultural machine corresponding to the trajectory point, the instantaneous speed of the agricultural machine, the mileage of the agricultural machine, and the time when the point is collected. Then, the GPS data is uploaded to the GPS server 130.

[0050] In some possible embodiments, the reporting mechanism of the GPS device 120 is as follows:

[0051] Each point data collected is reported to the GPS server 130; or

[0052] Based on a timer mechanism, the corresponding GPS data is reported to the GPS server 130 every predetermined time length; or

[0053] Based on the movement state of the agricultural machine 110, the corresponding GPS data is reported to the GPS server 130 every time the movement of the agricultural machine 110 is detected to stop, i.e., the longitude and latitude coordinates and the mileage of a plurality of trajectory points are the same, or the instantaneous speed of a plurality of trajectory points is 0; or

[0054] Based on a mileage mechanism, the corresponding GPS data is reported to the GPS server 130 every predetermined mileage; or

[0055] Based on a data stack mechanism, each point data collected is stored in a data stack with a predetermined length, and the corresponding GPS data is reported to the GPS server 130 every time the data stack is full, etc.

[0056] The GPS server 130 can be a server deployed remotely from the farming area, configured to generate a corresponding track map according to the GPS data. The track map includes the moving track of the agricultural machine 110. Then the track map is sent to the satellite map server 140 and the image processing server 150. And a track point sequence corresponding to the track map is generated according to the GPS data, and the track point sequence is sent to the image processing server 150.

[0057] The satellite map server 140 can also be a server deployed remotely from the farming area, configured to collect a satellite map adapted to the track map. The satellite map and the track map have the same size parameters and correspond to the same map area. Then the satellite map is sent to the image processing server 150.

[0058] In some possible embodiments, the size parameters of the track map and the satellite map can be set to the same parameters by the server. In addition, the satellite map corresponding to the track map can be determined in the following manner:

[0059] According to the size parameters, the satellite map is divided into multiple tiles, and different tiles correspond to different areas. Each track point in the track map can carry its corresponding longitude and latitude coordinates. Therefore, according to the longitude and latitude coordinates of each track point, the corresponding satellite map can be locked.

[0060] The image processing server 150 can also be a server deployed remotely from the farming area, configured to receive the track map and the track point sequence sent by the GPS server 130, and the satellite map sent by the satellite map server 140. Then the transformer model is used to perform corresponding digital image processing operations, and the information of the track point sequence is matched with the processed image features to obtain a satellite map marked with farmland areas and road areas. Then the marked satellite map is sent to the control center 160.

[0061] In some possible embodiments, the GPS server 130, the satellite map server 140 and the image processing server 150 can be independent servers deployed separately, or can be deployed together, or even be a large server integrating the functions.

[0062] The control center 160 is configured to perform statistical analysis on the farming data according to the marked satellite map, and perform corresponding farming planning based on the statistical analysis result, for example, to count the farming operation area of the agricultural machine, or to determine the spatial distribution of the crops, or to plan the farming operation route of the agricultural machine 110, etc.

[0063] Based on the system 100, the embodiment of the present application proposes a method for recognizing the track of the agricultural machine.

[0064] FIG. 2 is a schematic block diagram of a method 200 for identifying a track of an agricultural machine according to an embodiment of the present application.

[0065] S210: determining a first track map according to first track point information sent by a GPS device carried by the agricultural machine.

[0066] The first track map includes a first moving track connected by a plurality of track points, the first moving track is used to represent a moving track of the agricultural machine in a first area, the first track map is used to represent a relationship between the first moving track and first state information, and the first state information includes at least one of the following information: speed information, time information, and mileage information.

[0067] In some possible embodiments, the first track map described above can be an image based on a standard deviation (STD) of an image. The first track map is drawn according to a dotting time corresponding to each track point, and is a gray-scale image. Each track point constituting the first moving track corresponds to a different gray-scale value, which can be associated with the first state information described above.

[0068] FIG. 3 is a schematic diagram of a track map according to an embodiment of the present application.

[0069] (a) in FIG. 3 is a speed map, in which the gray-scale value of the track point is associated with the speed information, and the smaller the gray-scale value, the faster the speed;

[0070] (b) in FIG. 3 is a cumulative time map, in which the gray-scale value of the track point is associated with the time information, and the smaller the gray-scale value, the longer the cumulative time;

[0071] (c) in FIG. 3 is a travel mileage map, in which the gray-scale value of the track point is associated with the mileage information, and the smaller the gray-scale value, the longer the travel mileage.

[0072] S220: determining an initial point sequence according to the first track point information, the initial point sequence including a longitude and latitude coordinate of a location where the agricultural machine is located, an instantaneous speed of the agricultural machine, a travel mileage of the agricultural machine, and a dotting time.

[0073] It should be understood that, based on S220, since the initial point sequence is determined based on the first track point information, the first track point information can at least represent the information included in the initial point sequence. However, the data form of the first track point information and the initial point sequence is different. The former can be a common data information without a special data form, and the latter is a sequence form of the first track point information. Such a data form is more suitable for a neural network model required in a subsequent process, thereby facilitating subsequent matching of the track point and the image feature.

[0074] S230: determining a first satellite map, the first satellite map being used to represent the first region, the first satellite map and the first track map having the same size parameter and being in the same geodetic coordinate system.

[0075] According to S230, the first satellite map and the first track map need to have the same size parameter and be used to represent the same region, and need to be in the same geodetic coordinate system, so the map level and the size parameter of the two maps need to be determined in advance to accurately establish the correspondence between the two maps.

[0076] In some possible embodiments, for each first track map, a suitable map level needs to be selected. If the level is too low, the track may be too thin and lose a lot of details. If the level is too high, the image will be trapped in the local area and cannot obtain more global information. Therefore, before the present solution is executed, the level of the first track map can be determined through pre-experiment, and the level of the first track map can be level 18.

[0077] In some possible embodiments, since the first moving track can be long and the size of the first track map is limited, the first moving track often needs to be composed of multiple first track maps, and even up to thousands of first track maps. If the picture with too large size is directly input into the neural network model, the running speed of the model will be very slow, and a lot of invalid information will be included, such as a pure black background without track, which will cause waste of computing resources. Therefore, before the present solution is executed, the size parameter of the first track map can be determined through pre-experiment, and the size parameter can be a first length and a first width, which can be specifically represented based on the number of pixels, for example, can be 1024x1024. In addition, for each first track map, pixel information can be added at the edges thereof respectively, and the first length and the first width in the size parameter of the first track map are respectively extended to a second length and a second width, for example, can be 1280x1280, and then input into the model for calculation, so as to ensure that the model can obtain sufficient rich edge information of the image.

[0078] Similarly to the first track map, before the present solution is executed, the size parameter of the first satellite map can be determined through pre-experiment, and the size parameter is also a first length and a first width, and then the first length and the first width in the size parameter of the first satellite map are respectively extended to a second length and a second width. The specific operation is as follows:

[0079] Taking the size parameter of the first trajectory map as 1280x1280 as an example, first, a tile server is built on a satellite map of the same level as the first trajectory map; then, the corresponding tiles are locked according to the longitude and latitude coordinates of each trajectory point of the first trajectory; then, 4x4 256x256 tiles are combined into a large 1024x1024 tile; finally, pixel information is added to the edges of the satellite map corresponding to the large tile, and the size parameter is also expanded to 1280x1280, so as to ensure that the model can obtain sufficient rich image edge information.

[0080] S240: input the first trajectory map, the first satellite map and the initial point sequence into the first neural network model, and output the first labeled map.

[0081] The first neural network model includes an encoder and a processor, and both the encoder and the processor include n stages. The first neural network model is configured to:

[0082] S241: superimpose the first trajectory map and the first satellite map to determine the first input image.

[0083] It should be understood that since both are in the same geodetic coordinate system, the image orientations of the first trajectory map and the first satellite map can be consistent.

[0084] After obtaining the above first input image, the first input image is transmitted to the encoder with n stages, and the following n-stage feature extraction is performed.

[0085] S242: in each stage of the encoder, feature extraction is performed on the image at different down-sampling scales to determine the m-th feature map. Correspondingly, in each stage of the processor, a full connection operation is performed on the point sequence to determine the m-th fused point sequence, and based on the MSA mechanism, a self-attention operation is performed on the m-th feature map and the m-th point sequence to determine the m-th fused point sequence. S242 specifically operates as follows:

[0086] In the first stage of the encoder, feature extraction is performed on the first input image at the first down-sampling scale to determine the first feature map. In the first stage of the processor, a full connection operation is performed on the initial point sequence to determine the first point sequence, and the number of channels used to represent point information in the first point sequence is the same as the number of channels used to represent point information in the first feature map. Based on the mechanism, a self-attention operation is performed on the first feature map and the first point sequence to determine the first fused point sequence.

[0087] In the m-th stage of the encoder, feature extraction is performed on the (m-1)-th feature map at the m-th downsampling scale to determine the m-th feature map. In the m-th stage of the processor, a fully connected operation is performed on the (m-1)-th fusion point sequence to determine the m-th point sequence. The number of channels used to represent point information in the m-th point sequence is the same as the number of channels used to represent point information in the m-th feature map. Based on the MSA mechanism, a self-attention operation is performed on the m-th feature map and the m-th point sequence to determine the m-th fusion point sequence, where m∈[2,n] and is an integer.

[0088] It should be understood that S242 above is a traversal process, in which the first input image undergoes n stages of feature extraction operations in sequence.

[0089] In some possible embodiments, the above n stages are also called the n stages of the encoder, specifically four stages. In each stage, the input image needs to go through multiple transformer processing blocks to extract image features. The downsampling operation in each stage is as follows:

[0090] Taking the first input image as an example, which is a 24-channel image.

[0091] In stage 1, two blocks were used to downsample the input image by 1 / 4 scale. The input image had 24 channels and the output image had 32 channels, thus determining the first processing result image.

[0092] In stage 2, two blocks were used to downsample the first processed image by a scale of 1 / 8, with 32 input image channels and 64 output image channels, to determine the second processed image.

[0093] In stage 3, 6 blocks were used to downsample the second processed image by a scale of 1 / 16. The input image had 64 channels and the output image had 128 channels to determine the third processed image.

[0094] In stage 4, two blocks are used to downsample the third processing result image by a scale of 1 / 32. The input image has 128 channels and the output image has 256 channels to determine the fourth processing result image, which is the first feature image mentioned above.

[0095] When the number of image channels of the first input image is other than the number mentioned above, the number of blocks, input image channels and output image channels corresponding to each stage can be adaptively adjusted.

[0096] It should be understood that as the stage deepens, the size of the output processing result map will decrease, but the number of image channels will increase to save more information.

[0097] S243: In the case where m is equal to n, performing a full connection operation on the nth fusion point sequence to determine the target point sequence.

[0098] The target point sequence includes the longitude and latitude coordinates of each trajectory point constituting the first moving trajectory and the classification information for the candidate, and the classification information for the candidate at least includes farmland and road.

[0099] In some possible embodiments, when the agricultural machine is an agricultural machine running on land, the above-mentioned classification information for the candidate can be farmland and road. When the agricultural machine is an agricultural unmanned aerial vehicle for aerial fertilization, the above-mentioned classification information for the candidate can be farmland and aerial route. In addition, since the trajectory point also corresponds to the elevation information, the above-mentioned process can also take into account the elevation information of the trajectory point, so the above-mentioned classification information for the candidate can also include mountainous areas.

[0100] S244: Determine the classification information to which each trajectory point constituting the first moving trajectory in the target point sequence respectively belongs through a normalized exponential function (softmax) from the above-mentioned classification information for the candidate, and map the first moving trajectory and the classification information to which each trajectory point constituting the first moving trajectory respectively belongs to the above-mentioned first satellite map according to the longitude and latitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, and output a first marked map.

[0101] It should be understood that the above-mentioned classification information for the candidate should at least include three classifications: background, road, and farmland. Since each trajectory point is strongly associated with the first moving trajectory and is in the same plane as the first moving trajectory, the classification of background is not distinguishable, so this classification can be ignored. Therefore, the classification information for the candidate can at least include the two classifications of road and farmland.

[0102] It should be understood that the size parameters of the above-mentioned first marked map correspond to the first satellite map and the first trajectory map.

[0103] It should be understood that the classification information for the candidate should at least include three classifications: background, road, and farmland. Since each trajectory point is strongly associated with the first moving trajectory and is in the same plane as the first moving trajectory, the background classification is not distinguishable, and thus can be ignored. Therefore, the classification information for the candidate can include at least the road and farmland classifications. Based on this, the identification result for the first moving trajectory includes the location of each trajectory point of the first moving trajectory in the first satellite map and the classification information of each trajectory point, so as to determine whether the area covered by the first moving trajectory is farmland or a road.

[0104] Based on the above technical solution, the moving trajectory of the agricultural machine can be efficiently and accurately identified based on the trajectory point information collected by the GPS device and the corresponding satellite map, and the moving trajectory can be marked on the satellite map without manual on-site investigation. Moreover, the moving trajectory can represent the area type to which each part belongs, i.e., farmland or a road. Therefore, the distribution information of the farmland can be intuitively determined based on the part of the moving trajectory belonging to the farmland.

[0105] In some possible embodiments, based on the above S241 to S244 and the corresponding embodiments, the embodiment of the present application further proposes a neural network model architecture 400.

[0106] FIG. 4 is a schematic block diagram of the neural network model architecture 400 proposed by the embodiment of the present application.

[0107] As can be seen from FIG. 4, from a longitudinal perspective, the neural network model architecture 400 can include two branches, one being a sequence processing branch and the other being an image processing branch. In each stage, the point sequence information of the two branches is fully connected and the image feature extraction function is coordinated. After the processing of the n stages is completed, the two branches are merged into one. From a horizontal perspective, the neural network model architecture 400 can include an input layer 410, a feature extraction layer 420 having four stages, a feature point conversion layer 430, a feature fusion layer 440, a linear conversion layer 450, and an output layer 460.

[0108] The input layer 410 is configured to perform the above S241 to input the determined input image into the feature extraction layer 420.

[0109] The feature extraction layer 420 is configured to perform the full connection operation on the point sequence and the feature extraction on the input image at a specific down-sampling scale in the above S242.

[0110] The feature point conversion layer 430 is configured to perform a full connection operation on the point sequence, so that the number of channels used to represent point information in the point sequence is the same as the number of channels used to represent point information in the feature map of the stage.

[0111] The feature fusion layer 440 is configured to perform the self-attention operation on the point sequence and the feature map in S242.

[0112] The linear conversion layer 450 is configured to perform the full connection operation on the fused point sequence output after the n-stage processing, to determine the feature point set.

[0113] The output layer 460 is configured to perform S244 to determine the classification information corresponding to each trajectory point, mark the first moving trajectory with the classification information of each trajectory point on the first satellite map, and output the first marked map.

[0114] The specific operations of each layer are described in the corresponding embodiments above, and will not be repeated here.

[0115] In some possible embodiments, the first neural network model further includes a preprocessing layer between the input layer 410 and the feature extraction layer 420. In the preprocessing layer, downsampling operations are performed on the first trajectory map and the first satellite map, the second length and the second width in the size parameters of the first trajectory map and the first satellite map are reduced to the third length and the third width, and the number of image channels in the size parameters of the first trajectory map and the first satellite map is correspondingly expanded. Taking the first trajectory map and the first satellite map as 3-channel images as an example, a convolution kernel can be used in the preprocessing layer to perform downsampling operations on the first trajectory map and the first satellite map respectively, so that the length and the width of the two images are reduced, and the number of channels of the two images is expanded to 24.

[0116] Based on the above technical solutions, the size of the input image can be reduced, and the number of image features is also enriched, which helps to improve the accuracy of subsequent neural network models in extracting features of the input image.

[0117] In some possible embodiments, the first marked map includes the first satellite map and the first moving trajectory, the first moving trajectory is marked at the corresponding position of the first satellite map, the trajectory points with the classification information of farmland in the first moving trajectory are connected by first color lines, and the trajectory points with the classification information of road in the first moving trajectory are connected by second color lines. For example, the part of the first moving trajectory belonging to farmland is presented by green lines, and the part of the first moving trajectory belonging to road is presented by red lines.

[0118] Based on the above technical scheme, the classification of the moving track can be intuitively distinguished on the satellite map, the distribution information of the farmland is directly obtained, and subsequent statistical analysis of the distribution information of the farmland and determination of the corresponding farming plan are facilitated.

[0119] In some possible embodiments, since the size of the farmland is usually large, the span of the moving track of the agricultural machine is also wide, and based on the above embodiments, the size parameter of the input image is limited in order to guarantee the processing efficiency of the neural network model. A complete large-scale farmland can be completely presented by only one limited-size track map and satellite map. Therefore, for a whole piece of farmland, a plurality of track maps and a plurality of satellite maps can be used for presentation respectively. Correspondingly, the above first track map, first satellite map and first point sequence can be referred to as a set of data. In view of the above situation, the first neural network model needs to sequentially input Q sets of data, and then the first neural network model correspondingly outputs Q first marked maps, wherein Q is an integer greater than 1; then the Q first marked maps are spliced to determine a second marked map. The second marked map can include a complete large-scale farmland.

[0120] Based on the above technical scheme, while guaranteeing the processing efficiency of the neural network model, the second marked map that can completely present the large-scale farmland is obtained by splicing a plurality of first marked maps, and the intuitiveness of presenting the distribution information of the farmland is further increased.

[0121] In some possible embodiments, after receiving the above data set, the first neural network model can first determine the rationality of the data set: first, the coordinate points corresponding to the four corners of the first satellite map are obtained, and the coordinate region corresponding to the four coordinate points is determined. The coordinate points can be represented by longitude and latitude coordinates; then it is determined whether the coordinates of each track point in the first point sequence corresponding to the first track map are located in the above coordinate region. If all are located in the above coordinate region, it indicates that the data set is reasonable, and the subsequent data processing process can be performed. If there are track points not located in the above coordinate region, it indicates that the data set is abnormal, and feedback information can be sent to the upstream device to instruct the upstream device to resend the data set; or if the number of track points not located in the above coordinate region is less than the tolerable number, for example, 1, 5, etc., the track points are deleted from the first point sequence, and then the data processing is performed based on the data set after the track points are deleted; if the number of track points not located in the above coordinate region is greater than or equal to the tolerable number, feedback information is sent to the upstream device to instruct the upstream device to resend the data set.

[0122] Based on the above technical solution, the first neural network model can avoid subsequent data processing procedures with relatively large computational overhead for unreasonable data sets, thereby avoiding unnecessary large computational overhead.

[0123] In some possible embodiments, after the first mark map or the second mark map is acquired, a corresponding agricultural work is performed according to the first mark map or the second mark map based on a corresponding pre-trained model, which can be a neural network model, and the agricultural work includes at least one of the following:

[0124] Based on the first model, the farming area of the agricultural machine is counted; or

[0125] Based on the second model, the spatial distribution of the crops in the first area is determined; or

[0126] Based on the third model, the farming route of the agricultural machine is planned.

[0127] Based on the above technical solution, the manual participation in the above related agricultural work can be further reduced, and the full-automatic closed-loop management of the collection, analysis and planning of the farmland distribution information can be realized.

[0128] In some possible embodiments, the performance of the first neural network model can be evaluated based on a focal loss function (focal loss), and the parameters of the first neural network model can be adjusted based on the output result of the focal loss, wherein the focal loss includes a first weight and a second weight, the first weight corresponds to the road in the classification information, and the second weight corresponds to the farmland in the classification information. Through multiple tests, it is determined that when the first weight is set to 2.5 to 3.0, the first neural network model is more likely to converge.

[0129] In addition, the embodiments of the present application also provide a system for implementing any one of the above methods. FIG. 5 is a schematic block diagram of another system 500 for recognizing the track of the agricultural machine according to an embodiment of the present application.

[0130] The system 500 can be a subsystem of the system 100, and the system 500 includes a GPS server 510, a satellite map server 520 and an image processing server 530.

[0131] The GPS server 510 is configured to determine a first trajectory graph according to first trajectory point information sent by a GPS device carried by the agricultural machine, the first trajectory graph comprising a first movement trajectory connected by a plurality of trajectory points, the first movement trajectory being used to represent a movement trajectory of the agricultural machine in a first region, the first trajectory graph being used to represent a relationship between the first movement trajectory and first state information, the first state information comprising at least one of the following information: speed information, time information, and mileage information; and determine an initial point sequence according to the first trajectory point information, the initial point sequence comprising, for each trajectory point, a latitude and longitude coordinate of a position of the agricultural machine, an instantaneous speed of the agricultural machine, a travel mileage of the agricultural machine, and a dotting time.

[0132] The satellite map server 520 is configured to determine a first satellite map, the first satellite map being used to represent the first region, the first satellite map and the first trajectory graph having the same size parameter and being in the same geodetic coordinate axis.

[0133] The image processing server 530 is configured to input the first trajectory graph, the first satellite map, and the initial point sequence into a first neural network model, the first neural network model comprising an encoder and a processor, both the encoder and the processor comprising n stages, the first neural network model being configured to:

[0134] superimpose the first trajectory graph and the first satellite map to determine a first input image;

[0135] at a first stage of the encoder, perform feature extraction on the first input image at a first down-sampling scale to determine a first feature map, and at a first stage of the processor, perform a full connection operation on the initial point sequence to determine a first point sequence, a number of channels used to represent point information in the first point sequence being the same as a number of channels used to represent point information in the first feature map; and based on a MSA mechanism, perform a self-attention operation on the first feature map and the first point sequence to determine a first fused point sequence;

[0136] at an mth stage of the encoder, perform feature extraction on an (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and at an mth stage of the processor, perform a full connection operation on an (m-1)th fused point sequence to determine an mth point sequence, a number of channels used to represent point information in the mth point sequence being the same as a number of channels used to represent point information in the mth feature map; and based on the MSA mechanism, perform a self-attention operation on the mth feature map and the mth point sequence to determine an mth fused point sequence, m being an integer and belonging to a range of 2 to n;

[0137] in a case where m is equal to n, perform a full connection operation on the nth fused point sequence to determine a target point sequence, the target point sequence comprising latitude and longitude coordinates of each trajectory point constituting the first movement trajectory and classification information for a candidate, the classification information for the candidate at least comprising farmland and road;

[0138] The classification information to which each trajectory point constituting the first moving track respectively belongs in the target point sequence is determined from the classification information for the candidate by a normalized exponential function, and the first moving track and the classification information to which each trajectory point constituting the first moving track respectively belongs are mapped to the first satellite map according to the longitude and latitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, and a first marked map is output.

[0139] In some possible embodiments, before the image processing server 530 inputs the first trajectory map and the first satellite map into the first neural network model, the image processing server 530 is further configured to: respectively expand a first length and a first width in size parameters of the first trajectory map and the first satellite map to a second length and a second width.

[0140] In some possible embodiments, before the image processing server 530 superimposes the first trajectory map and the first satellite map, the image processing server 530 is further configured to: perform a downsampling operation on the first trajectory map and the first satellite map, reduce the second length and the second width in the size parameters of the first trajectory map and the first satellite map to a third length and a third width, and correspondingly expand the number of image channels in the size parameters of the first trajectory map and the first satellite map.

[0141] In some possible embodiments, the image processing server 530 is further configured to: input, to the first neural network model, Q groups of data sets including the first trajectory map, the first satellite map and the first point sequence in sequence, the first neural network model correspondingly outputs Q first marked maps, and Q is an integer greater than 1; and splice the Q first marked maps to determine a second marked map.

[0142] In some possible embodiments, the present application further provides a device for identifying a track of a farm machine, which comprises a processor and a memory, wherein the processor and the memory are connected, the memory is configured to store program code, and the processor is configured to call the program code to execute the method in any of the possible implementation manners described above.

[0143] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system, device and unit described above can refer to the corresponding process in the foregoing method embodiments, which will not be described herein.

[0144] In the multiple embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the described device embodiments are merely schematic. For example, the division of the units is only a logical function division. There can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0145] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0146] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be a physically independent unit, or two or more units can be integrated into a unit.

[0147] If the functions are implemented in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory, a magnetic disk or an optical disk, and various media that can store program codes.

[0148] The above is merely specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for agricultural machinery trajectory recognition, characterized in that, The method comprises: determining a first trajectory graph according to first trajectory point information sent by a global positioning system GPS device carried by the agricultural machine, the first trajectory graph comprising a first moving trajectory connected by a plurality of trajectory points, the first moving trajectory being used to represent a moving trajectory of the agricultural machine in a first region, the first trajectory graph being used to represent a relationship between the first moving trajectory and first state information, the first state information comprising at least one of the following information: speed information, time information and mileage information; determining an initial point sequence according to the first trajectory point information, the initial point sequence comprising, for each trajectory point, a latitude and longitude coordinate of a position where the agricultural machine is located, an instantaneous speed of the agricultural machine, a travel mileage of the agricultural machine and a dotting time; determining a first satellite map, the first satellite map being used to represent the first region, the first satellite map and the first trajectory graph having the same size parameter and being in the same geodetic coordinate axis; inputting the first trajectory graph, the first satellite map and the initial point sequence into a first neural network model to output a first marked map, the first marked map comprising an identification result for the first moving trajectory, the first neural network model comprising an encoder and a processor, the encoder and the processor each comprising n stages, the first neural network model being configured to: superimpose the first trajectory graph and the first satellite map to determine a first input image; in the 1st stage of the encoder, performing feature extraction on the first input image at a 1st down-sampling scale to determine a 1st feature map, and in the 1st stage of the processor, performing a full connection operation on the initial point sequence to determine a 1st point sequence, the number of channels for representing point information in the 1st point sequence being the same as the number of channels for representing point information in the 1st feature map; and based on a multi-head self-attention MSA mechanism, performing a self-attention operation on the 1st feature map and the 1st point sequence to determine a 1st fused point sequence; in the mth stage of the encoder, performing feature extraction on an (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and in the mth stage of the processor, performing a full connection operation on an (m-1)th fused point sequence to determine an mth point sequence, the number of channels for representing point information in the mth point sequence being the same as the number of channels for representing point information in the mth feature map; and based on the MSA mechanism, performing a self-attention operation on the mth feature map and the mth point sequence to determine an mth fused point sequence, m being an integer and belonging to [2, n]; in the case where the m is equal to n, performing a full connection operation on the nth fused point sequence to determine a target point sequence, the target point sequence comprising latitude and longitude coordinates of each trajectory point constituting the first moving trajectory and classification information for candidates, the classification information for candidates at least comprising farmland and road; and determining a first trajectory graph according to first trajectory point information sent by a global positioning system GPS device carried by the agricultural machine, the first trajectory graph comprising a first moving trajectory connected by a plurality of trajectory points, the first moving trajectory being used to represent a moving trajectory of the agricultural machine in a first region, the first trajectory graph being used to represent a relationship between the first moving trajectory and first state information, the first state information comprising at least one of the following information: speed information, time information and mileage information; determining an initial point sequence according to the first trajectory point information, the initial point sequence comprising, for each trajectory point, a latitude and longitude coordinate of a position where the agricultural machine is located, an instantaneous speed of the agricultural machine, a travel mileage of the agricultural machine and a dotting time; determining a first satellite map, the first satellite map being used to represent the first region, the first satellite map and the first trajectory graph having the same size parameter and being in the same geodetic coordinate axis; inputting the first trajectory graph, the first satellite map and the initial point sequence into a first neural network model to output a first marked map, the first marked map comprising an identification result for the first moving trajectory, the first neural network model comprising an encoder and a processor, the encoder and the processor each comprising n stages, the first neural network model being configured to: superimpose the first trajectory graph and the first satellite map to determine a first input image; in the 1st stage of the encoder, performing feature extraction on the first input image at a 1st down-sampling scale to determine a 1st feature map, and in the 1st stage of the processor, performing a full connection operation on the initial point sequence to determine a 1st point sequence, the number of channels for representing point information in the 1st point sequence being the same as the number of channels for representing point information in the 1st feature map; and based on a multi-head self-attention MSA mechanism, performing a self-attention operation on the 1st feature map and the 1st point sequence to determine a 1st fused point sequence; in the mth stage of the encoder, performing feature extraction on an (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and in the mth stage of the processor, performing a full connection operation on an (m-1)th fused point sequence to determine an mth point sequence, the number of channels for representing point information in the mth point sequence being the same as the number of channels for representing point information in the mth feature map; and based on the MSA mechanism, performing a self-attention operation on the mth feature map and the mth point sequence to determine an mth fused point sequence, m being an integer and belonging to [2, n]; in the case where the m is equal to n, performing a full connection operation on the nth fused point sequence to determine a target point sequence, the target point sequence comprising latitude and longitude coordinates of each trajectory point constituting the first moving trajectory and classification information for candidates, the classification information for candidates at least comprising farmland and road; and determine, from the classification information for the candidate, classification information to which each trajectory point constituting the first moving track in the target point sequence respectively belongs through a normalized exponential function, and map the first moving track and the classification information to which each trajectory point constituting the first moving track respectively belongs to the first satellite map according to the longitude and latitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, and output the first marked map.

2. The method of claim 1, wherein, Before inputting the first trajectory map and the first satellite map into the first neural network model, the method further comprises: expanding a first length and a first width in size parameters of the first trajectory map and the first satellite map to a second length and a second width, respectively.

3. The method of claim 2, wherein, Before superimposing the first trajectory map and the first satellite map to determine the first input image, the method further comprises: performing a downsampling operation on the first trajectory map and the first satellite map to reduce a second length and a second width in size parameters of the first trajectory map and the first satellite map to a third length and a third width, and correspondingly expand the number of image channels in size parameters of the first trajectory map and the first satellite map.

4. The method of claim 1, wherein, The trajectory points of the first moving track with classification information as farmland are connected by first color lines, and the trajectory points of the first moving track with classification information as roads are connected by second color lines.

5. The method according to any one of claims 1 to 4, characterized in that, The method further comprises: inputting Q sets of data sets comprising the first trajectory map, the first satellite map and the initial point sequence into the first neural network model in sequence, and the first neural network model correspondingly outputs Q first marked maps, wherein Q is an integer greater than 1; stitching the Q first marked maps to determine a second marked map.

6. A system for recognizing the trajectory of agricultural machinery, characterized in that, The device comprises a global positioning system (GPS) server, a satellite map server and an image processing server. The GPS server is configured to determine a first trajectory map according to first trajectory point information sent by a GPS device carried by a farm machine, the first trajectory map comprising a first moving track connected by a plurality of trajectory points, the first moving track being used to represent a moving track of the farm machine in a first area, the first trajectory map being used to represent a relationship between the first moving track and first state information, the first state information comprising at least one of the following information: speed information, time information and mileage information; and determine an initial point sequence according to the first trajectory point information, the initial point sequence comprising longitude and latitude coordinates of a position where the farm machine is located, instantaneous speed of the farm machine, travel mileage of the farm machine and a dotting time; The satellite map server is configured to determine a first satellite map, the first satellite map being used to represent the first area, size parameters of the first satellite map being the same as those of the first trajectory map, and being in a same geodetic coordinate axis; The image processing server is configured to input the first trajectory graph, the first satellite map and the initial point sequence into a first neural network model, output a first labeled map, the first labeled map comprising an identification result for the first moving trajectory, the first neural network model comprising an encoder and a processor, both the encoder and the processor comprising n stages, the first neural network model being configured to: superimpose the first trajectory graph and the first satellite map to determine a first input image; at stage 1 of the encoder, performing feature extraction on the first input image at a first down-sampling scale to determine a first feature map, and at stage 1 of the processor, performing a full connection operation on the initial point sequence to determine a first point sequence, the number of channels in the first point sequence for representing point information being the same as the number of channels in the first feature map for representing point information; and based on a MSA mechanism, performing a self-attention operation on the first feature map and the first point sequence to determine a first fused point sequence; at stage m of the encoder, performing feature extraction on an (m-1)th feature map at an mth down-sampling scale to determine an mth feature map, and at stage m of the processor, performing a full connection operation on an (m-1)th fused point sequence to determine an mth point sequence, the number of channels in the mth point sequence for representing point information being the same as the number of channels in the mth feature map for representing point information; and based on a MSA mechanism, performing a self-attention operation on the mth feature map and the mth point sequence to determine an mth fused point sequence, m∈[2,n] and being an integer; in a case where the m is equal to n, performing a full connection operation on an nth fused point sequence to determine a target point sequence, the target point sequence comprising the latitude and longitude coordinates of each trajectory point constituting the first moving trajectory and classification information for candidates, the classification information for candidates at least comprising farmland and road; by a normalized exponential function, determining, from the classification information for candidates, classification information to which each trajectory point constituting the first moving trajectory respectively belongs in the target point sequence, and according to the latitude and longitude coordinates of each trajectory point and the classification information to which each trajectory point respectively belongs, mapping the first moving trajectory and the classification information to which each trajectory point constituting the first moving trajectory respectively belongs to the first satellite map, and outputting the first labeled map.

7. The system of claim 6, wherein, Before the image processing server inputs the first trajectory graph and the first satellite map into the first neural network model, the image processing server is further configured to: expand a first length and a first width in size parameters of the first trajectory graph and the first satellite map to a second length and a second width, respectively.

8. The system of claim 7, wherein, Before the image processing server superimposes the first trajectory graph and the first satellite map, the image processing server is further configured to: performing a downsampling operation on the first trajectory map and the first satellite map, reducing a second length and a second width in size parameters of the first trajectory map and the first satellite map to a third length and a third width, and extending a number of image channels in the size parameters of the first trajectory map and the first satellite map correspondingly.

9. The system of claim 8, wherein, The image processing server is further configured to: inputting, into the first neural network model, a data set including the first trajectory map, the first satellite map, and the initial point sequence in sequence, the first neural network model outputting Q first labeled maps correspondingly, Q being an integer greater than 1; stitching the Q first labeled maps to determine a second labeled map.

10. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer instructions, and the computer instructions are executed by the processor to implement the method in any one of claims 1 to 5.

Images