Land area measurement method based on image segmentation network, drone, medium

Abstract

The present application discloses a land area measurement method, drone, and medium based on an image segmentation network. The method includes: determining a target land area including multiple monitoring nodes; determining a target inspection path based on the multiple monitoring nodes, and capturing a land image at each monitoring node based on the target inspection path; determining a target GSD based on target distance information and target pixel number, wherein the target distance information is the actual distance between the reference object and the monitoring node, and the target pixel number is the number of pixels from the reference point to the image center point of the land image; inputting the land image into an image segmentation network model for image segmentation processing to obtain a target segmented image; and obtaining the land area of the target land area based on the target segmented image and the target GSD. According to the solution provided in the embodiment of the present application, the land area of the target land area can be measured by combining the drone and the image segmentation network model, thereby effectively improving the efficiency of measuring the land area.

Description

Technical Field

[0001] The present application relates to, but is not limited to, the field of image processing technology, and in particular to a land area measurement method, drone, and medium based on an image segmentation network. Background Art

[0002] In order to better realize the scientific and rational planning of land resources, it is crucial to obtain land area data. At present, the main method for measuring land area is the field measurement method, such as manually pulling a tape measure to measure the area of the target land area. Although the field measurement method has high data collection accuracy, it has problems such as low degree of informatization, slow measurement speed, and high investment in manpower and material resources, resulting in low efficiency of land area measurement. Summary of the Invention

[0003] The embodiments of the present application provide a land area measurement method, drone, and medium based on an image segmentation network, which can effectively improve the efficiency of measuring land area.

[0004] In a first aspect, an embodiment of the present application provides a land area measurement method based on an image segmentation network, which is applied to a drone. The method includes:

[0005] determining a target land area, wherein the target land area includes a plurality of monitoring nodes;

[0006] Determining a target inspection path according to the plurality of monitoring nodes, and photographing a land image at each of the monitoring nodes according to the target inspection path, wherein the monitoring node is a location node corresponding to an image center point of the land image;

[0007] Acquiring target distance information, where the target distance information is the actual distance between a preset reference object and the monitoring node;

[0008] Obtaining a target pixel number, where the target pixel number is the number of pixels between a reference point corresponding to the reference object in the land image and an image center point of the land image;

[0009] Determine a target GSD based on the target distance information and the target pixel number;

[0010] Inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image;

[0011] The land area of the target land region is obtained according to the target segmentation image and the target GSD.

[0012] In some embodiments, the image segmentation network model is trained according to the following steps:

[0013] Determining location information of the land image, and acquiring a map remote sensing image corresponding to the location information, wherein the land image and the map remote sensing image have the same image size;

[0014] Acquire a first annotated image corresponding to the land image, and acquire a second annotated image corresponding to the map remote sensing image;

[0015] Inputting the land image, the map remote sensing image, the first annotated image, and the second annotated image into a preset cross-modal alignment network model for image alignment to obtain a target training image;

[0016] An initial image segmentation network model is obtained, and the initial image segmentation network model is trained according to the target training image to obtain the image segmentation network model.

[0017] In some embodiments, the cross-modal alignment network model includes a first encoder, a second encoder, a third encoder, a fourth encoder, and a first decoder. Inputting the land image, the map remote sensing image, the first annotated image, and the second annotated image into a preset cross-modal alignment network model for image alignment to obtain a target training image includes:

[0018] Inputting the first annotated image and the second annotated image into the first encoder for encoding to obtain a first feature vector;

[0019] Inputting the map remote sensing image into the second encoder for encoding processing to obtain a second feature vector;

[0020] Inputting the land image into the third encoder for encoding to obtain a third feature vector;

[0021] Inputting the first eigenvector, the second eigenvector, and the third eigenvector into the fourth encoder for encoding respectively to obtain a new first eigenvector, a new second eigenvector, and a new third eigenvector;

[0022] Performing feature alignment processing on the new first eigenvector and the new second eigenvector to obtain first intermediate data;

[0023] Performing feature alignment processing on the new second eigenvector and the new third eigenvector to obtain second intermediate data;

[0024] Inputting the first intermediate data and the second intermediate data into the first decoder for decoding respectively to obtain an aligned target annotation image and a target land image;

[0025] An image merging process is performed on the target annotated image and the target land image to obtain the target training image.

[0026] In some embodiments, the image segmentation network model includes a fifth encoder and a second decoder, the fifth encoder includes a feature extraction network and a first convolutional layer, the second decoder includes a second convolutional layer, a third convolutional layer, a first upsampling layer, and a second upsampling layer, and inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image includes:

[0027] Inputting the land image into the feature extraction network to perform feature extraction to obtain a fourth feature vector;

[0028] Inputting the fourth eigenvector into the first convolutional layer for convolution operation to obtain a fifth eigenvector, and inputting the fifth eigenvector into the first upsampling layer for upsampling to obtain a sixth eigenvector;

[0029] Inputting the fourth eigenvector into the second convolutional layer for convolution operation to obtain a seventh eigenvector;

[0030] Performing feature splicing processing on the sixth eigenvector and the seventh eigenvector to obtain an eighth eigenvector;

[0031] The eighth eigenvector is sequentially input into the third convolutional layer and the second upsampling layer to obtain the target segmented image.

[0032] In some embodiments, the target segmentation image includes patches of multiple land types, and obtaining the land area of the target land region according to the target segmentation image and the target GSD includes:

[0033] Determining the number of pixels of each of the image patches;

[0034] The product of the number of pixels of each of the map patches and the target GSD is determined as the land area of the land type corresponding to the map patch.

[0035] In some embodiments, the land area measurement method based on the image segmentation network further includes:

[0036] Calculating the longitude and latitude information of each of the map patches according to a preset longitude and latitude conversion rule;

[0037] Calculating the graphic size information of each of the image patches according to a preset edge detection algorithm;

[0038] The latitude and longitude information and the graphic size information are saved in a preset database.

[0039] In some embodiments, the image segmentation network model is trained according to a cross entropy loss function.

[0040] In a second aspect, an embodiment of the present application provides a drone, comprising:

[0041] a target land area determination module, the target land area determination module being used to determine a target land area, the target land area including a plurality of monitoring nodes;

[0042] a land image acquisition module, the land image acquisition module being configured to determine a target inspection path based on the plurality of monitoring nodes, and capture a land image at each monitoring node according to the target inspection path, wherein the monitoring node is a location node corresponding to an image center point of the land image;

[0043] A target distance information acquisition module, wherein the target distance information acquisition module is used to acquire target distance information, wherein the target distance information is the actual distance between a preset reference object and the monitoring node;

[0044] a target pixel number acquisition module, the target pixel number acquisition module being used to acquire the target pixel number, the target pixel number being the number of pixels from the reference point corresponding to the reference object in the land image to the image center point of the land image;

[0045] A target GSD acquisition module is used to determine a target ground sampling distance value GSD according to the target distance information and the number of target pixels;

[0046] A target segmentation image acquisition module is used to input the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmentation image;

[0047] A land area acquisition module is used to obtain the land area of the target land region according to the target segmented image and the target GSD.

[0048] In a third aspect, an embodiment of the present application provides a drone, comprising: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the land area measurement method based on the image segmentation network as described in the first aspect is implemented.

[0049] In a third aspect, an embodiment of the present application further provides a computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the land area measurement method based on the image segmentation network as described in the first aspect.

[0050] The embodiment of the present application provides a land area measurement method, drone, and medium based on an image segmentation network, wherein the method includes: determining a target land area, wherein the target land area includes multiple monitoring nodes; determining a target inspection path based on the multiple monitoring nodes, and photographing a land image at each of the monitoring nodes according to the target inspection path, wherein the monitoring node is a position node corresponding to the image center point of the land image; obtaining target distance information, wherein the target distance information is the actual distance between a preset reference object and the monitoring node; obtaining a target pixel number, wherein the target pixel number is the number of pixels from the reference point corresponding to the reference object in the land image to the image center point of the land image; determining a target ground sampling distance value GSD based on the target distance information and the target pixel number; inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image; and obtaining the land area of the target land area based on the target segmented image and the target GSD. According to the solution provided by the embodiment of the present application, the land area of the target land area can be measured by combining the drone and the image segmentation network model, thereby effectively improving the efficiency of measuring the land area. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 This is a flowchart of the steps of a land area measurement method based on an image segmentation network provided by an embodiment of the present application;

[0052] Figure 2 This is a flowchart of the steps for training an image segmentation network model provided by another embodiment of the present application;

[0053] Figure 3 is a flowchart of the steps for obtaining a target training image provided by another embodiment of the present application;

[0054] Figure 4 is a flowchart of the steps for obtaining a target segmented image provided by another embodiment of the present application;

[0055] Figure 5 is a flowchart of steps for obtaining the land area of a target land area provided by another embodiment of the present application;

[0056] Figure 6 This is a flowchart of the steps of saving data to a database provided by another embodiment of the present application;

[0057] Figure 7 This is a schematic diagram of a module of a drone provided by another embodiment of the present application;

[0058] Figure 8 is a network structure diagram of a cross-modal alignment network model provided by another embodiment of the present application;

[0059] Figure 9 is a network structure diagram of an image segmentation network model provided by another embodiment of the present application;

[0060] Figure 10 This is a structural diagram of a drone provided in another embodiment of the present application. DETAILED DESCRIPTION

[0061] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0062] It is understood that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and the like in the specification, claims, or accompanying drawings are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0063] The present application provides a land area measurement method, drone, and medium based on an image segmentation network, wherein the method includes: determining a target land area, wherein the target land area includes multiple monitoring nodes; determining a target inspection path based on the multiple monitoring nodes, and photographing a land image at each of the monitoring nodes according to the target inspection path, wherein the monitoring node is a position node corresponding to the image center point of the land image; obtaining target distance information, wherein the target distance information is the actual distance between a preset reference object and the monitoring node; obtaining a target pixel number, wherein the target pixel number is the number of pixels from the reference point corresponding to the reference object in the land image to the image center point of the land image; determining a target ground sampling distance value GSD based on the target distance information and the target pixel number; inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image; and obtaining the land area of the target land area based on the target segmented image and the target GSD. According to the solution provided in the embodiment of the present application, the land area of the target land area can be measured by combining the drone and the image segmentation network model, thereby effectively improving the efficiency of measuring the land area.

[0064] The embodiments of the present application are further described below with reference to the accompanying drawings.

[0065] like Figure 1 As shown, Figure 1This is a flowchart of the steps of a land area measurement method based on an image segmentation network provided by an embodiment of the present application. The embodiment of the present application provides a land area measurement method based on an image segmentation network, which is applied to a drone. The method includes but is not limited to the following steps:

[0066] Step S110, determining a target land area, where the target land area includes a plurality of monitoring nodes;

[0067] Step S120, determining a target inspection path based on the plurality of monitoring nodes, and capturing a land image at each monitoring node according to the target inspection path, wherein the monitoring node is a location node corresponding to the image center point of the land image;

[0068] Step S130, obtaining target distance information, where the target distance information is the actual distance between the preset reference object and the monitoring node;

[0069] Step S140, obtaining the target pixel number, which is the number of pixels between the reference point corresponding to the reference object in the land image and the image center point of the land image;

[0070] Step S150, determining the target GSD according to the target distance information and the number of target pixels;

[0071] Step S160: inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image;

[0072] Step S170 , obtaining the land area of the target land region according to the target segmented image and the target GSD.

[0073] It should be noted that the drone of this embodiment can be any model of drone, as long as the endurance is sufficient to complete the inspection. In order to obtain land images, remote sensing equipment can be installed in the drone. This embodiment does not make any improvements to the specific hardware structure, as long as the function can be realized.

[0074] It should be noted that the embodiments of the present application do not limit the specific method for determining the target inspection path, which can be obtained by performing path planning on multiple monitoring nodes in the target land area through a genetic algorithm.

[0075] It should be noted that this embodiment does not limit the specific image type of the land image. The land image in this embodiment is an orthographic image, and can also be an oblique image. Those skilled in the art can select it according to actual needs.

[0076] It should be noted that the embodiments of the present application do not limit the specific number of reference objects, and those skilled in the art can determine it according to actual needs.

[0077] It should be noted that the specific implementation method of determining the target GSD according to the target distance information and the number of target pixels in the embodiment of the present application can be obtained according to the following formula:

[0078] GSD=γ4×σ1 4 +γ3×σ3 3 +γ2×σ2 2 +γ1×σ1+γ0;

[0079] Where i is the number of reference objects, i∈(1,2,...,N), σ i is the coefficient factor, σ i =h i / m i , h i is the target distance information, m i is the target pixel number, γ is a preset coefficient. It can be understood that obtaining the target GSD can provide an effective data basis for obtaining the land area of the target land area.

[0080] It can be understood that the target land area can be set according to actual needs. After the UAV determines the target inspection path and receives the inspection instruction, the land image can be captured according to the target inspection path and the preset height at each monitoring node in the target land area. The land image is input into the image segmentation network model pre-trained by the UAV for image segmentation processing to obtain the target segmented image, and the land area of the target land area is obtained according to the target GSD and the target segmented image; according to the solution provided in the embodiment of the present application, the land area of the target land area can be measured in combination with the UAV and the image segmentation network model. Compared with the traditional technical solution of measuring the land area by actual measurement, the efficiency of measuring the land area can be more effectively improved.

[0081] In addition, refer to Figure 2 In one embodiment, Figure 1 The image segmentation network model in the illustrated embodiment is trained according to the following steps:

[0082] Step S210, determining the location information of the land image, and obtaining a map remote sensing image corresponding to the location information, wherein the land image and the map remote sensing image have the same image size;

[0083] Step S220, obtaining a first annotated image corresponding to the land image, and obtaining a second annotated image corresponding to the map remote sensing image;

[0084] Step S230: inputting the land image, the map remote sensing image, the first annotated image, and the second annotated image into a preset cross-modal alignment network model for image alignment to obtain a target training image;

[0085] Step S240: Obtain an initial image segmentation network model, and train the initial image segmentation network model according to the target training image to obtain an image segmentation network model.

[0086] It should be noted that the embodiments of the present application do not limit the specific method of determining the location information of the land image and obtaining the map remote sensing image corresponding to the location information. The drone can obtain the current location information of the drone by calling the API interface of the map navigation software while shooting the land image, and obtain the map remote sensing image corresponding to the location information that is consistent with the size of the land image.

[0087] It should be noted that the embodiment of the present application does not limit the land type annotation information contained in the first annotated image and the second annotated image. The land type annotation information may include annotation information such as homestead, cultivated land, woodland or fish pond.

[0088] It can be understood that the image segmentation network model obtained by training the initial image segmentation network model based on the land image, map remote sensing image, first annotated image and second annotated image can effectively improve the accuracy of image segmentation, thereby improving the data availability of land area.

[0089] In addition, refer to Figure 3 In one embodiment, the cross-modal alignment network model includes a first encoder, a second encoder, a third encoder, a fourth encoder and a first decoder. Figure 2 Step S230 in the illustrated embodiment further includes but is not limited to the following steps:

[0090] Step S310: Input the first annotated image and the second annotated image into a first encoder for encoding to obtain a first feature vector;

[0091] Step S320: inputting the map remote sensing image into a second encoder for encoding to obtain a second feature vector;

[0092] Step S330: Input the land image into a third encoder for encoding to obtain a third eigenvector;

[0093] Step S340: Input the first eigenvector, the second eigenvector, and the third eigenvector into a fourth encoder for encoding, to obtain a new first eigenvector, a new second eigenvector, and a new third eigenvector;

[0094] Step S350: performing feature alignment processing on the new first eigenvector and the new second eigenvector to obtain first intermediate data;

[0095] Step S360: performing feature alignment processing on the new second eigenvector and the new third eigenvector to obtain second intermediate data;

[0096] Step S370: Input the first intermediate data and the second intermediate data into a first decoder for decoding, to obtain an aligned target annotation image and a target land image;

[0097] Step S380 , performing image merging processing on the target annotation image and the target land image to obtain a target training image.

[0098] It should be noted that the specific structure of the cross-modal alignment network model in the embodiment of the present application can be referred to Figure 7 The following describes the specific steps of obtaining the target training image in this embodiment using a specific example: Figure 7 The cross-modal alignment network model shown includes a first encoder, a second encoder, a third encoder, a fourth encoder and a first decoder. The first annotated image and the second annotated image are input to the first encoder for encoding processing to obtain a first feature vector A i , input the map remote sensing image into the second encoder for encoding processing, and obtain the second eigenvector B i , input the land image into the third encoder for encoding processing, and obtain the third eigenvector C i , i∈(1,2,...,S), S is A i 、B i 、C i The total number of the first eigenvector A i , the second eigenvector B i and the third eigenvector C i are respectively input into the fourth encoder for encoding processing to obtain a new first feature vector a i , the new second eigenvector b i and the new third eigenvector c i ; Perform feature alignment processing on the new first eigenvector and the new second eigenvector to obtain first intermediate data, perform feature alignment processing on the new second eigenvector and the new third eigenvector to obtain second intermediate data, input the first intermediate data and the second intermediate data into the first decoder for decoding processing respectively, obtain the aligned target annotation image and the target land image, perform image merging processing on the target annotation image and the target land image to obtain the target training image, in order to improve the accuracy of the first intermediate data and the second intermediate data, the embodiment of the present application combines a preset alignment loss function to realize feature alignment processing between the new first eigenvector and the new second eigenvector, and between the new second eigenvector and the new third eigenvector, and the expression of the alignment loss function is as follows:

[0099] L match =L align +L reson ;

[0100] Among them, L align The expression is:

[0101] L align =α×L align1 +β×L align2 ;

[0102] Among them, L align1 , L align2 The expressions are:

[0103] L align1 =max(1+Dis(a i ,b i )-Dis(a i ,b j ),0);

[0104] L align2 =max(1+Dis(a i ,b i )-Dis(a i ,b j ),0);

[0105] Among them, L reson The expression is:

[0106] L reson =α×L reson1 +β×L reson2 ;

[0107] Among them, L reson1 , L reson2 The expressions are:

[0108]

[0109]

[0110] in, Annotate the aligned target image, is the aligned target land image, α and β are coefficient factors that can balance the two alignment weights. The specific values can be determined according to actual conditions. In the embodiment of the present application, the value can be 0.5.

[0111] It can be understood that obtaining target training images can provide an effective data basis for training the initial image segmentation network model.

[0112] In addition, in one embodiment, the image segmentation network model includes a fifth encoder and a second decoder, the fifth encoder includes a feature extraction network and a first convolution layer, and the second decoder includes a second convolution layer, a third convolution layer, a first upsampling layer, and a second upsampling layer. Figure 4 , Figure 1 Step S160 in the illustrated embodiment further includes but is not limited to the following steps:

[0113] Step S410: inputting the land image into a feature extraction network to extract features and obtain a fourth feature vector;

[0114] Step S420: Input the fourth eigenvector to the first convolutional layer for convolution operation to obtain a fifth eigenvector, and input the fifth eigenvector to the first upsampling layer for upsampling to obtain a sixth eigenvector;

[0115] Step S430: input the fourth eigenvector into the second convolutional layer for convolution operation to obtain a seventh eigenvector;

[0116] Step S440, performing feature concatenation processing on the sixth eigenvector and the seventh eigenvector to obtain an eighth eigenvector;

[0117] Step S450: input the eighth eigenvector into the third convolutional layer and the second upsampling layer in sequence to obtain a target segmented image.

[0118] It should be noted that the specific structure of the image segmentation network model in the embodiment of the present application can be referred to Figure 8 ,like Figure 8 As shown, the image segmentation network model includes a fifth encoder and a second decoder. The fifth encoder includes a feature extraction network and a first convolutional layer. The second decoder includes a second convolutional layer, a third convolutional layer, a first upsampling layer, and a second upsampling layer. The embodiments of the present application do not limit the specific structure of each network layer in the image segmentation network model. For example, the feature extraction network may include: a convolutional layer with a 1×1 convolution kernel; three convolutional layers with a 3×3 convolution kernel and a dilation rate of 6; and a pooling layer. For example, the convolution kernels of the first and second convolutional layers are 1×1, the convolution kernel of the third convolutional layer is 3×3, and the number of upsampling times of the first and second upsampling layers is 4.

[0119] It should be noted that in order to improve the accuracy of the image segmentation network model in obtaining the target segmented image, the embodiment of the present application trains the image segmentation network model through the cross entropy loss function. The formula of the cross entropy loss function is as follows:

[0120]

[0121] Where C is the identifier of the land type, p∈(p0,…,p C-1) is a probability distribution, each element p i Indicates the probability that each patch in the target segmentation image belongs to the i-th land type. C-1 ) is the onehot representation of the category to which the target annotation image belongs. When the patch in the target segmentation image belongs to the i-th category, y i =1, otherwise y i =0.

[0122] It can be understood that obtaining the target segmented image can provide an effective data basis for obtaining the land area of the target land region.

[0123] In addition, refer to Figure 5 In one embodiment, the target segmented image includes patches of multiple land types. Figure 1 Step S170 in the illustrated embodiment further includes but is not limited to the following steps:

[0124] Step S510, determining the number of pixels of each image patch;

[0125] Step S520: The product of the number of pixels of each patch and the target GSD is determined as the land area of the land type corresponding to the patch.

[0126] It can be understood that by determining the number of pixels of each patch of different land types in the target segmentation image, and determining the product of the number of pixels of each patch and the target GSD as the land area of the land type corresponding to the patch, the intelligent calculation of the land area can be realized, which improves the efficiency of measuring land area compared with the technical solution of manual measurement.

[0127] In addition, in one embodiment, referring to Figure 6 The land area measurement method based on image segmentation network provided in the embodiment of the present application also includes but is not limited to the following steps:

[0128] Step S610, calculating the longitude and latitude information of each image patch according to a preset longitude and latitude conversion rule;

[0129] Step S620, calculating the graphic size information of each image patch according to a preset edge detection algorithm;

[0130] Step S630: Save the latitude and longitude information and the graphic size information to a preset database.

[0131] It should be noted that the embodiment of the present application does not limit the file storage format when saving the latitude and longitude information and graphic size information to a preset database. It can be a vector file format, raster file format or compressed raster file format corresponding to the geographic information system GIS, etc., and technical personnel in this field can adjust it according to actual needs.

[0132] It can be understood that the longitude and latitude information of each map patch is calculated according to the preset longitude and latitude conversion rules, the graphic size information of each map patch is calculated according to the preset edge detection algorithm, and the longitude and latitude information and graphic size information are saved to the preset database. This can facilitate land planning-related workers to read the information and provide an effective data basis for subsequent land planning work.

[0133] In addition, refer to Figure 9 , Figure 9 : is a schematic diagram of a module of a drone provided by another embodiment of the present application. In one embodiment, the present application provides a drone 900, which includes:

[0134] A target land area determination module 910 is used to determine a target land area, where the target land area includes multiple monitoring nodes;

[0135] The land image acquisition module 920 is used to determine a target inspection path based on multiple monitoring nodes, and capture land images at each monitoring node according to the target inspection path, wherein the monitoring node is a location node corresponding to the image center point of the land image;

[0136] The target distance information acquisition module 930 is used to acquire target distance information, which is the actual distance between the preset reference object and the monitoring node;

[0137] The target pixel number acquisition module 940 is used to acquire the target pixel number, which is the number of pixels from the reference point corresponding to the reference object in the land image to the image center point of the land image;

[0138] A target GSD acquisition module 950 is used to determine a target ground sampling distance value GSD according to target distance information and target pixel number;

[0139] The target segmentation image acquisition module 960 is used to input the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmentation image;

[0140] The land area acquisition module 970 is used to obtain the land area of the target land region according to the target segmentation image and the target GSD.

[0141] The specific implementation method of the drone obtaining the land area of the target land area is basically the same as the specific embodiment of the land area measurement method based on the image segmentation network mentioned above, and will not be repeated here.

[0142] In addition, reference Figure 10 , Figure 10 This is a structural diagram of a drone provided by another embodiment of the present application. One embodiment of the present application also provides a drone 1000, which includes: a memory 1010, a processor 1020, and a computer program stored in the memory 1010 and executable on the processor 1020.

[0143] The processor 1020 and the memory 1010 may be connected via a bus or other means.

[0144] The non-transient software program and instructions required to implement the land area measurement method based on the image segmentation network of the above embodiment are stored in the memory 1010. When executed by the processor 1020, the land area measurement method based on the image segmentation network applied to the drone 1000 of the above embodiment is executed, for example, the above-described Figure 1 Steps S110 to S170 of the method, Figure 2 Steps S210 to S240 of the method, Figure 3 Steps S310 to S380 of the method, Figure 4 Steps S410 to S450 of the method, Figure 5 Steps S510 to S520 of the method, Figure 6 Method steps S610 to S630.

[0145] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0146] In addition, an embodiment of the present application further provides a computer-readable storage medium, which stores computer-executable instructions. The computer-executable instructions are executed by a processor or controller, for example, by a processor in the above-mentioned UAV 1000 embodiment, so that the above-mentioned processor can execute the land area measurement method based on image segmentation network applied to the UAV in the above-mentioned embodiment, for example, executing the above-mentioned Figure 1 Steps S110 to S170 of the method, Figure 2 Steps S210 to S240 of the method, Figure 3 Steps S310 to S380 of the method, Figure 4 Steps S410 to S450 of the method, Figure 5Steps S510 to S520 of the method, Figure 6 Method steps S610 to S630 in the method disclosed above. It will be appreciated by those skilled in the art that all or some of the steps and systems in the method disclosed above can be implemented as software, firmware, hardware, and appropriate combinations thereof. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or temporary media). As known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical disk storage, magnetic cassettes, magnetic tapes, magnetic disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and can be accessed by a computer. Furthermore, as is well known to those skilled in the art, communication media typically embodies computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transport mechanism, and may include any information delivery media.

Claims

1. A land area measurement method based on image segmentation network, applied to UAV, characterized by: The method comprises: determining a target land area, wherein the target land area includes a plurality of monitoring nodes; Determining a target inspection path according to the plurality of monitoring nodes, and photographing a land image at each of the monitoring nodes according to the target inspection path, wherein the monitoring node is a location node corresponding to an image center point of the land image; Acquiring target distance information, where the target distance information is the actual distance between a preset reference object and the monitoring node; Obtaining a target pixel number, where the target pixel number is the number of pixels between a reference point corresponding to the reference object in the land image and an image center point of the land image; Determine a target ground sampling distance value GSD according to the target distance information and the target pixel number; Inputting the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image; The land area of the target land region is obtained according to the target segmentation image and the target GSD.

2. The land area measurement method based on image segmentation network according to claim 1, characterized in that: The image segmentation network model is trained according to the following steps: Determining location information of the land image, and acquiring a map remote sensing image corresponding to the location information, wherein the land image and the map remote sensing image have the same image size; Acquire a first annotated image corresponding to the land image, and acquire a second annotated image corresponding to the map remote sensing image; Inputting the land image, the map remote sensing image, the first annotated image, and the second annotated image into a preset cross-modal alignment network model for image alignment to obtain a target training image; An initial image segmentation network model is obtained, and the initial image segmentation network model is trained according to the target training image to obtain the image segmentation network model.

3. The land area measurement method based on image segmentation network according to claim 2, characterized in that: The cross-modal alignment network model includes a first encoder, a second encoder, a third encoder, a fourth encoder, and a first decoder. Inputting the land image, the map remote sensing image, the first annotated image, and the second annotated image into a preset cross-modal alignment network model for image alignment processing to obtain a target training image includes: Inputting the first annotated image and the second annotated image into the first encoder for encoding to obtain a first feature vector; Inputting the map remote sensing image into the second encoder for encoding processing to obtain a second feature vector; Inputting the land image into the third encoder for encoding to obtain a third feature vector; Inputting the first eigenvector, the second eigenvector, and the third eigenvector into the fourth encoder for encoding respectively to obtain a new first eigenvector, a new second eigenvector, and a new third eigenvector; Performing feature alignment processing on the new first eigenvector and the new second eigenvector to obtain first intermediate data; Performing feature alignment processing on the new second eigenvector and the new third eigenvector to obtain second intermediate data; Inputting the first intermediate data and the second intermediate data into the first decoder for decoding respectively to obtain an aligned target annotation image and a target land image; An image merging process is performed on the target annotated image and the target land image to obtain the target training image.

4. The land area measurement method based on image segmentation network according to claim 1, characterized in that: The image segmentation network model includes a fifth encoder and a second decoder, the fifth encoder includes a feature extraction network and a first convolutional layer, the second decoder includes a second convolutional layer, a third convolutional layer, a first upsampling layer, and a second upsampling layer, and the land image is input into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmented image, including: Inputting the land image into the feature extraction network to perform feature extraction to obtain a fourth feature vector; Inputting the fourth eigenvector into the first convolutional layer for convolution operation to obtain a fifth eigenvector, and inputting the fifth eigenvector into the first upsampling layer for upsampling to obtain a sixth eigenvector; Inputting the fourth eigenvector into the second convolutional layer for convolution operation to obtain a seventh eigenvector; Performing feature splicing processing on the sixth eigenvector and the seventh eigenvector to obtain an eighth eigenvector; The eighth eigenvector is sequentially input into the third convolutional layer and the second upsampling layer to obtain the target segmented image.

5. The land area measurement method based on image segmentation network according to claim 1, characterized in that: The target segmentation image includes a plurality of land type patches, and obtaining the land area of the target land region according to the target segmentation image and the target GSD includes: Determining the number of pixels of each of the image patches; The product of the number of pixels of each of the map patches and the target GSD is determined as the land area of the land type corresponding to the map patch.

6. The land area measurement method based on image segmentation network according to claim 5, characterized in that: The method further comprises: Calculating the longitude and latitude information of each of the map patches according to a preset longitude and latitude conversion rule; Calculating the graphic size information of each of the image patches according to a preset edge detection algorithm; The latitude and longitude information and the graphic size information are saved in a preset database.

7. The land area measurement method based on image segmentation network according to claim 1, characterized in that: The image segmentation network model is trained according to the cross entropy loss function.

8. A drone, characterized in that: include: a target land area determination module, the target land area determination module being used to determine a target land area, the target land area including a plurality of monitoring nodes; a land image acquisition module, the land image acquisition module being configured to determine a target inspection path based on the plurality of monitoring nodes, and capture a land image at each monitoring node according to the target inspection path, wherein the monitoring node is a location node corresponding to an image center point of the land image; A target distance information acquisition module, wherein the target distance information acquisition module is used to acquire target distance information, wherein the target distance information is the actual distance between a preset reference object and the monitoring node; a target pixel number acquisition module, the target pixel number acquisition module being used to acquire the target pixel number, the target pixel number being the number of pixels from the reference point corresponding to the reference object in the land image to the image center point of the land image; A target GSD acquisition module is used to determine a target ground sampling distance value GSD according to the target distance information and the number of target pixels; A target segmentation image acquisition module is used to input the land image into a pre-trained image segmentation network model for image segmentation processing to obtain a target segmentation image; A land area acquisition module is used to obtain the land area of the target land region according to the target segmented image and the target GSD.

9. A drone comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the land area measurement method based on the image segmentation network as claimed in any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing computer-executable instructions, wherein the computer-executable instructions are used to execute the land area measurement method based on image segmentation network according to any one of claims 1 to 7.

Images