Method and device for evaluating the planting area of ​​staple food crops

Abstract

The present invention relates to the field of remote sensing image technology, and provides a method and device for evaluating the planting area of ​​staple crops. The method comprises: using a first target extraction network to extract targets from a first medium-resolution remote sensing image of a region to be investigated, and stratifying according to the extracted planting distribution information and cultivated land distribution information of the first staple crop, the terrain and planting structure of the region to be investigated; determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; according to the first sub-meter remote sensing image corresponding to the sample frame of each layer, using a second target extraction network to extract the area to obtain the first staple crop planting area in the sample frame corresponding to each layer, and determining the estimated value of the total planting area of ​​staple crops in the region to be investigated according to the first staple crop planting area in the sample frame. The present invention improves the accuracy of the evaluation of the planting area of ​​staple crops.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing image technology, and in particular to a method and device for evaluating the planting area of ​​staple crops. Background Art

[0002] With the popularization of remote sensing technology, in the existing technology, multiple medium-resolution scale images of the survey area can usually be spliced ​​to obtain the remote sensing image of the survey area, and crop classification can be carried out for the entire province based on the remote sensing image of the survey area. Then, the sampling frame is combined with the crop classification results of the remote sensing image to determine the planting area of ​​staple crops in the sampling frame. Furthermore, the planting area of ​​staple crops in the survey area can be evaluated by reverse calculation based on the planting area of ​​staple crops in the sampling frame.

[0003] However, due to its resolution limitations, medium-resolution images are not rich and detailed enough in terms of spectral details, which in turn affects the precision and accuracy of the crop classification process, resulting in low data accuracy on the planting area of ​​staple food crops in the sample frame and low accuracy in assessing the planting area of ​​staple food crops in the survey area. Summary of the invention

[0004] The present invention provides a method and device for evaluating the planting area of ​​staple crops, which are used to solve the defect of low evaluation accuracy of the planting area of ​​staple crops in the prior art and improve the evaluation accuracy of the planting area of ​​staple crops.

[0005] In a first aspect, the present invention provides a method for evaluating the planting area of ​​staple crops, the method comprising the following steps:

[0006] Using a first target extraction network, target extraction is performed on a first medium-resolution remote sensing image of the area to be investigated, so as to obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the area to be investigated, and the area to be investigated is stratified according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the area to be investigated, the terrain features and the planting structure of the area to be investigated, so as to obtain at least one layer of grids to be sampled;

[0007] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0008] According to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0009] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the first target extraction network is obtained by training an initial first extraction network through the following steps, including:

[0010] Determine a plurality of second medium-resolution remote sensing images of the area to be investigated within a first preset time period, and planting distribution information of second staple crops and second cultivated land distribution information corresponding to each of the areas to be investigated as a training data set;

[0011] Obtaining hyperparameters during training; the hyperparameters include at least one of a learning rate, a batch size, and a number of iterations;

[0012] Input each batch of training data into the initial first extraction network, forward propagate the extraction results, and calculate the loss function value corresponding to each batch;

[0013] According to the loss function values ​​corresponding to each batch, the model parameters of the initial first extraction network are updated to obtain updated model parameters;

[0014] Based on the updated model parameters, the first object extraction network is determined.

[0015] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the area to be investigated is layered according to the planting distribution information of the first staple crop and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated, and at least one layer of grids to be sampled is obtained, including:

[0016] Based on the terrain characteristics of the area to be investigated, the area to be investigated is divided into a layer with relatively flat terrain and a layer with relatively large terrain fluctuations using the standard deviation of the digital elevation model, and the layer with relatively flat terrain and the layer with relatively large terrain fluctuations are determined as the first-stage stratification result;

[0017] Determine the second-stage stratification result according to the first-stage stratification result, the planting distribution information of the first staple food crop corresponding to the area to be investigated, the first cultivated land distribution information and the planting structure;

[0018] Determining a target stratification result based on the first-stage stratification result and the second-stage stratification result;

[0019] According to the target stratification results, the grids to be sampled in each layer are constructed.

[0020] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the method of determining the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer comprises:

[0021] Determine the sample frame for each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer;

[0022] Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

[0023] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, determining an estimated value of the total planting area of ​​staple crops in the area to be investigated according to the planting area of ​​the first staple crop in the sample frame corresponding to each layer, comprises:

[0024] For any of the layers, according to the planting area of ​​the first staple food crop in the sample frame corresponding to the layer, the number of grids in the sample frame corresponding to the layer, and the number of grids of the to-be-sampled grids corresponding to the layer, determine the planting area of ​​the first staple food crop corresponding to the layer;

[0025] According to the planting area of ​​the first staple food crops corresponding to each layer, an estimated value of the total planting area of ​​the staple food crops in the area to be investigated is calculated.

[0026] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, before determining the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the method further includes:

[0027] Determining the target size of the grid to be sampled according to a preset correlation rule;

[0028] Among them, the preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of each grid to be sampled, the better the effect of the division size of each grid to be sampled, and the spatial data is the planting area of ​​staple food crops within the second preset time period; the target size is used to characterize that there is no spatial autocorrelation between each grid to be sampled under the target size.

[0029] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, before determining the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the method further includes:

[0030] Determine the key areas of concern within the area to be investigated;

[0031] The determining, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the first sub-meter remote sensing image corresponding to the sample frame of each layer comprises:

[0032] According to the key focus area and the grids to be sampled of the layer, determining the grids to be sampled in the grids to be sampled of the layer that cover the key focus area;

[0033] Determine the sample frame corresponding to each layer based on the target size of the grid to be sampled, the target ratio within the grid to be sampled of each layer, and the grid to be sampled covering the key focus area;

[0034] Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

[0035] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the method further comprises:

[0036] Based on the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area, a second estimated value of the total planting area of ​​staple crops in the key focus area is determined.

[0037] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, before using a first target extraction network to extract a target from a first medium-resolution remote sensing image of a region to be investigated and obtaining the planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the region to be investigated, the method further includes:

[0038] Acquire at least one second medium resolution remote sensing image;

[0039] Selecting the third medium-resolution remote sensing images that meet the preset time phase condition and cloud coverage rate condition from among the second medium-resolution remote sensing images;

[0040] The third medium-resolution remote sensing image is subjected to image preprocessing to obtain a first medium-resolution remote sensing image of the area to be investigated.

[0041] In a second aspect, the present invention further provides a device for evaluating the planting area of ​​staple crops, the device comprising the following modules:

[0042] a stratification module, configured to use a first target extraction network to perform target extraction on a first medium-resolution remote sensing image of a region to be investigated, obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the region to be investigated, and stratify the region to be investigated according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the region to be investigated, the terrain features of the region to be investigated, and the planting structure, to obtain at least one layer of grids to be sampled;

[0043] An evaluation module is used to determine the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer; the target ratio is determined based on the proportion of the staple food crop planting area within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0044] According to the first sub-meter remote sensing image corresponding to each of the layers, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each of the layers, and based on the first staple food crop planting area in the sample frame corresponding to each of the layers, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0045] In a third aspect, the present invention further provides an electronic device 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 method for evaluating the planting area of ​​staple food crops as described above is implemented.

[0046] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements a method for evaluating the planting area of ​​staple food crops as described in any one of the above.

[0047] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned methods for evaluating the planting area of ​​staple food crops.

[0048] The present invention provides a method and device for evaluating the planting area of ​​staple crops. First, a first target extraction network is used to perform target extraction on a first medium-resolution remote sensing image of a survey area to obtain planting distribution information and first cultivated land distribution information of a first staple crop corresponding to the survey area, and the survey area is layered according to the planting distribution information and first cultivated land distribution information of the first staple crop corresponding to the survey area, the terrain characteristics and the planting structure of the survey area to obtain at least one layer of grids to be sampled; then, based on the target size of the grids to be sampled and the target ratio within the grids to be sampled of each layer, the first sub-meter remote sensing image corresponding to the sampled frame of each layer is determined, wherein the target ratio is determined based on the proportion of the planting area of ​​staple crops in the grids to be sampled to the internal cultivated land area of ​​the grids to be sampled; further, according to the first sub-meter remote sensing image corresponding to the sampled frame of each layer, a second target extraction network is used to perform area extraction to obtain the planting area of ​​the first staple crop in the sampled frame corresponding to each layer, and according to the planting area of ​​the first staple crop in the sampled frame corresponding to each layer, an estimated value of the total planting area of ​​staple crops in the survey area is determined.

[0049] In the present invention, when the region to be investigated is stratified, that is, when the resolution requirements for remote sensing images are relatively low, a first medium-resolution remote sensing image of the region to be investigated is obtained, and a first target extraction network is used to extract the target of the first medium-resolution remote sensing image of the region to be investigated, and the planting distribution information and the first cultivated land distribution information of the first staple food crop corresponding to the region to be investigated are obtained. Then, the region to be investigated is stratified in combination with the terrain characteristics and planting structure of the region to be investigated, and the efficiency of stratification is high and the reliability of stratification results is high; further, when operations such as crop classification and area extraction that require relatively high resolution of remote sensing images are performed, the area is extracted according to the first sub-meter remote sensing image corresponding to the sample frame of each layer, and the second target extraction network is used to extract the area to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and then the estimated value of the total planting area of ​​staple food crops in the region to be investigated is obtained by reverse deduction. In the present invention, two remote sensing images of different scales are scientifically used, and combined with the traditional geostatistical sampling method, the total planting area of ​​staple food crops in the region to be investigated is finally evaluated, and the accuracy of the evaluation of the planting area of ​​staple food crops is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0051] Figure 1This is one of the flow charts of the method for evaluating the planting area of ​​staple food crops provided by the present invention.

[0052] Figure 2 This is the second flow chart of the method for evaluating the planting area of ​​staple food crops provided by the present invention.

[0053] Figure 3 It is a structural schematic diagram of the staple food crop planting area evaluation device provided by the present invention.

[0054] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first" and "second" are generally a class, and the number of objects is not limited. For example, the first node can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally represents that the front and back associated objects are in an "or" relationship.

[0057] Combine the following Figure 1-Figure 4 The present invention describes a method and device for evaluating the planting area of ​​staple food crops.

[0058] Figure 1 This is one of the flow diagrams of the method for calculating the planting area of ​​staple crops provided by the present invention, such as Figure 1 As shown, the method includes the following:

[0059] Step 101: using a first target extraction network, extracting targets from a first medium-resolution remote sensing image of a region to be investigated, obtaining planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the region to be investigated, and stratifying the region to be investigated according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the region to be investigated, the terrain features of the region to be investigated, and the planting structure, to obtain at least one layer of grids to be sampled;

[0060] Specifically, it should be noted that the execution subject of this embodiment is an electronic device, which is used to evaluate the total planting area of ​​staple crops in the area to be investigated, thereby improving the accuracy of the evaluation of the planting area of ​​staple crops.

[0061] The evaluation process in this embodiment can be implemented by the following steps:

[0062] Firstly, the first target extraction network is used to extract targets from the first medium-resolution remote sensing image of the area to be investigated, and the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated are obtained.

[0063] Among them, staple crops are, for example, corn, rice, wheat, cotton, and the like. The main difference between medium-resolution remote sensing images and sub-meter remote sensing images is their spatial resolution. The spatial resolution of medium-resolution remote sensing images is usually between 10 meters and 100 meters. This type of image has a moderate spatial resolution, a high revisit period, and a large width, which can achieve large-scale earth observation and is the core remote sensing data source for obtaining information on the earth's surface. An example of a medium-resolution image is the Landsat series of satellites, which have a spatial resolution of about 30 meters. Among them, the first medium-resolution remote sensing image of the area to be investigated is obtained, for example, from the following public websites: National Remote Sensing Data and Application Service Platform, Geospatial Data Cloud, China Remote Sensing Data Sharing Network, etc. The spatial resolution of sub-meter remote sensing images is below 1 meter, and can even reach the centimeter level. These images can provide a clear view of surface features, greatly improving the accuracy of tasks such as urban planning, land use surveys, environmental monitoring, and disaster assessment. An example of sub-meter remote sensing imagery is China's Gaofen-2 satellite, which has a spatial resolution of 0.8 meters for panchromatic images and 3.2 meters for multispectral images.

[0064] Among them, the area to be investigated can be determined according to actual assessment needs, such as Province A, Province B, Province C, etc.

[0065] After obtaining the medium-resolution remote sensing image of the area to be investigated, the first target extraction network can be used to perform target recognition, and the recognition result is, for example, the planting distribution information of the first staple food crop and the first cultivated land distribution information. The first cultivated land distribution information is, for example, the area where the cultivated land is located in the area to be investigated, and the planting distribution information of the staple food crop is, for example, the planting area where the staple food crop is located in the area to be investigated, and the first cultivated land distribution information includes the planting distribution information of the first staple food crop.

[0066] The first target extraction network is, for example, a feature extraction network such as VGG16, or a target detection network based on deep learning such as YOLO, SSD, Faster R-CNN, etc. It can be understood that the first target extraction network is a model with the best performance obtained by training using historical data and labels corresponding to the historical data, so that the accuracy of target extraction using the first target extraction network is higher.

[0067] Furthermore, after identifying the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the area to be investigated can be further stratified in combination with the terrain characteristics and planting structure of the area to be investigated to obtain at least one layer of grids to be sampled. The purpose of regional stratification is to subsequently extract representative samples from each layer, so that the final sample is more representative and the area assessment result of the area to be investigated is more accurate.

[0068] Step 102: determining the first sub-meter remote sensing image corresponding to the selected sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0069] Specifically, after stratifying the area to be investigated and dividing the grids based on the target size of the grids to be sampled, the next step can be to perform sampling in combination with the principles of geostatistics. For example, the probability proportional scale sampling method is used to sample each layer. The target ratio within the grids to be sampled in each layer is, for example, the ratio of the planting area of ​​staple food crops in the grid to the cultivated land area in the grid. The target ratio is used as auxiliary information for sample sampling. In this way, both the planting area information of staple food crops and the distribution of cultivated land are taken into account, the sampling reliability is higher, and the representativeness of the samples drawn is higher.

[0070] Furthermore, after obtaining the selected samples of each layer, higher-resolution remote sensing images can be used for crop classification and area extraction in the subsequent process of area extraction of the selected sample frames of each layer. For example, the first sub-meter remote sensing image corresponding to the selected sample frames of each layer is obtained. The extraction effect is better based on determining the first sub-meter remote sensing image corresponding to the selected sample frames of each layer.

[0071] Step 103: Based on the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and based on the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0072] Specifically, after obtaining the first sub-meter remote sensing image corresponding to the sample frame covering each layer, the process of extracting the staple food crop area using the first sub-meter remote sensing image can be implemented based on the following method:

[0073] The second target extraction network is used to extract the area to obtain the planting area of ​​the first staple food crop in the sample frame corresponding to each layer. The second target extraction network uses a plurality of second sub-meter remote sensing images of the area to be investigated within a third preset time period and the planting area of ​​the second staple food crop corresponding to each of the second sub-meter remote sensing images.

[0074] Furthermore, according to the planting area of ​​the first staple food crop in the sample frame corresponding to each layer, the estimated value of the total planting area of ​​staple food crops in the area to be investigated is determined. For example, the total planting area of ​​staple food crops in the investigated area is inferred.

[0075] The method provided in this embodiment firstly uses a first target extraction network to perform target extraction on a first medium-resolution remote sensing image of the area to be investigated, obtains the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, and stratifies the area to be investigated according to the planting distribution information and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated, and obtains at least one layer of grids to be sampled; then, based on the target size of the grids to be sampled and the target ratio within the grids to be sampled of each layer, determines the first sub-meter remote sensing image corresponding to the sampled frame of each layer, wherein the target ratio is determined based on the planting area of ​​the staple food crop in the grid to be sampled to the internal cultivated land area of ​​the grid to be sampled; further, based on the first sub-meter remote sensing image corresponding to the sampled frame of each layer, uses a second target extraction network to perform area extraction to obtain the planting area of ​​the first staple food crop in the sampled frame corresponding to each layer, and determines the estimated value of the total planting area of ​​the staple food crop in the area to be investigated according to the planting area of ​​the first staple food crop in the sampled frame corresponding to each layer.

[0076] In the present invention, when the region to be investigated is stratified, that is, when the resolution requirements for remote sensing images are relatively low, a first medium-resolution remote sensing image of the region to be investigated is obtained, and a first target extraction network is used to extract the target of the first medium-resolution remote sensing image of the region to be investigated, and the planting distribution information and the first cultivated land distribution information of the first staple food crop corresponding to the region to be investigated are obtained. Then, the region to be investigated is stratified in combination with the terrain characteristics and planting structure of the region to be investigated, and the efficiency of stratification is high and the reliability of stratification results is high; further, when operations such as crop classification and area extraction that require relatively high resolution of remote sensing images are performed, the area is extracted according to the first sub-meter remote sensing image corresponding to the sample frame of each layer, and the second target extraction network is used to extract the area to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and then the estimated value of the total planting area of ​​staple food crops in the region to be investigated is obtained by reverse deduction. In the present invention, two remote sensing images of different scales are scientifically used, and combined with the traditional geostatistical sampling method, the total planting area of ​​staple food crops in the region to be investigated is finally evaluated, and the accuracy of the evaluation of the planting area of ​​staple food crops is improved.

[0077] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the first target extraction network is obtained by training an initial first extraction network through the following steps, including:

[0078] Determine a plurality of second medium-resolution remote sensing images of the area to be investigated within the first preset time period, and the planting distribution information of the second staple food crop and the second cultivated land distribution information corresponding to each area to be investigated as a training data set;

[0079] Obtain hyperparameters during training; hyperparameters include at least one of the learning rate, batch size, and number of iterations;

[0080] Input each batch of training data into the initial first extraction network, forward propagate the extraction results, and calculate the loss function value corresponding to each batch;

[0081] According to the loss function values ​​corresponding to each batch, the model parameters of the initial first extraction network are updated to obtain updated model parameters;

[0082] Based on the updated model parameters, a first target extraction network is determined.

[0083] Specifically, in some embodiments, the first target extraction network is obtained after training. Exemplarily, the process of training the initial first extraction network to obtain the first target extraction network is as follows:

[0084] First, multiple second medium-resolution remote sensing images of the area to be investigated within a first preset time period, and the planting distribution information of the second staple food crop and the second cultivated land distribution information corresponding to each area to be investigated are determined as training data. The first preset time period usually takes the remote sensing image data within a historical time period, and the labels corresponding to the remote sensing image data within the historical time period, and the labels represent the planting distribution information of the second staple food crop and the second cultivated land distribution information corresponding to each area to be investigated.

[0085] Furthermore, hyperparameters in the training process are obtained, wherein the hyperparameters include at least one of a learning rate, a batch size, and a number of iterations.

[0086] Then, the training data of each batch is input into the initial first extraction network, the extraction results are forward propagated, and the loss function values ​​corresponding to each batch are calculated; according to the loss function values ​​corresponding to each batch, the model parameters of the initial first extraction network are updated to obtain the updated model parameters. For example, the model parameters with the smallest final loss function are determined as the updated model parameters.

[0087] Then, based on the updated model parameters, the first target extraction network is determined. The updated model parameters are determined as the model parameters of the first target extraction network to implement model training. Finally, the accuracy of crop classification and extraction based on the first target extraction network is high.

[0088] In the method provided in this embodiment, the first target extraction network is obtained after training, and the accuracy of crop classification and extraction based on the first target extraction network is finally high.

[0089] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the area to be investigated is stratified according to the planting distribution information of the first staple crop and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated, and at least one layer of grids to be sampled is obtained, including:

[0090] Based on the terrain characteristics of the area to be investigated, the area to be investigated is divided into a layer with relatively flat terrain and a layer with large terrain fluctuations using the standard deviation of the digital elevation model. The layer with relatively flat terrain and the layer with large terrain fluctuations are determined as the first-stage stratification results.

[0091] Determine the second-stage stratification results based on the first-stage stratification results, the planting distribution information of the first staple food crop corresponding to the area to be investigated, the first cultivated land distribution information and the planting structure;

[0092] Determine the target stratification results based on the first-stage stratification results and the second-stage stratification results;

[0093] According to the target stratification results, the grids to be sampled in each layer are constructed.

[0094] Specifically, in some embodiments, the implementation process of stratifying the area to be investigated in step 101 according to the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated includes the following steps:

[0095] First, based on the terrain characteristics of the area to be investigated, the area to be investigated is divided into a layer with relatively flat terrain and a layer with large terrain undulations using the standard deviation of the digital elevation model (DEM), and the layer with relatively flat terrain and the layer with large terrain undulations are determined as the first-stage stratification results. For example, based on the DEM data, the area to be investigated is divided into a flat terrain layer with a standard deviation below a certain threshold and a complex terrain layer with a standard deviation above the threshold.

[0096] Furthermore, the second-stage stratification results are determined based on the first-stage stratification results, the planting distribution information of the first staple food crop corresponding to the area to be investigated, the first cultivated land distribution information, and the planting structure. For example, on the basis of the first-stage terrain stratification, the second-stage stratification is performed according to the crop planting scale, planting structure, planting fragmentation, and planting structure and fragmentation indicators to improve the homogeneity within the layer. Planting scale: The remote sensing classification results of the target crops in the statistical sampling frame are used as indicator parameters. Planting structure: Since some crops of the same period are difficult to distinguish from remote sensing images, the planting structure indicator is introduced to reflect the differences in regional planting structures. Planting fragmentation: The planting fragmentation indicator is introduced to quantitatively analyze the heterogeneity of plant spatial distribution caused by mixed pixels. Comprehensive indicator of planting structure and fragmentation: Comprehensively consider the planting structure and planting fragmentation to more objectively reflect the regional differences in crop planting within the provincial scope.

[0097] Further, based on the first-stage stratification results and the second-stage stratification results, the target stratification results are determined. Further, according to the target stratification results, the grids to be sampled in each layer are constructed to achieve regional stratification and grid division. For example, according to the target stratification results, the grids to be sampled are constructed, and the areas in each layer can be further subdivided into smaller grid units to facilitate sampling surveys. Geographical Information System (GIS) technology can be used to draw the grids to be sampled in the GIS according to the stratification results.

[0098] The method provided in this embodiment uses medium-resolution remote sensing images to obtain sampling auxiliary information such as the approximate distribution of staple crops and cultivated land in the survey area, and based on the planting distribution information of the first staple crop and the first cultivated land distribution information corresponding to the survey area, the terrain characteristics of the survey area, and the planting structure, the survey area is stratified and gridded, and the stratification results are more objective. The present invention adopts stratification based on topography and planting habits, and after stratification, a probability proportional scale sampling method is used to ensure that the probability of each sampling unit being selected is proportional to its size in the population (such as the planting area). In this way, large-scale units are more likely to be selected, thereby improving the representativeness of the sample, so that the sample can more accurately reflect the characteristics of the population, and can ensure that resources (such as investigators, time, funds, etc.) are reasonably allocated.

[0099] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, the first sub-meter remote sensing image corresponding to the sample frame of each layer is determined, including:

[0100] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, the sample frame of each layer is determined;

[0101] Based on the sample frames corresponding to each layer, the first sub-meter remote sensing image corresponding to the sample frames completely covering each layer is determined.

[0102] Specifically, in some embodiments, the specific implementation process of determining the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer in step 102 includes the following steps:

[0103] Firstly, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, the sample frame of each layer is determined.

[0104] The target size of the grid to be sampled represents the grid division method in each layer, and the division of the target size can affect the final area assessment result to a certain extent. The target ratio in the grid to be sampled in each layer is, for example, the ratio of the planting area of ​​staple food crops in the grid to be sampled to the internal cultivated land area in the grid to be sampled.

[0105] In this embodiment, the probability proportional scale sampling method can be used to sample each layer. When sampling, the ratio of the planting area of ​​staple crops in the grid to the cultivated land area in the grid (target ratio) is used as auxiliary information. In this way, both the planting area of ​​staple crops and the distribution of cultivated land are taken into account. Therefore, when determining the total sample size When the total cost is given ( ) is used to determine the sample size.

[0106]

[0107] in, It is the total cost. It is the constant expenditure required to carry out the work. It is The sampling cost of each sample is Indicates the sample size to be drawn from each stratum.

[0108]

[0109]

[0110] in, is the intermediate parameter, For the The overall layer weight of the layer, It is Layer standard deviation.

[0111]

[0112] After determining the total sample size, the optimal allocation method under given costs, namely Neyman allocation, is used to determine the sample size to be drawn from each layer. .

[0113]

[0114]

[0115]

[0116] in, For the The number of grids divided by the layer, Indicates the sample size to be drawn from each stratum.

[0117] Furthermore, after the selected sample frames corresponding to each layer are determined, the first sub-meter remote sensing image corresponding to the selected sample frames that completely cover each layer can be determined based on the selected sample frames corresponding to each layer. For example, the first sub-meter remote sensing image that can cover the selected samples is obtained, and according to the distribution of objects in the image sampling frame and the classification needs, field surveys are carried out to collect ground object samples, and ground object classification work is carried out for all the selected sample frames in each layer to determine the planting area of ​​staple food crops in the sample frame.

[0118] The method provided in this embodiment uses a probability proportional scale sampling method to sample each layer. When sampling, the ratio of the planting area of ​​staple food crops in the grid to the cultivated land area in the grid (target ratio) is used as auxiliary information. In this way, both the planting area information of staple food crops and the distribution of cultivated land are taken into account. The results of regional stratification and sampling are more objective, the representativeness of the selected samples is higher, and the subsequent process of inferring the planting area of ​​staple food crops in the area to be investigated based on the selected samples is more accurate.

[0119] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, an estimated value of the total planting area of ​​staple crops in the area to be investigated is determined according to the planting area of ​​the first staple crop in the sample frame corresponding to each layer, including:

[0120] For any layer, the planting area of ​​the first staple food crop corresponding to the layer is determined according to the planting area of ​​the first staple food crop in the sample frame corresponding to the layer, the number of grids in the sample frame corresponding to the layer, and the number of grids to be sampled corresponding to the layer;

[0121] According to the planting area of ​​the first staple food crop corresponding to each layer, the estimated value of the total planting area of ​​staple food crops in the area to be investigated is calculated.

[0122] Specifically, in some embodiments, the specific implementation process of determining the estimated value of the total planting area of ​​staple crops in the area to be investigated according to the planting area of ​​the first staple crop in the sample frame corresponding to each layer in step 103 includes the following steps:

[0123] First, for any layer, according to the planting area of ​​the first staple food crop in the sample frame corresponding to the layer, the number of grids in the sample frame corresponding to the layer And the number of grids to be sampled corresponding to the layer , determine the planting area of ​​the first staple food crop corresponding to the layer.

[0124] For example, the process of inferring the total area of ​​staple food crops planted in the survey area is as follows:

[0125] Estimated total area of ​​staple food crops in tier j The calculation method is as follows:

[0126]

[0127] in, represents the estimated total area of ​​staple food crops in the jth level, It is The number of grid layers, It is The number of grids in the sample frame corresponding to the layer, It is the planting area of ​​the first staple food crop in the sample frame obtained after area extraction, that is, Tier Cotton planting area in each grid.

[0128] Furthermore, the total planting area of ​​staple crops in the area to be investigated can be estimated based on the first staple crop planting area corresponding to each layer. For example, the planting area of ​​staple crops in the entire investigation area can be obtained by adding up the reversed data of each layer.

[0129] The method provided in this embodiment determines the planting area of ​​the first staple crop corresponding to each layer according to the planting area of ​​the first staple crop in the sample frame corresponding to each layer, the number of grids in the sample frame corresponding to each layer, and the number of grids to be sampled corresponding to each layer. Then, according to the planting area of ​​the first staple crop corresponding to each layer, the estimated value of the total planting area of ​​staple crops in the area to be investigated is calculated, so as to realize the evaluation of the total planting area of ​​staple crops in the area to be investigated. Because the planting area of ​​the first staple crop in the sample frame corresponding to each layer is obtained by the second target extraction network based on sub-meter remote sensing image extraction, the recognition accuracy is high, and then the accuracy of the evaluation of the total planting area of ​​staple crops in the area to be investigated is also high.

[0130] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, before determining the first sub-meter remote sensing image corresponding to the sample frame of each layer, the method further includes:

[0131] According to the preset correlation rules, the target size of the grid to be sampled is determined;

[0132] Among them, the preset correlation rule is used to characterize the similarity and aggregation degree between the spatial data of each grid to be sampled, and the spatial data is the planting area of ​​staple crops within the second preset time period; the target size is used to characterize that there is no spatial autocorrelation between the grids to be sampled under the target size.

[0133] Specifically, in some embodiments, before determining the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, it also includes determining the target size of the grid to be sampled, and then, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer.

[0134] The process of determining the target size of the grid to be sampled can be achieved through the following steps:

[0135] According to the preset correlation rule, the target size of the grid to be sampled is determined. The target size is used to characterize that there is no spatial autocorrelation between the grids to be sampled under the target size. The preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of the grids to be sampled, the better the effect of the division size of each grid to be sampled.

[0136] Among them, the preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of each grid to be sampled, the better the effect of the division size of each grid to be sampled, that is, the higher the independence between the grids to be sampled (i.e., samples), the higher the representativeness of the grids to be sampled. Exemplarily, Moran's I can be introduced to represent the similarity and aggregation degree between the spatial data of each grid to be sampled (i.e., the autocorrelation of the grid to be sampled), wherein the spatial data can select the planting area of ​​staple food crops in a second preset time period (e.g., a historical time period, within 1 year, 2 years, or 3 years). The calculation process of Moran's I is as follows:

[0137]

[0138] in, It is used to characterize the Moran index between grid i and grid j. Represents the grid and Grid The spatial weight relationship between them. and Grid and Grid The observed value is the planting area of ​​staple food crops in the grid extracted using medium-resolution remote sensing images. It is the average of all observations, that is, the average staple food crop planting area of ​​the entire grid extracted using medium-resolution remote sensing images. Moran's I ranges from -1 to 1, where values ​​close to 1 indicate that similar values ​​in the grid to be sampled are clustered together, which is positive correlation, values ​​close to -1 indicate that values ​​of different spatial units are clustered together, which is negative correlation, and values ​​close to 0 indicate that there is no spatial autocorrelation. However, in actual operations, factors such as the staple food planting structure in the survey area and the labor cost of ground survey sample collection must be considered to adjust the grid size, and the Moran index cannot be used as an absolute reference.

[0139] Here, when Moran's I is equal to or close to 0, it means that the spatial correlation between sampling units is equal to or close to 0. Whether it is a positive number or a negative number, it means that the samples have a high spatial correlation, so as long as it is not close to or equal to 0, it is fine.

[0140] In the method provided in this embodiment, the target size is used to characterize that there is no spatial autocorrelation between the grids to be sampled at the target size, and the target size of the grids to be sampled is determined according to a preset correlation rule, wherein the preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of the grids to be sampled, the better the effect of the division size of each grid to be sampled. The grids to be sampled and the stratification effect divided based on this embodiment are more reliable.

[0141] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, before determining the first sub-meter remote sensing image corresponding to the sample frame of each layer, the method further includes:

[0142] Identify key areas of concern within the area to be investigated;

[0143] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled in each layer, the first sub-meter remote sensing image corresponding to the sample frame of each layer is determined, including:

[0144] According to the key focus area and the grids to be sampled in the layer, determine the grids to be sampled in the layer that cover the key focus area;

[0145] Based on the target size of the grid to be sampled, the target ratio within the grid to be sampled in each layer, and the grid to be sampled covering the key focus area, the sample frame corresponding to each layer is determined;

[0146] Based on the sample frames corresponding to each layer, the first sub-meter remote sensing image corresponding to the sample frames completely covering each layer is determined.

[0147] Specifically, in some embodiments, there are usually some areas that need to be focused on when evaluating the staple food crops in the area to be investigated, for example, the counties and districts that are the focus of the year in the survey area. Based on this, this embodiment can adjust the samples based on the key focus areas, so that the subsequent evaluation of the staple food crop planting area in the key focus areas can be based on the evaluation results of the staple food crop planting area in the overall area to be investigated, which can reduce the repetitive workload to a certain extent.

[0148] For example, the subsequent process of determining the high-resolution remote sensing image (the first sub-meter remote sensing image) of the sample can take the factors of the key focus area into consideration, which can be achieved through the following steps:

[0149] First, determine the key focus areas within the area to be investigated. For example, County B in Province A is the key focus district and county (key focus area); for example, County C and District D in Province A are the key focus districts and counties (key focus areas).

[0150] Then, according to the key focus area and the grids to be sampled in the layer, the grids to be sampled in the layer that cover the key focus area are determined, that is, the key focus area is mapped to the grids to be sampled in each layer.

[0151] Furthermore, based on the target size of the grid to be sampled, the target ratio within the grid to be sampled in each layer, and the grid to be sampled covering the key focus area, the corresponding sample frame of each layer is determined. For example, the key focus counties are taken as a sub-population of the entire survey area, and the number of additional samples required for the sub-population is determined based on the traditional sampling survey theory and method, and the probability proportional size sampling method is used to extract the sample frame.

[0152] Furthermore, the first sub-meter remote sensing image corresponding to the sample frames corresponding to each layer can be determined based on the sample frames corresponding to each layer. In particular, the first sub-meter remote sensing image needs to completely cover the sample frames corresponding to each layer (including the grids corresponding to the key focus areas).

[0153] The method provided in this embodiment determines the key focus area within the area to be investigated, and determines the grids to be sampled in the layer that cover the key focus area according to the key focus area and the grids to be sampled in the layer; determines the sample frames corresponding to each layer based on the target size of the grids to be sampled, the target ratios within the grids to be sampled in each layer, and the grids to be sampled covering the key focus area; determines the first sub-meter remote sensing image corresponding to the sample frames that completely cover each layer based on the sample frames corresponding to each layer, and subsequently evaluates the area of ​​staple food crops based on the first sub-meter remote sensing image, and the evaluation has a high accuracy.

[0154] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, the method further comprises:

[0155] Based on the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area, a second estimated value of the total planting area of ​​staple crops in the key focus area is determined.

[0156] Specifically, in some embodiments, after the key focus area is determined, the staple food crop planting area in the key focus area can be evaluated based on the evaluation results of the staple food crop planting area in the overall area to be investigated, which can reduce repeated workload and improve evaluation efficiency.

[0157] For example, the process of assessing the area of ​​staple food crops in the focus area includes the following steps:

[0158] The second estimated value of the total planting area of ​​staple crops in the key focus area is determined by the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area.

[0159] Among them, since the key focus areas (such as key focus counties) have added samples other than those based on overall sampling, it is necessary to make weighted estimates of the existing and subsequent added samples to estimate the total area of ​​staple food crops planted in the key focus counties. for:

[0160]

[0161] in,

[0162] in, It is an estimate of the total area of ​​staple food crops in key counties and districts calculated without adding samples. is the total estimate based on the samples drawn from the key counties using the probability proportional to size sampling method (i.e. the estimate of the total area of ​​staple food crops in the key counties calculated based on the additional samples), which are calculated as follows: and The estimator of the variance is and , is the preset weight parameter.

[0163] The method provided in this embodiment is based on the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area, and determines the second estimated value of the total planting area of ​​staple crops in the key focus area. This method makes full use of the advantage that when the size or structure of the population changes, the counties that are the focus of the year are introduced, and the probability proportional scale sampling method can relatively easily adjust the sampling strategy to adapt to the new situation. Samples are added for the counties that are the focus of the year without affecting the overall estimation, and the planting area of ​​staple crops in the counties that are the focus of the year can be accurately obtained. This embodiment constructs a reverse inference model based on existing and subsequent added samples for weighted estimation to estimate the total planting area of ​​staple crops in the counties that are the focus of the year, rather than directly using the original sampling system for estimation, so that the reverse inference results of the counties that are the focus of the year are more accurate.

[0164] According to a method for evaluating the planting area of ​​staple crops provided by the present invention, a first target extraction network is used to extract targets from a first medium-resolution remote sensing image of a survey area, and before obtaining the planting distribution information of the first staple crop and the first cultivated land distribution information corresponding to the survey area, the method further includes:

[0165] Acquire at least one second medium resolution remote sensing image;

[0166] Selecting third medium-resolution remote sensing images that meet preset time phase conditions and cloud coverage conditions from among the second medium-resolution remote sensing images;

[0167] The third medium-resolution remote sensing image is subjected to image preprocessing to obtain a first medium-resolution remote sensing image of the area to be investigated.

[0168] Specifically, in some embodiments, the first medium-resolution remote sensing image is a processed medium-resolution remote sensing image, that is, before using the first target extraction network to extract the target from the first medium-resolution remote sensing image of the area to be investigated, the original medium-resolution remote sensing image may be preprocessed to obtain the first medium-resolution remote sensing image of the area to be investigated. An example of the process of preprocessing the original medium-resolution remote sensing image is as follows:

[0169] First, obtain at least one second medium-resolution remote sensing image. The second medium-resolution remote sensing image is also the original medium-resolution remote sensing image. There are many ways to obtain the original medium-resolution remote sensing image, such as from the following public websites: National Remote Sensing Data and Application Service Platform, Geospatial Data Cloud, China Remote Sensing Data Sharing Network, etc.

[0170] Furthermore, the third medium-resolution remote sensing images that meet the preset time phase conditions and cloud coverage conditions can be screened out from the second medium-resolution remote sensing images, wherein the preset time condition setting principle can be to make the remote sensing images of staple food crops and other food crops easy to distinguish, and the cloud coverage condition setting principle can be that clouds cannot cover the area where the food crops are located.

[0171] Furthermore, after the third medium-resolution remote sensing image that meets the requirements is screened, in order for the first target extraction network to better understand the third medium-resolution remote sensing image, the third medium-resolution remote sensing image can be further image preprocessed to obtain the first medium-resolution remote sensing image of the area to be investigated. The preprocessing operation is, for example, calibration of coordinate information, orthorectification, radiation correction, and the like.

[0172] Furthermore, the sample size and quality of the first medium-resolution remote sensing image obtained through the above operation are more convenient for the first target extraction network to understand, and to perform more accurate target identification, so as to facilitate more accurate regional stratification and grid division in the subsequent process.

[0173] The method provided in this embodiment can preprocess the original medium-resolution remote sensing image before using the first target extraction network for target recognition. The sample size and quality of the first medium-resolution remote sensing image of the area to be investigated obtained after the preprocessing are more convenient for the first target extraction network to understand, so that more accurate target recognition can be performed, which is convenient for subsequent more accurate regional stratification and grid division, thereby improving the division efficiency.

[0174] Figure 2 This is the second flow chart of the method for evaluating the planting area of ​​staple crops provided by the present invention, such as Figure 2 As shown, the method includes:

[0175] First, the medium-resolution remote sensing images in the survey area are screened, and after screening, medium-resolution images that meet the requirements can be obtained.

[0176] Furthermore, the medium-resolution images that meet the requirements are preprocessed.

[0177] Furthermore, the distribution of staple food planting and cultivated land is extracted.

[0178] Furthermore, the Moran index and planting habits were introduced to divide the grid and stratify the survey area.

[0179] Furthermore, probability proportional to size sampling determines the sampling box.

[0180] Furthermore, we will determine the key districts and counties for the year; increase the sample boxes; screen sub-meter remote sensing images; and determine the ground survey samples;

[0181] Furthermore, the planting area of ​​staple food crops within the sample frame is extracted based on sub-meter remote sensing images.

[0182] Furthermore, the staple food crop area of ​​the sample box is inferred based on the sample box data.

[0183] Furthermore, the data on the staple food planting area in the survey area can be inferred; the data on the staple food planting area in the key counties and districts for the year can also be determined based on the staple food crop area in the sample box.

[0184] The following is a description of the staple crop planting area evaluation device provided by the present invention. The staple crop planting area evaluation device described below and the staple crop planting area evaluation method described above can correspond to each other.

[0185] Figure 3 : is a schematic diagram of the structure of the staple food crop planting area evaluation device provided by the present invention, such as Figure 3 As shown, the staple food crop planting area evaluation device 300 includes a stratification module 310 and an evaluation module 320; wherein,

[0186] The stratification module 310 is used to perform target extraction on the first medium-resolution remote sensing image of the area to be investigated using the first target extraction network, obtain the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, and stratify the area to be investigated according to the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated, and obtain at least one layer of grids to be sampled;

[0187] The evaluation module 320 is used to determine the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer; the target ratio is determined based on the proportion of the staple food crop planting area within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0188] According to the first sub-meter remote sensing image corresponding to each of the layers, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each of the layers, and based on the first staple food crop planting area in the sample frame corresponding to each of the layers, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0189] The device provided in this embodiment includes a stratification module 310 and an evaluation module 320, wherein the stratification module 310 is used to use a first target extraction network to perform target extraction on a first medium-resolution remote sensing image of a to-be-surveyed area, obtain the planting distribution information of a first staple food crop and the first cultivated land distribution information corresponding to the to-be-surveyed area, and stratify the to-be-surveyed area according to the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the to-be-surveyed area, the terrain features of the to-be-surveyed area, and the planting structure, to obtain at least one layer of to-be-sampled grids; then, the evaluation module 320 is used to extract the target of the to-be-sampled grids based on the target of the to-be-sampled grids. The target size and the target ratio in each layer of the grid to be sampled are used to determine the first sub-meter remote sensing image corresponding to the sample frame of each layer, wherein the target ratio is determined based on the proportion of the staple food crop planting area in the grid to be sampled to the internal cultivated land area in the grid to be sampled; further, according to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0190] In the present invention, when the region to be investigated is stratified, that is, when the resolution requirements for remote sensing images are relatively low, a first medium-resolution remote sensing image of the region to be investigated is obtained, and a first target extraction network is used to extract the target of the first medium-resolution remote sensing image of the region to be investigated, and the planting distribution information and the first cultivated land distribution information of the first staple food crop corresponding to the region to be investigated are obtained. Then, the region to be investigated is stratified in combination with the terrain characteristics and planting structure of the region to be investigated, and the efficiency of stratification is high and the reliability of stratification results is high; further, when operations such as crop classification and area extraction that require relatively high resolution of remote sensing images are performed, the area is extracted according to the first sub-meter remote sensing image corresponding to the sample frame of each layer, and the second target extraction network is used to extract the area to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and then the estimated value of the total planting area of ​​staple food crops in the region to be investigated is obtained by reverse deduction. In the present invention, two remote sensing images of different scales are scientifically used, and combined with the traditional geostatistical sampling method, the total planting area of ​​staple food crops in the region to be investigated is finally evaluated, and the accuracy of the evaluation of the planting area of ​​staple food crops is improved.

[0191] According to the present invention, a staple food crop planting area assessment device 300 is provided, wherein the device further comprises a model training module;

[0192] The model training module is used to train the initial first extraction network to obtain a first target extraction network; the model training module is specifically used to:

[0193] Determine a plurality of second medium-resolution remote sensing images of the area to be investigated within a first preset time period, and planting distribution information of second staple crops and second cultivated land distribution information corresponding to each of the areas to be investigated as a training data set;

[0194] Obtaining hyperparameters during training; the hyperparameters include at least one of a learning rate, a batch size, and a number of iterations;

[0195] Input each batch of training data into the initial first extraction network, forward propagate the extraction results, and calculate the loss function value corresponding to each batch;

[0196] According to the loss function values ​​corresponding to each batch, the model parameters of the initial first extraction network are updated to obtain updated model parameters;

[0197] Based on the updated model parameters, the first object extraction network is determined.

[0198] According to the staple crop planting area evaluation device 300 provided by the present invention, the stratification module 310 is specifically used for:

[0199] Based on the terrain characteristics of the area to be investigated, the area to be investigated is divided into a layer with relatively flat terrain and a layer with relatively large terrain fluctuations using the standard deviation of the digital elevation model, and the layer with relatively flat terrain and the layer with relatively large terrain fluctuations are determined as the first-stage stratification result;

[0200] Determine the second-stage stratification result according to the first-stage stratification result, the planting distribution information of the first staple food crop corresponding to the area to be investigated, the first cultivated land distribution information and the planting structure;

[0201] Determining a target stratification result based on the first-stage stratification result and the second-stage stratification result;

[0202] According to the target stratification results, grids to be sampled in each layer are constructed.

[0203] According to the staple crop planting area evaluation device 300 provided by the present invention, the evaluation module 320 is specifically used for:

[0204] Determine the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer;

[0205] Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

[0206] According to the staple crop planting area evaluation device 300 provided by the present invention, the evaluation module 320 is further used for:

[0207] For any of the layers, according to the planting area of ​​the first staple food crop in the sample frame corresponding to the layer, the number of grids in the sample frame corresponding to the layer, and the number of grids of the to-be-sampled grids corresponding to the layer, determine the planting area of ​​the first staple food crop corresponding to the layer;

[0208] According to the planting area of ​​the first staple food crops corresponding to each layer, an estimated value of the total planting area of ​​the staple food crops in the area to be investigated is calculated.

[0209] According to the staple crop planting area evaluation device 300 provided by the present invention, the stratification module 310 is further used for:

[0210] Determining the target size of the grid to be sampled according to a preset correlation rule;

[0211] Among them, the preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of each grid to be sampled, the better the effect of the division size of each grid to be sampled; the target size is used to characterize that there is no spatial autocorrelation between each grid to be sampled under the target size.

[0212] According to the staple crop planting area evaluation device 300 provided by the present invention, the evaluation module 320 is further used for:

[0213] Determine the key areas of concern within the area to be investigated;

[0214] According to the key focus area and the grids to be sampled of the layer, determining the grids to be sampled in the grids to be sampled of the layer that cover the key focus area;

[0215] Determine the sample frame corresponding to each layer based on the target size of the grid to be sampled, the target ratio within the grid to be sampled of each layer, and the grid to be sampled covering the key focus area;

[0216] Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

[0217] According to the staple crop planting area evaluation device 300 provided by the present invention, the evaluation module 320 is further used for:

[0218] Based on the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area, a second estimated value of the total planting area of ​​staple crops in the key focus area is determined.

[0219] According to the present invention, a staple food crop planting area assessment device 300 is provided, the device further comprising a data preprocessing module;

[0220] The data preprocessing module is used to:

[0221] Before extracting targets from the first medium-resolution remote sensing image of the area to be investigated using the first target extraction network to obtain the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the method further includes:

[0222] Acquire at least one second medium resolution remote sensing image;

[0223] Selecting the third medium-resolution remote sensing images that meet the preset time phase condition and cloud coverage rate condition from among the second medium-resolution remote sensing images;

[0224] The third medium-resolution remote sensing image is subjected to image preprocessing to obtain a first medium-resolution remote sensing image of the area to be investigated.

[0225] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communications interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute the method for evaluating the planting area of ​​staple crops, which method includes:

[0226] Using a first target extraction network, target extraction is performed on a first medium-resolution remote sensing image of the area to be investigated, so as to obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the area to be investigated, and the area to be investigated is stratified according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the area to be investigated, the terrain features and the planting structure of the area to be investigated, so as to obtain at least one layer of grids to be sampled;

[0227] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0228] According to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0229] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0230] On the other hand, the present invention further provides a computer program product, the computer program product comprising a computer program, the computer program can be stored in a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the method for evaluating the planting area of ​​staple food crops provided by the above methods, the method comprising:

[0231] Using a first target extraction network, target extraction is performed on a first medium-resolution remote sensing image of the area to be investigated, so as to obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the area to be investigated, and the area to be investigated is stratified according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the area to be investigated, the terrain features and the planting structure of the area to be investigated, so as to obtain at least one layer of grids to be sampled;

[0232] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0233] According to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0234] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the method for evaluating the planting area of ​​staple food crops provided by the above methods, the method comprising:

[0235] Using a first target extraction network, target extraction is performed on a first medium-resolution remote sensing image of the area to be investigated, so as to obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the area to be investigated, and the area to be investigated is stratified according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the area to be investigated, the terrain features and the planting structure of the area to be investigated, so as to obtain at least one layer of grids to be sampled;

[0236] Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled;

[0237] According to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

[0238] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0239] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0240] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for evaluating the planting area of ​​staple crops, characterized in that: include: Using a first target extraction network, target extraction is performed on a first medium-resolution remote sensing image of the area to be investigated, so as to obtain planting distribution information of a first staple food crop and first cultivated land distribution information corresponding to the area to be investigated, and the area to be investigated is stratified according to the planting distribution information of the first staple food crop and first cultivated land distribution information corresponding to the area to be investigated, the terrain features and the planting structure of the area to be investigated, so as to obtain at least one layer of grids to be sampled; Based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, determining the first sub-meter remote sensing image corresponding to the sample frame of each layer; the target ratio is determined based on the proportion of the planting area of ​​staple food crops within the grid to be sampled to the internal cultivated land area within the grid to be sampled; According to the first sub-meter remote sensing image corresponding to the sample frame of each layer, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each layer, and according to the first staple food crop planting area in the sample frame corresponding to each layer, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

2. The method for evaluating the planting area of ​​staple food crops according to claim 1, characterized in that: The first target extraction network is obtained by training the initial first extraction network through the following steps, including: Determine a plurality of second medium-resolution remote sensing images of the area to be investigated within a first preset time period, and planting distribution information of second staple crops and second cultivated land distribution information corresponding to each of the areas to be investigated as a training data set; Obtaining hyperparameters during training; the hyperparameters include at least one of a learning rate, a batch size, and a number of iterations; Input each batch of training data into the initial first extraction network, forward propagate the extraction results, and calculate the loss function value corresponding to each batch; According to the loss function values ​​corresponding to each batch, the model parameters of the initial first extraction network are updated to obtain updated model parameters; Based on the updated model parameters, the first object extraction network is determined.

3. The method for evaluating the planting area of ​​staple food crops according to claim 1, characterized in that: The step of stratifying the area to be investigated according to the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the terrain characteristics and the planting structure of the area to be investigated to obtain at least one layer of grids to be sampled includes: Based on the terrain characteristics of the area to be investigated, the area to be investigated is divided into a layer with relatively flat terrain and a layer with relatively large terrain fluctuations using the standard deviation of the digital elevation model, and the layer with relatively flat terrain and the layer with relatively large terrain fluctuations are determined as the first-stage stratification result; Determine the second-stage stratification result according to the first-stage stratification result, the planting distribution information of the first staple food crop corresponding to the area to be investigated, the first cultivated land distribution information and the planting structure; Determining a target stratification result based on the first-stage stratification result and the second-stage stratification result; According to the target stratification results, grids to be sampled in each layer are constructed.

4. The method for evaluating the planting area of ​​staple food crops according to claim 1, characterized in that: The determining, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the first sub-meter remote sensing image corresponding to the sample frame of each layer comprises: Determine the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer; Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

5. The method for evaluating the planting area of ​​staple food crops according to claim 1, characterized in that: Determining the estimated value of the total planting area of ​​staple crops in the area to be investigated according to the planting area of ​​the first staple crops in the sample frame corresponding to each layer includes: For any of the layers, according to the planting area of ​​the first staple food crop in the sample frame corresponding to the layer, the number of grids in the sample frame corresponding to the layer, and the number of grids of the to-be-sampled grids corresponding to the layer, determine the planting area of ​​the first staple food crop corresponding to the layer; According to the planting area of ​​the first staple food crops corresponding to each layer, an estimated value of the total planting area of ​​the staple food crops in the area to be investigated is calculated.

6. The method for evaluating the planting area of ​​staple food crops according to any one of claims 1 to 5, characterized in that: Before determining the first sub-meter remote sensing image corresponding to the sampled sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the method further includes: Determining the target size of the grid to be sampled according to a preset correlation rule; Among them, the preset correlation rule is used to characterize that the smaller the similarity and aggregation degree between the spatial data of each grid to be sampled, the better the effect of the division size of each grid to be sampled, and the spatial data is the planting area of ​​staple food crops within the second preset time period; the target size is used to characterize that there is no spatial autocorrelation between each grid to be sampled under the target size.

7. The method for evaluating the planting area of ​​staple crops according to any one of claims 1 to 5, characterized in that: Before determining the first sub-meter remote sensing image corresponding to the sampled sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the method further includes: Determine the key areas of concern within the area to be investigated; The determining, based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer, the first sub-meter remote sensing image corresponding to the sample frame of each layer comprises: According to the key focus area and the grids to be sampled of the layer, determining the grids to be sampled in the grids to be sampled of the layer that cover the key focus area; Determine the sample frame corresponding to each layer based on the target size of the grid to be sampled, the target ratio within the grid to be sampled of each layer, and the grid to be sampled covering the key focus area; Based on the sample frames corresponding to the layers, a first sub-meter remote sensing image corresponding to the sample frames that completely cover the layers is determined.

8. The method for evaluating the planting area of ​​staple food crops according to claim 7, characterized in that: The method further comprises: Based on the estimated value of the total planting area of ​​staple crops in the area to be investigated and the first estimated value of the total planting area of ​​staple crops in the grid to be sampled in the key focus area, a second estimated value of the total planting area of ​​staple crops in the key focus area is determined.

9. The method for evaluating the planting area of ​​staple crops according to any one of claims 1 to 5, characterized in that: Before using the first target extraction network to extract targets from the first medium-resolution remote sensing image of the area to be investigated to obtain the planting distribution information of the first staple food crop and the first cultivated land distribution information corresponding to the area to be investigated, the method further includes: Acquire at least one second medium resolution remote sensing image; Selecting the third medium-resolution remote sensing images that meet the preset time phase condition and cloud coverage rate condition from among the second medium-resolution remote sensing images; The third medium-resolution remote sensing image is subjected to image preprocessing to obtain a first medium-resolution remote sensing image of the area to be investigated.

10. A device for evaluating the planting area of ​​staple crops, characterized in that: include: a stratification module, configured to use a first target extraction network to perform target extraction on a first medium-resolution remote sensing image of a region to be investigated, obtain planting distribution information of a first staple crop and first cultivated land distribution information corresponding to the region to be investigated, and stratify the region to be investigated according to the planting distribution information of the first staple crop and first cultivated land distribution information corresponding to the region to be investigated, the terrain features of the region to be investigated, and the planting structure, to obtain at least one layer of grids to be sampled; An evaluation module is used to determine the first sub-meter remote sensing image corresponding to the sample frame of each layer based on the target size of the grid to be sampled and the target ratio within the grid to be sampled of each layer; the target ratio is determined based on the proportion of the staple food crop planting area within the grid to be sampled to the internal cultivated land area within the grid to be sampled; According to the first sub-meter remote sensing image corresponding to each of the layers, the second target extraction network is used to perform area extraction to obtain the first staple food crop planting area in the sample frame corresponding to each of the layers, and based on the first staple food crop planting area in the sample frame corresponding to each of the layers, the estimated value of the total staple food crop planting area in the area to be investigated is determined.

Images