A method and device for evaluating the quality of cultivated land, a computer device and a medium

Abstract

The application belongs to the technical field of cultivated land quality evaluation, and discloses a cultivated land quality evaluation method, device, computer equipment and medium. The method obtains land grid data of the cultivated land to be evaluated, compares and learns the characteristics of the land grid data by a backbone network to obtain context features; converts the land grid data into three-dimensional tensor data; inputs the three-dimensional tensor data into a CNN module to extract potential local features in the three-dimensional tensor data and obtain spatial features; combines the spatial features with the context features to obtain comprehensive features, inputs the comprehensive features into a full connection layer to obtain a spatial prediction value; extracts topological attribute features in the three-dimensional tensor data by using a GNN module to obtain a topological prediction value; and fuses the spatial prediction value and the topological prediction value to obtain an evaluation result of the quality of the cultivated land to be evaluated. The accuracy of soil quality evaluation can be improved.

Description

Technical Field

[0001] This application relates to the field of farmland quality assessment technology, and in particular to a method, apparatus, computer equipment and medium for assessing farmland quality. Background Technology

[0002] The quality assessment of arable land resources is crucial for achieving arable land protection. Currently, arable land quality assessments are mostly conducted manually, with domain experts comprehensively analyzing and judging the condition of each plot to determine its quality. Other methods utilize mathematical models and techniques such as the Delphi method and entropy weighting to evaluate the importance of various factors influencing arable land quality, thereby assigning a quality score to each plot. Furthermore, some methods based on deep learning models and employing end-to-end training paradigms have emerged for evaluating arable land quality.

[0003] However, these technologies currently have some limitations. Manual judgment is cumbersome and inefficient, and the evaluation results are easily influenced by subjective factors. Mathematical model-based methods lack complex nonlinear modeling capabilities. In the task of farmland quality assessment, the relationship between influencing factors and farmland quality is not linear, so mathematical models cannot effectively model this relationship, leading to inaccurate quality evaluation results. Currently, deep learning-based methods ignore the common problem of data imbalance in practical applications. The uneven distribution of data makes it difficult to train deep learning classification models because the models often show bias towards categories with more training data, while performing poorly on categories with scarce data, such as high-quality farmland. Summary of the Invention

[0004] Therefore, embodiments of this application provide a method, apparatus, computer equipment, and medium for assessing arable land quality, thereby improving the assessment performance of arable land quality and obtaining more accurate assessment results.

[0005] Firstly, this application provides a method for assessing the quality of arable land.

[0006] This application is achieved through the following technical solution:

[0007] A method for assessing arable land quality, based on the deep learning model SAFE, wherein the deep learning model SAFE includes a backbone network, a CNN module, and a GNN module, comprising:

[0008] The land grid data of the cultivated land to be evaluated is obtained, and the land grid data is input into the backbone network. The backbone network performs comparative learning on the features of the land grid data to obtain the contextual features of the land grid data.

[0009] The land grid data is associated with geographic attributes to convert the land grid data into three-dimensional tensor data;

[0010] The three-dimensional tensor data is input into the CNN module to extract potential local features from the three-dimensional tensor data, thereby obtaining spatial features;

[0011] The spatial features are combined with the contextual features to obtain comprehensive features. The comprehensive features are then input into the fully connected layer to obtain the spatial prediction value of the cultivated land to be evaluated.

[0012] The three-dimensional tensor data is input into the GNN module, the topological attribute features in the three-dimensional tensor data are extracted, the topological attribute features are predicted, and the topological prediction value of the cultivated land to be evaluated is obtained.

[0013] The spatial and topological prediction values ​​are fused to obtain the assessment results of the quality of the cultivated land to be evaluated.

[0014] In a preferred embodiment of this application, the method may be further configured such that, before inputting the three-dimensional tensor data into the CNN module, the method further includes:

[0015] The three-dimensional data tensor is divided into grids x i Centered on the center, it is divided into local matrix slices of size C×C, where the clipping parameter C is set to 30~35.

[0016] In a preferred example of this application, the land grid data can be further configured to be input into a backbone network, and the backbone network performs comparative learning on the features of the land grid data, including:

[0017] In the backbone network, cosine similarity is used to measure the distance between input samples and anchor samples. The loss function for contrastive learning is set as follows:

[0018]

[0019] Specifically, when the input sample and the anchor sample belong to the same category, the loss function is minimized to make the input sample and the anchor sample closer together:

[0020]

[0021] When the input sample and the anchor sample belong to different categories, the loss function is minimized to separate the input sample and the anchor sample:

[0022]

[0023] in, Represents the loss function. Represents cosine similarity. This represents the contrastive learning loss function value for the i-th sample. This represents the contrastive learning loss function value for the i-th sample whose predicted value is 1. This represents the contrastive learning loss function value for the i-th sample whose predicted value is 0. This represents the feature vector of the i-th input sample. The feature vector representing the anchor sample. denoted as the predicted value of the i-th input sample, and ∈ represents the threshold controlling whether to perform comparative learning loss calculation on input samples and anchor samples of different categories.

[0024] In a preferred example of this application, the land grid data can be further configured to associate with geographic attributes, thereby converting the land grid data into three-dimensional tensor data, including:

[0025] Obtain the longitude and latitude sets of all grids from the geohash attribute of the land grid data;

[0026] Sort the longitudes in the longitude set and the latitudes in the latitude set, and add longitude and latitude indices to each grid.

[0027] Based on longitude and latitude indexes, the attribute vectors of land grid data are mapped to the corresponding grid locations to obtain three-dimensional tensor data.

[0028] In a preferred example of this application, the method can be further configured to input the comprehensive features into a fully connected layer to obtain the spatial prediction value of the cultivated land to be evaluated, including:

[0029]

[0030] in, This indicates that the CNN module extracts spatial features, where D2 is the dimension of the extracted spatial features. Indicates a fully connected layer. Representing three-dimensional tensor data, X represents the contextual features extracted by the backbone network. i This represents the land grid data of the i-th grid.

[0031] In a preferred embodiment of this application, the three-dimensional tensor data can be further configured to be input into a GNN module to extract topological attribute features from the three-dimensional tensor data, including:

[0032] The three-dimensional tensor data is input into the GNN module, and the features of adjacent nodes are aggregated based on the aggregation function to obtain aggregated features;

[0033] The aggregated features are nonlinearly transformed based on the update function to generate updated node representations. The attribute interaction information in the updated node representations is then extracted to obtain topological attribute features.

[0034] In a preferred embodiment of this application, the method may be further configured to predict the topological attribute features to obtain a topological prediction value for the cultivated land to be evaluated, including:

[0035] The topological attribute features are predicted using a multilayer perceptron, with the ReLU function used as the activation function.

[0036] Secondly, this application provides an apparatus for assessing the quality of arable land.

[0037] This application is achieved through the following technical solution:

[0038] An apparatus for assessing arable land quality, the apparatus being used to perform the assessment method described in the first aspect above, the apparatus comprising:

[0039] The data acquisition module is used to acquire land grid data of the cultivated land to be evaluated; and to associate the land grid data with geographic attributes so that the land grid data is converted into three-dimensional tensor data.

[0040] The feature extraction module is used to perform comparative learning on the features of the land grid data through the backbone network to obtain the context features of the land grid data; to extract potential local features from the three-dimensional tensor data through the CNN module to obtain spatial features, and to combine the spatial features with the context features to obtain comprehensive features; and to extract topological attribute features from the three-dimensional tensor data through the GNN module.

[0041] The quality assessment module is used to obtain the spatial prediction value of the cultivated land to be assessed from the comprehensive features; predict the topological attribute features to obtain the topological prediction value of the cultivated land to be assessed; and fuse the spatial prediction value and the topological prediction value to obtain the quality assessment result of the cultivated land to be assessed.

[0042] In a preferred example of this application, it can be further set as follows:

[0043] Thirdly, this application achieves its goal through the following technical solutions:

[0044] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described methods for assessing arable land quality.

[0045] Fourthly, this application provides a computer-readable storage medium.

[0046] This application is achieved through the following technical solution:

[0047] A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the above-described methods for assessing arable land quality.

[0048] In summary, compared with the prior art, the beneficial effects of the technical solution provided in this application include at least the following:

[0049] This application employs a spatially aware deep learning model, SAFE, which uses a convolutional neural network (CNN) to integrate and extract the influence paradigm of spatial local features in land grid data. The CNN effectively encodes the influence fusion paradigm of neighboring regions within each feature dimension. Furthermore, a graph neural network (GNN) is used to mine the topological attribute interaction effects between related nodes, capturing the mutual influence relationships between each land grid attribute in the topological structure of the land grid data. In addition, contrastive learning is used in the backbone network to enhance its classification performance under imbalanced data conditions, enabling it to minimize the distance between similar data and maximize the distance between different categories to optimize feature embedding, thus making the land quality assessment results more accurate. Attached Figure Description

[0050] Figure 1 A schematic flowchart illustrating a method for assessing arable land quality according to an embodiment of this application;

[0051] Figure 2 A schematic diagram of the structure of a farmland quality assessment device provided in an embodiment of this application;

[0052] Explanation of reference numerals in the attached figures:

[0053] Data acquisition module-101, feature extraction module-201, quality assessment module-301. Detailed Implementation

[0054] This specific embodiment is merely an explanation of this application and is not intended to limit it. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they fall within the scope of the claims of this application.

[0055] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] Furthermore, the term "and / or" in this application is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this application, unless otherwise specified, generally indicates that the preceding and following related objects have an "or" relationship.

[0057] In this application, the terms "first," "second," etc., are used to distinguish identical or similar items with essentially the same function. It should be understood that there is no logical or temporal dependency between "first," "second," and "nth," nor are there any restrictions on quantity or execution order.

[0058] The method for assessing farmland quality in this application is based on a deep learning model, SAFE (Spatial Awareness Framework for Evaluating Farmland), which is a binary classification model comprising a backbone network, a CNN module, and a GNN module. The backbone network is a TabTransformer model used to assist modeling by utilizing the mutual influence information of neighboring regions. The CNN module extracts potential spatial local feature fusion paradigms from the grid data used to assess farmland quality. The GNN module extracts topological relationships and attribute interactions from the grid data. To enhance SAFE's ability to handle imbalanced data, self-supervised regularization techniques, such as contrastive learning, are introduced into the backbone network to improve its discriminative power for feature embeddings. SAFE can predict whether a land grid is high-quality farmland or has the potential to become high-quality farmland.

[0059] The embodiments of this application will now be described in further detail with reference to the accompanying drawings. Figure 1 As shown, this application proposes a method for assessing arable land quality, comprising:

[0060] S10: Obtain land grid data of the cultivated land to be evaluated, input the land grid data into the backbone network, and the backbone network compares and learns the features of the land grid data to obtain the contextual features of the land grid data.

[0061] Specifically, the farmland to be assessed is divided into several grids of the same size. The land grid data is a tabular feature matrix, represented as X∈R. N×D , where N represents the number of land grids, and D is the attribute / factor dimension of each grid.

[0062] Each land grid possesses complex influencing factor attributes, which can be broadly categorized into three types based on attribute classification: physical conditions, environmental impacts, and land type. For example, physical conditions include, but are not limited to, factors such as soil thickness and soil texture; environmental impacts include, but are not limited to, altitude and slope; and land type includes, but is not limited to, areas designated for construction or areas not designated for construction. From a numerical perspective, attributes can be divided into two categories: continuous numerical attributes and discrete numerical attributes. For instance, soil thickness, altitude, and slope are continuous numerical attributes, while soil texture and soil type are discrete numerical attributes.

[0063] The backbone network aims to capture the latent features of each land network and analyze its spatial context information. The backbone network is flexible and can adopt various architectures, such as MLP, DeepFM, and Transformer frameworks. Preferably, in this application, to transform these sparse attribute vectors into dense latent features, TabTransformer is used as the backbone structure of the deep learning model SAFE. TabTransformer excels at utilizing the contextual features of categorical attributes.

[0064] S20: Associate land grid data with geographic attributes to convert the land grid data into three-dimensional tensor data.

[0065] S30: Input the 3D tensor data into the CNN module, extract the potential local features in the 3D tensor data, and obtain the spatial features.

[0066] S40: Combine spatial features with contextual features to obtain comprehensive features. Input the comprehensive features into the fully connected layer to obtain the spatial prediction value of the cultivated land to be evaluated.

[0067] S50: Input the three-dimensional tensor data into the GNN module, extract the topological attribute features from the three-dimensional tensor data, predict the topological attribute features, and obtain the topological prediction value of the cultivated land to be evaluated.

[0068] S60: The spatial prediction values ​​and topological prediction values ​​are fused to obtain the assessment results of the quality of the cultivated land to be evaluated.

[0069] Among them, the assessment result is the predicted probability value of high-quality arable land or land with the potential to become high-quality arable land.

[0070] SAFE, a spatially aware deep learning model, uses a convolutional neural network (CNN) to integrate and extract the influence paradigm of spatial local features in land grid data. The CNN effectively encodes the influence fusion paradigm of neighboring regions within each feature dimension. Furthermore, a graph neural network (GNN) is used to mine the topological attribute interaction effects between related nodes, capturing the mutual influence relationship between each land grid attribute in the topological structure of the land grid data. In addition, contrastive learning is used in the backbone network to enhance the classification performance under imbalanced data conditions, so that it can minimize the distance between similar data and maximize the distance between different categories to optimize feature embedding, making the land quality assessment results more accurate.

[0071] Specifically, the evaluation results It is calculated by combining the spatial predicted value of spatial attributes and the topological predicted value of topological attributes, and the formula is expressed as:

[0072] in Represented as spatial prediction values, Represented as topological prediction value, The result represents the probability that the farmland to be assessed is of high quality or has the potential to become high quality land.

[0073] It should be noted that, given the challenges posed by imbalanced datasets, this application employs a balanced softmax loss to train the classifier. In the model of this application, let n... l y represents the number of samples in category l∈{0,1}. i Represents land grid x i If the true category is determined, then the loss function... It can be represented as:

[0074]

[0075] The backbone network TabTransformer excels at utilizing contextual features of categorical attributes, where the self-attention mechanism is represented as a function. This enables it to capture interaction patterns between influencing factors, thereby enhancing the understanding of the internal dependencies in the input land grid data:

[0076]

[0077] Among them, f MLP (·) represents the multilayer sensing function, a vector and These represent the continuous numerical attributes and categorical attributes of each instance, respectively. The context embedding of the categorical attributes is determined by... Obtain, where f Emb (·) is an embedded function. Operator This indicates a join operation.

[0078] The TabTransformer function with parameter θ1 is defined as follows:

[0079]

[0080] Where D1 represents the dimension of the feature representation. Therefore, a dense latent feature matrix Calculated as

[0081] Land grid data is input into the backbone network, which then learns by comparing the features of the land grid data. First, binary data sets are constructed, and these binary data sets are used to improve the comparative learning ability of the backbone network. For a dataset {x} containing N samples... i y i} i＝1...N This forms a training dataset containing N tuples. Where i = 1...N, x a It is a randomly selected positive sample as the anchor point. x represents i Does it belong to x? a The same category. Positive samples indicate that the plot is or has the potential to become high-quality arable land, while negative samples indicate that the plot is not or has no potential to become high-quality arable land.

[0082] In some embodiments, the distance between input samples and anchor points is measured based on cosine similarity in the backbone network, and the loss function for contrastive learning is set as follows:

[0083]

[0084] Where the input sample and the anchor sample belong to the same category (i.e.) (At time), by minimizing the loss function to make the input sample and the anchor sample closer:

[0085]

[0086] When the input sample and the anchor sample belong to different categories (i.e.) (At time), the input sample and anchor sample are separated by minimizing the loss function:

[0087]

[0088] in, Represents the loss function. Represents cosine similarity. This represents the contrastive learning loss function value for the i-th sample. This represents the contrastive learning loss function value for the i-th sample whose predicted value is 1. This represents the contrastive learning loss function value for the i-th sample whose predicted value is 0. This represents the feature vector of the i-th input sample. The feature vector representing the anchor sample. denoted as the predicted value of the i-th input sample, and ∈ represents the threshold controlling whether to perform comparative learning loss calculation on input samples and anchor samples of different categories.

[0089] Contrastive learning has demonstrated significant effectiveness in addressing the imbalanced data problem. By integrating this approach into the feature extraction process of the backbone network, the feature mapping function can be optimized. This process clusters similar data points together while dissimilar data points are dispersed, thereby improving the performance of the classifier.

[0090] In some embodiments, associating land grid data with geographic attributes to convert the land grid data into three-dimensional tensor data includes: obtaining the longitude set and latitude set of all grids from the geohash attribute of the land grid data; sorting the longitude in the longitude set and the latitude in the latitude set, and adding a longitude index and a latitude index to each grid; and mapping the attribute vector of the land grid data to the corresponding grid location based on the longitude index and the latitude index to obtain three-dimensional tensor data.

[0091] Specifically, the tabular format of raw land grid data is not suitable for training and feature extraction of convolutional neural networks (CNNs) and generative neural networks (GNNs). To address this issue, this application proposes a data transformation strategy that converts land grid data into grid data by representing the geographic latitude and longitude coordinates of each grid cell using its geohash attribute. To illustrate this process, the dataset is represented as: X = {x1, x2, ..., x...} i , ..., x N},

[0092] The longitude set is represented as follows:

[0093]

[0094] The latitude set is represented as:

[0095]

[0096] The longitude and latitude sets correspond to the geographic coordinates needed to transform land grid data into a grid format suitable for CNN and GNN processing. After independently sorting the longitude and latitude sets, the land grid x is represented. i Longitude and latitude grid indexes were exported. in Here, W and H represent the total number of grid indices in the longitude dimension and the total number of grid indices in the latitude dimension, respectively. Using these indices, the original land grid data is reconstructed into three-dimensional tensor data. Each attribute vector is mapped to its corresponding grid position. This mapping relationship allows the reconstruction of a two-dimensional matrix corresponding to each feature dimension.

[0097] Considering the use of the entire three-dimensional tensor data Predicting the grid classification of farmland to be assessed is computationally intensive. Furthermore, considering the reduced influence between geographically distant areas, this application introduces a hyperparameter C to define the relevant spatial extent before inputting the 3D tensor data into the CNN module. Specifically: the 3D data tensor is divided into grids x... i Centered on the data, the data is divided into local matrix slices of size C×C, where the clipping parameter C is set to 30–35. After this processing, the land grid data is transformed into tensor data. Preferably, in this application, the trimming parameter C is 32.

[0098] Using the transformed 3D tensor data, a Convolutional Neural Network (CNN) is employed to extract and integrate local features through convolution operations. CNNs exhibit higher analytical efficiency in fine-grained local analysis and are better able to capture the spatial hierarchical structure of the data. The extraction process of the backbone network is represented as follows: D2 is the dimension of the extracted spatial features.

[0099] In some embodiments, comprehensive features are input into a fully connected layer to obtain spatial predictions of the arable land to be evaluated, including:

[0100]

[0101] in, This indicates that the CNN module extracts spatial features, where D2 is the dimension of the extracted spatial features. Indicates a fully connected layer. Representing three-dimensional tensor data, This represents the contextual features extracted by the backbone network.

[0102] Given that the distribution of high-quality arable land is often uneven and irregular, Graph Neural Networks (GNNs) are suitable for capturing the potential attribute interactions between plots. By leveraging the inherent topological structure of graphs, GNNs can efficiently learn node attributes through edge updates, providing substantial benefits in attribute interaction and extraction. Compared to traditional grid-based and sequence-based structures, GNNs exhibit better performance due to their ability to integrate more complex and flexible structural information.

[0103] In some embodiments, inputting three-dimensional tensor data into a GNN module and extracting topological attribute features from the three-dimensional tensor data includes: inputting three-dimensional tensor data into a GNN module, performing aggregation operations on the features of adjacent nodes based on an aggregation function to obtain aggregated features; performing a nonlinear transformation on the aggregated features based on an update function to generate an updated node representation, and extracting attribute interaction information from the updated node representation to obtain topological attribute features.

[0104] Specifically, the converted 3D data (Local matrix slicing) and dividing the data into N1 blocks. h and w represent the height and width of the block, respectively. Each block is considered as a block with characteristic representation v. i ∈R D×hw The set of unordered nodes can be further represented as: For edge initialization, we will initialize each node v i Connecting to its eight nearest neighbors forms a path from the neighboring nodes to node v. i Given directed edges, construct a graph. In the construction graph Subsequently, graph convolutional layers facilitate information exchange between nodes.

[0105] The process of predicting topological attribute features to obtain the topological prediction value of the cultivated land to be evaluated includes: using a multilayer perceptron to predict the topological attribute features and using the ReLU function as the activation function.

[0106]

[0107] in, Indicates v i The set of adjacent nodes. It should be noted that... The algorithm is dynamically adjusted in each iteration (each graph convolutional layer), using a sparse k-nearest neighbor (k-NN) function to measure distance during the iteration process. After the iterative adjustment process is complete, node v... i The topological properties and MLP functions Topology prediction u iThis represents the vertex feature aggregation vector under the max pooling operation. This represents the update operation of the graph convolutional layer. This represents the activation vector after the graph convolution update layer and activation function.

[0108] In the task of assessing arable land quality, the overall objective loss function is used. Minimize to complete the optimization of the entire model:

[0109] Where γ is a weighting coefficient, which can be set to 1 in this application.

[0110] Another embodiment of this application also provides a farmland quality assessment device, which is used to perform the above-described assessment method, such as... Figure 2 As shown, the evaluation device includes:

[0111] The data acquisition module 101 is used to acquire land grid data of the cultivated land to be evaluated; and to associate the land grid data with geographic attributes so as to convert the land grid data into three-dimensional tensor data.

[0112] The feature extraction module 201 is used to perform comparative learning on the features of the land grid data through the backbone network to obtain the context features of the land grid data; to extract potential local features from the three-dimensional tensor data through the CNN module to obtain spatial features; to combine the spatial features with the context features to obtain comprehensive features; and to extract topological attribute features from the three-dimensional tensor data through the GNN module.

[0113] The quality assessment module 301 is used to obtain the spatial prediction value of the cultivated land to be assessed from the comprehensive features; predict the topological attribute features to obtain the topological prediction value of the cultivated land to be assessed; and fuse the spatial prediction value and the topological prediction value to obtain the quality assessment result of the cultivated land to be assessed.

[0114] The specific limitations of the farmland quality assessment device provided in this embodiment can be found in the embodiments of the farmland quality assessment method described above, and will not be repeated here. Each module in the above-mentioned farmland quality assessment device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in a computer device in hardware form, or stored in the memory of a computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0115] This application provides a computer device that may include a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface communicates with external terminals via a network connection. When the computer program is executed by the processor, it causes the processor to perform the steps of the farmland quality assessment method as described in any of the above embodiments.

[0116] The working process, working details, and technical effects of the computer equipment provided in this embodiment can be found in the embodiment of the farmland quality assessment method described above, and will not be repeated here.

[0117] This application provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the steps of the farmland quality assessment method as described in any of the above embodiments. The computer-readable storage medium refers to a data storage medium, which may include, but is not limited to, floppy disks, optical disks, hard disks, flash memory, USB flash drives, and / or memory sticks. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0118] The working process, working details, and technical effects of the computer-readable storage medium provided in this embodiment can be found in the embodiment of the cultivated land quality assessment method described above, and will not be repeated here.

[0119] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0120] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0121] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system described in this application can be divided into different functional units or modules to complete all or part of the functions described above.

Claims

1. A method of evaluating the quality of a ploughed field, characterized in that, The method is executed based on a deep learning model SAFE, the deep learning model SAFE comprises a backbone network, a CNN module and a GNN module, and the method comprises the following steps: Land grid data of the cultivated land to be evaluated is acquired, and the land grid data is input into the backbone network; the backbone network performs contrastive learning on features of the land grid data, and context features of the land grid data are obtained; Longitude sets and latitude sets of all grids are obtained from geohash attributes of the land grid data; longitudes in the longitude sets and latitudes in the latitude sets are sorted; longitude indexes and latitude indexes are added to each grid; attribute vectors of the land grid data are mapped to corresponding grid positions based on the longitude indexes and the latitude indexes, and three-dimensional tensor data are obtained; The three-dimensional tensor data are input into the CNN module, potential local features in the three-dimensional tensor data are extracted, and spatial features are obtained; The spatial features are combined with the context features, comprehensive features are obtained, the comprehensive features are input into a full connection layer, and spatial prediction values of the cultivated land to be evaluated are obtained, comprising the following steps: wherein, represents a spatial prediction value, represents a CNN module extracting spatial features, the feature change in the extraction process being represented as D2 is the dimension of the extracted spatial features, represents a fully connected layer, the feature change in the fully connected layer being represented as represents a context feature extracted by the backbone network, the feature change in the extraction process being represented as X i represents land grid data of the i-th grid, represents three-dimensional tensor data, The three-dimensional tensor data are input into the GNN module; adjacent node features are aggregated based on an aggregation function, aggregated features are obtained; the aggregated features are nonlinearly transformed based on an update function, updated node representations are generated; attribute interaction information in the updated node representations is extracted, topological attribute features are obtained; the topological attribute features are predicted using a multilayer perceptron, and a ReLU function is used as an activation function; The spatial prediction values and topological prediction values are fused, and an evaluation result of quality of the cultivated land to be evaluated is obtained.

2. The method of evaluating the quality of ploughing according to claim 1, characterized in that, Before the three-dimensional tensor data are input into the CNN module, the following steps are further included: The three-dimensional tensor data is cut into local matrix slices of size C x C with the center of each grid x i as the center, where the clipping parameter C is set to 30-35.

3. The method of evaluating the quality of ploughing according to claim 1, characterized in that, The land grid data are input into the backbone network; the backbone network performs contrastive learning on features of the land grid data, comprising the following steps: Distances between input samples and anchor samples are measured based on cosine similarity in the backbone network; a loss function of the contrastive learning is set as: When the input samples and the anchor samples belong to the same category, the input samples and the anchor samples are made close to each other by minimizing the loss function: When the input samples and the anchor samples belong to different categories, the input samples and the anchor samples are separated by minimizing the loss function: wherein, represents a loss function, represents a cosine similarity, represents a contrastive learning loss function value of the i-th sample, represents a contrastive learning loss function value of a sample whose i-th prediction value is 1, represents a contrastive learning loss function value of a sample whose i-th prediction value is 0, represents a feature vector of the i-th input sample, represents a feature vector of an anchor sample, represents a prediction value of the i-th input sample, and represents a threshold value for controlling whether to calculate a contrastive learning loss value for different categories of input samples and anchor samples.

4. An evaluation device for the quality of a ploughed field, characterized in that The evaluation method is executed, comprising the following steps: A data acquisition module is configured to acquire land grid data of the cultivated land to be evaluated, and to associate the land grid data with geographical attributes, so that the land grid data are converted into three-dimensional tensor data; A feature extraction module is configured to perform contrastive learning on features of the land grid data by the backbone network, to obtain context features of the land grid data; to extract potential local features in the three-dimensional tensor data from the three-dimensional tensor data by the CNN module, to obtain spatial features; to combine the spatial features with the context features, to obtain comprehensive features; and to extract topological attribute features in the three-dimensional tensor data from the three-dimensional tensor data by the GNN module. The quality evaluation module is used for obtaining a spatial prediction value of the cultivated land to be evaluated from the overall feature; a topological prediction value of the cultivated land to be evaluated is obtained by predicting the topological attribute feature; and the spatial prediction value and the topological prediction value are fused to obtain an evaluation result of the quality of the cultivated land to be evaluated.

5. A computer device, comprising: The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the steps of the method in any one of claims 1 to 3.

Images