Land property detection method based on space-time adaptive fusion and geographic features

Abstract

The invention provides a land property detection method based on space-time adaptive fusion and geographic features, and relates to the technical field of image processing, and the method comprises the steps: obtaining a sampling image of a to-be-detected land parcel, land parcel attribute data and digital elevation model data; a Swin Transform model is combined with a land texture adaptive fusion module GTAF to extract image features from a sampling image, land parcel attribute data is converted into node and edge forms by constructing a land parcel relation graph to obtain attribute features, and land parcel topographic features are extracted from digital elevation model data by using a MobileNetV3 network; the image features include plot texture features, plot color features and plot structure features, and the attribute features include plot area features, plot coordinate features and plot type features; the terrain features of the land parcels comprise gradient features and slope direction features; and carrying out weighted fusion on the image features, the attribute features and the land parcel topographic features by using an adaptive mutual increase attention module Ada-MBA, and determining the land property of the to-be-detected land parcel based on the weighted fusion features.

Claims

1. A land property detection method based on spatiotemporal adaptive fusion and geographical features, characterized in that: include: Obtaining sampling images, land attribute data and digital elevation model data of the land parcel to be detected; The Swin Transformer model is combined with the land texture adaptive fusion module GTAF to extract image features from the sampled image. The land parcel attribute data is converted into nodes and edges by constructing a land parcel relationship graph to obtain attribute features. The MobileNetV3 network is used to extract land parcel terrain features from the digital elevation model data. The image features include land parcel texture features, land parcel color features, and land parcel structure features. The attribute features include land parcel area features, land parcel coordinate features, and land parcel type features. The land parcel terrain features include slope features and slope aspect features. The adaptive mutual attention module Ada-MBA is used to perform weighted fusion on the image features, the attribute features and the land terrain features, and the land nature of the land to be detected is determined based on the weighted fusion features.

2. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The step of obtaining a sampled image, land parcel attribute data, and digital elevation model data of a land parcel to be detected includes: Using external acquisition equipment to obtain sampling images of the land to be inspected at different time points and different shooting angles; Acquire the area data, coordinate data, and type data of the plot to be detected from the cadastral database, normalize the area data and the coordinate data, and mark the type data using one-hot encoding to obtain plot attribute data; Generate digital elevation model data based on surface elevation data collected by drones.

3. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The Swin Transformer model is combined with the ground texture adaptive fusion module GTAF to extract image features from the sampled image, including: The sampled image is converted into an RGB three-channel image and divided into four 4-sized blocks, resulting in a shape of 56 56 First image feature of 96; A four-stage structure is used to extract multi-scale features from the first image features, and the output shape is 7 7 Second image feature of 768; The second image feature is processed by global average pooling and a fully connected layer to generate the land parcel texture feature, the land parcel color feature and the land parcel structure feature.

4. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The step of converting the land parcel attribute data into the form of nodes and edges to obtain attribute features by constructing a land parcel relationship graph includes: The unique identifier of the plot to be detected is used as a node, and edges are obtained based on spatial adjacency and semantic similarity to construct a plot relationship graph; The GraphSAGE model is used to extract the plot area feature, the plot coordinate feature, and the plot type feature from the plot relationship graph.

5. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The method of extracting land terrain features from the digital elevation model data using the MobileNetV3 network includes: The digital elevation model data of different scales are normalized and data enhanced, and the number of channels is adjusted before inputting into the MobileNetV3 network to obtain the terrain features of the plot.

6. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The method of using an adaptive mutual augmentation attention module Ada-MBA to weightedly fuse the image features, the attribute features, and the land parcel terrain features, and determining the land nature of the land parcel to be detected based on the weighted fusion features, includes: Performing linear transformation and normalization on the image features, the attribute features, and the land parcel terrain features, and generating corresponding attention weights for each of them using a fully connected layer; The image features, the attribute features and the land parcel terrain features are weightedly fused based on the attention weights, and the land properties of the land parcel to be detected are determined based on the weighted fusion features.

7. The land property detection method based on spatiotemporal adaptive fusion and geographical features according to claim 1, characterized in that: The method of using an adaptive mutual augmentation attention module Ada-MBA to weightedly fuse the image features, the attribute features, and the land parcel terrain features, and determining the land nature of the land parcel to be detected based on the weighted fusion features, includes: Obtaining a first weighted fusion feature at a first time point and a second weighted fusion feature at a second time point; The land property change information of the to-be-detected land parcel is determined based on the norm difference between the first weighted fusion feature and the second weighted fusion feature.

8. A land property detection device based on spatiotemporal adaptive fusion and geographical features, characterized in that: include: Data acquisition module, feature extraction module and feature fusion module; among them, The data acquisition module is configured to acquire a sample image, land parcel attribute data and digital elevation model data of the land parcel to be detected; The feature extraction module is configured to extract image features from the sampled image using a Swin Transformer model combined with a land texture adaptive fusion module GTAF, convert the land parcel attribute data into nodes and edges to obtain attribute features by constructing a land parcel relationship graph, and extract land parcel terrain features from the digital elevation model data using a MobileNetV3 network; the image features include land parcel texture features, land parcel color features, and land parcel structure features; the attribute features include land parcel area features, land parcel coordinate features, and land parcel type features; and the land parcel terrain features include slope features and slope aspect features; The feature fusion module is configured to use an adaptive mutual attention module Ada-MBA to perform weighted fusion on the image features, the attribute features and the land terrain features, and determine the land nature of the land to be detected based on the weighted fusion features.

9. An electronic device comprising a processor and a memory; the memory stores a computer program, wherein: When the computer program is executed by the processor, the land property detection method based on spatiotemporal adaptive fusion and geographic features according to any one of claims 1 to 7 is implemented.

10. A computer storage medium, characterized in that A computer program is stored thereon, wherein when the computer program is executed by a processor, the land property detection method based on spatiotemporal adaptive fusion and geographic features as claimed in any one of claims 1 to 7 is implemented.

Images