Remote sensing image fishpond extraction method based on row-column self-attention full convolutional neural network

Abstract

The invention discloses a remote sensing image fishpond extraction method based on a self-attention full convolutional neural network, and the method comprises the basic steps: calculating an NDWI through employing a remote sensing image, carrying out the preliminary classification through employing the NDWI, and then manufacturing a remote sensing image sample set and a calibration sample set; constructing a self-attention full convolutional neural network and training a model, and predicting the remote sensing image by using the trained model to obtain probability distribution of each pixelcategory of the remote sensing image; and finally, fusing the probability distribution of each pixel category with an NDWI classification result to obtain a classification result. On the basis of using a full convolutional neural network, a row-column bidirectional GRU self-attention model is designed, the method has the characteristics of high accuracy and complete edge for fishpond extraction ina remote sensing image, and the method can be applied to fishpond automatic extraction, change detection and the like.

Description

technical field [0001] The invention belongs to the field of remote sensing target recognition, in particular to a method for extracting fish ponds from remote sensing images based on row-column self-attention full convolutional neural networks. Background technique [0002] In the interpretation of remote sensing images, the most important thing is the identification of image information. Human perception of external vision is a unified whole, including the complete perception of the shape, size, color, distance and other properties of each object in the scene according to the precise time and space orientation. Remote sensing information mainly includes the spectral information of ground objects. However, different objects are also clearly reflected in remote sensing images due to their different structural shapes and spatial positions. When people make visual judgments, in addition to feeling the difference in hue and hue, they can also obtain a lot of information throug...

AI Technical Summary

Problems solved by technology

Therefore, the boundary of fish ponds has a more regular shape, and due to various factors such as narrow edges and indeterminate shapes, traditional classification algorithms extract fish pond edges that are blurred and incomplete, and cannot automatically obtain a more regular geometric shape

Images