Method of automatic construction of remote-sensing image segmentation reference library and system thereof

Abstract

The invention relates to the field of remote-sensing image segmentation, and discloses a method of automatic construction of a remote-sensing image segmentation reference library. The method of automatic construction of the remote-sensing image segmentation reference library includes: step 1, screening public-source vector data to remove ground objects of which ground object areas are smaller thana preset area; step 2, carrying out registration superposition on a remote-sensing image and public-source vector data after screening to obtain A; step 3, clipping the remote-sensing image in the Ainto a plurality of preset-pixel-size small-range image blocks, setting the small-scale image blocks as units, judging, through image coordinates, whether the public-source vector data are located inimage block ranges, and generating masks of which sizes are the same as sizes of the original image blocks; and step 4, combining the masks and the original image to obtain the remote-sensing image segmentation reference library. According to the method, problems that in the prior art, efficiency of constructing reference libraries by manual labeling is low, and scales are smaller at the same timeare overcome, and automatic construction of the remote-sensing image segmentation reference library is realized.

Description

technical field [0001] The invention relates to remote sensing image segmentation, in particular to a method for automatically constructing a remote sensing image segmentation reference library and a system thereof. Background technique [0002] Remote sensing image segmentation is the basis for automatic extraction of image information, middle and high-level image analysis and understanding, and is of great significance. Traditional remote sensing segmentation technology usually uses artificially designed feature extractors and thresholds to semi-automatically extract target information. The obvious defect lies in the uncertainty and limitations of artificially designed features. With the popularity of high-resolution remote sensing data and the high-precision recognition rate of image recognition models based on deep learning, how to effectively combine deep learning models with remote sensing data for ground object recognition has been a hot issue in the field of remote s...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to provide a method for automatically constructing a reference library for remote sensing image segmentation. This method overcomes the problems of low efficiency and small scale in the prior art of manual annotation and construction of a reference library, and realizes a remote sensing image segmentation reference library. automatic build

Images