A remote sensing image retrieval method and system based on unsupervised feature learning

Abstract

The invention provides a remote sensing image retrieval method and system based on unsupervised feature learning, including extracting salient maps for each image in the retrieval image library, and obtaining corresponding binarized salient maps according to the segmentation threshold of the salient maps; In each image, according to the corresponding binarized saliency map, the salient area is segmented by mask operation; image blocks of the same size are extracted from the salient area of ​​the image to construct training samples, and the samples are trained by unsupervised feature learning method to learn the features of the image ; Finally, perform image retrieval. The invention extracts image blocks of the same size from the salient areas of the image to construct training samples for unsupervised feature learning, which makes up for the traditional defect of directly performing random sampling on the original image, not only conforms to the visual attention characteristics of the human eye, but also can directly It reflects people's retrieval needs, and avoids the complicated feature extraction process while ensuring retrieval accuracy.

Description

technical field [0001] The invention belongs to the technical field of image processing, and relates to a remote sensing image retrieval method and system based on unsupervised feature learning. Background technique [0002] With the development of remote sensing technology and sensor technology, the spatial resolution of available remote sensing images is increasing day by day, and the amount of data is increasing at an alarming rate. Massive remote sensing data provide rich data sources for social and economic development and scientific research, but on the other hand, it also brings great challenges to people, that is, how to realize the effective management and efficient utilization of massive image data. Due to the current image data processing and analysis technology is still in the development stage, processing and analysis capabilities are limited, making the effective management of remote sensing image data far behind the speed of data growth. In addition, when tar...

AI Technical Summary

Problems solved by technology

The existing methods usually randomly select a certain number of image blocks from the original large image to construct training samples, but due to the random selection position, it is difficult for the image blocks to completely include the specific retrieval objects on the image.

Images