A Multiscale Object Detection in Remote Sensing Images based on Human Visual Attention Mechanism

Bing He, Weihua Dong, Huiping Liu, Tianyu Yang, Yilong Liu

Keywords: Remote sensing images; Object detection; Eye tracking; EEG; Deep learning

TL;DR: We investigate the visual attention of experts using eye tracking and EEG, and propose a multi-scale object detection method based on human visual attention mechanism.

Abstract: With the development of automatic remote sensing image interpretation methods, many remote sensing target detection methods based on deep learning have been proposed. However, multi-scale remote sensing target detection still suffers from slow training computation speed, low detection accuracy, and difficult feature extraction. In order to solve these problems, we investigate the visual attention and cognitive process of remote sensing interpretation experts using eye tracking (ET) and electroencephalogram (EEG), and propose a multi-scale object detection method based on human visual attention mechanism. We redesign the candidate frame extractor, including top-down and bottom-up modules. Experimental results on three public datasets show that our method is more accurate than the baseline algorithm and effective for multi-scale remote sensing images.

Submission Number: 12