FESSD: Feature Enhancement Single Shot MultiBox Detector Algorithm for Remote Sensing Image Target Detection

Abstract

Automatic target detection of remote sensing images (RSI) plays an important role in military surveillance and disaster monitoring. The core task of RSI target detection is to judge the target categories and precise location. However, the existing target detection algorithms have limited accuracy and weak generalization capability for RSI with complex backgrounds. This study presents a novel feature enhancement single shot multibox detector (FESSD) algorithm for remote sensing target detection to achieve accurate detection of different categories targets. The FESSD introduces feature enhancement module and attention mechanism into the convolution neural networks (CNN) model, which can effectively enhance the feature extraction ability and nonlinear relationship between different convolution features. Specifically, the feature enhancement module is used to extract the multi-scale feature information and enhance the model nonlinear learning ability; the self-learning attention mechanism (SAM) is used to expand the convolution kernel local receptive field, which makes the model extract more valuable features. In addition, the nonlinear relationship between different convolution features is enhanced using the feature pyramid attention mechanism (PAM). The experimental results show that the mAP value of the proposed method reaches 81.9% and 81.2% on SD-RSI and DIOR datasets, which is superior to other compared state-of-the-art methods.