A Multi-Scale Spatial Attention Region Proposal Network for High-Resolution Optical Remote Sensing Imagery

Ruchan Dong,Yan Zhang,Weiyan Shen,Licheng Jiao,Jin Zhao

doi:10.3390/rs13173362

Ruchan Dong, Yan Zhang + Show 3 more

Open Access

https://doi.org/10.3390/rs13173362

Copy DOI

Abstract

Deep convolutional neural networks (DCNNs) are driving progress in object detection of high-resolution remote sensing images. Region proposal generation, as one of the key steps in object detection, has also become the focus of research. High-resolution remote sensing images usually contain various sizes of objects and complex background, small objects are easy to miss or be mis-identified in object detection. If the recall rate of region proposal of small objects and multi-scale objects can be improved, it will bring an improvement on the performance of the accuracy in object detection. Spatial attention is the ability to focus on local features in images and can improve the learning efficiency of DCNNs. This study proposes a multi-scale spatial attention region proposal network (MSA-RPN) for high-resolution optical remote sensing imagery. The MSA-RPN is an end-to-end deep learning network with a backbone network of ResNet. It deploys three novel modules to fulfill its task. First, the Scale-specific Feature Gate (SFG) focuses on features of objects by processing multi-scale features extracted from the backbone network. Second, the spatial attention-guided model (SAGM) obtains spatial information of objects from the multi-scale attention maps. Third, the Selective Strong Attention Maps Model (SSAMM) adaptively selects sliding windows according to the loss values from the system’s feedback, and sends the windowed samples to the spatial attention decoder. Finally, the candidate regions and their corresponding confidences can be obtained. We evaluate the proposed network in a public dataset LEVIR and compare with several state-of-the-art methods. The proposed MSA-RPN yields a higher recall rate of region proposal generation, especially for small targets in remote sensing images.

Highlights

Deep convolutional neural networks have promoted tremendous advances in computer vision, in the field of object detection
This paper proposes the selective strong attention maps model (SSAMM) that can adaptively select the appropriate number of windows based on the feedback of the system
We introduced a novel multi-scale spatial attention region proposal network (MSA-RPN) network for region proposal generation of high-resolution remote sensing images

Summary

Introduction

Deep convolutional neural networks have promoted tremendous advances in computer vision, in the field of object detection. Two networks have attracted the most attentions of object detection researchers: the Faster RCNN [1], the representative of two-stage algorithms, and YOLO [2], a typical case of one-stage algorithms. The two-stage algorithms usually present high recognition rates, and have been extensively used for detection on remote sensing images. High-resolution remote sensing images usually cover vast lands and oceans with targets in various sizes, so identifying targets in them is quite challenging. The two-stage algorithms usually generate object proposals first, that is, to localize target areas. Object proposals have become the critical issue in detection tasks on remote sensing images

Methods

Results

Conclusion