Generating Anchor Boxes Based on Attention Mechanism for Object Detection in Remote Sensing Images

Zhuangzhuang Tian,Ronghui Zhan,Jiemin Hu,Zhiqiang He,Wei Wang,Zhaowen Zhuang

doi:10.3390/rs12152416

Zhuangzhuang Tian, Ronghui Zhan + Show 4 more

Open Access

https://doi.org/10.3390/rs12152416

Copy DOI

Abstract

Nowadays, object detection methods based on deep learning are applied more and more to the interpretation of optical remote sensing images. However, the complex background and the wide range of object sizes in remote sensing images increase the difficulty of object detection. In this paper, we improve the detection performance by combining the attention information, and generate adaptive anchor boxes based on the attention map. Specifically, the attention mechanism is introduced into the proposed method to enhance the features of the object regions while reducing the influence of the background. The generated attention map is then used to obtain diverse and adaptable anchor boxes using the guided anchoring method. The generated anchor boxes can match better with the scene and the objects, compared with the traditional proposal boxes. Finally, the modulated feature adaptation module is applied to transform the feature maps to adapt to the diverse anchor boxes. Comprehensive evaluations on the DIOR dataset demonstrate the superiority of the proposed method over the state-of-the-art methods, such as RetinaNet, FCOS and CornerNet. The mean average precision of the proposed method is 4.5% higher than the feature pyramid network. In addition, the ablation experiments are also implemented to further analyze the respective influence of different blocks on the performance improvement.

Highlights

As an important tool of earth observation, remote sensing has been widely used in civil and military fields, such as geological monitoring, climatic forecast, ecological environment monitoring, and battle reconnaissance
The experimental results show that the proposed method has a significant improvement in the object detection task for remote sensing images compared with the conventional methods
All the experiments are conducted on a computer with a central processing unit (CPU) of Intel 6700K, a graphics processing unit (GPU) of NVIDIA GTX 1080Ti, and random access memory (RAM) of 32 GB

Summary

Introduction

As an important tool of earth observation, remote sensing has been widely used in civil and military fields, such as geological monitoring, climatic forecast, ecological environment monitoring, and battle reconnaissance. With the continuous development of remote sensing technology, rapidly increasing images make manual interpretation a tremendous task, and as a result, automatic interpretation has drawn considerable research attention. For optical remote sensing images, researchers have done a lot of work on object detection [1,2,3,4]. With the advancement of computer vision and machine learning, especially deep learning, object detection for natural images is more and more mature. Many researchers have drawn on the successful experience of natural images, and applied deep-learning-based methods to remote sensing images [5,6,7,8]

Methods

Results

Discussion

Conclusion