Detection and Recognition of SAR Small Ship Objects Using Deep Neural Network

Qiuzhuang Yuan,Songjie Wei,Meilin Liu,Pengfei Jiang

doi:10.1051/jnwpu/20193730587

Abstract

Synthetic aperture radar(SAR) ship target detection plays an increasingly important role in marine monitoring. Aimed at the problems of recognizing small size of ship targets in SAR images and the inability of traditional methods to extract fine target features due to external disturbances, we propose an improved SAR small target detection model based on the deep learning technology. The proposed model mainly consists of two parts:region proposal network(RPN) and object detection network. Firstly, a CNN model is designed and trained to accurately identify small ship targets. Then, the model is used to initialize the parameters of the shared feature extraction layer. Last, we train the proposed object detection model using a self-collected Sentinel-1 SAR small target dataset. The experimental results show that the proposed target detection model has better detection and recognition performance and anti-interference ability for small ship scalable targets in SAR images, and has certain reference value for the research of small target detection in SAR images.

Highlights

快速、准确地检测,本文基于 faster R⁃CNN 设计了一个改进的目标检测模型。图 2 是本文设计的目标检测模型的网络结构,该模型由一个基于全卷积的区域建议网络( region proposal network, RPN) 和一个基于区域的目标检测网络 fast R⁃CNN 两部分组成。其中,区域建议网络主要用来高效地生成不同尺度和长宽比的候选区域,目标检测网络对区域建议网络提取的这些候选区域进行分类和位置的回归,最终通过非极大值抑制算法筛选出一个置信度最高的候选框作为检测结果。该模型使用本文设计的 CNN 分类
Aimed at the problems of recognizing small size of ship targets in SAR images and the inability of traditional methods to extract fine target features due to external disturbances, we propose an improved SAR small target detec⁃ tion model based on the deep learning technology
The proposed model mainly consists of two parts: region proposal network( RPN) and object detection network

Summary

Introduction

觉任务中取得了巨大成功。许多专家和学者开始利用深度神经网络来完成计算机视觉任务中的目标检测问题。 Girshick 等人将目标检测问题转化为分类问题,相继提出了 R⁃CNN[6] 和 fast R⁃CNN[7] 目标检测方法。由于 fast R⁃CNN 候选区域的提取仍然是在 CPU 上进行的,速度较慢,Ren 等[8] 提出了 faster R⁃CNN 方法,引入了区域建议网络( region proposal network, RPN) 取代传统的区域提取方法,借助 CNN 来完成候选区域的提取,并且将 RPN 和 fast R⁃CNN 共享全图卷积特征,大大提高了模型的检测效率。快速、准确地检测,本文基于 faster R⁃CNN 设计了一个改进的目标检测模型。图 2 是本文设计的目标检测模型的网络结构,该模型由一个基于全卷积的区域建议网络( region proposal network, RPN) 和一个基于区域的目标检测网络 fast R⁃CNN 两部分组成。其中,区域建议网络主要用来高效地生成不同尺度和长宽比的候选区域,目标检测网络对区域建议网络提取的这些候选区域进行分类和位置的回归,最终通过非极大值抑制算法 ( non⁃maximum suppression, NMS) 筛选出一个置信度最高的候选框作为检测结果。该模型使用本文设计的 CNN 分类为了解决多尺度检测的问题,文献[ 11] 引入了特征金字塔网络( feature pyramid networks, FPN) , 将 FPN 作为多尺度目标特征提取器,融合进 faster R⁃CNN 中,实现了对多尺度目标尤其是小目标的有效检测。因此,本文首次将 FPN 与 faster R⁃CNN 融合的算法应用到 SAR 图像舰船小目标检测中,以提高对舰船小目标的检测准确率。

Results

Conclusion