| 基于通道剪枝的SAR图像舰船检测优化算法 |
| |
| 引用本文： | 周舟,王海鹏,徐丰,张志俊,王志诚.基于通道剪枝的SAR图像舰船检测优化算法[J].上海航天,2020,37(4):48-54. |
| |
| 作者姓名： | 周舟 王海鹏 徐丰 张志俊 王志诚 |
| |
| 作者单位： | 复旦大学 电磁波信息科学教育部重点实验室，上海200433;上海无线电设备研究所，上海201109 |
| |
| 基金项目： | 国家自然科学基金资助项目(61571132);上海航天科技创新基金资助项目(SAST2017-078，SAST2016-061) |
| |
| 摘 要： | 近几年,随着深度学习的发展,基于深度学习的目标检测算法开始应用于合成孔径雷达(SAR)图像中的舰船检测。但深度学习模型结构复杂,参数量与计算量巨大,无法应用到星载处理器的实时处理中。本文提出一种结合了Faster-RCNN和卷积通道剪枝的舰船检测方法,在保证检测精度不受较大影响的情况下,剪除卷积层中的部分参数,提高检测效率。实验表明:经过剪枝优化的Faster-RCNN舰船检测模型中的参数量降低了约56%,而推理时间减少了约51%,同时精度下降仅有1.9%。这给未来在星载处理器上部署舰船检测算法提供了新的思路。
|
| 关 键 词： | 合成孔径雷达图像 舰船检测 深度学习 通道剪枝 |
| 收稿时间： | 2019/10/12 0:00:00 |
| 修稿时间： | 2019/10/28 0:00:00 |
Optimization Algorithm for Ship Detection in SAR Images Based on Channel Pruning |
| |
| ZHOU Zhou,WANG Haipeng,XU Feng,ZHANG Zhijun,WANG Zhicheng.Optimization Algorithm for Ship Detection in SAR Images Based on Channel Pruning[J].Aerospace Shanghai,2020,37(4):48-54. |
| |
| Authors: | ZHOU Zhou WANG Haipeng XU Feng ZHANG Zhijun WANG Zhicheng |
| |
| Affiliation: | Key Laboratory of Electromagnetic Wave and Information Science, Fudan University, Shanghai 200433, China;Shanghai Radio Equipment Research Institute, Shanghai 201109, China |
| |
| Abstract: | With the development of deep learning in recent years, the target detection algorithm based on deep learning has been applied to the ship detection in Synthetic Aperture Radar (SAR) images. However, the deep learning model cannot be applied to the real-time processing of space onboard processors due to its complex structure, a large number of parameters, and a huge amount of calculation. In this paper, a ship detection method combining faster-RCNN and convolutional channel pruning is proposed. In order to improve the detection efficiency, without a drop in the accuracy, parts of the convolution filters are cut off. The experimental results show that the number of parameters in the faster-RCNN ship detection model after channel pruning optimization decreases by about 56%, the reasoning time decreases by about 51%, while the detection accuracy decreases by only 1.9%. This provides a new idea for the deployment of ship detection algorithms on space borne processors in the future. |
| |
| Keywords: | Synthetic Aaperture Radar (SAR) image ship detection deep learning channel pruning |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《上海航天》浏览原始摘要信息 |
|
点击此处可从《上海航天》下载全文 |
|
