科研之友
微信
新浪微博
Facebook
分享链接
摘要
Faster RCNN is a region proposal based object detection approach. It integrates the region proposal stage and classification stage into a single pipeline, which has both rapid speed and high detection accuracy. However, when the model is applied to the target detection of remote sensing imagery, faced with multi-scale targets, its performance is degraded. We analyze the influences of pooling operation and target size on region proposal, then a modified solution for region proposal is introduced to improve recall rate of multi-scale targets. To speed up the convergence of the region proposal networks, an improved generation strategy of foreground samples is proposed, which could suppresses the generation of non-effective foreground samples. Extensive evaluations on the remote sensing image dataset show that the proposed model can obviously improve detection accuracy for multi-scale targets, moreover the training of the model is rapid and high-efficient.
- 出版日期2019-4
- 单位中国人民解放军国防科学技术大学
