Ship Detection Based on Faster R-CNN Network in Optical Remote Sensing Images
A ship detection model based on Faster R-CNN is proposed for ship detection tasks in optical remote sensing images.Deep convolutional neural network could replace traditional manual design feature to extract ship features automatically and quickly from makes the detection performance of ship no longer relying on the design of artificial features.This paper proposes a strategy that combines the model with two different size of convolution neural networks respectively.Experiments on datasets HRSC16 verify the models detection capabilities and the mean average precision can achieve 78.2%.For the problem of low recall rate in the detection of adjacent vessels,this paper adopts the Soft-NMS method.Compared with the traditional NMS,the Soft-NMS method can electively improve the model detection performance to 80.1%.At the same time,it also shows that the model we proposed is a robust model and has a certain degree of generalization ability.
Ship detection Faster R-CNN Deep learning Soft-NMS
Min Zhai Huaping Liu Fuchun Sun Yan Zhang
The State Key Laboratory of Astronautic Dynamics,Xian Satellite Control Center,Xian,China;The Depa The Department of Computer Science and Technology,Tsinghua University,Beijing 100084,China Xian Satellite Control Center,Xian 710043,China
国际会议
江苏镇江
英文
22-31
2019-09-20（万方平台首次上网日期，不代表论文的发表时间）