Maritime Target Detection Of Intelligent Ship Based On Faster R-CNN

摘要

Maritime target detection is important part of intelligent ship’s perceptual system, Traditional method of extracting artificial features are inefficient and has poor generalization, this paper proposes deep learning to automatically acquire deep features of other targets, Faster R-CNN is adopted for target’s recognition and location; Resnet will replace VGG16 as the main framework of detection algorithm; What’s more, in order to improve the detection effect of the model in complex marine environment, it combines hard example mining. Then the model was trained and tested by self-made Pascal VOC2007 dataset. The experimental results show that the method can effectively identify the targets of different types of ships and has higher accuracy of detection.