A Convolutional Neural Network for Ship Targets Detection and Recognition in Remote Sensing Images

摘要

According to the target direction, target size, shooting angle and scene diversity in remote sensing image, the target detection and recognition accuracy of remote sensing image is not high. This paper adopts a convolutional neural network(CNN) for ship targets detection and recognition in Remote Sensing Images, which is based on Faster R-CNN model. In our model, the network is optimized for the characteristics of ship targets in remote sensing images, use the multi-level proposal region extraction and multi-level feature fusion technology to construct the CNN model for ship target detection and recognition. The experimental results show that the network model has strong robustness and has a high accuracy at ship targets detection and recognition in remote sensing image.