Remote Sensing Image Analysis of Regional Ship Targets Using a Deep Learning Algorithm
Bing Ou and Jingjing Yang
Quickly identifying ships in the dock with remote sensing images enables more effective management of the dock. This paper briefly introduced the detection of ship targets in remote sensing images and used the Faster-regional convolutional neural network (RCNN) algorithm to identify ships in remote sensing images. The Faster-RCNN algorithm was improved to improve its performance of target detection. When using a regional proposal network (RPN) to calculate the candidate target frame, the improved algorithm selected it from feature maps of convolution layers at different scales. The improved Faster-RCNN algorithm was compared with traditional Faster-RCNN and support vector machine (SVM) algorithms in the simulation experiment. The results showed that the improved Faster-RCNN detection algorithm was more accurate in positioning and recognizing categories; the improved Faster-RCNN detection algorithm had higher accuracy in identifying ships and showed a smaller decrease with the increase of the threshold value of IOU .
Keywords: deep learning, ship recognition, target detection, remote sensing image