Detection of Small Targets in Remote Sensing Images Using Improved Full Convolutional Neural Network
A novel method based on improved full convolutional neural network is proposed for the detection of small targets in re-mote sensing images.Firstly,the basic concepts and models of hierarchical probability graph model and deep learning are analyzed.Then,the recursive steps of obtaining the posterior marginal mode of the layered Markov random field in the layered probability graph model are proposed.In addition,the full convolutional neural network and hierarchical probability graph model are combined to realize the improvement of the full convolutional neural network.Finally,a novel method for small target detection in remote sens-ing images is constructed based on this improved full convolutional neural network.In addition,random forest technique is used to es-timate the posterior probability of each class and resolution from the classification learning samples.The proposed method is com-pared with the other four methods based on the processing of satellite data set in a certain area.The simulation results show that com-pared with other methods,the proposed method has higher detection accuracy for small targets in remote sensing images such as low vegetation,vehicles and trees.
万方数据
维普
