Remote Sensing Image Target Detection And Application Based On Hash Learning

As an important part of remote sensing image(RSI)processing,object detection is of great significance in both military and civil fields.However,due to the recent advances in satellite technology,the resolution of RSI has increased dramatically,which brings a great challenge to the efficiency of object detection in remote sensing images.By mapping data into binary codes,hashing can significantly reduce the storage cost and improve the computing efficiency.This paper intensive studies the general process of object detection in RSI and the principle of current mainstream learning th hash methods,and then introduces hashing concept for object detection in RSI to improve the detection speed.Moreover,this paper makes improvements for the current mainstream learning to hash methods according to the characteristics of remote sensing object.The major innovative works are as follows:(1)This paper proposes a novel hashing approach called Rotation-invariant Discrete Hashing(RIDISH)to effectively handle the problem of object rotation variations in RSI.While various hashing methods have been proposed,most of them only demonstrate their effectiveness on nature scene images.Considering the characteristics of the remote sensing objects,in this paper,we improve the current mainstream hashing method and propose a new hashing method called Rotation-invariant Discrete Hashing(RIDISH),in which way the learned binary codes can be rotation-invariant.Empirical results demonstrate the accuracy and robustness of our approach.What's more.Our RIDISH can achieve satisfactory classification accuracy with shorter hash codes.(2)This paper proposes a novel object detection method based on RIDISH to significantly improve the detection speed.This method can be divided into four stages:region proposal,RIDISH classification,SVM classification and non-maximum suppression,in which the core step is RIDISH classification.The experimental results show that RIDISH can quickly exclude about 98%of the non-object proposals,while remaining about 2%of the proposals for further classification,so that the detection speed can be significantly improved on the premise of ensuring the accuracy of detection.(3)This paper designs and implements an object detection system based on learning to hash methods.The system is divided into three functional modules:data selection module,result display module and performance evaluation module,which are used for selection of remote sensing image information and hashing information,visualization of detection results at various stages,and performance evaluation of test results respectively.Using this system,we can clearly show the detection results in different stages of detection procedure in remote sensing images,and make a comparative analysis of the performance of object detection in remote sensing images based on different hashing methods.