Remote Sensing Image Target Detection And Application Based On Optimal Projected Discrete Hashing

Remote sensing images object detection is an important part of remote sensing images task,and it has been well applied in many scenes.With the development and progress of remote sensing images acquisition technology,the amount of remote sensing images data increases rapidly and the resolution of the acquired remote sensing images is getting higher and higher,nowadays,the object detection of remote sensing images is facing the storage and real-time problems.With its high computational efficiency,hash learning is used to deal with large-scale data.It maps high-dimensional data into binary code,and then performs other tasks,which is very effective in improving the speed of big data tasks.This thesis mainly improves the classification method of hash learning based on the characteristics of remote sensing images,and proposed the object classification of remote sensing images based on optimized projection supervised discrete hashing,and further improved the method according to the problems encountered,and then applied it to object detection task of remote sensing images.The work of this thesis mainly includes:（1）The thesis proposed the remote sensing object classification based on optimized projection discrete hashing,considering the supervised hashing in the learning process is not fully using the original remote sensing image characteristics information,this thesis adds an optimized projection constraint when using supervised information to learn hash code and hash function,so that keep the generated hash code preserve the similarity of label information and the characteristics of the original image information.It is verified through experiments that the optimized projection discrete hash method（OPSDH）can improve the classification precision of remote sensing images compared with the previous supervised hashing methods.（2）The thesis studied an improved object classification method of remote sensing images based on optimized projection discrete hashing,the object classification accuracy improved as hash code length increases and short hash code in learning process of information loss serious,improved OPSDH method using hash code and the original space characteristic information reconstruction to reduce information loss in the process of generating short hash code.Compared with the current efficient hashing method,the improved OPSDH can significantly improve the classification accuracy of short hash code while ensuring the classification effect of long hash code.（3）The thesis studied object detection method of remote sensing images based on optimized projection discrete hashing,the improved OPSDH method was applied to object detection of remote sensing images to improve the speed of the remote sensing object detection,studied the object detection process of remote sensing images,we use OPSDH classification to remove most of the background information,thereby greatly accelerated the speed of detection.Experimental verification results show that OPSDH also has obvious advantages in object detection compared with other hash learning classification methods.（4）The thesis implemented an object detection system of remote sensing images based on optimized projection discrete hashing,the system puts the proposed object detection method of remote sensing images into practical application,and it includes the data selection input module,the detection result presentation module and detection performance evaluation module,and we introduced and demonstrated the functional realization of each module in detail.