Research On Remote Sensing Image Target Detection Method Based On Local Feature

In recent years,with the development of remote sensing satellite technology,the remote sensing images in military and civilian fields of application gradually expanded,researching on remote sensing image target detection method is more and more important to scholars.At the same time,the image resolution has gradual increased,the information of remote sensing image increased significant with the size become bigger.However,people in the entire image or the whole video is usually interested in a small part,use of the global features to detect the target in the practical application has many limitations,and the local features has the advantage of performance the information for sensing target detection under complex background,and provides an effective way.In order to enhance the efficiency and reliability of the automated processing to remote sensing information,and to strengthen the ability of military investigation,this paper focuses on remote sensing images in the detection and identification of several typical targets,such as mass target,array target,and port target.In order to target visual features and the local characteristics of the spatial relations as the breakthrough point,the key technology of system is researching for the detection of target recognition.Based on the proposed aiming at different types of sensitive target detection in remote sensing image recognition method,improve the efficiency of the remote sensing detection target recognition system and adaptability,and valuable research results have been achieved.This article mainly research of remote sensing image target detection method based on local visual characteristics,carried out the following work.1.First of all,the article analyzes the characteristics of human visual physiological structure and remote sensing target,the target of remote sensing image interpretation for the research of the system,and the elements of interpretation,the method and development trend are summarized.Then,this paper proposes a remote sensing image target edge feature detection based on visual perception.Through the study of the theory of visual perception system research and analysis of its characteristics,digging out the potential relationship between the visual perception and remote sensing target edges.Combining with the characteristics of remote sensing image first,and enhance the edge character of remote sensing image target;Then,to the traditional edge detection algorithm was improved,with the introduction of high and low threshold enhanced target edge and the edge points in the contrast,preliminary by using image segmentation,edge detection,and pseudo edge again by removing operation;Finally,the remote sensing image edge character description is made.According to the image enhancement methods and simulation experiment,verified the effectiveness of the improved image edge detection method,and the effectiveness of edge feature descriptor.2.This paper proposes a mass target detection method based on Quasi-circular characteristics,emphasis was on the detection of tank target identification problem.The experimental result shows that to compare with other methods,the proposed detection method to improve the accuracy greatly,and can quickly positioning terminal,based on the test result,it is suitable for remote sensing images of different resolution.3.This paper proposes an array detection target recognition method based on invariant features,focuses on the background of land the plane target detection identification problem.The experimental result shows that the method is simple in implementation process,the target background interference has high robustness,less computation and better effect.4.Automatic inshore ship detection from remote sensing imagery has many important applications,such as ship change detection and harbor dynamic surveillance.In this paper,we put forward a method on inshore ship detection using local salient characteristics.Experimental results indicate that using the proposed method to detect the inshore ship is more effective and robustness to adapt the complex background and the mooring orientation,the detection result is more accuracy compared with traditional methods,and the recognition rate is very high.