Research On Ship Target Detection Technology Based On Visible-light-remote Sensing Image

With the improvement of strip width and image resolution of optical remote sensing satellites and the shortening of acquisition period of remote sensing image,how to extract valuable information from marine remote sensing images,promote economic development and ensure national defense has become research hotspots in remote sensing image understanding and analysis.The use of technical means to detect manmade ship targets from massive image data has enormous application and strategic needs in the civil and military fields.Fast and accurate positioning and detection of ships from complex remote sensing image scenes remains a problem in the context of background disturbances such as waves,clouds and islands.Aiming at the existing data resources,this paper focuses on the detection methods and existing problems of ship targets in visible-remote-sensing images,analyzes the characteristics of ship targets in high-resolution remote sensing images,and studies and expounds the key technologies of ship detection.Under the background interference of sea waves and cloud,the target localization and detection of suspected ships in largescale and high-resolution visible-remote-sensing images were completed,and some research results and engineering reference value were obtained.The main research work of this paper is as follows:1.Focusing on the detection of ship targets,the characteristics of visible-lightremote sensing images and ship targets have been studied and analyzed.The research of key technologies such as region of interest localization,remote sensing image segmentation,targets slice feature description and classification and identification have been completed,which serve as the basic principle support and theoretical basis of the ship detection algorithm in ocean area.2.Aiming at the problem of screening and localization of suspected ship targets in visible-light-remote sensing images in the background of clouds and waves,a fast localization method for candidate regions based on visual attention calculation model is proposed.Among them,the candidate target localization method based on spectral residual model solves the problem of rapid positioning and incomplete segmentation of ship targets under cloud cover.The candidate target localization method based on maximum symmetric surround model solves the problem of rapid positioning and incomplete segmentation of ship targets under the background of ocean waves.3.Aiming at the problem of ship target detection in visible-light-remote sensing image under ocean background.According to the shape characteristics of ship,the gradient direction histogram features are improved,and the gradient direction feature are designed.Meanwhile,the variance of the uniform local binary pattern(LBP)and the gray level co-occurrence matrix feature are used to describe the texture information of the ship.Furthermore,a fusion feature detection method based on AdaBoost classification is proposed,which effectively eliminates the interference of non-ship targets in target slice,reduces the dimension of the feature vector,reduces the computational complexity,and solves the problem of low recognition accuracy.The experimental results show that the proposed algorithm can accurately extract the quantity and position information of ship targets in remote sensing images,and can meet the actual engineering needs in terms of detection time and accuracy.