Research On Detection Algorithm Of Long-range Target At Sea Based On Panoramic Vision

In response to China 's call to build marine power country,resolutely protect China 's maritime areas,timely completion of maritime rights and surface rescue mission,so there is a need to improve our ability to monitor and control the ocean,focusing on the development of sea target monitoring equipment,it is of great theoretical and practical significance to study the algorithm of surface target detection.Catadioptric panoramic vision system can obtain high-resolution scene information at one time,and meet the requirements of maritime monitoring field with long distance and large viewing angle.However,panoramic sea images collected by the system are visual images.Detection technology of small target at sea based on visual images is not mature currently.So research on detection algorithm of long-range target at sea based on panoramic vision is proposed in this paper.The existence of panoramic equipment area increases the complexity of panoramic sea images' background;The particularity of the imaging principle of panoramic vision system leads to sea-sky-line to be elliptical and causes the existing linear sea-sky-line detection methods being not used directly.Considering the above two factors,the basic idea of this paper is: Firstly,panoramic sea images collected by the panoramic vision system is processed,including extracting and breaking up panoramic equipment area;then extract panoramic sea-sky-line;finally long-range small target at sea is detected on the basis of determining the sea-sky-line zone.Firstly,the preprocessing of panoramic images is used to eliminate the interference information in the images.This paper takes full of median filter to suppress noise which is introduced in the process of panoramic image acquisition,then panoramic images are sharped by the Laplace operator,which makes the edge information of sea-sky-line and small target more obvious.Panoramic equipment area is more obvious and easier to cause human visual attention than other area in images,therefore panoramic equipment area interference detection based on saliency detection is designed,and the four most classical detection methods--FT method,LC method,HC method and AC method are used to extract and segment panoramic equipment area.The effectiveness of the proposed algorithms are proved by experimental results.Secondly,the method of detecting sea-sky-line in panoramic images is designed.In order to apply the existing linear sea-sky-line detection method,Progressive Probability Hough Linear Transform is used to detect sea-sky-line on cylinder images of panoramic images.Because of the limitation of the above method,this paper proposes two methods to directlydetect sea-sky-line in panoramic sea images.One is that panoramic sea-sky-line is approximately circular,and sea-sky-line detection based on improved random Hough transform is proposed.The other is that panoramic sea-sky-line is oval in fact,based on this characteristic,this paper designs an improved Snake method by setting adaptive initial contour line to detect sea-sky-line directly from the panoramic images.The effectiveness of the proposed algorithms are proved by large amount of experimental results.Finally,long-range small targets at sea are detected in the determined sea-sky-line zone.The detection of small targets in the sea-sky-line zone can reduce the search range,reduce computation load and improve detection efficiency,so the detection of small target based on visual saliency and long-range small targets at sea detection based on 9/7 lifting wavelet are designed in this paper.A large number of experimental results show that the two designs are able to successfully detect long-range small targets.Long-range small targets detection system based on panoramic vision is build based on Visual C++6.0 platform.