Small Target Detection Based On The Extraction Of Panoraminc Sea-sky-line

As there is a long coastline and vast ocean territory in our country, marine rights enforcement and maritime search mission are heavy, so it's very import to develop advanced sea surveillance equipment and study sea targets detection method. Using panoramic vision system in the sea surveillance, on the one hand, it can meet the demand of marine environment with large field view, full range and remote monitoring. On the other hand, it can effectively reduce the number of monitoring devices and hardware costs. However, this system is currently lack of technical support to target detection. Due to that the targets coming from far to near are first appear on the sea-sky-line, so detecting the small targets in sea-sky-line area can provide more reaction time for shooting forensics. Therefore, small target detection algorithm based on panoramic sea-sky-line extraction is studied in this paper.Because the imaging principle of catadioptric panoramic vision system is special, so it not only changes the straight sea-sky-line in conventional vision system to oval sea-sky-line,but also increase a lot of panoramic equipment imaging. As to the particularity of panoramic sea image and complexity of background, in this paper, it fist extracts and eliminates the panoramic equipment zone interference which has an adverse effect on sea-sky-line extraction and target detection, then extracts the panoramic sea-sky-line, and finally detects small targets at sea-sky-line area. So there are three core contents in this paper, which are panoramic equipment interference excluding, panoramic sea-sky-line extraction and small target detection at sea-sky-line area.Firstly, excluding method of panoramic equipment interference is studied. According to the panoramic equipment belongs to man-made object, whose texture feature is different from sea and sky and other natural scenery, an interference extraction and excluding algorithm of panoramic equipment based on fractal dimension is designed in this paper; according to the panoramic equipment interference in panoramic image is more obvious and easier to cause visual attention, an extraction algorithm based on visual saliency map is designed. And finally the effectiveness of these algorithms is verified by experiments.Secondly, panoramic sea-sky-line extraction algorithm is studied. On the basis of eliminating the interference caused by panoramic equipment area, aiming to the feature that panoramic sea-sky-line is approximately a circle, a panoramic sea-sky-line extraction algorithm based on improved gradient Hough circle transform is designed in this paper;aiming to the feature that the panoramic sea-sky-line has an obvious edge outline, a panoramic sea-sky-line extraction algorithm based on improved active contour model is proposed in this paper; aiming to the feature that the gradient energy of panoramic sea-sky-line is relatively large in panoramic image, a panoramic sea-sky-line extraction algorithm based on improved Seam Carving is put forward in this paper. And finally the effectiveness of this three algorithms are verified by experiments, and are compared with the existing algorithms.Finally, the small target detection algorithm is studied. On the basis of the panoramic sea-sky-line extraction, when the small targets are regarded as the singular point of image signal, a small target detection algorithm at sea-sky-line area based on mutual lifting wavelet energy is designed in this paper; using the dark channel theory which is zoomed in small target, a small target detection algorithm at sea-sky-line area based on priori dark channel theory is put forward in this paper. Then a large number of experiments are done to verify the effectiveness of this algorithm, and the superiority of the algorithm is also be proved by comparing with the existing algorithm.