The Research On Lifejacket Detection And Tracking In Complex Ferry Environment

Ferries are the essential form of the travel for masses, especially peasants, in Hunan province. Whether the ferry equipment is complete is directly related to the life and property safety of the masses. In recent years, the government strengthens the ferries management, but traffic accidents still happened often. Water Transit Authority of Hunan province, therefore, put forward to need for the equipment can automatically analyze the video of illegal phenomenon, detect it and warn early. In this thesis, the author proposed a research on the lifejacket detection and tracking in complex environment. And combined it with pedestrian and boat detection and tracking algorithm to detect and judge illegal phenomenon.Lifejacket as a target, can be researched with the method of target detection and tracking. This thesis introduced the general target motion analysis method by four steps. Firstly, you need to make the judgment of the target representation in specific situations. There are many types of the representation method of target, this thesis mainly introduced methods based on shape and appearance. Secondly, specific features of image is extracted in representation method. And this paper introduced the common characteristics, such as, color, edge, optical flow, texture, etc. Thirdly, the thesis introduced how to detect the target in each frame. And introduces the detection methods in common use, such as, point detection, background deduction, image segmentation, supervised learning, etc. Finally, we introduced the target tracking method in common use.Proposed a lifejacket detection algorithm based on HSV color feature and contour area. Lifejacket is a special target, its color feature is very obvious. So in this thesis we covers the lifejacket detection algorithm combine with color feature and contour area. The color range of a number of lifejackets is statistic to find the sample color range of the lifejacket. To filter the image color which is not in the range of the sample color of lifejacket by establishing histogram, and obtain a connected binary image. Then the lifejacket is selected by calculating contour area, filtering the small area.Proposed a nearest neighbor matching tracking algorithm of the lifejacket. Firstly, a virtual test box is created in image in order to processing the image quicker. The virtual test box create on the road which need be the only way that ships enter ferry. Find the pair of nearest neighbors in adjacent frame as the same one, and track it until it move out the virtual test box. Judge the move direction by the direction of lifejacket moving out. Through this method, we can track the lifejacket and count it.