Study of multiple moving targets’ detection in fisheye video based on the moving blob model

Abstract

This paper discussed some improved algorithms for multiple moving targets detection and tracking in fisheye video sequences which based on the moving blob model. The view field of fisheye lens achieved 183 degree which used in our system, so it has more effective in the no blind surveillance system. However, the fisheye image has a big distortion that makes it difficult to achieve an intelligent function. In this paper we try to establish a moving blob model to detect and track multiple moving targets in the fisheye video sequences, in order to achieve the automation and intelligent ability for no blind surveillance system. It is divided into three steps. Firstly, the distortion model of fisheye lens was established, we are discussing the character of the imaging principle of fisheye lens, and calculate the distortion coefficient which can be used in the moving blob model. Secondly, the principle of the moving blob model was analyzed in detail which based on the fisheye distortion model. It was included four main algorithms, which the first is the traditional algorithm of background extraction; and the background updating algorithm; the algorithm of the fisheye video sequence with the background subtracted in order to get the moving blobs; the algorithm of removing the shadow of blobs in RGB space. Thirdly, we determined that every extracted blob is a real moving target by calculating the pixels with a threshold, which can discard the faulty moving targets. Lastly, we designed the algorithm for tracking the moving targets based on the moving blobs selected through calculating the geometry center. The experiment indicated that every algorithm has a better processing efficiency of multiple moving targets in fisheye video sequences. Compared the traditional algorithm, the improved algorithm can be detected the moving target in a circular fisheye image effectively and stably.