Efficient Image Sequence Analysis Using Fuzzy Techniques

Abstract

Machine Vision Systems (MVSs) used, for example in mobile robots, have to function in real-time and there is a need to develop more efficient ways of processing frame sequences. The problem can be addressed by mimicking aspects of human vision [1–4]. Previous work by Griffiths et al described a process for examining scenes using a camera mounted on a pan and tilt unit. This process used stochastic scanpaths that were tuned using a Fuzzy Inference System (FIS) to determine the next orientation of the camera. The development of this technique for scanning image frames, captured offline by a digital video camera, is described in Allen et al [6]. This paper introduces the concept of Fuzzy tuned scanpaths in Section 2. Section 3 examines the implementation and operation of an adaptive technique. Experimental work using this adaptive technique is described in Section 4 and conclusions are given in Section 5.