Abstract

With the recent development of various kinds of spatial light modulator, many real time tracking techniques have been proposed. In this paper we present a technique for adaptive object tracking based on a joint transform correlator using a single LCTV in a video feedback architecture. The basic algorithm is to correlate the object in the current frame with itself in the previous frame of sequential video images. Based on the location of the output correlation peak, the relative positions of the object in these two frames can be determined, and the object's actual location can be updated. It must be pointed out that in a JTC architecture, the reference image and the input scene to be correlated with are placed side by side on a SLM. Therefore, a JTC can be made adaptive by constantly updating the reference image to correlate with the dynamic target. Once an object in an input scene is identified as the target and its position is determined, the tracking system can run autonomously. For multiobject tracking, there are N pairs of correlation peaks on the output plane corresponding to the Nobjects to be tracked in an input scene. To associate each peak correctly with its host, we propose a tree-search decision-making algorithm. Using this hybrid system, multiple objects can be tracked simultaneously.

© 1989 Optical Society of America