We have proposed a tracker which is based on multiple hypothesis model and iterative closest point algorithm. The algorithm successfully tracks multiple targets by using multiple hypothesis model matching approach. Hypothesis models are formed by using track information; the models are updated periodically whenever a new track is added. Although there are number of algorithms proposed so far in this area, our approach addresses number of issues, which others fail to do. The algorithm works well in variable number of targets, as soon as target appears in the track; it is automatically added and when the target leaves the track is dropped. The detection accuracy observed on the video tracks is noticeable.