Abstract. This paper describes a probabilistic integrated object recognition and tracking framework called PIORT, together with two specific methods derived from it, which are evaluated experimentally in several test video sequences. The first step in the proposed framework is a static recognition module that provides class probabilities for each pixel of the image from a set of local features. These probabilities are updated dynamically and supplied to a tracking decision module capable of handling full and partial occlusions. The two specific methods presented use RGB colour features and differ in the classifier implemented: one is a Bayesian method based on maximum likelihood and the other one is based on a neural network. The experimental results obtained have shown that, on one hand, the neural net based approach performs similarly and sometimes better than the Bayesian approach when they are integrated within the tracking framework. And on the other hand, our PIORT methods have achieved better results when compared to other published tracking methods in video sequences taken with a moving camera and including full and partial occlusions of the tracked object.

This paper presents an extension of a previously reported method for object tracking in video sequences [9] to handle object occlusion. The new tracking method is embedded in a system that integrates recognition and tracking in a probabilistic framework. Our system uses object recognition results provided by a neural net that are computed from colour features of image regions for each frame. The location of tracked objects is represented through probability images that are updated dynamically using both recognition and tracking results. From these probabilities and a simple prediction of the apparent motion of the object in the image, a binary decision is made for each pixel and object. The new features of the proposed tracking method include the automated detection of occlusion and the adaptation of the motion prediction to the cases of entering occlusion, full occlusion and exiting occlusion. Experimental results show the effectiveness of the method except when the target object is occluded by an object with a similar appearance.

This paper presents a new method for object tmcking in video sequences that is especially suitable in very noisy environments. In such situations, segmented images from one frame to the nexr one are usually so dzfferent that it is very hard or even impossible to match the corresponding regions or contours of both images. With the aim of tracking objects in these situations, our approach has two main characterjstics. On one hand we assume that the tracking approaches based on contours cannot be applied and therefore, our system uses object recognition results computedfiom regions {specflcally, colour spots $+om segmented images). On the other hrmd, we discard to march rhe spots of consecurive segmented images and, consequently, the metho& that represent the objects by snuctures such as graphs or skeletons, since the slnrsrures obruiwd m q be roo ClrJiewnl in conseculive fiumw. Thus, we wpwsenr rhe loculion ~J~rucked objects through irrtuges of yrubobiliries thut are updated dynumically using botA recugrtiriort urd tmckirrg result3 in previous steps. From these probabilities and a simple pr.ediction of the apparent motion of the object in the image, a bi~lary decision cat? be made for each pixel rmd object.

No abstract

This paper presents a comparison of two classifiers that are used as a first step within a probabilistic object recognition and tracking framework called PIORT. This first step is a static recognition module that provides class probabilities for each pixel of the image from a set of local features. One of the implemented classifiers is a Bayesian method based on maximum likelihood and the other one is based on a neural network. The experimental results show that, on one hand, both classifiers (although they are very different approaches) yield a similar performance when they are integrated within the tracking framework. And on the other hand, our object recognition and tracking framework obtains good results when compared to other published tracking methods in video sequences taken with a moving camera and including total and partial occlusions of the tracked object.