w4 is a real time visual surveillance system for detecting and tracking multiple people and monitoring their activities in an outdoor environment. It operates on monocular gray-scale video imagery, or on video imagery from an infrared camera. W4 employs a combination of shape analysis and tracking to locate people and their parts (head, hands, feet, torso) and to create models of people's appearance so that they can be tracked through interactions such as occlusions. It can determine whether a foreground region contains multiple people and can segment the region into its constituent people and track them. W4 can also determine whether people are carrying objects, and can segment objects from their silhouettes, and construct appearance models for them so they can be identified in subsequent frames. W4 can recognize events between people and objects, such as depositing an object, exchanging bags, or removing an object. It runs at 25 Hz for 320x240 resolution images on a 400 Mhz dual-Pentium II PC.

Gait is a biometric which is subject to increasing interest. Current approaches include modelling gait as a spatio-temporal sequence and as an articulated model. By considering legs only, gait can be considered to be the motion of interlinked pendula. We describe how the Hough transform is used to extract the lines which represent legs in sequences of video images. The change in inclination of these lines follows simple harmonic motion; this motion is used as the gait biometric. The method of least squares is used to smooth the data and to infill for missing points. Then, Fourier transform analysis is used to reveal the frequency components of the change in inclination of the legs. The transform data is then classified using the k-nearest neighbour rule. Experimental analysis shows how phase-weighted Fourier magnitude spectra afford an improved classification rate over use of just magnitude spectra. Accordingly, it appears that it is not just the frequency content which m...

Abstract: In this paper we address the problem of tracking feature points along image sequences. To analyze the undergoing movement we use a common approach based on Kalman filtering which performs the estimation and correction of the feature point’s movement in every image frame. The criterion proposed to establish correspondences, between the group of estimates in each image and the new data to include, minimizes the global matching cost based on the Mahalanobis distance. In this paper, along with the movement tracking, we use a management model which is able to deal with the occlusion and appearance of feature points and allows objects tracking in long sequences. We also present some experimental results obtained that validate our approach.