Fee-Lee Lim

Non-overlapping Distributed Tracking using Particle Filter

Tracking people or objects across multiple cameras is a challenging research area in visual computing especially when these cameras have non-overlapping field-of-views. The important task is to associate a target of interest with its previous appearances across time and space within the camera network. In this paper, we propose a unified tracking framework using particle filter to efficiently switch between track prediction (to deal with non-overlapping region tracking) and visual tracking. The particle filter tracking system uses a map to provide the possible trajectory information of the target as it moves within the non-overlapping regions. We implemented and tested this tracking approach in an in-house multiple cameras system. Promising results were obtained which suggested the feasibility of such an approach

Non-overlapping Distributed Tracking System Utilizing Particle Filter

Tracking people across multiple cameras is a challenging research area in visual computing, especially when these cameras have non-overlapping field of views. The important task is to associate a current subject with other prior appearances of the same subject across time and space in a camera network. Several known techniques rely on Bayesian approaches to perform the matching task. However, these approaches do not scale well when the dimension of the problem increases; e.g. when the number of subject or possible path increases. The aim of this paper is to propose a unified tracking framework using particle filters to efficiently switch between visual tracking (field of view tracking) and track prediction (non-overlapping region tracking). The particle filter tracking system utilizes a map (known environment) to assist the tracking process when targets leave the field of view of any camera. We implemented and tested this tracking approach in an in-house multiple cameras system as well as using on-line data. Promising results were obtained which suggested the feasibility of such an approach.

Resolution consideration in spatially variant sensors

Log polar transformations for space variant systems have been proposed and used in active vision research. The idea is to generate an image with a varying resolution over a wide angle field of view. The fovea is of high resolution and the periphery is of exponentially reduced resolution. The justifications for such a sensor are: (i) it provides high resolution and a wide viewing angle; (ii) feature invariance in the fovea simplifies foveation; (iii) it allows multi-resolution analysis; and (iv) it is cheaper and more efficient to build a variable resolution sensor over a particular field of view rather than a uniform high resolution sensor. The receptor density of the human retina is very high, i.e. of the order of 108 receptors at the fovea. The question is, what resolution should space variant active vision systems have? Real visual sensors have been implemented but is the resolution produced high enough? This paper investigates the resolution requirements of a space variant sensor by simulation for a tracking system using raytracing.

Human saccadic eye movements and tracking by active foveation in log polar space

One of the possible models of the human visual system (HVS) in the computer vision literature has a high resolution fovea and exponentially decreasing resolution periphery. The high resolution fovea is used to extract necessary information in order to solve a vision task and the periphery may be used to detect motion. To obtain the desired information, the fovea is guided by the contents of the scene and other knowledge to position the fovea over areas of interest. These eye movements are called saccades and corrective saccades. A two stage process has been implemented as a mechanism for changing foveation in log polar space. Initially, the open loop stage roughly foveates on the best interest feature and then the closed loop stage is invoked to accurately iteratively converge onto the foveation point. The open loop stage developed for the foveation algorithm is applied to saccadic eye movements and a tracking system. Log polar space is preferred over Cartesian space as: (1) it simultaneously provides high resolution and a wide viewing angle; and (2) feature invariance occurs in the fovea which simplifies the foveation process.

Situation Awareness System in a Maritime Port

The Automatic Identification System (AIS) is a transponder-based system that broadcasts the ownships data (e.g. position, speed, course, ship identification number, and dimensions) on a regular basis. Any other AIS receivers can receive these data. This paper investigate the use of AIS data to develop a system that is able to learn the location of legal fairways around a port given the AIS data collected over a given period of time. The learned fairways information is subsequently used to pre-empt abnormal activities along the fairway. Such a system is therefore capable of better preventing port disasters through better real-time situation awareness.

A Solution for Illumination Challenged Face Recognition using Exemplar-based Synthesis Technique

This paper addresses the problem of illumination variations affecting the performance of real world face recognition systems. We propose an exemplar (or appearance)-based technique to compute the canonical representation of the human face. The canonical image is equivalent to the reflectance field of the face that is invariant to illumination. We subsequently use the canonical face to perform the following two tasks: (1) Extract the series of independent lighting model which interacts with the canonical face to form the novel faces. (2) Synthesize novel face appearances using these lighting models. We demonstrate the ability of the approach to generate novel views of the face and show how they help improve the performance of illumination challenged face recognition problem

An illumination invariant face recognition approach using exemplar-based synthesis technique

This paper proposes a new method to solve the problem of face recognition under varying illumination conditions. We introduce an exemplar-based technique to decouple and subsequently recover the canonical face and the illumination functions from the intensity images. The canonical image is equivalent to the reflectance field of the face that is invariant to illumination. We subsequently use the canonical face to synthesize novel face appearances together with a set of lighting models. We then demonstrate the ability of the synthesis approach to improve the performance of the face recognition task.