Computer Engineering Commons^™

Comparing Discrimination And Cfa For Selecting Tracking Features, Damian M. Lyons, D. Frank Hsu

Faculty Publications

The ability of a tracker to isolate the foreground target from the background of an image is crucially dependent on the set of features selected for tracking. Collins & Liu [2] propose an on-line, adaptive approach to selecting the set of features based on the insight that the set of features that best discriminate between target and background classes is the best set to use for tracking. In previous work [10], we have proposed an approach based on Combinatorial Fusion Analysis for selecting features for Real-Time tracking. We discuss the relative merits of the two methods and motivate their combination …

Extended Abstract Rotopod: A Novel Approach To Efficient Legged Locomotion, Damian M. Lyons

Faculty Publications

A number of attempts have been made to integrate the efficiency of wheeled locomotion with the terrain versatility of legged locomotion, e.g., Univ.Michigan’s Rhex platform and Case Western’s Whegs. Those platforms cast legs as rotating spokes placed traditionally at the corners of a rectangular platform. In this paper, we present an alternate approach, with three legs radiating down from a central hub. The energy to move the platform is generated by a rotating reaction mass mounted at the hub and, at rest, rotating parallel to the ground plane.

Our approach is to construct a platform whose natural, uncontrolled motion is …

Evaluation Of A Parallel Architecture And Algorithm For Mapping And Localization, Damian M. Lyons, Giselle R. Isner

Faculty Publications

The Beowulf cluster approach to parallel computation offers a potentially cheap and robust source of computational power for high complexity algorithms in robotics. The challenge is to integrate this approach with the mobility and time critical response constraints of many robotic algorithms. The key contributions of this paper are: (1) introduction of a computational architecture for integrating a cluster into the control architecture of one or more robots, (2) a cluster implementation of Thrun et al’s Concurrent Localization and Mapping (CML) algorithm, and (3) presentation of results to illustrate the performance of the implemented CML algorithm and validate the architectural …

Combinatorial Fusion Criteria For Robot Mapping, Damian M. Lyons, D. Frank Hsu, Qiang Ma, Liang Wang

Faculty Publications

We address the problem of sensor fusion for stereo and ultrasound depth measurements for map building for a robot operating in a cluttered environment. In such a situation it’s difficult to make useful and realistic assumpt ions about the sensor or environment statistics. Combinatorial Fusion Analysis is used to develop an approach to fusion with unknown sensor and environment statistics. A metric is proposed that shows when fusion from a set of fusion alternatives will produ ce a more accurate estimation of depth than either sonar or stereo alone and when not. The metric consists of two crit eria: (a) …

Selection Of Fusion Operations Using Rank-Score Diversity For Robot Mapping And Localization, Damian M. Lyons, D. Frank Hsu, Qiang Ma, Liang Wang

Faculty Publications

In this paper, we evaluate the use of a rank-score diversity measure for selecting sensory fusion operations for a robot localization and mapping application. Our current application involves robot mapping and navigation in an outdoor urban search and rescue situation in which we have many similar and mutually occluding landmarks. The robot is a 4- wheel direct drive platform equipped with visual, stereo depth and ultrasound sensors. In such an application it’s difficult to make useful and realistic assumptions about the sensor or environment statistics. Combinatorial Fusion Analysis(CFA) is used to develop an approach to fusion with unknown sensor and …

Combining Multiple Scoring Systems For Video Target Tracking Based On Rank-Score Function Variation, D. Frank Hsu, Damian M. Lyons, Jizhou Ai

Faculty Publications

Tracking of video targets is the process of estimating the current and predicting the future state of a target from a sequence of video sensor measurements. Multitarget video tracking is complicated by the fact that targets can occlude one another and affect video feature measurements in a highly non-linear and difficult to model fashion., Tracking multiple targets that undergo repeated mutual occlusions is a challenging problem with several issues to be addressed. In this paper we propose a multisensory fusion approach to the problem of multitarget video tracking with occlusion. Each sensory cue is treated as a scoring system on …

Combinatorial Fusion Criteria For Real-Time Tracking, D. Frank Hsu, Damian M. Lyons, Jizhou Ai

Faculty Publications

We address the problem of automated video tracking of targets when targets undergo multiple mutual occlusions. Our approach is based on the idea that as targets are occluded, selection of feature subsets and combinations of those features are effective in identifying the target and improving tracking performance. We use Combinatorial Fusion Analysis to develop a metric to select which subset of features will produce the most accurate tracking. In particular we show that the combination of a pair of features A and B will improve the accuracy only if (a) A and B have relative high performance, and (b) A …

Feature Selection For Real-Time Tracking, D. Frank Hsu, Damian M. Lyons, Jizhou Ai

Faculty Publications

We address the problem of selecting features to improve automated video tracking of targets that undergo multiple mutual occlusions. As targets are occluded, different feature subsets and combinations of those features are effective in identifying the target and improving tracking performance. We use Combinatorial Fusion Analysis to develop a metric to dynamically select which subset of features will produce the most accurate tracking. In particular we show that the combination of a pair of features A and B will improve the accuracy only if (a) A and B have relative high performance, and (b) A and B are diverse. We …

Analysis Of Gaits For A Rotating Tripedal Robot, Damian M. Lyons, Kiran Pamnany

Faculty Publications

A goal of robotics has been to develop mechanisms that have the efficiency and speed of wheeled robots with the terrain flexibility of legged robots. In previous work, we have proposed a unique three-legged mechanism, the rotopod, designed to integrate these two useful approaches to locomotion. In this paper, we present an analysis of the variety of gaits that can be exhibited by the rotopod. We present examples of many of these gaits and discuss their potential use. A trajectory generation algorithm is presented that can be used to generate point to point trajectories using one of three different styles …

Rotational Legged Locomotion, Damian M. Lyons, Kiran Pamnany

Faculty Publications

In this paper we presented a novel, agile robot mechanism, which we call a rotopod, which combines aspects of wheeled and legged locomotion. A general description of how a tripedal rotopod can be made to step, rotating the mechanism about one leg, and moving the center of the mechanism, is presented. The concept of a gait for this mechanism is defined, and is used to show how extremely agile the mechanism can be. Specific resistance is employed as a way to explore the relative efficiency of this mechanism versus a wheel. Finally, we describe our first prototype rotopod and report …

Rank-Based Multisensory Fusion In Multitarget Video Tracking, Damian M. Lyons, D. Frank Hsu

Faculty Publications

An attractive approach to improve tracking performance for visual surveillance is to use information from multiple visual sensory cues such as position, color, shape, etc. Previous work in fusion for tracking has tended to focus on fusion by numerically combining the scores assigned by each cue. We argue that for video scenes with many targets in a crowded situation, the splitting and merging of regions associated with targets, and the subsequent dramatic changes in cue values and reliabilities, renders this form of fusion less effective. In this paper we present experimental results showing that use of cue rank information in …

A Dynamic Pruning And Feature Selection Strategy For Real-Time Tracking, D. Frank Hsu, Damian M. Lyons

Faculty Publications

Automated video tracking is useful in a number of applications such as surveillance, multisensor networks, robotics and virtual reality. In this paper we investigate an approach to tracking based on fusing the output of a collection of video trackers, each attending to a different feature or cue on the target. We show both theoretically and experimentally that the method used to prune the growth of target hypotheses can have a great impact on the trackers performance, and indirectly, change the benefit of using linear score combination as opposed to a non-linear rank combination for fusion. We also show that the …