Thorough ZDF-Based Localization for Binocular Tracking

W e present an eficient method to fixate a binocular gaze point on a target object moving around in a complicated environment. Assuming the target is in the riicinity of t1t.e gaze point, the image features of the target can be isolated by using zero disparity feature from others which suffer some disparities. This assumption is proper while successful binocular tracking, and the zero disparity filtering (ZDF) is computationally very cheap enough to realize quick visual feedback on an ordinal image processing hardware. The position estimation obtained from such isolated features, however, is restricted i n zero disparity fields, resulting in no depth cue. In order to get depth information simultaneously by ZDF processing, we introduce the novel localization technique based on the idea of "virtual horopter". The proposed method is implemented on our active vision system and the total performance of binocular tracking is demonstrated by actual tracking experiments.