HFR-Video-Based Stereo Correspondence Using High Synchronous Short-Term Velocities

Area of Science:

• Computer Vision
• Robotics
• Artificial Intelligence

Background:

• Distinguishing between multiple, similarly-appearing moving objects in stereo video presents a significant challenge for current correspondence algorithms.
• Traditional methods often fail when objects lack unique textural or color features, hindering applications in dynamic scene analysis.

Purpose of the Study:

• To develop a novel method for solving the multi-object correspondence problem in stereoscopic videos, particularly for objects with similar appearances.
• To leverage high-speed camera synchronization and object motion patterns for improved correspondence accuracy.

Main Methods:

• The study proposes using short-term velocities (STVs) of objects as distinctive motion features.
• It calculates the similarity of STVs to establish object-level cross-camera multiple object correspondence, integrating pixel-level and feature-level stereo correspondences.
• Experiments utilized markers on a metronome and natural hand movements to simulate diverse motion scenarios.

Main Results:

• The proposed method demonstrated good performance in stereo correspondence tasks.
• The approach effectively addressed the challenge of distinguishing between similar-looking objects based on their motion dynamics.

Conclusions:

• Short-term velocities (STVs) are effective motion features for solving the multi-object correspondence problem in stereoscopic vision.
• This technique enhances the robustness of stereo vision systems dealing with visually similar dynamic objects.