Ernest Cheung

Papers

Total Citations

H-Index

About

Ernest Cheung’s research centers on advancing 3D sensing and robotic manipulation, with a particular focus on stereo vision and object pose estimation. His most impactful work, “Optimization-based automatic parameter tuning for stereo vision” (2015, 6 citations), addresses a critical bottleneck in robotics: the manual calibration of stereo cameras. By automating parameter tuning, Cheung’s method enables robots to generate dense, high-resolution point clouds more efficiently—a key requirement for reliable navigation and object handling. This contribution is especially valuable for developing smaller, cheaper, and lower-power robotic systems that rely on passive sensing. Cheung further advanced robotic dexterity through his work on initial pose estimation. In “Initial pose estimation using cross-section contours” (2014, 2 citations) and “Multi-contour initial pose estimation for 3D registration” (2015, 2 citations), he devised novel algorithms to approximate an object’s 6-DOF pose from stable, planar configurations. These methods provide robust initialization for the Iterative Closest Point algorithm, overcoming challenges like sensor noise and occlusion—critical for service robots manipulating everyday household objects. Though his citation counts are modest, Cheung’s contributions to practical, optimization-driven 3D perception lay essential groundwork for autonomous systems operating in cluttered, real-world environments.