Tianyu Huang

Tianyu Huang

Auto-generated profile, based on publicly available information

Tianyu Huang

Department of Mechanical and Automation Engineering, The Chinese University of Hong Kong, Hong Kong, China

Research Focus & Metrics

Academic Career Length

3Years (2024-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 April 2024 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
THTianyu Huang
Tianyu Huang
HLHaoang Li
Haoang Li
LPLiangzu Peng
Liangzu Peng
YLYinlong LIU
Yinlong LIU
YLYunhui Liu
Yunhui Liu
Estimating the rigid transformation with 6 degrees of freedom based on a putative 3D correspondence set is a crucial procedure in point cloud registration. Existing correspondence identification methods usually lead to large outlier ratios (>95% is common), underscoring the significance of robust registration methods. Many researchers turn to parameter search-based strategies (e.g., Branch-and-Bround) for robust registration. Although related methods show high robustness, their efficiency is limited to the high-dimensional search space. This paper proposes a heuristics-guided parameter search strategy to accelerate the search while maintaining high robustness. We first sample some correspondences (i.e., heuristics) and then just need to sequentially search the feasible regions that make each sample an inlier. Our strategy largely reduces the search space and can guarantee accuracy with only a few inlier samples, therefore enjoying an excellent trade-off between efficiency and robustness. Since directly parameterizing the 6-dimensional nonlinear feasible region for efficient search is intractable, we construct a three-stage decomposition pipeline to reparameterize the feasible region, resulting in three lower-dimensional sub-problems that are easily solvable via our strategy. Besides reducing the searching dimension, our decomposition enables the leverage of 1-dimensional interval stabbing at all three stages for searching acceleration. Moreover, we propose a valid sampling strategy to guarantee our sampling effectiveness, and a compatibility verification setup to further accelerate our search. Extensive experiments on both simulated and real-world datasets demonstrate that our approach exhibits comparable robustness with state-of-the-art methods while achieving a significant efficiency boost.

2 October 2024 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
THTianyu Huang
Tianyu Huang
YLYinlong LIU
Yinlong LIU
BYBohan Yang
Bohan Yang
YLYunhui Liu
Yunhui Liu
Given a set of outlier-contaminated 2D–3D line correspondences between the scene and a captured image, we aim to recover the absolute camera pose. This is a fundamental problem in computer vision and robotics, for which many methods have been developed and shown impressive performance, but they fail to simultaneously guarantee high robustness and efficiency. In this letter, we propose an approach that can process high outlier ratios (e.g., 90%) as the state-of-the-art method while achieving a significant efficiency boost, namely, dozens of times faster. The high robustness and efficiency of our approach benefit from a globally optimal camera orientation estimation module, in which we embed an interval stabbing strategy into a customized Branch-and-Bound (BnB) solving framework. While BnB is widely known to be inefficient, this does not apply to our method thanks to the searching acceleration brought by fast interval stabbing. In addition, we investigate the special case where the camera vertical direction is given as priors; we show that this case can be solved by interval stabbing with both high robustness and real-time efficiency. Experiments on both simulated and real-world data demonstrate the robustness and efficiency of our approach.

27 March 2025 in IEEE/ASME Transactions on Mechatronics
Verified
Journal Article
Closed access
YSYanshu Song
Yanshu Song
JZJie Zhao
Jie Zhao
THTianyu Huang
Tianyu Huang
DLDarwin Lau
Darwin Lau
YLYunhui Liu
Yunhui Liu
Loading and unloading goods in narrow cuboid spaces (e.g., truck carriages) is a critical task in logistics transportation. It is often time-consuming, labor-intensive, risky, and therefore is valuable and necessary to be automated. However, due to the complexity of operational scenarios, most existing robots struggle to simultaneously integrate: small size and weight, sufficient load capacity, high robustness, high picking flexibility and efficiency, and good adaptability to narrow cuboid spaces. In this work, we present a light-weight robotic system capable of achieving flexible, efficient, and heavy-load palletizing in constrained cuboid spaces. We propose a novel adsorption-tray hybrid gripper which can significantly reduce the required adsorption area (by 64%) while maintaining high picking flexibility without affecting load capacity. Based on it, we propose the lift-then-pick picking method, and provide the design method of an adjustable picking trajectory, as well as an optimization strategy for the quantity of suction nozzles. Furthermore, we successfully introduce the time-optimal control idea into the picking process based on its unique dynamic constraints, which improves efficiency by 46%. We also implement a compliance control mechanism that significantly enhances the robustness (i.e., nearly doubling the picking success rate). In addition, we develop a specially designed manipulator to carry the gripper. The horizontally arranged degree of freedom configuration and the new rotational joint construction method enable the arm to maintain light-weight while sufficient load capacity. And its coverage of cuboid spaces far exceeds that of traditional industrial arms (100% vs. 53%). Finally, we manufacture a prototype and validate the above advantages through both simulation and real-world tests, confirming its potential for applications.

1 November 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
LGLinhai Gui
Linhai Gui
GSGuangli Sun
Guangli Sun
THTianyu Huang
Tianyu Huang
ZZZhisong Zhou
Zhisong Zhou
GWGang Wang
Gang Wang
YLYunhui Liu
Yunhui Liu

12 March 2026 in IEEE Transactions on Image Processing
Verified
Journal Article
Closed access
JLJunfeng Li
Junfeng Li
JHJunjie He
Junjie He
WLWenjie Liu
Wenjie Liu
THTianyu Huang
Tianyu Huang
SZShunbo Zhou
Shunbo Zhou
HLHaoang Li
Haoang Li
Generating dynamic scenes from images has gained increasing attention. Existing methods have two major limitations: (1) They can hardly handle sparse images which exhibit limited geometry constraints and insufficient motion; (2) They struggle to maintain spatial-temporal consistency when rendering multi-view videos. To address these limitations, we propose SCSV, a spatial-temporal consistent dynamic scene generation method from sparse views. Our method consists of two stages: scene reconstruction and scene expansion, both of which decouple background and foreground. In the scene reconstruction stage, we first interpolate a set of images between the input images based on a video generation model, followed by the optimization of the scene Gaussian from the interpolated and input images. To improve the spatial-temporal consistency of the reconstructed scene, we propose an uncertainty-aware Gaussian training approach, which introduces adaptive weights of images and pixels. In the scene expansion stage, for background, we render novel views and refine them with a geometry-aware diffusion process. These refined images are then used to incrementally add the Gaussians. As to foreground, we generate human motion according to previous motion, enabling temporal coherent generation of motion. To further enhance the physical plausibility, we integrate the expanded foreground into the background using a gravity-aware alignment. Experiments on NeuMan, Bonn, and EMDB datasets demonstrate that our SCSV achieves superior performance compared to state-of-the-art methods. The code will be released upon acceptance.