ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Journal Article Revisiting direct methods for monocular model-based 3D object tracking: A hierarchical optimization
Cited 0 time in scopus Download 12 time Share share facebook twitter linkedin kakaostory
Authors
Byung-Kuk Seo
Issue Date
2026-11
Citation
Image and Vision Computing, v.175
ISSN
0262-8856
Publisher
Elsevier
Language
English
Type
Journal Article
DOI
https://dx.doi.org/10.1016/j.imavis.2026.106161
Abstract
Monocular model-based 3D object tracking has steadily progressed and has continued to draw interest across diverse applications. In particular, region-based methods have recently become the predominant approach, consistently demonstrating state-of-the-art performance. In this work, we explore a new direction in monocular model-based 3D object tracking by revisiting direct methods and introducing a novel energy function and optimization strategy that achieve significant performance improvements. Specifically, the energy function is formulated as a multi-constraint cost function that effectively combines photometric and geometric constraints in a complementary manner. The optimization strategy is designed as a hierarchical framework with stage-tailored energy functions and is further enhanced by an efficient second-order approximation of the photometric cost function. To validate our method, we conduct comprehensive performance evaluations and analyses, including detailed comparative and ablation studies. Experimental results demonstrate that our method is highly competitive with current top-performing methods on challenging objects and scenes.
Keyword
Hierarchical optimization, Direct method, Model-based approach, 3D object tracking, 6DoF pose estimation
KSP Keywords
3D object, Art performance, Comprehensive performance, Cost Function, Energy Function, Hierarchical framework, Model-based approach, Multi-constraint, Object Tracking, Optimization strategy, Performance evaluation
This work is distributed under the term of Creative Commons License (CCL)
(CC BY)
CC BY