Integrate Sparse Depth Information into Pedestrians Detection
Integrate Sparse Depth Information into Pedestrians Detection
复制标题
DOI：
--
复制
发表时间：
2011
期刊：
影响因子：
--
通讯作者：
Yu Wang;Jien Kato;K. Ishii
中科院分区：
文献类型：
--
作者：
Yu Wang;Jien Kato;K. Ishii
在本文中，我们提出将稀疏的3D深度信息集成到行人检测任务中，以实现性能的快速提升。我们提出的方法采用概率方法将基于图像特征的检测和稀疏深度估计相结合。深度信息被用作线索，并为检测提供额外的判别能力。本文主要有两个方面的贡献：1)简化了图形模型，将深度线索有效地整合到检测中；2)稀疏深度估计方法，可以提供快速可靠的深度信息估计。实验表明，我们的方法可以在最小的额外时间内提供比基线检测器有希望的增强。
In this paper, we propose to integrate sparse 3D depth information into pedestrian detection task, in order to achieve a fast boost in performance. Our proposed method uses a probabilistic way to integrate image-feature-based detection and sparse depth estimation together. The depth information is used as a cue, and provides additional discriminative ability for the detection. There are two contributions in this paper: 1) a simplified graphical model which could efficiently integrate depth cue into detection; and 2) a sparse depth estimation method which could provide fast and reliable estimation of depth information. The experiment shows that our method could provide promising enhancement over baseline detector with minimal additional time.