A5026431700- Video Surveillance and Tracking Methods
- Advanced Neural Network Applications
- Human Pose and Action Recognition
- Visual Attention and Saliency Detection
- Advanced Image and Video Retrieval Techniques
- Hand Gesture Recognition Systems
- Anomaly Detection Techniques and Applications
Fudan University
2015-2018
Human detection has received great attention during the past few decades, which is yet still a challenging problem. In this paper, we focus on problem of 3-D human detection, i.e., finding bodies and determining their coordinates in complex space using depth data only. Since traditional sliding-window-based approaches for target localization are time-consuming recent deep-learning-based object detectors generate too many region proposals, propose to utilize candidate head-top locating stage...
Abstract Concret-filled-steel-tube arch bridges often employ solid trial-assembly for the ribs to confirm matching accuracy and overall alignment. However, these methods suffer from issues such as large site occupation, multiple assembly cycles, prolonged construction periods. This paper proposes a virtual technology of steel-pipe-arch based on limited perception, which achieves rapid without need physical segment matching. By obtaining joint control point data through measurement method...
Real-time human detection is important for a wide range of applications. In this paper, two-staged method has been developed real-time in cluttered and dynamic environments with depth data. We start generating set possible head-tops to ensure all locations are included. To end, novel physical radius-depth (PRD) detector proposed quickly detect candidates. The second stage applies convolutional neural network (CNN), aiming at extracting feature upper body automatically instead hand-crafting,...
This paper proposes a two-staged approach to real-time human detection in cluttered environments using RGB-D camera. The first stage is novel physical blob (P-Blob) that can quickly find plausible heads. second uses combination of upper-body features filter out false positives. Experiment results on three publicly available datasets show the proposed method reliably detect people video real time.
Real-time human detection in crowded and dynamic environments poses a significant challenge, due to complex background, occlusion different poses. In this paper, we propose two-staged approach using color depth data taken by an RGB-D camera. The first stage is find plausible head-top locations quickly image. second extract effective discrimination features from discard the false positives with support vector machine. experiments on publicly available office dataset, mobile platform dataset...
Efficient and robust detection of humans has received great attention during the past few decades. This paper presents a two-staged approach for human in RGB-D images. As traditional sliding window-based methods target localization are often time-consuming, we propose to use super-pixel method depth data efficiently locate plausible head-top locations first stage. In second stage, Random Ferns seek features by combining information from different image spaces, which can select most...