研究目的
To propose a bodyprint extraction method for people re-identi?cation using omnidirectional cameras.
研究成果
The method shows promising results with a performance of 87% in re-identifying people under various conditions. Future work includes evaluating the method on a larger database and different scenarios, and improving the matching method.
研究不足
The database is limited to 5 people. The method assumes a fixed person height and camera pointing vertically to the ground. The performance could be affected by the track duration and the position of people in the image.
1：Experimental Design and Method Selection:
The study uses omnidirectional cameras with fisheye lenses for bodyprint extraction. The method involves head detection, camera model understanding, bodyprint extraction, and comparison metrics.
2：Sample Selection and Data Sources:
A database of videos with 5 different people, totaling 31 tracks, was used. The scenario included areas with different illumination, frontal, lateral, and rear views of people, and occlusions.
3：List of Experimental Equipment and Materials:
AXIS omnidirectional camera model M3007 with a fisheye lens.
4：Experimental Procedures and Operational Workflow:
The process includes head detection, tracking, temporal signature extraction, alignment, bodyprint extraction, and matching.
5：Data Analysis Methods:
The similarity of bodyprints is measured using the maximum euclidean distance of the RGB components.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
