2018 IEEE Fourth International Conference on Multimedia Big Data (BigMM)
Download PDF

Abstract

Accurate multi-human detection is very important for intelligent video surveillance, especially for lots of human-centered applications, e.g. people counting and crowd detection. Currently, most existing algorithms can not solve this problem well due to the heavy occlusions between the humans. To overcome this difficult problem, we propose a new method based on RGB-D information. In the proposed algorithm, we view “Occlusion Region” which consists of multiple humans as a key issue for the whole multi-human detection task. First, we search the candidate occlusion regions with hash searching scheme on the original color image. Then, we segment each searched occlusion region with head-shoulder template matching on the corresponding depth image. Extensive experimental results show that our method can achieve significant improvement comparing to the state-of-the-art methods.
Like what you’re reading?
Already a member?
Get this article FREE with a new membership!

Related Articles