High-speed human detection in top-view images by feature scaling for informed filters
Ryusuke Miyamoto, Shuhei Aoki, Takuro Oki · 2017
To monitor vital signs in real-time during exercise, a novel routing scheme called “image assisted routing” is proposed, for which high-speed and accurate human detection executed on embedded systems is indispensable. We propose feature scaling for informed filters with the aim of speeding up of human detection without degrading accuracy. Experimental results obtained by using a CG-based dataset show that the computation speed becomes about 2.77 times faster with the proposed feature scaling.