A Skeleton and Visual Tracking Fusion Based Person-Following System for Mobile Service Robots

To overcome the problems of human skeleton loss and body occlusion in a person following system, this paper proposes a real-time and reliable control method for mobile service robots, which combines human skeleton tracker and the discriminative correlation filter with channel and spatial reliability (CSR-DCF) visual tracker. In particular, the CSR-DCF visual tracker is responsible for correcting and locking the special person when the human skeleton tracker provided by Orbbec SDK becomes unstable under the moving mobile service robot. While the skeleton tracker is responsible for taking over the tracking task when the CSR-DCF visual tracker drifts to non-skeleton regions. Experiment results show that the total accuracy has risen to 84.3%, which verifies the effectiveness of the proposed method.