Kernel Particle Filter for Real-Time 3D Body Tracking in Monocular Color Images.

摘要: This paper presents the application of a kernel particle filter for 3D body tracking in video stream acquired from single uncalibrated camera. Using intensity-based and color-based cues as well an articulated model with shape represented by cylinders, real-time monocular cluttered image sequences has been realized. The algorithm runs at 7.5 Hz on laptop computer tracks upper human two arms. First, experimental results show that proposed approach good recovering capabilities despite using small number particles. is intended use mobile robot to improve interaction.