NEURAL NETWORK PREDICTION OF PERSON POSITION FOR HUMAN FOLLOWING MOBILE ROBOT PLATFORM

This paper addresses an important and challenging problem of real-time vision-based human tracking to enable mobile robots to follow a human and help him as a co-worker in hazardous environment. A novel approach that combines intelligent stereo vision-based human detection with neural network estimator that helps tracking is presented. Stereo vision-based detection combines features extracted from 2D stereo images with intelligent classification algorithms to detect humans in a robot's environment. Prediction of the 3D coordinates of a human in the robot's camera coordinate system based on recurrent neural network reduces image region of interest and ensures more reliable human tracking. Within a working scenario of a mobile robot intended to follow a human co-worker in indoor applications, collected video data will be collected and compared with simulation results obtained from the recurrent neural network.