Pedestrian detection for underground mine vehicles using thermal images

Abstract

Mine vehicles are a leading cause of mining fatalities. A reliable anti-collision system is needed to prevent vehicle-personnel collisions. The proposed collision detection system uses the fusion of a three-dimensional (3D) sensor and thermal infrared camera for human detection and tracking. In addition to a thermal camera, a distance sensor will provide depth information and allow the calculation of the vehicle and pedestrian velocities. The results of subsystem tests show that a simple temperature range is sufficient for segmentation and a neural network shows the best classification results in terms of speed and accuracy. Results of initial tests performed on two different 3D sensors show a significant disadvantage to the use of time of flight cameras in a mine environment.