A pedestrian detection system for underground mine vehicles is being developed that requires the segmentation of people from thermal images in underground mine tunnels. A number of thresholding techniques are outlined and their performance on a number of thermal images is investigated. The thresholding techniques are evaluated on images in various ambient conditions and it is shown that a minimum error thresholding technique is the most effective.
Reference:
Dickens, JS and Green, JJ. Segmentation techniques for extracting humans from thermal images. 4th Robotics & Mechatronics Conference of South Africa, CSIR International Convention Centre, CSIR, Pretoria, South Africa, 23-25 November 2011
Dickens, J., & Green, J. (2011). Segmentation techniques for extracting humans from thermal images. RobMech. http://hdl.handle.net/10204/5553
Dickens, JS, and JJ Green. "Segmentation techniques for extracting humans from thermal images." (2011): http://hdl.handle.net/10204/5553
Dickens J, Green J, Segmentation techniques for extracting humans from thermal images; RobMech; 2011. http://hdl.handle.net/10204/5553 .
