Temperature-based alternate perception method for human-motion detection with visually impaired user applications

Abstract

This paper presents an alternate perception method to improve the perception ability of the blind/visually impaired people who are generally familiar with stationary objects but less confident in congested environment where human motion is unpredictable. This method, taking advantages of the fact that the human body is essentially a natural heat source, utilizes temperature fields and their gradients from infrared (IR) images to locate individual humans in a relatively congested area and determine their face orientation and motion states. Unlike conventional illumination-sensitive visual images that often include many unrelated details, the normalized IR images in this method are naturally pre-filtered, effectively removing unwanted background ''noise'', which greatly simplifies the subsequent image processing tasks. The temperature-based method has been illustrated with three realistic scenes commonly encountered in daily walking. Practical implementation issues, which include the effects of low image resolution, overlapping/occlusions, reflection and non-human heat sources on the face orientation detection, are discussed; and their effects on the effectiveness and robustness of the method are experimentally investigated. The experimental findings offer physically intuitive insights into the development of alternate perception methods based on temperature field and its potential applications in intelligent space, smart city and smart cars.