The aim of the presented work is the development of a natural scene visual analysis system for a travel assistance tool for the visually impaired. The purpose of such a system is the creation of a simplified scene description, suitable for auditory presentation. System under development is intended to aid a blind user in avoiding obstacles in a typical urban environment. A Bayesian classifier is utilised for object detection and tracking in a stereoscopic image sequence captured from a pair of head mounted cameras. The obtained results show the method's potential to track objects with complicated shapes solely on the basis of their visual features. Scene depth, represented in the classifier by the image disparity, plays an essential role in identifying objects in textured image areas. Lack of a texture does not allow for reliable disparity estimation, lowering classification performance. The author combines colour, disparity and pixel coordinates to produce a feature vector for an image classifier. The proposed classifier update technique allows for object tracking in an image sequence of a dynamic scene.
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.