An incremental-learning-by-navigation approach to vision-basedautonomous land vehicle guidance in indoor environments using verticalline information and multiweighted generalized Hough transform technique

摘要

An incremental learning by navigation approach to vision basednautonomous land vehicle (ALV) guidance in indoor environments isnproposed. The approach consists of three stages: initial learning,nnavigation, and model updating. In the initial learning stage, the ALVnis driven manually, and environment images and other status data arenrecorded automatically. Then, an offline procedure is performed to buildnan initial environment model. In the navigation stage, the ALV movesnalong the learned environment automatically, locates itself by modelnmatching, and records necessary information for model updating. In thenmodel updating stage, an offline procedure is performed to refine thenlearned model. A more precise model is obtained after eachnnavigation-and-update iteration. Used environment features are verticalnstraight lines in camera views. A multiweighted generalized Houghntransform is proposed for model matching. A real ALV was used as thentestbed, and successful navigation experiments show the feasibility ofnthe proposed approach