科研之友
微信
新浪微博
Facebook
分享链接
摘要
This paper describes an automatic road-tracking approach for rural areas using air borne image sequences Due to the relatively small image size, small-frame camera systems are ideal platforms for the development of automatic image analysis approaches which make use of redundant image information available through the obligatory image overlap In the proposed approach an initial road network extracted in a first Image or manually digitized by an operator is automatically continued in overlapping adjacent images First a supervised classification is performed The existing road net work is used as the training region in the overlap ping area in order to optimally account for radio metric and illumination properties of the individual images Second the segments obtained in the first step are combined with extracted edges in order to derive an optimal road tracking The combination of areal and linear features also enables to extract road crossings Third a post processing step closes gaps between extracted sections to build a complete road network Results demonstrate that a reliable and accurate tracking of main roads including junctions and field-paths is possible with the proposed approach given a clear color separation of path and background An open issue concerns the closing of large gaps which needs further attention in future research
- 出版日期2010
