Straight-line-based primitive extraction in grey-scale object recognition

摘要

A method for extraction of straight-line-based primitives is proposed. It is based on a bottom-up approach that extracts linear primitives first, then line segments, that are grouped in straight lines. Based on the extracted lines, quadrangles are detected and further approximated by rectangles. The extracted primitives are used in object recognition. The proposed approach provides a good relation between the complexity of extracted primitives and the complexity of algorithm itself, that allows it to be used in industrial applications.