Detecting Line-Road and Road-Intersection Patterns at Various Angles.

摘要

An extended Walsh transform technique for detecting and recognizing a set of cartographic features, such as linear roads and road intersections, at a variety of angles with the axes of the inspecting window is described. The method was implemented as software for an existing sensor array minicomputer system, which consists of a solid state array to convert optical images into electronic signals, a minicomputer as signal processor, and a computer-controlled translational stage as the imagery holder. Each angular orientation of the cartographic features is defined as a class of pattern and classified by using two minimum distance classifiers together with some pretesting algorithms. Typical experimental results are presented and conclusions are given.