This paper presents a fast and memory saving method to estimate a measuring point by making use of vertical edge patterns. It is especially useful to locate a mobile robot's position in a room where many those patterns are observed at corners of wall and edges of objects. They are detected with horizontal angle measurement using an optical system, which consists of a TV camera and a turning mechanism.
A matching method is used to understand detected patterns, where positions of some edges are assumed to be known. The most probable measuring point is estimated from Baysian rule under the invariants conservation, similar to projective invariant, which is derived in this paper. This method is a searching process for an optimal correspondence between known edges and detected edge patterns. For each correspondence, it is possible to use analytical equations to obtain the most probable point.
Some pruning rules are introduced to deduce searching efforts and to improve reliability of the answer obtained. They are constraints on visual angles of two edges, the number of noise patterns between edges, probability value and estimated position.
An experimental result is shown to confirm effectiveness of this method.