Extraction of road traffic sign information based on a vehicle-borne mobile photogrammetric system

摘要

A new approach to automatically extracting road traffic sign information based on a vehicle-borne mobile photogrammetric system (VMPS) is proposed in this paper. The method can be divided into four steps: (a) traffic sign detection in a single image; (b) geometric information computation in a stereopair; (c) semantic information recognition; and (d) stereoscopic information consistency verification. In order to enhance the robustness and adaptability of the approach, colour recognition probabilistic neural networks (PNNs) based on pixel vectors, together with shape-identification PNNs based on central projected vectors, are used during the traffic sign detection and recognition steps. The proposed approach is applied to many real-scene stereopairs taken by VMPS at different times and places. Experimental results demonstrate its feasibility and effectiveness.