This paper addresses the design of a system for automatic sign interpretation (ASI) with specific applications to road sign imagery for driver assistance. Other applications include intelligent highway systems, and sign inventory systems for transportation departments. This paper extends past work in which a prototype system for the detection of road signs was designed and tested on a limited set of images and sign types [1]. The specific enhancements reported in this paper include the following: (i) enhanced dual-band spectral analysis in the hue-saturation-intensity (HSI) and RGB domains, (ii) the introduction of a relational analysis scheme to form the identification stage, (iii) the creation of a larger database of road sign images encompassing all road sign types encountered in the United States, and (iv) testing under adverse conditions with non-ideal road sign images. The proposed method can be adapted for implementation in real time as this was one of the design considerations. The human cognition of signs under adverse conditions is one of the basic research questions that is yet to be answered. The ASI system designed will help to shed some light on the human cognitive process.
展开▼
