A robust and efficient technique for detection and recognition of American road signs.

摘要

Due to advances in technology, road safety has improved tremendously and efforts are now focused on problems caused by the driver error or human error. Drivers are opting to continue driving well into their senior years even with poor eyesight and slow reflexes; and, there are now more distractions than ever, such as cell phones, which can lead to more accidents. In this thesis, a Driver Assistance System is presented. The system detects and recognizes American traffic Signs using Principle Component Analysis (PCA). It is a multi-layered hierarchical scheme for object pre-processing including a color based object segmentation scheme. The system is composed of three parts: colour segmentation, shape recognition and road sign classification. The shape recognition and sign classification are both based on principle component analysis. The proposed system has a correct classification rate of 99.2%. Currently, the system uses road signs found on United States roads, but it can be generalized to other countries' traffic road signs.