Recognition of traffic signs using a multilayer neural network
Si Wei Lu · 1994
Conventional pattern recognition systems require an input pattern to be presented in a predetermined position. Its orientation and size have to be in such a position for better recognition. However, for a camera in a moving car, the traffic signs may be shifted in position, rotated slightly, or scale changed in relation to the standard position. Furthermore, the traffic signs can be partially occluded and maybe noise-corrupted. Such situations cause lots of difficulties for recognition.>