Traffic sign recognition
The Traffic Sign Recognition System or TSRS is based on colour vision.
Only traffic signs that are highly important with respect to the driving context are shown to the driver, and different levels of emphasis, tied to the urgency of the warning, are applied to the signal to the driver.
For example, the speed limit sign will be displayed only if the vehicle exceeds the maximum speed allowed, or a traffic sign will be shown if the vehicle is approaching it at a potentially dangerous speed and, in both cases, there will be a different degree of importance (visual, acoustic, visual and acoustic, flashing, etc.) in line with the danger of the situation.
Operation
This solution consists in a detection and classification of traffic signs based on a three-level algorithm and consisting of: colour segmentation, shape recognition and neural network.
The final objective of the algorithm is to detect and classify just about all the traffic signs along the way.
Colour segmentation was included for the purpose of obtaining a reply in real time, since segmentation based on colour is faster than the one based on shape.
Two different methods are used to detect the shape; one is based on diagrams that overlap simple shape models while the other is based on the detection of the outline and of the geometric contour.
The set of traffic signs taken into consideration was subdivided into different categories, in accordance with their shape and colour.
Finally, a neural network was built and instructed for each group of traffic signs.
Special devices are used to reduce dependence on external lighting conditions: this is extremely important in terms of good performances in the early morning and late afternoon hours, when sunlight presents a considerable deviation towards the red spectrum.