Technologies Of Real-time Traffic Sign And Light Detection And Recognition

With the rapid development of the socio-economic, the cars are becoming moreand more popular. However, when enjoying the convenience of the cars, people alsofaced with traffic jams，frequent traffic accidents and low road transport efficiency. Atthis point, intelligent transportation systems (ITS) was born， and has been thewidespread of many scholars. Traffic Sign Recognition(TSR) and Traffic LightRecognition（TLR） based on image processing technology, as important parts of theIntelligent Transportation Systems (ITS), has gradually become a research focus athome and abroad.Technologies of traffic signs and lights detection and recognition play an importantrole in the intelligent vehicle and driver assistance systems. Detection and recognitionof traffic signs and lights in a natural road environment, affected by a variety of roadconditions, weather changes and other factors, are facing greater difficulties.This paper studies the three main types of traffic signs: prohibition signs, warningsigns and direction signs, as well as circular and arrow-shaped traffic lights. This paperused different algorithms to detect and recognize traffic sign and light.Avoid the large amount of computation of the conversion from RGB to HSV, thispaper detected the traffic signs in RGB space and depending on the relationship of threechannels. This method basically solved the problem of illumination changes. Themethod that detected on the small picture and recognized on the original picture cansave a lot of time.In the stage of recognition of traffic signs, three level identification strategies wasused. According to the relationship between color and shape of traffic signs, we firstlyused the boundary distance （DtB） features to class the shapes. Then extracted the ringprojection features of the internal signs and got further identified. Finally, recognizedthe signs by multi-scale, multi-angle template matching.The traffic lights are prone to overexposure. In the detection of traffic lights, weused the color variance in RGB space. We used the color of traffic lights used therelations of three type of colors. In the recognition of traffic lights, we used templatematching to recognize the shape of traffic lights. Finally, we extracted the black lightbox background to confirm the results.A large number of experimental results show that this algorithm of traffic signs and lights detection and recognition has high accuracy and good real-time effects.