Road Traffic Sign Identification in Weak Illumination for Intelligent Vehicle Based on Machine Vision
Zhaofei Huang,Shian Qiu,Jun Li,Yibing Zhao,Pan Cui,Ronghui Zhang
DOI: https://doi.org/10.2174/2212797611666180309162545
2018-01-01
Recent Patents on Mechanical Engineering
Abstract:Background: For life, cars have brought convenience to people, besides causing environmental pollution, traffic congestion, and road safety. One of the main reasons for traffic accidents is that drivers cannot respond to road state and information timely. Intelligent vehicle and safety driving assistance technology are one of the ways to reduce traffic accidents. Many papers and patents are studying how to reduce the rate of traffic accidents. Objective: Traffic sign recognition is an important part of the intelligent vehicle system and advanced driver assistance system. To reduce the rate of traffic accidents in weak illumination, this paper proposes a detection and recognition method for intelligent vehicle traffic sign based on machine vision. Methods: Firstly, the traffic signs are reclassified from the color and shape characteristics of traffic signs. Then, the concept of color shape pairs is put forward, and the color-geometric model is established. To achieve the enhancement effect, the image is preprocessed by the uniformity of the histogram. We mainly use the fast algorithm of invariant moments and Zernike moments to extract the sign features. Finally, template matching and support vector machine are used to traffic sign image recognition. Results: The experimental results show that the technology has good robustness in complex environments such as weak light, occlusion and shadow and it can improve the image recognition rate effectively. Conclusion: Compared with the traditional method, it provides a method for intelligent vehicle traffic sign identification under weak illumination conditions. Keywords: Advanced driver assistance system, intelligent vehicle, machine vision, support vector machine, template matching, traffic sign, weak light.