FPGA-based Real-time Road Alarm System

With the rapid development of economy in recent years, the highway system is larger, and the number of vehicles is growing exponentially. The car brings us convenience, and plays the more important role in people’s daily life. Due to the deterioration of traffic accidents every year, the car also causes a great loss to people’s life and property, as well as national economy. According to analysis, unconscious deviation from the lane and running red lights are the important reasons leading to traffic accidents. To reduce the number of traffic accidents led by deviation from the lane and running red lights, the vehicle assistant driving system based on machine vision and pattern recognition technique become an important research area of intelligent transportation, is also one of the key and difficult points in the research field of the driverless field.At present, computer technology and artificial intelligence technology are in rapid development, there are many new target location and identification methods. The vehicle auxiliary systems remain having insurmountable obstacles brought by the changeable weather background and complicated road environment. The existing research methods are mostly based on PC algorithm research, because this model can use and assess different processing algorithms rapidly. However, due to large amount of path data information, the serial port of PC is hard to realize real-time processing and feedback, and vehicle auxiliary system based on PC is also difficult to achieve economic and efficient mass production.According to above problems, this paper presents a lane departure warning system and a traffic light automatic identification system based on FPGA. The system using FPGA parallel features can effectively achieve the real-time processing system and feedback and design reasonable electronic solutions to achieve effective insurances for mass production. Around these problems, this paper has research mainly as follows:1. On a large number of lane departure warning system research, this paper put forward a complete set of lane departure warning system based on FPGA, and elaborated how to implement Gaussian Smoothing Algorithm on FPGA, not only make full use of the excellent characteristics of the FPGA parallel, also saves precious resources. In the lane line detection part, used the simple linear model to guarantee the lane line detection effect, and also proposed an early-warning model based on linear model, greatly reduced the complexity of the system and made great contributions to the real-time requirement.2. In the automatic recognition of traffic light, color and geometry characteristics are the most obvious characteristics of the traffic light. This system fully combined the features above together and segment color image first then extract the candidate areas of the traffic lights. Had filtered and geometrical features to obtain the stability of the candidate region and then to circle fitting based on areas extracted, greatly improved the detection accuracy of the traffic light. Analyzed the advantages and disadvantages of Hough and least square method in the circle fitting process, and finally realized the method of least square circle fitting approach on FPGA.