운전자보조시스템을 위한 오픈소스 하드웨어 기반 차선 및 차량인식에 관한 연구

Abstract

Recently automobile technology is a key element of the car has been changed into an electronic system computerization. Such aautomobile, depending on the development of the core IT system that combines the intelligent car technology has evolved. As intelligent vehicle technology is increasingly common trend that more and more electronic devices are plugged in and connected electronic devices are coupled via the interoperability. These electronic devices are becoming more service and convenience, the kind used in engagement increases with the increasing use of advanced models thus numerous electronic devices in the car has emerged from the need of a new electronic control unit. However, the modules and wiring connecting the electronic device and an electronic device mounted on the vehicle will increase. This not only increases the weight and take up a lot of space is generated and expensive. In this paper, to solve this problem by using the open source hardware conducted a study on the more advanced driver assistance systems integration software features and techniques of the lane recognized by the vehicle at least the electronic device. Lane recognition is to reduce the operation time by setting multiple regions of interest, increased the recognition performance. To reduce erroneous detection rate of the other lanes in the lane are detected by the pre-processing to extract the lane boundary line, and recognizes the lane through the Hough transformation. Using information on the slope of the lane change were to occur, such as a lane, a warning signal to the driver when lane departure. Vehicle recognition information is the outline of the vertical and horizontal components of the vehicle in a knowledge-based method for extracting by using a shape or color, and the step of forming the vehicle with the hypothesis of the Haar-like features were used. After learning the vehicle and non-vehicle in the verification stage of the appearance-based method it was used for hypothesis using the learned results for the vehicle recognized by the Adaboost classifier. Experimental results were confirmed the smooth real-time performance and potential of the system, an electronic device that could solve the problems of space, weight and costs accounted in the vehicle.