Design & Implementation of Realtime Vehicle Safety Image System based on Embedded System

• Main Authors: CHEN,SHIH-SING, 陳士興
• Other Authors: Ching-Lung Su
• Format: Others
• Language: zh-TW
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/e249a8

碩士 === 國立雲林科技大學 === 電子工程系 === 102 === With rapid advancement of science and technology, Automobile Electronics has flourished recently. Products such as GPS, EDR, digital television…etc., are highly praised by customers. There are many kinds of products released, and automobile safety system is the product that every international automobile company are researching and developing for. High-level cars, such as Mercedes Benz, Volvo, can choose this system. Because of the expensive price, most of the customers can’t afford it. Moreover, automobile safety system uses radar, infrared ray and laser for sensors. These sensors not only cost high, but also are influenced easily by the atmosphere, weather and the angle of detection. Therefore, I propose using double cameras as the main sensors. Compared with radar, infrared and laser, using camera as sensor could lower the cost and not influenced by atmosphere, weather and there’s no limit of angle of detection. For range detection, I combine single lens with twin lens in order to improve the accuracy of distance measurement of the system. I proposed the detections of automobile safety system in this thesis include Multi-Lane Detection, Curve Detection, Lane Departure Warming System, Vehicle Detection, and distance measurement by single and twin lens subtense methods. Multi-Lane Detection Algorithm uses luminance difference to search for the information of lane strips in the screen. Differ from using Sobel filter to search for lane strips, this detection can decrease the computational complexity. There are two conditions for detecting cars, daytime and nighttime. We search the shadow of the car when in the daytime, and search taillights when in the nighttime. In order to increase the speed of daytime and nighttime determination, using the feature that we can’t find taillight in the daytime to find taillight first. Then, searching for car’s shadow if we can’t find the taillight. Last but not least, in order to avoid car accident, I use distance measurement by single and twin lens subtense methods as the distance between cars. When the distance minor than safe distance, there will be warnings.