All Day Front Vehicle Distance Estimation in Highway

• Main Authors: Ming-Chan Sung, 宋明展
• Other Authors: Ying-Kuei Yang
• Format: Others
• Language: zh-TW
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/utz888

碩士 === 國立臺灣科技大學 === 電機工程系 === 106 === In recent years, people have not only pursues automobile performance but also increases their awareness of safety. Major car manufacturers have also been eager to develop Advanced Driver Assistance Systems (ADAS) to enhance the safety of automobiles. The most important function of an ADAS is to issue warnings and assist driving before a dangerous situation occurs in order to avoid traffic accidents. In this thesis, the image processing methods are used to estimate the distance of the preceding vehicle and a method is proposed to decide whether the front has sufficient light. The lane line is used to find the sky block and then determine if there is enough light in the sky block. In vehicle detection, when lighting is sufficient, the shadow of a vehicle and the rear horizontal bumper are used for detection. The vehicle's two red rear lights are used as a base in low light conditions Finally, in distance estimation, this thesis proposes a novel method different from previous methods based on the ratio of lane width. The proposed method is based on the proportion of height position of the vehicle in a image in order to improve system accuracy and processing speed. The experiment conducted in this thesis is based on 19 driving images taken in highways under various circumstances. In the experiment, the light recognition error took place only when highway sign board and windshield wiper are both located in the light detection area when it is raining.. Fog causes unclear images and therefore results in vehicle recognition errors either.