A study on Intersection Warning System with Artificial Intelligence Technology

• Main Authors: TU,YEN-YU, 凃彥羽
• Other Authors: LIN,LIANG-TAY
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/rn853m

碩士 === 逢甲大學 === 建設碩士在職學位學程 === 107 === There are many reasons that cause the traffic accidents, such as not paying attention to the situation in front of the cars, not following the traffic signs, not going after the cars leave, not making a proper turn and the carelessness of driving etc. And the highest frequency is not paying attention to the situation in front of the cars, which means that the drivers are distracting. First of all, this article will focus on the previous research on how to reduce the factors of the risk, including how to reduce the unattended situation, blind spot detection, etc. This research focuses on the analysis of immediate road conditions and the identification of vehicle types through YOLO image analysis applications. It classifies four categories, large vehicles, medium-sized vehicles, scooters and pedestrians. Analyze the direction of vehicles at the intersection by exhaustive methods to reduce false judgements. Publish the classified signal information to the CMS immediate board to remind people to pay attention to the intersection status and reduce the accident rate at the intersections. At present, the intersections which are not signalized in Taichung City accounts for about 48% of the total accident rate. It is expected that the accident rate will be reduced by 30% in this way. In addition to the YOLO image analysis method, this article also explores how the environmental factors of the camera affect the recognition, such as illumination, rainy day, cloudy day, tail lights, ground reflection, etc.. It is expected that through such classification, it won’t take too much time to analysis recognition or reduce the readability due to the angle problem of camera setting.