| Summary: | 碩士 === 國立屏東科技大學 === 資訊管理系所 === 101 === In the last decade, the amount of traffic vehicles is increased accompanying with the development of advanced technology, which causes the significant increase of traffic accidents. According to the statistics from the National Police Agency, the cases of annual car accidents are at least more than 100,000. The major reasons among these cases are improper driving including improper lane change. In order to avoid the accident resulted from the carelessness of paying attention to the surrounding circumstances and blind spot issue of the rear-view mirror, the safety detecting systems of traffic lane change is provided for reducing the incidence of traffic accidents. The system contains (1) a pre-processing module which can convert the two-side video of car into YCbCr image. The threshold is empolyed to remove the nonessential information; (2) a characteristic processing module which use connected component analysis, edge-detection and the corner as characteristic selection; (3) the car-sided detection module, a trend analysis based on the characteristic information can generate an alert to remind drivers for avoiding traffic lane change while the rear car is approaching. The system meets the real-time requirements, the spending time of each two-sided detection of car is around 34 ~ 45ms, and more than 20 images can be obtained per second. The accuracy rate of recognition is 94%. These results show that the system fit’s the requirement of safe detection of traffic lane change.
|
|---|
