Traffic Detection System from Real-time Images

• Main Authors: Shun-Huang Hong, 洪順煌
• Other Authors: Chung-Cheng Chiu
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/04403921285855317142

碩士 === 國防大學中正理工學院 === 電子工程研究所 === 95 === This thesis proposes a real-time system to recognize and track multiple vehicles on roadway images. The system mounted over the expressway to snap a sequence of images (image size : 320 240 full color). Then, the system uses a moving object segmentation method to separate the moving vehicles from the image sequences. After the object segmentation, the objects can be classified and counted by the proposed recognition and tracking methods respectively. Because the CCD camera is mounted on a far location from the roadway, we will encounter the occlusion problems. In this system, the occlusive problems are solved by the proposed occlusive segmentation method and then each segmented vehicle is recognized according to their outlines and tracked by a tracking algorithm at the same time. Even though the occluded vehicles appearing in the images have been keeping merging or having more than two vehicles to merge together all the time, the system can still segment the vehicles. The proposed recognition method uses the visual length, visual width, and roof of vehicles to classify the vehicles in to vans, utility vehicles, sedans, mini trucks, or large vehicles. In this thesis, we use test video data more than 7 days that were collected under various weather and lighting conditions to evaluate the system performance. Experiments obtained by using complex road scenes are reported, which demonstrate the validity of the method in terms of robustness, accuracy, and time responses.