Date of Award
Doctor of Philosophy in Computing Sciences - (Ph.D.)
Automatic traffic accident detection is an important task in traffic video analysis due to its key applications in developing intelligent transportation systems. Reducing the time delay between the occurrence of an accident and the dispatch of the first responders to the scene may help lower the mortality rate and save lives. Since 1980, many approaches have been presented for the automatic detection of incidents in traffic videos. In this dissertation, some challenging problems for accident detection in traffic videos are discussed and a new framework is presented in order to automatically detect single-vehicle and intersection traffic accidents in real-time.
First, a new foreground detection method is applied in order to detect the moving vehicles and subtract the ever-changing background in the traffic video frames captured by static or non-stationary cameras. For the traffic videos captured during day-time, the cast shadows degrade the performance of the foreground detection and road segmentation. A novel cast shadow detection method is therefore presented to detect and remove the shadows cast by moving vehicles and also the shadows cast by static objects on the road.
Second, a new method is presented to detect the region of interest (ROI), which applies the location of the moving vehicles and the initial road samples and extracts the discriminating features to segment the road region. After detecting the ROI, the moving direction of the traffic is estimated based on the rationale that the crashed vehicles often make rapid change of direction. Lastly, single-vehicle traffic accidents and trajectory conflicts are detected using the first-order logic decision-making system.
The experimental results using publicly available videos and a dataset provided by the New Jersey Department of Transportation (NJDOT) demonstrate the feasibility of the proposed methods. Additionally, the main challenges and future directions are discussed regarding (i) improving the performance of the foreground segmentation, (ii) reducing the computational complexity, and (iii) detecting other types of traffic accidents.
Ghahremannezhad, Hadi, "Advanced traffic video analytics for robust traffic accident detection" (2023). Dissertations. 1679.