Traffic Accident Detection Method Using Trajectory Tracking and Influence Maps

With the development of artificial intelligence, techniques such as machine learning, object detection, and trajectory tracking have been applied to various traffic fields detect accidents analyze their causes. However, detecting using closed-circuit television (CCTV) an emerging subject in learning remains challenging because complex environments limited vision. Traditional research has limitations deducing trajectories accident-related objects extracting spatiotemporal relationships among objects. This paper proposes a accident detection method that helps determine whether each frame shows by generating considering influence maps convolutional neural network (CNN). The with were enhanced improve accidents. A CNN is utilized extract latent representations from produced trajectories. Car Accident Detection Prediction (CADP) was experiments train our model, which achieved accuracy approximately 95%. Thus, proposed attained remarkable results terms performance improvement compared methods only rely on CNN-based detection.