Scholarlink

This paper introduces OptiTrackEx, a novel method for detecting vehicular collisions using dashcam footage. OptiTrackEx leverages the advanced YOLOv8 model combined with localized Optical Flow analysis to extract spatial features critical for identifying vehicle collisions. These features undergo thresholding for anomaly detection and are further processed using a Convolutional Neural Network (CNN) for frame-by-frame collision prediction. The method’s universality is a key advantage, enabling effective generalization across various vehicle types and facilitating real-time application. OptiTrackEx outperforms existing video classification methods achieving an F1 score of 0.947. This superiority stems from its efficient feature extraction and reduced computational requirements, positioning OptiTrackEx as a significant innovation in traffic safety and real-time collision detection technology.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

OptiTrackEx: A Deep Learning Approach to Real Time Vehicle Collision Detection

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Soumick Sarker,
  • Anima Pramanik,
  • Sobhan Sarkar,
  • Indranil Bose

摘要

This paper introduces OptiTrackEx, a novel method for detecting vehicular collisions using dashcam footage. OptiTrackEx leverages the advanced YOLOv8 model combined with localized Optical Flow analysis to extract spatial features critical for identifying vehicle collisions. These features undergo thresholding for anomaly detection and are further processed using a Convolutional Neural Network (CNN) for frame-by-frame collision prediction. The method’s universality is a key advantage, enabling effective generalization across various vehicle types and facilitating real-time application. OptiTrackEx outperforms existing video classification methods achieving an F1 score of 0.947. This superiority stems from its efficient feature extraction and reduced computational requirements, positioning OptiTrackEx as a significant innovation in traffic safety and real-time collision detection technology.