Abstract
In the process of traffic accident forensic appraisal, the analysis of accident footage is a critical component. In practice, analysts frequently encounter a vast amount of “non-standard” video footage. Common issues include unstable local frame rates, frame dropping phenomena, and the absence of on-screen timestamps. Consequently, accurately determining the frame rate for such footage becomes the primary task in video analysis. To address this practical challenge, this study utilizes a video playback kernel integrated with multiple SDKs and combines it with a camera self-calibration method to develop a video player designed for image analysis and speed estimation. This player possesses robust capabilities for stable frame rate analysis, as well as flexible and accurate calibration and distance measurement functions. Experimental validation demonstrates that this player enables rapid speed estimation from monocular fixed dashcam footage on statutory standard roads featuring broken lane lines.
Key words: Video analysis, Monocular camera, Player, Speed estimation, Self-calibration, Frame rate