Implementation of a Real-Time Intersection Accident Detection System (Phase 1)

摘要

The focus of this research is the feasibility study for the implementation of a real-time accident detection system at intersections. After reviewing accident detection algorithms investigated in the prior phase of the research, we explored schemes to improve the algorithm, conducted algorithm tests both in the lab and in the field, investigated system architecture and communications mechanisms, and evaluated the feasibility of implementation. Based on the original optimal algorithm with Discrete Wavelet Transform (DWT) feature extraction method identified from the previous project, two algorithm refinement schemes, a lifting scheme and an overlapping scheme, were investigated. Lab and field test results consistently show that the overlapping scheme has a higher detection rate than the original optimal algorithm, and the lifting scheme has the advantage only in computation time.