A Traffic Congestion Detection Method for Surveillance Videos Based on Macro Optical Flow Velocity

摘要

An intelligent and automatic traffic congestion detection method is proposed to reduce labor intensive monitoring work. The developed approach, applied on expressways, is to gather traffic congestion information using surveillance videos. The speed of traffic flow, defined as "macro optical flow velocity, is calculated from video pixel's motion directly. The calculation progress includes two steps: 1) extract the feature points, which meeting three conditions: corner, strong motion and main direction consistency; and 2) calculate optical flow vectors of feature points using LK algorithm. The average value is determined as the macro optical flow velocity. According its continuous time feature and other characteristic, the traffic congestion status is automatically and directly detected. The test results of surveillance videos from Guangzhou-Shenzhen expressway have shown that the proposed method could automatically detect traffic congestion in 10 seconds.