Research And Application Of Vehicle Detection In Traffic Video Based ViBE

The traffic monitoring video contains a large amount of information and complete types.Vehicle detection and location algorithm based on video image frame has been widely used in intelligent traffic monitoring.In recent years,with the rapid development of image processing technology,multi-moving detection algorithm based on traffic video has gradually become a research hotspot.This paper first studies the target motion information detection algorithm and improves the ViBE algorithm in two aspects.Firstly,the sampling point is expand from the original neighborhood to the and different weights was allocated to the sample point to reduce the repetition rate.Second,the original fixed distance threshold was improve to the adaptive distance threshold.Then obtain the marked moving target possible region by morphological closing operation,the intercepting the same region in the original image.According to the nested cascade classifier algorithm combined with Haar-Like and MB-LBP features design a classier.Finally,the vehicle position information extracted from the video frame by using the NMS(nonmaximum suppression)algorithm for constraint screening detection.The experimental results of the improved algorithm show that the precision,recall and F1 values is higher 1.5%,5.4%,and 3.8% than the original algorithm.Based on the improved ViBE multi-feature NCGAB classifier,the detection accuracy on CDW2014 and WF paper datasets were 76.2% and 84.7%,respectively,and the false detection rates were 12.4% and 8.1%,the missed detection rates were 11.4% and 7.29%.