Video-based distance estimation for a stream of vehicles

Abstract

A novel measuring method of a distance between vehicles in a vehicle stream is presented in the paper. The presented method utilizes vision data for vehicle detection. Input images are converted into binary images, Conversion is performed with the use of analysis of small gradient magnitudes in the input images. Two detection fields, initial and final, are defined. Vehicle detection is carried out by analysis of the state of detection fields. The changes of the state of both detection fields, caused by passing vehicles, are indicate by appropriate image ordinal numbers from the sequence of input images. Application of two detection fields allows determination of a time period between successive vehicles and vehicle speed. The distance between two consecutive vehicles is calculated as a relation of ordinal image numbers. Processing quantities and experimental results are provided.