Robust Method for Analyzing the Various Speeds of Multitudinous Vehicles in Nighttime Traffic Based on Thermal Images

Abstract

This paper presents a robust method for analyzing the various speeds of multitudinous vehicles by utilizing thermal images as the source of operation. The objective of research is to analyze the speeds of vehicles when encountering the light constraints of a nighttime traffic. For the operational steps, three spontaneous thermal features of a vehicle consist of the windscreen, radiated heat from the engine , and elsewhere will be initially detected; the mentioned features that imply a vehicle type, then, will be classified and tracked. The final process is to analyze the speeds of multitudinous vehicles. Firstly, the locus of each tracked vehicle will be continually searched from frame to frame in order to utilize for computing a displacement. Hereafter, the displacement was known, the averaged speed of vehicle can be found by dividing the displacement with an interval time between the consecutive frames. With this rational behind, for all detected vehicles, the various speeds of multitudinous vehicles are able to be identified by repeating the same operational pattern to all vehicles that appeared in a frame. It is found that, when implementing the analysis method to sample thermal image frames recorded from a nighttime traffic of Rangsit-Nakornayok road, various vehicle types and speeds could be continually indentified when changing from a frame to a frame. Meanwhile, the correction of a speed analysis of multitudinous vehicles is over 59 percents, considering on the situations that has one vehicle, two and three consecutive vehicles. However, the percentage is relatively low for the situations that over three consecutive vehicles are appeared, and it is a challenge for undertaking further research.