Automatic traffic data extraction tool for mixed traffic conditions using image processing techniques

This study addresses challenges related to extracting detailed data about vehicle movements in diverse traffic situations, such as those in India. Manual data extraction with the desired precision requires considerable human resources and time, posing a significant obstacle to obtaining the vast amounts of data needed for analysis. To overcome this issue, the study introduces a computer-based offline tool that utilizes advanced technologies, including YOLOv4 deep learning and the SORT algorithm, to analyse recorded traffic videos. This tool can identify vehicle types, track their paths, and calculate speed. The final output of the tool is presented in the form of a spreadsheet. The tool’s accuracy was confirmed by comparing its outputs with manually collected data, demonstrating its reliability in various traffic situations. It is anticipated to be effective in similar traffic conditions in other Asian countries.