A Novel Approach for Vehicle Classification and Counting at an Unsignalized Intersection.
Authors :- Bhatt, K., Kshirsagar, V., Shah, J., Bhalerao, R.
Publication :- Innovation in Smart and Sustainable Infrastructure, Volume 2. ISSI 2022. Lecture Notes in Civil Engineering, vol 485. Springer, 2024
raffic data extraction for intersections is one of the major problems challenged by the researchers. The analysis of traffic behavior is dependent on the traffic data collected using videography. Although the data extraction from the videography for developing countries like India under heterogeneous traffic conditions is a quite challenging and time-taking task. Evaluating the classified traffic count for each approach at the intersection is quite time-consuming as well as enhances the chance of human error during manual counting. Till now, the tools developed are used to analyze the straight movement classified traffic count. However, in this paper, data extraction methodology classifies and counts the vehicles at an uncontrolled intersection by the recorded video. The algorithm enables the classification of vehicle counts for each approach to a three-legged uncontrolled intersection. The time duration for data extraction for 30 min of video is reduced to 5–6 times as compared to manual extraction. The results have been compared for manual extraction to algorithm results. The accuracy percentage obtained for vehicle counting is approximately 86%, and error is observed due to the wrong side driving behavior. Additionally, the accuracy for vehicle classification is recorded as 95%, and the error for classification is due to the area misinterpretation during turning movements; the camera placed at an angle from the location deals to misinterpret a few vehicle classes between the three-wheeler and four-wheeler. Moreover, this algorithm is useful for extracting the traffic data from the videography, and this method will be further modified for the real-time videos of the traffic survey for different types of intersections.