Vehicle Speed Detection using Haar Cascade Classifier and Correlation Tracking

Abstract

Objectives: The aim of this study is to develop an efficient and cost-effective solution for predicting vehicle speeds using recorded video data. Methods: The proposed system employs a combination of image processing techniques and computer vision to calibrate cameras for traffic simulation, enabling the extraction of information on average vehicle speeds. It utilizes the Haar Cascade Classifier for object detection, followed by a correlation tracker for vehicle tracking. Speed estimation is achieved through the frame differencing method. The dataset comprises 90 minutes of recorded data from highway cameras, showcasing diverse traffic scenarios with various vehicle types (trucks, trailers, cars, buses, and bikes) at varying speeds. Predicted values are compared with ground truth data obtained from a GPS-equipped car, using Mean Absolute Error (MAE) as the evaluation metric. Findings: The algorithm's performance is evaluated, resulting in an average error rate of 1.72 km/h (2.07%). These findings are compared with state-of-the-art data. Novelty: This study introduces a novel system that combines the Haar Cascade Classifier, correlation tracker, and frame differencing method to track vehicle positions, incorporating bike detection into the analysis, and calculate their moving speeds. A relative analysis underscores the system's performance, emphasizing its effectiveness in real-world applications and demonstrating refinement in accuracy assessment. Keywords: Image processing, Vehicle speed estimation, Haar Cascade Classifier, Correlation tracker, Error rate calculation, Computer vision