A novel algorithm for vehicle motion detection, classification and tracking based on aerial surveillance video in computer vision

Moving vehicle detection is a critical component of intelligent traffic monitoring systems. To address the limitations of traditional frame-subtraction methods, this paper proposes a novel Vehicle Motion Detection, Classification, and Tracking (VMDCT) algorithm. The approach utilizes the Horn and Schunck method to compute optical flow magnitudes for detecting moving objects. In parallel, a binary Support Vector Machine (SVM) classifier with Histogram of Oriented Gradients (HOG) features was used to identify and classify vehicles. The framework integrates multiple techniques to provide a seamless solution for detecting and classifying moving vehicles from aerial cameras, accounting for dynamic variables such as camera altitude and the projected size of vehicles in the scene, while also adaptively measuring motion magnitude across the field of view. To further optimize detection, Pre-processing addresses parallax motion through frame registration and video stabilization, using linear stack alignment combined with Scale-Invariant Feature Transform (SIFT) and point feature matching. The proposed system was evaluated on two aerial video datasets, comprising grayscale frames at standard frame rates, with Gaussian and Contrast-Limited Adaptive Histogram Equalization (CLAHE) filters applied to reduce noise and enhance frame quality. Experimental results demonstrate that the algorithm effectively mitigates the impact of parallax motion and achieves robust vehicle detection and classification under complex conditions, with an average detection accuracy of 95.5%. Quantitative assessments of each processing phase and qualitative comparisons with related systems confirm the effectiveness of the proposed framework. Overall, the system achieves an average success rate of 96.77% for moving object detection and 98.57% for vehicle classification, highlighting its practical applicability and flexibility for intelligent traffic monitoring and future research.