Road Object Detection and Tracking Research Using Kalman Filter

This research focuses on road object detection and tracking using Kalman filter to improve traffic safety. In our implementation, we employ the Interactive Multiple Model (IMM) algorithm, which integrates multiple distinct motion models to accommodate various target movement patterns. The algorithm implementation involves probability-weighted combination of different Kalman filters, where each filter corresponds to a specific motion model (such as constant velocity or constant acceleration models). Through comparative analysis of multiple models managed by IMM's Markovian switching mechanism, the system dynamically selects the most appropriate model for the current target, thereby significantly improving detection and tracking accuracy. Key functions in our implementation include model probability calculation, filter interaction, and state estimation fusion. The research outcomes are expected to provide more accurate information and data support for traffic management and road safety decision-making, with practical applications in real-time tracking systems requiring robust performance under varying motion conditions.