A Bidirectional 8-Neighborhood Border Tracking and Kalman-Based Centerline Reconstruction Method for Line-Following Robots

Vision-based line-following robots need highly robust and accurate centerline extraction to ensure effective autonomous navigation. Traditional methods suffer from performance degradation when encountering lane situations like crossing, X shaped with diagonal horizon line and etc. This paper introduces a bidirectional 8-neighborhood border tracking algorithm, enhanced with least squares border completion and Kalman-filter-based smoothing. A dual-pass strategy detects both bottom-up and top-down borders, allowing recovery of the missing line part via a combination of the top-down reconstructed centerline segment and the original bottom-up segment before the breakpoint. The top-down reconstructed centerline is filtered using a one-dimensional Kalman filter to improve smoothness and reduce noise and jitter, ensuring stability in the steering response. Experimental evaluation demonstrates a 94% success rate in reconstructing broken lanes and solving the breaks when meeting cross-type discontinuities. These results confirm the applicability of the method in real-time embedded systems for intelligent vehicle steering. Moreover, the algorithm is computationally lightweight and suitable for deployment on low-power embedded platforms.