<p>The effectiveness of Lane Centering Assist Systems (LCAS) relies heavily on accurately detecting lane markings across diverse road surface conditions. Generally, lane markings are segmented using fixed threshold values based on the color of the markings (WHITE/YELLOW) and this method does not perform well for all road conditions. To address the limitations of fixed threshold methods, this study proposes a novel dynamic thresholding approach based on an adaptive confident road region estimation (DT-ACRRE) algorithm to enhance the robustness and accuracy of lane detection. The approach leverages a computationally efficient Fast Normalized Cross Correlation (FNCC) technique and Connected Component Analysis (CCA) for lane marking edge detection and grouping. By incorporating virtual vanishing point estimation and applying the least-squares approach, left and right lane markings are effectively detected. The developed algorithm has been validated using tuSimple and KITTI benchmark datasets, as well as real-time experiments with a scale-down prototype vehicle equipped with the developed LCAS, demonstrating significantly improved accuracy in detecting lane markings in indoor and outdoor conditions. With an average accuracy of 96.8% which is nearly a 4% improvement over existing methods at a computational cost of 25.3&#xa0;ms per frame, the proposed (DT-ACRRE) based lane detection algorithm shows notable promise for enhancing the adaptability and accuracy of LCAS in diverse road conditions. These findings underscore the algorithm's potential to make LCAS more effective and reliable in real-world driving scenarios, ultimately contributing to enhanced road safety.</p>