LDCCAES: A Concomitant Perception Methodology Facilitating Real-Time Detection and Estimation of Median-Lane Positioning for Prototype Autonomous Vehicle

In the domain of autonomous vehicle technology, the call for advanced, dependable methods to ensure secure, proficient road navigation is of utmost importance. This paper elucidates an innovative approach to mid-lane estimation, exploiting the strength of contour-based approaches. Our procedure incorporates image processing techniques and contour analysis to spot, estimate lane markings, culminating in a trustworthy trajectory prediction for autonomous vehicles. The proposed methodology commences with preprocessing of the image to bolster the contrast of lane markings and suppress noise interference. Subsequently, we employed our proposed method of Least Distance to the Centroid of a Contour Area in Euclidean Space (LDCCAES) for the precise identification of mid-lane marking segments. The detected contours are subjected to filtration based on their morphological characteristics, thereby eliminating possible false positives. In the following stage, a distance-driven clustering algorithm is harnessed to detect, associate the closest centroids of the sieved contours, paving a cohesive, continuous mid-lane trajectory. The final stage extracts the most significant contour to symbolize the estimated mid-lane, ensuring a robust, dependable output. Our empirical results corroborate the effectiveness of the proposed method. The resilient performance of our mid-lane estimation approach underscores its potential for assimilation into comprehensive autonomous vehicle applications. This groundbreaking research, firmly anchored in the field of computer vision, lays a solid base for enhancing safety and lane-following precision of autonomous vehicle navigation in real-world conditions.