Enhanced Identification of Curving Lanes Using a Neural Network Model with Minimal Complexity

It’s widely acknowledged that lane detection is a vital feature for the autonomous cars that will be used in the transportation systems of the future. In more recent research investigations, attention has been paid to the challenges of extracting and fitting lane markings from photographs of roads. These, however, continue to be useless when the lane is curved and the light is colored. In the case of a curving road, changes in illumination and the structure of the road are the primary factors that influence the effectiveness of lane identification, which might result in accidents. For the purpose of estimating the complete lane by integrating the lane markers under a variety of diverse road circumstances, this study proposes a unique technique that is based on an operational the FLANN paradigm of linked artificial neural networks, that is both low in complexity and efficient. To estimate the curved shape of the streets, we apply trigonometric, polynomial, exponential in nature, and Chebyshev well-designed development-based FLANN models with lane markers retrieved from each area in the road image. We evaluate and verify each model's efficiency on road photos using three standard datasets. The Chebyshev FLANN model outpaces the other three FLANN techniques in terms of computation time and mean accuracy, making it the best of the four. Results from the remaining three models were surpassed by the CFLANN model, which achieved an accurate detection rate of 95.8%.