Unstructured Road Detection using a Superpixel-Based Image Segmentation Clustering for Autonomous Vehicles

In real-world applications, such as autonomous driving, image-based road recognition is both critical and challenging. This research paper focuses on detecting unstructured roads using an enhanced feature-based method. The algorithm we propose uses the image segmentation method. The two best methods, called K-means and Fuzzy C-means, are being selected, and to improve the result of road detection, we combine with Superpixel Linear Iterative Clustering (SLIC) that converts pixels into superpixels that accurately depict the structure of the objects in the picture. According to the experimental findings, the most effective proposed algorithm for detecting potholes and road prediction for unstructured roads is the modified K -Means algorithm, which employs superpixels (SP-Kmeans) as clustering objects instead of pixels since it provides the maximum accuracy, reaching an average accuracy of 92.47 percent.