Enhanced Asphalt Crack Detection Using Unet++: Combining Multi-Class Segmentation and Crack Measurement

Road safety and the longevity of infrastructure depend on the identification and examination of asphalt cracks. Three gradually enhanced Faces for automatic crack segmentation and analysis are designed and evaluated in this study using the Unet + + architecture. First Face uses binary segmentation with simple picture preprocessing. Advanced preprocessing and crack size measuring are integrated in Second Face. To differentiate between places that are cracked and those that are not, Third Face offers a more thorough study of road defects by using multi-class segmentation. The results show that the improved Faces are superior, with Third Face attaining the highest accuracy (99.09%) and IoU (0.91). These developments highlight how automated road maintenance systems can be enhanced by incorporating advanced pretreatment and post-processing approaches.