Automated Asphalt Crack Detection Using Convolutional Neural Networks

Road damage caused by asphalt cracks is a significant issue in the civil engineering industry as it poses threat to road and highway safety. Detecting classifying difficult undertaking because of intricate pavement conditions created various factors such shadows, oil stains, water spots. These can create challenges differentiating from surrounding pavement. The focus our study was put forward architecture deep convolutional neural network (DCNN) that automatically detect categorize cracks. To train DCNN, we utilized RGB images were captured manually with resolution 1024x768 pixels. then segmented into patches measuring 32x32 During training employed two filter sizes, which 3x3 5x5. Our presented approach achieved recall 98%, precision 99%, accuracy successfully detecting presence images. DCNN also capable With fair classification for both sizes no noticeable difference between cracks, transverse, longitudinal, alligator. In contrast bigger smaller required greater processing time during training. Overall, 94.5% while using suggested method classify different kinds Key Words: Crack detection, Faster R-CNN, Unet, YOLO, ResNet-50