earticle

영어

This study focuses on the design and performance evaluation of a lightweight object detection model, YOLOv11BiFormer, for the efficient detection of road surface defects such as alligator cracks, longitudinal cracks, potholes, and transversal cracks. To enhance computational efficiency and improve the detection of small-scale objects, the proposed model integrates the BiFormer block and C2f module into the existing YOLOv11 architecture. The dataset used for training and evaluation consists of 7,238 highresolution images, which were evenly divided into 5,065 training images and 2,137 validation images across the four defect categories. Experimental results show that the YOLOv11BiFormer model outperforms the original YOLOv11 in multiple metrics: mAP 0.5 improved from 0.522 to 0.546, mAP 0.5:0.95 increased from 0.691 to 0.703, and precision rose from 0.462 to 0.497. Furthermore, the number of parameters and model size were reduced from 2,582,932 to 2,464,956 and from 5.5MB to 5.2MB, respectively. Visual analysis also demonstrated superior detection accuracy and clearer boundary identification with the BiFormer-enhanced model.These findings suggest that the proposed YOLOv11 BiFormer model is well-suited for real-time road defect detection in mobile devices and edge computing environments, offering a promising solution for intelligent transportation systems and automated infrastructure inspection.