Deep Discriminant Learning-based Asphalt Road Cracks Detection via Wireless Camera Network

The detection of cracks on asphalt pavement is an important task to ensure the safety of driving and the service life of pavement. Based on a visual camera-based asphalt pavement crack detection system introduced in this paper, the asphalt pavement image data can be wirelessly transmitted and the detection of cracks within the image data can be carried out in real-time. Main steps of the processing are as follows. First, the transmitted images are preprocessed via histogram equalization, Gaussian filtering, edge detection, morphological gradient, data augmentation, etc. Then, this paper proposes three deep discriminant learning-based models as classifiers, whose performance are also compared with those of classifiers based on popular deep generative adversarial learning-based models and classic machine learning models. Experimental results demonstrate that the detection method based on the VGG19 classifier has the most satisfactory performance, in which an average $F_{1}$ score of 0.8886 is reported. Also, the introduced method is robust for various situations in this asphalt pavement crack detection study.