The CrackNet, an efficient architecture based on the Convolutional Neural Network (CNN), is proposed in this article for automated pavement crack detection on 3D asphalt surfaces with explicit objective of pixel‐perfect accuracy. Unlike the commonly used CNN, CrackNet does not have any pooling layers which downsize the outputs of previous layers. CrackNet fundamentally ensures pixel‐perfect accuracy using the newly developed technique of invariant image width and height through all layers. CrackNet consists of five layers and includes more than one million parameters that are trained in the learning process. The input data of the CrackNet are feature maps generated by the feature extractor using the proposed line filters with various orientations, widths, and lengths. The output of CrackNet is the set of predicted class scores for all pixels. The hidden layers of CrackNet are convolutional layers and fully connected layers. CrackNet is trained with 1,800 3D pavement images and is then demonstrated to be successful in detecting cracks under various conditions using another set of 200 3D pavement images. The experiment using the 200 testing 3D images showed that CrackNet can achieve high Precision (90.13%), Recall (87.63%) and F‐measure (88.86%) simultaneously. Compared with recently developed crack detection methods based on traditional machine learning and imaging algorithms, the CrackNet significantly outperforms the traditional approaches in terms of F‐measure. Using parallel computing techniques, CrackNet is programmed to be efficiently used in conjunction with the data collection software.
CrackNet is the result of an 18-month collaboration within a 10-person team to develop a deep learning-based pavement crack detection software that demonstrated successes in terms of consistency for both precision and bias. This paper proposes an improved architecture of CrackNet called CrackNet II for enhanced learning capability and faster performance. The proposed CrackNet II represents two major modifications on the original CrackNet. First, the feature generator, which provides handcrafted features through fixed and nonlearnable procedures, is no longer used in CrackNet II. Consequently, all layers in CrackNet II have learnable parameters that are tuned during the learning process. Second, CrackNet II has a deeper architecture with more hidden layers but fewer parameters. Such an architecture yields five times faster performance compared with the original CrackNet. Similar to the original CrackNet, CrackNet II still uses invariant image width and height through all layers to place explicit requirements on pixel-perfect accuracy. In addition, the combination of a convolution layer and a 1 × 1 convolution layer was repeated in CrackNet II to learn local motifs with different sizes of local receptive fields. CrackNet II was trained with 2,500 diverse example images and then demonstrated to outperform the original CrackNet. The experiment using 200 testing images showed that CrackNet II performs generally better than the original CrackNet in terms of both precision and recall. The overall precision, recall, and F-measure achieved by CrackNet II for the 200 testing images were 90.20, 89.06, and 89.62%, respectively. Compared with the original CrackNet, CrackNet II is capable of detecting more fine or hairline cracks, while eliminating more local noises and maintaining much faster processing speed.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.