DAUNet: Deep Augmented Neural Network for Pavement Crack Segmentation

Abstract

Crack detection and measurement are essential tasks for maintaining and ensuring safety. Accurate crack detection is very challenging because of non-uniform intensity, poor continuity, and irregular patterns of cracks. The complexity of the background and variability in the data acquisition process also complicate the problem. Many approaches to crack detection have been proposed, but the accuracy of the detection leaves much to be desired. The aim of this study is to develop a practical crack detection method for real-time maintenance. We focus on a deep end-to-end and pixel-wise crack segmentation. We propose a lightweight U-Net-based network architecture with emphasis on the learning process. In order to verify the effectiveness of the proposed method, we conduct tests on publicly available pavement crack datasets and compare our model with state-of-the-art crack detection methods. Extensive experiments show that the proposed method effectively detects cracks in a complex environment, and achieves superior performance.

A Crack Segmentation Model Combining Morphological Network and Multiple Loss Mechanism

“…The objective evaluation indicators used in this paper were the segmentation performance evaluation indicators ODS (Optimal Dataset Scale), OIS (Optimal Image Scale), AIU (Average Intersection over Union) [ 12 ] and sODS (simplified versions of ODS) and sOIS (simplified versions of OIS) [ 18 ]. OIS is the aggregate F measure of the best threshold in each image in the dataset, and ODS is the best F-measure on the fixed threshold dataset.…”

“…Therefore, the U-Net network has been widely used in image target segmentation, since it was proposed for medical image segmentation. Due to its excellent segmentation performance, U-Net network and its extension have also been widely used in crack segmentation [ 18 , 19 , 20 , 21 , 22 ] and rust detection in metallic constructions [ 23 ]. Hong et al [ 20 ] introduced an attention module to the U-Net network and fused the features of skip connections to segment cracks on UAV aerial photography pavement.…”

“…Crack segmentation methods based on machine vision are mainly divided into traditional algorithms [ 1 , 2 , 3 , 4 , 5 , 6 , 7 , 8 , 9 ] and deep-learning-based methods [ 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 , 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 , 26 , 27 , 28 , 29 , 30 , 31 , 32 , 33 ].…”

“…Depth-learning-based methods can automatically extract the shallow and deep characteristics of cracks, so they can greatly improve segmentation and detection accuracy with regard to cracks. The backbone networks commonly used for crack segmentation include the multi-scale pyramid structure network [ 10 , 11 , 12 , 13 , 14 , 15 , 16 , 17 ] and the U-Net network [ 18 , 19 , 20 , 21 , 22 ]. Methods based on multi-scale features can extract more abundant crack image features, but the existing methods make it difficult to detect cracks similar to the background.…”

“…Methods based on multi-scale features can extract more abundant crack image features, but the existing methods make it difficult to detect cracks similar to the background. Because of its good performance in medical image segmentation, the U-Net network and its expansion are widely used in crack detection [ 18 , 19 , 20 , 21 , 22 ]. From the experimental results of many studies, it can be seen that the methods based on deep learning have a significant improvement in the effects of crack segmentation.…”