Article
Detection Based on Crack Key Point and Deep Convolutional
Neural Network
Dejiang Wang * , Jianji Cheng and Honghao Cai
Citation: Wang, D.; Cheng, J.; Cai, H.
Detection Based on Crack Key Point
and Deep Convolutional Neural
Network. Appl. Sci. 2021, 11, 11321.
https://doi.org/10.3390/app112311321
Academic Editors: Nikos D. Lagaros
and Vagelis Plevris
Received: 20 October 2021
Accepted: 25 November 2021
Published: 29 November 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Department of Civil Engineering, Shanghai University, Shanghai 200444, China; jianjicheng@shu.edu.cn (J.C.);
seekercai@shu.edu.cn (H.C.)
* Correspondence: djwang@shu.edu.cn
Abstract:
Based on the features of cracks, this research proposes the concept of a crack key point
as a method for crack characterization and establishes a model of image crack detection based on
the reference anchor points method, named KP-CraNet. Based on ResNet, the last three feature
layers are repurposed for the specific task of crack key point feature extraction, named a feature
filtration network. The accuracy of the model recognition is controllable and can meet both the
pixel-level requirements and the efficiency needs of engineering. In order to verify the rationality
and applicability of the image crack detection model in this study, we propose a distribution map of
distance. The results for factors of a classical evaluation such as accuracy, recall rate, F1 score, and
the distribution map of distance show that the method established in this research can improve crack
detection quality and has a strong generalization ability. Our model provides a new method of crack
detection based on computer vision technology.
Keywords:
crack detection; deep convolutional neural network; object detection; crack key point;
fusion and filtration of features
1. Introduction
Cracks are critical flaws that affect the behavior and durability of structures, which
can have a negative effect on structural safety. Due to the inevitability and general of cracks
on the surface of concrete structures, the search for efficient and low-cost crack detection
of concrete has been important in structural damage identification. There are two main
directions for the research on crack detection methods: the one is through sensors to test a
static and dynamic response of the structure, based on which, the position and depth of a
crack are identified [
1
–
3
]; the other is through image processing techniques to provide the
position and other information about a crack [4,5].
Image-based methods are simple and effective, so they have gained extensive attention.
Computer image processing and vision technology, as well as the upgrading of computing
hardware and image-based crack detection methods, especially those based on deep
convolutional neural networks, have undergone unprecedented development.
Classical image crack detection methods, such as segmentation by a threshold [
6
], the
edge detection algorithm [
7
,
8
], and the morphological filtering method [
9
], not only identify
cracks effectively but also assess parameters such as crack length and width. However,
their main work is focused on image processing. Crack detection remains a manual process
with low efficiency.
To improve the efficiency of detection, researchers have introduced machine learning
to deal with crack features and have established a classifier to realize automatic crack
detection [
10
–
12
]. Crack detection methods of traditional machine learning algorithms
combined with image processing techniques have been applied in this area.
Machine learning has broadened the idea of applying computer vision methods
for defect detection and condition assessment in civil engineering [
13
] and has brought
about new research directions for all types of detection, including crack detection. Many
Appl. Sci. 2021, 11, 11321. https://doi.org/10.3390/app112311321 https://www.mdpi.com/journal/applsci
