AUTOMATIC DAMAGE DETECTION OF

HISTORIC MASONRY BUILDINGS BASED ON

DENSENET DEEP LEARNING MODEL
 Introduction
Abstract
Vision-based manual inspection technology for identifying and assessing
superficial damage of historic buildings is time- and labor-consuming. At
present, the surface damage detection method for historic heritage sites relies
mainly on in-situ visual inspection techniques supplemented by professional
equipment . Although this method has significant advantages in operation, when
inspectors lack experience, they will assess the damage inaccurately, which has
a serious impact on the later structural safety assessment and repair. Large-
scale inspection using this method is an inefficient and tedious process.
Moreover, the later data analysis and processing requires many experienced
professionals, which is time-consuming and labor-intensive. Therefore, this
method cannot meet the needs of the rapid damage detection of historic
buildings. To overcome these limits, this work proposed a novel automatic
damage detection technique using Faster R-CNN model based on DenseNet
framework to detect two categories of damage (efflorescence and spalling) for
historic masonry structures. The proposed defect detection model verified using
MATLAB Simulink model.
 Literature Survey
Title of the Author & Methodology Findings
paper Date
A New • Zhening Evaluates the performance Study of state of the art
pavement crack Huang of the state of the art crack crack detection
dataset and a • Weiwei detection algorithms. algorithms, including
comparison Chen 1.DeepCrack
study of crack • Abir Al- 2.VGG-16/UNet
detection Tabbaa 3.Resnet34/Unet
algorithms • Ioannis on the proposed dataset,
Brilakis to objectively evaluate
July 2022 their performance
Recognition of • Jose E. The use of object-oriented Multispectral classification
materials and Merono based classification and with visible and near
damage on • Alberto J. recording non-visible light, infrared bands increase
historical Perea in order to detect damage recognition among
buildings using • Maria J. in materials employed in different damages,
digital image Aguilera the construction of integrating data in a 3D
classification • Ana M. historical buildings used in model allows us to obtain
Laguna the restoration and its thematic maps with the
June 2014 quantification. size and position of the
damage.
 Literature Survey
Title of the paper Author & Methodology Findings
Date
Monitoring • Manlio Uses IoT, Structural Inspect the critical
Ancient Bacco health monitoring, VR, structural damage, patterns
Buildings: Real • Paolo unmanned aerial of cracks in the structural
Deployment of an Barsocchi vehicles, condition components of the building
IoT System • Pietro monitoring, being monitored. Marker-
Enhanced by Cassara ArUco Coded Markers , based SLAM algorithm
UAVs and Virtual • Danlia Close range provides a complete and
Reality Germane photogrammetry robust 3D information
se about specific critical
March 2020 points along the crack.
Automated • Byunghy Mask R-CNN, Feature Explains the architecture of
Multiple Concrete un Kim Extraction by CNN and Mask R-CNN that is used
Damage Detection • Soojin Feature Pyramid for general detection
Using Instance Cho Network Region purposes, can be modified
Segmentation November Proposal Network, to detect long and slender
Deep Learning 2020 Object Classification and shapes of cracks, rebar
Model Bounding Box exposure, and efflorescence
Refinement in further research.
 Literature Survey
Title of the paper Author & Methodology Findings
Date
Detection of Surface • Nhat-Duc Gray intensity Capable of identifying crack
Crack in Building Hoang adjustment method, objects and analyzing their
Structures Using May 2011 called Min-Max Gray characteristics. The M2GLD
Image Processing Level Discrimination indeed can improve the
Technique with an (M2GLD) to performance of the Otsu
Improved Otsu preprocess the image method.
Method for Image thresholded by the
Thresholding. Otsu method.
A Comparative • M.R.S. • Otsu It is observed that a general
Review of Image Zawad1 • Statistical structure of CNN-based
Processing Based • M.F.S. Approach algorithm with camera
CrackDetection Zawad2 • CNN images for crack detection
Techniques. • M.A. • Fuzzy could be an efficient
Rahman • Thresholding approach with higher
• S.N. Method accuracy
Priyom
July 2021
 Existing Methodologies

• Discrete cosine transform (DCT)

• Set Partitioning in Hierarchical Trees (SPIHT) algorithm
• AexNet
 Drawbacks of existing
methodology

• If the training data set is limited and the parameters are too many, overfitting is
likely to occur.
• The larger the network model structure is, the more complex the calculation will
be, making it difficult to carry out practical application.
• The deeper the network model structure is, the more easily the gradient is lost in
the deep layer, and the optimization model is difficult to be optimized.
• Degradation: Accuracy first increases with depth, and when saturation is
reached, the increase of model depth leads to the decrease of accuracy
 Proposed Methodology

• This work proposed a novel automatic

damage detection technique using Faster R-
CNN model based on DenseNet framework to
detect two categories of damage (efflorescence
and spalling) for historic masonry structures.
 Block diagram

Input Preproces DenseNet

Filtering
Image sing

Performa Severity
nce Result detection
analysis
DenseNet Architecture
 Components of DenseNet include:

 Connectivity

 DenseBlocks

 Growth Rate

 Bottleneck layers
• Connectivity
• In each layer, the feature maps of all the previous layers are not summed, but
concatenated and used as inputs. Consequently, DenseNets require fewer
parameters than an equivalent traditional CNN, and this allows for feature reuse as
redundant feature maps are discarded. So, the lth layer receives the feature-maps of
all preceding layers, x0,...,xl-1, as input:

• where [x0,x1,...,xl-1] is the concatenation of the feature-maps, i.e. the output

produced in all the layers preceding l (0,...,l-1). The multiple inputs of H l are
concatenated into a single tensor to ease implementation
 DenseBlocks

•The use of the concatenation operation is not feasible when the size of feature maps changes.

•However, an essential part of CNNs is the down-sampling of layers which reduces the size of
feature-maps through dimensionality reduction to gain higher computation speeds.

•To enable this, DenseNets are divided into DenseBlocks, where the dimensions of the feature maps
remains constant within a block, but the number of filters between them is changed.

•The layers between the blocks are called Transition Layers which reduce the the number of
channels to half of that of the existing channels.

•For each layer, from the equation above, Hl is defined as a composite function which applies three
consecutive operations: batch normalization (BN), a rectified linear unit (ReLU) and a convolution (Conv).
 Bottleneck layers

• Although each layer only produces k output feature-maps, the number of inputs
can be quite high, especially for further layers. Thus, a 1x1 convolution layer can be
introduced as a bottleneck layer before each 3x3 convolution to improve the
efficiency and speed of computations
DenseNet-121 Architecture
 Software requirements
• MATLAB 2020b
 Advantage of proposed
methodology
• Less training needed
• Temporal features considered
 References
C.S. Nadith Pathirage, J. Li, L. Li, H. Hao, W. Liu, P. Ni, Structural damage identification based on

autoencoder neural networks and deep learning, Eng. Struct. 172 (2018) 13–28,

https://doi.org/10.1016/j.engstruct.2018.05.109.

C.S. Nadith Pathirage, J. Li, L. Li, H. Hao, W. Liu, R. Wang, Development and application of a deep

learning–based sparse autoencoder framework for structural damage identification, Struct. Health Monit.

(2018), https://doi.org/10.1177/ 1475921718800363

. M.H. Rafiei, W.H. Khushefati, R. Demirboga, H. Adeli, Supervised deep restricted Boltzmann machine for

estimation of concrete, ACI Mater. J. 114 (2) (2017) 237–244, https://doi.org/10.14359/51689560

 Software/Hardware Tools
Used

MATLAB 2020b
Input image
Input image after enhancement
gray scale conversion
Binary segmentation
Mask RCNN wit denoising
layer mapping
Results
Evaluation
 Conclusion
•Cracks are typical line structures that are of interest in many computer-vision
applications. In practice, many cracks, e.g., pavement cracks, show poor continuity
and low contrast, which bring great challenges to image-based crack detection by
using low-level features. This work proposed a novel automatic damage detection
technique using Mask R-CNN model based on Dense Net framework to detect two
categories of damage (efflorescence and spalling) for historic masonry structures.
The proposed defect detection model shows better performance in terms of all the
parameters

•
Thank You

Queries??