There is an increasing interest in detecting heat loss through buildings using unmanned aerial vehicles (UAVs) and thermal sensors. The present study constitutes an attempt to develop a system that can detect heat loss in different inaccessible portions of building structures, such as roofs and high-rise facades. Traditionally, inspectors have conducted surveys to investigate insulation performance and detect heat loss through various portions of buildings. However, these kinds of surveys tend to be time-consuming, costly, and risky. To mitigate risks, a small, low-cost Adafruit thermal infrared sensor and a small, onboard Raspberry Pi microcontroller were mounted on a UAV to detect heat loss through buildings. The lightweight Raspberry Pi microcontroller and Adafruit thermal sensor were powered by additional batteries. A lightweight battery was selected based on the maximum payload and power demand of the microcontroller and thermal sensor. The Raspberry Pi was controlled remotely by a portable computer. The UAV flight plan was controlled remotely by FreeFlight Pro software. Several experimental tests were conducted in both indoor and outdoor environments. Both video and image data were obtained remotely from the thermal sensor and microcontroller. A standard FLIR thermal camera with a very high resolution was also used to ensure the accuracy of the results obtained from the UAV-based thermal sensor. All the images captured by the Adafruit thermal sensor were compared with the standard thermal camera images. The results showed that the presently developed system can detect heat loss through inaccessible locations in buildings with modifications only in sensor resolution.
A new computer vision-based method is proposed for concrete crack detection in tunnel structures using multi-spectral dynamic imaging (MSX). The MSX images were collected from a tunnel in the University of Manitoba, Canada. A total of 3600 MSX images (299 × 299 pixels) were used to train the modified deep inception neural network (DINN), and an additional 300 MSX images (299 × 299 pixels) were employed for validation purposes. The MSX images were examined by the trained neural network for concrete crack detection. The main purpose of this research was to examine the potential of the neural network to distinguish between noise and concrete surface cracks in the MSX images. A fully connected layer and a softmax layer were added to the DINN network in the transfer learning section to reduce the network computation cost. The proposed network used green bounding boxes to detect the portions with cracks in the MSX images. A training accuracy of 95.5% and a validation accuracy of 94% were achieved at 1600 iterations. The optimum training steps obtained from the training and validation were used for testing purposes. The robustness of the trained network was evaluated using an additional 96 MSX images (640 × 480 pixels). A maximum testing accuracy of 94% was recorded when the prediction probability was limited to 90%.
Structural health monitoring has become an outstanding tool to perform structural condition assessments, once performed solely by trained experts. In this study, a methodology utilizing an inexpensive depth sensor to detect and quantify volumetric damages within concrete surfaces is proposed. To allow automatic damage detection, a Faster Region-based Convolutional Neural Network (Faster R-CNN)-based method is implemented. A database of 444 images with resolution of 853×1440 pixels annotated for concrete spalling is developed. The network is modified, trained and validated using the proposed database. Damage quantification is automatically performed using the depth data output by the sensor. The surface of the analyzed element is extracted by merging the bounding boxes output by the Faster R-CNN onto the depth map. A polystyrene test rig containing damage simulations of known volume was utilized to test the accuracy of volume calculation. In addition to that, a concrete beam was also used to test the entire system. The Faster R-CNN yielded an average precision (AP) of 77.97% for damage detection. Damage quantification routine presents error of 9.45% in volume quantification of samples located within 100 cm and 250 cm away from the sensor plane. On top of that, maximum depth measurements of damages show a mean precision error (MPE) of 3.24% considering the same distance range. The implemented method allows for damage segmentation and quantification regardless of the distance between the sensor and the analyzed element.
The primary method of structural health monitoring is human-based visual inspection, which—despite its limitations of consistency and accessibility—can warn about changes in a bridge’s condition. To improve the visual inspection of civil infrastructure and address these drawbacks of human-oriented inspection, computer vision-based techniques have been developed to detect structural damage in images. Most of these methods, however, detect only specific types of damage, such as cracks in concrete or steel. Another drawback is that the traditional convolutional neural network-based damage detection method is not able to provide the location of the detected damage. To provide quasi-realtime simultaneous detection and localization of multiple types of damage, a structural damage detection method based on Faster Regionbased Convolutional Neural Network (Faster R-CNN) is proposed. The original architecture of Faster R-CNN is modified, trained, validated, and tested for this study. The robustness of the trained Faster R-CNN is evaluated and demonstrated using seven new images taken of various structures.
Many convolutional neural networks (CNN) –based approaches were proposed and applied to detect damage in various civil structures in recent years. Usually, the training process of the classical CNN requires a large number of labeled data which is from the monitored structure in undamaged and various damaged scenarios. However, it is impractical to acquire sufficient data that can be exactly labeled with damaged from the infrastructures in service as training data. Thus, we propose a novel unsupervised CNN-based approach to automatically extract optimal feature representations from the unlabeled data in a single class. In the case study, a known dataset from an undamaged scenario is used to train CNN and a dataset from an unknown scenario is used to test the trained CNN. The proposed approach in unsupervised learning is capable of extracting feature representations from the raw acceleration signals that are sensitive to the presence of damage. Then, the extracted damage-sensitive features are fed into a one-class support vector machine (OC-SVM) for novelty detection. The feature set from the undamaged dataset is taken as training dataset to train the OC-SVM, and the extracted features from the unknown dataset are used for testing. In order to verify the effectiveness of the proposed approach in structural damage localization, a number of accelerometers are used to acquire sufficient raw acceleration data from a lab-scale steel bridge, and the preliminary experimental results show that the proposed novel CNN-based approach performs very well in damage localization.
This paper presents an existing face detection method using cascade features updated for determining the cracks on concrete surfaces. The main goal of structural health monitoring (SHM) is to safeguard our existing structures from cracks, corrosion, delamination, and spalls due to incessant use of structures. Cracks are the foremost defect that will occur in the structures, and they require quick attention before they lead to structural failure; it is a laborious job to detect the cracks using personnel (visual inspection) practices, which produce highly unreliable results. The results of contact sensor-based crack detection techniques, however, mainly depend on parameters such as temperature, sensitivity, accessibility, etc. Recently there has been high expansion in computer vision (image processing) techniques that facilitate the detection of cracks. In this study, a modified cascade face detection technique based on the Viola-Jones algorithm is proposed to detect cracks in concrete walls. Cascade features calculated from the Viola-Jones algorithm are trained on positive and negative datasets of images with and without cracks. Once training is completed, the Viola-Jones algorithm spots the cracks on test images with bounding boxes drawn around the region of the cracks.
Civil infrastructure is important to ensure the ongoing functionality of human living environments. However, in North America, much of the infrastructure is aging and requires continuous monitoring and maintenance to ensure the safety of people. Traditionally, visual inspection has been carried out to monitor the health of such structures. However, assessments require trained inspectors, and monitoring methods are difficult due to the size and location of the infrastructure. Recently, data acquisition using unmanned aerial vehicles (UAVs) equipped with cameras has been growing in popularity, and research has been conducted concerning the use of UAVs for the visual inspection of infrastructure. However, UAV inspection requires skilled pilots and the use of a global positioning system (GPS) for autonomous flight. Unfortunately, for some locations, a GPS signal cannot be reached for autonomous flight of the UAV. For example, the GPS signal on the inside of a building or underneath a bridge deck is unreliable, but these locations also require inspections to ensure structural health. In order to address this issue, autonomous UAV methods using ultrasonic beacons have been proposed. Beacons are able to provide positional data allowing UAVs to perform the autonomous mission. As an example of structural damage, we report the successful detection of concrete cracks using a deep convolutional neural network by processing the video data collected from an autonomous UAV.
Online structural health monitoring of large-scale models of infrastructures under hazardous environmental loadings— like earthquakes—has been a vital research topic during recent years. A linear Kalman filter has been employed in many cases in which the desired parameters are extracted in a propagated state vector during a recursive regime. Also, many other kinds of nonlinear filters have been developed for nonlinear systems identification following the linear Kalman filter concept, such as the unscented Kalman filter and the cubature Kalman filter. The main contribution of these two Kalman filtering techniques relies on the propagation of a covariance matrix instead of nonlinear transition and measurement functions. Our extensive literature review shows that divergence of estimated states for large degree-offreedom (DoF) models is the main drawback of these techniques. To overcome this weakness, these two filters’ predefined points, sigma points, are combined—with some modifications—to have more predetermined points for the propagation of states and output of covariance matrices. The proposed technique was developed to be used for large DoF systems with a high level of noisy measured data, which indicates a robust identification system. To evaluate the proposed method, a numerical model (10 DoF linear system) with high levels of noise in the measured response data are employed to evaluate the robustness of the proposed method. The results show that the proposed method is significantly superior to the traditional UKF for noisy measured data in systems with large degrees of freedom.
The integration of sensing technology with structural health monitoring (SHM) has lead to advancements in how structures are monitored and investigated. One of the issues that has accompanied advancement in the industry is the time required to carry out testing on large-scale concrete reinforced structures using methods like impact-echo and ground penetrating radar (GPR). Back end processing and automation of testing systems are two means of addressing time consuming testing programs. This study proposes a semi-autonomous testing setup to carry out impact-echo testing on a lab specimen and a full-scale field structure. The testing method is coupled with artificial neural network (ANN) processing to decrease the need for user-interactions to produce results from the testing. The use of the semi-autonomous testing method and ANN processing is postulated to decrease the time needed for testing and improve the repeatability and accuracy of the impact-echo testing.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.