Options
Schoefs, Franck
Preferred name
Schoefs, Franck
Official Name
Schoefs, Franck
Research Output
Now showing 1 - 10 of 12
- PublicationDamage Assessment of the Built Infrastructure using Smartphones(University College Dublin, 2018-08-30)
; ; ; The use of image-processing and machine-learning algorithms for road condition monitoring has attracted considerable interest in recent years. This surge in popularity has been propelled by advances in camera technology and the emergence of state-of-the-art deep learning techniques which have allowed inspectors to obtain high-quality imagery on a consistent basis, and then use efficient techniques to recognise road defects with credibility. A wide variety of road defect detection techniques have been proposed, however, the influence that different image acquisition devices have on the accuracy of defect detection has not been studied despite being a key component. The use of smartphone cameras as an inspection tool is of particular interest as they have become ubiquitous in recent times and the built-in cameras have progressed significantly. These cameras are now capable of producing perfectly acceptable images, yet they are still not well compared against established benchmarks. In this paper, the qualities of smartphones are explored and compared against a dedicated DSLR (Digital Single-Lens Reflex) camera. An experiment was designed that involved capturing video footage of a road surface using a smartphone (Samsung Galaxy S7) and a dedicated imaging device (Canon 600D DRSL). A deep learning based crack detection method was applied to imagery from the smartphone and the DRSL cameras and the performances levels were subsequently compared. The results indicate that smartphones are a viable, low-cost method for executing quick assessments of the road integrity. The evaluation of smartphone cameras also addresses the ongoing uncertainty around the level of performance that can be achieved using cheaper sensors for quantitative purposes. Such findings provide reassurance for inspectors wishing to use their smartphones for simple monitoring tasks.244 - PublicationEffects of Turbidity and Lighting on the Performance of an Image Processing based Damage Detection Technique(Taylor & Francis, 2014-02-10)
; ; ; Measuring the true extent of damage in a structure remains a difficult task for inspectors. For visual inspections, an accurate assessment of the damage state is often subjective in nature and prone to error, especially when an inspection is conducted in hostile surroundings or when there are challenging environmental conditions present. While incorporating some form of Non-Destructive Technique (NDT) is generally useful for the inspection process, its performance may similarly degrade in the presence of environmental conditions. It is thus of great practical importance to have a measure of the performance of an NDT for a host of varying conditions, thereby allowing the inspector to determine whether it could be successfully applied in a given situation. In this paper, a measure of the effectiveness of an NDT is probabilistically determined for various environmental conditions through the use of Receiver Operating Characteristic (ROC) curves. ROC curves offer a convenient way of characterizing and comparing the performance of an NDT under various conditions. The NDT considered in this paper is an image processing based damage detection technique which uses texture information in conjunction with Support Vector Machine (SVM) classification to identify damaged regions. The variability of this technique is evaluated for various damage forms that are subjected to two changing parameters; turbidity and lighting. There were three set levels (low, medium, high) for each parameter. The conditions that were conducive to good detection were isolated and ranked using the α-δ method as part of the ROC analysis. The technique is applied to standard dynamic range (SDR) images and high dynamic range (HDR) images in order to assess their respective sensitivities to the changing parameters.275 - PublicationROC-based Performance Analysis and Interpretation of Image-based Damage Diagnostic Tools for Underwater Inspections(Taylor & Francis, 2018-10-10)
; ; ; It is of practical importance for inspectors to have knowledge of the efficiency of Non-Destructive Testing (NDT) tools when applied commercially. It has become common practice to model the performance of NDT tools in a probabilistic manner in terms of Probability of Detection (PoD), Probability of False Alarm (PFA) and eventually by Receiver Operating Characteristic (ROC) Curves. Traditionally, these quantities are estimated from training data, however, there are often doubts about the validity of these estimates when the sample size is small. In the case of underwater inspections, the scarcity of good quality training data means that this scenario arises more often than not. Comprehensive studies around the on-site performance of image-based damage diagnostic tools have only recently been made possible through the availability of online resources such as the Underwater Lighting and Turbidity Image Repository (ULTIR), which contains photographs of various damages forms captured under controlled visibility conditions. This paper shows how meaningful information can be extracted from this repository and used to construct ROC curves that can be related to the on-site performance of image-based NDT methods for detecting various damage forms and under a range of environmental conditions. The ability to draw connections between image-based techniques applied in real underwater inspections with ROC curves that can be constructed on-demand provide the engineer/ inspector with a clear and systematic route for assessing the reliability of data obtained from image-based methods. As a case study, the general approach has been applied to characterise the performance of image-based techniques for identifying instances of corrosion and cracks on marine structures. A discussion around how the results can be used for further analysis is provided. This includes looking at how the results can be fed into in the decision chain and can be used for risk analysis, intervention and work scheduling, and eventually understanding the value of information.175 - PublicationA comparison of image based 3D recovery methods for underwater inspectionsOffshore structures can be subjected to millions of variable amplitude load cycles during their service life which is the primary cause of structural deterioration. Such fatigue loading is exacerbated by marine growth colonization which changes the surface roughness characteristics and increases the diameter of structural members. Having an accurate knowledge of these parameters is essential for analyzing the increased hydrodynamic forces acting on the structure. This paper addresses the issue of acquiring shape information by comparing two popular classes of image based shape recovery techniques; stereo photography and Structure from Motion (SfM). Stereo photography utilises a dual camera set-up to simultaneously photograph an object of interest from slightly different viewpoints, whilst SfM methods generally involve a single camera moving in a static scene. In this paper, these techniques are performed on a controlled shape in an underwater setting, as well as synthetic data which allows for an irregular shape typical of marine growth to be tested whilst still having knowledge of the exact geometrical shape. The results reveal that the self-calibrated stereo approach fared well at getting an appropriately scaled full metric reconstruction, whilst the SfM approach was more susceptible to breaking down.
78 - PublicationRegionally enhanced multiphase segmentation technique for damaged surfaces(Wiley Online Library, 2014-09-15)
; ; ; Imaging‐based damage detection techniques are increasingly being utilized alongside traditional visual inspection methods to provide owners/operators of infrastructure with an efficient source of quantitative information for ensuring their continued safe and economic operation. However, there exists scope for significant development of improved damage detection algorithms that can characterize features of interest in challenging scenes with credibility. This article presents a new regionally enhanced multiphase segmentation (REMPS) technique that is designed to detect a broad range of damage forms on the surface of civil infrastructure. The technique is successfully applied to a corroding infrastructure component in a harbour facility. REMPS integrates spatial and pixel relationships to identify, classify, and quantify the area of damaged regions to a high degree of accuracy. The image of interest is preprocessed through a contrast enhancement and color reduction scheme. Features in the image are then identified using a Sobel edge detector, followed by subsequent classification using a clustering‐based filtering technique. Finally, support vector machines are used to classify pixels which are locally supplemented onto damaged regions to improve their size and shape characteristics. The performance of REMPS in different color spaces is investigated for best detection on the basis of receiver operating characteristics curves. The superiority of REMPS over existing segmentation approaches is demonstrated, in particular when considering high dynamic range imagery. It is shown that REMPS easily extends beyond the application presented and may be considered an effective and versatile standalone segmentation technique.Scopus© Citations 52 592 - PublicationA Stereo-Matching Technique for Recovering 3D Information from Underwater Inspection Imagery(Wiley Online Library, 2018-03)
; ; ; Underwater inspections stand to gain from using stereo imaging systems to collect three-dimensional measurements. Although many stereo-matching algorithms have been devised to solve the correspondence problem, that is, find the same points in multiple images, these algorithms often perform poorly when applied to images of underwater scenes due to the poor visibility and the complex underwater light field. This article presents a new stereo-matching algorithm, called PaLPaBEL (Pyramidal Loopy Propagated BELief) that is designed to operate on challenging imagery. At its core, PaLPaBEL is a semiglobal method based on a loopy belief propagation message passing algorithm applied on a Markov random field. A pyramidal scheme is adopted that enables wide disparity ranges and high-resolution images to be handled efficiently. For performance evaluation, PaLPaBEL is applied to underwater stereo images captured under various visibility conditions in a laboratory setting, and to synthetic imagery created in a virtual underwater environment. The technique is also demonstrated on stereo images obtained from a real-world inspection. The successful results indicate that PaLPaBEL is well suited for underwater application and has value as a tool for the cost-effective inspection of marine structures.790Scopus© Citations 22 - PublicationAn underwater lighting and turbidity image repository for analysing the performance of image-based non-destructive techniques(Taylor & Francis, 2017-05-27)
; ; ; Image processing-based methods, capable of detecting and quantifying cracks, surface defects or recovering 3D shape information are increasingly being recognised as a valuable tool for inspecting underwater structures. It is of great practical importance for inspectors to know the effectiveness of such techniques when applied in conditions. This paper considers an underwater environment characterised by poor visibility chiefly governed by the lighting and turbidity levels, along with a range of geometry and damage conditions of calibrated specimens. The paper addresses the relationship between underwater visibility and the performance of image-based methods through the development and calibration of a first open-source underwater lighting and turbidity image repository (ULTIR). ULTIR contains a large collection of images of submerged specimens that have been photographed under controlled lighting and turbidity levels featuring various forms of geometry and damage. ULTIR aims to facilitate inspectors when rationalising the use of image processing methods as part of an underwater inspection campaign and to enable researchers to efficiently evaluate the performance of image-based methods under realistic operating conditions. Stakeholders in underwater infrastructure can benefit through this first large, standardised, well-annotated, and freely available database of images and associated metadata.827Scopus© Citations 31 - PublicationProtocols for Image Processing based Underwater Inspection of Infrastructure ElementsImage processing can be an important tool for inspecting underwater infrastructure elements like bridge piers and pile wharves. Underwater inspection often relies on visual descriptions of divers who are not necessarily trained in specifics of structural degradation and the information may often be vague, prone to error or open to significant variation of interpretation. Underwater vehicles, on the other hand can be quite expensive to deal with for such inspections. Additionally, there is now significant encouragement globally towards the deployment of more offshore renewable wind turbines and wave devices and the requirement for underwater inspection can be expected to increase significantly in the coming years. While the merit of image processing based assessment of the condition of underwater structures is understood to a certain degree, there is no existing protocol on such image based methods. This paper discusses and describes an image processing protocol for underwater inspection of structures. A stereo imaging image processing method is considered in this regard and protocols are suggested for image storage, imaging, diving, and inspection. A combined underwater imaging protocol is finally presented which can be used for a variety of situations within a range of image scenes and environmental conditions affecting the imaging conditions. An example of detecting marine growth is presented of a structure in Cork Harbour, Ireland.
260Scopus© Citations 5 - PublicationHigh dynamic range image processing for non-destructive-testingThis paper proposes High Dynamic Range (HDR) imaging as a protocol for nondestructive-testing for the first time. HDR imaging protocol has been experimentally validated on images of pitting corrosion in this paper and has been further applied to isolate a range of image backgrounds that arise out of a number of environmental conditions. The superiority of HDR imaging over a standard imaging process has been demonstrated. The results indicate that the proposed protocol is not limited by the examples considered in this paper and can readily be applied to a number of infrastructure maintenance management related applications. The protocol complements and improves standard visual inspection techniques and expert opinions. This approach immediately lends itself to further mathematical and statistical analyses, both qualitative and quantitative.
436Scopus© Citations 5 - PublicationSemantic Segmentation of Underwater Imagery Using Deep Networks Trained on Synthetic ImageryRecent breakthroughs in the computer vision community have led to the emergence of efficient deep learning techniques for end-to-end segmentation of natural scenes. Underwater imaging stands to gain from these advances, however, deep learning methods require large annotated datasets for model training and these are typically unavailable for underwater imaging applications. This paper proposes the use of photorealistic synthetic imagery for training deep models that can be applied to interpret real-world underwater imagery. To demonstrate this concept, we look at the specific problem of biofouling detection on marine structures. A contemporary deep encoder–decoder network, termed SegNet, is trained using 2500 annotated synthetic images of size 960 × 540 pixels. The images were rendered in a virtual underwater environment under a wide variety of conditions and feature biofouling of various size, shape, and colour. Each rendered image has a corresponding ground truth per-pixel label map. Once trained on the synthetic imagery, SegNet is applied to segment new real-world images. The initial segmentation is refined using an iterative support vector machine (SVM) based post-processing algorithm. The proposed approach achieves a mean Intersection over Union (IoU) of 87% and a mean accuracy of 94% when tested on 32 frames extracted from two distinct real-world subsea inspection videos. Inference takes several seconds for a typical image.
372Scopus© Citations 42