Recent years have been fruitful in the development of computer vision methods for a wide variety of applications. Despite the successful results achieved in the segmentation of cracks on concrete surfaces, poor results are still persisting during onsite application, mainly due to noise caused by biological colonization, which is present in most of historical heritage buildings. The authors have been working on this problematic and developed the SC-Crack method previously, however it still relies on the cumbersome task of acquiring sets of 17 channels images to compose an hyperspectral cube and still requires case-wise hyperparameter optimization. Consequently, it is important to define which spectral information mostly defines the success of the method, enabling to optimize both, the acquisition procedure and model processing. Following, a study aiming at the selection of the more informative channels was carried and the hyperparameter-free model is evaluated.In this scope, images of concrete specimens were acquired sequentially to compose a 17 channel hyperspectral image cube. These were sere compute allowing to define the most informative channels sets that are processed using the SC-Crack+ method, presented in this work. The reduced image cubes of cracking on clean concrete surfaces and on surfaces with biological colonization were processed and analyzed. Relevant and improved results were achieved for crack segmentation, following this SC-Crack+ model. This enables the possibility of mounting cameras with sensors and lenses particularly adapted for prone acquisition targeting only the most relevant hyperspectral information for crack segmentation and still using traditional feature engineering image processing methods.

The new Construction 4.0 paradigm takes advantage of existing technologies. In this scope, the development and application of image-based methods for evaluating and monitoring the state of conservation of buildings has shown significant growth, including support for maintenance plans. Recently, powerful algorithms have been applied to automatically evaluate the state of conservation of buildings using deep learning frameworks, which are utilised as a black-box approach. The large amount of data required for training, the difficulty in generalising, and the lack of parameters to assess the quality of the results often make it difficult for non-experts to evaluate them. For several applications and scenarios, simple and more intuitive image-based approaches can be applied to support building inspections. This paper presents the StainView, which is a fast and reliable method. The method is based on the classification of the mosaic image, computed from a systematic acquisition, and allows one to (i) map stains in facades; (ii) locate critical areas; (iii) identify materials; (iv) characterise colours; and (v) produce detailed and comprehensive maps of results. The method was validated in three identical buildings in Bairro de Alvalade, in Lisbon, Portugal, that present different levels of degradation. The comparison with visual inspection demonstrates that StainView enables the automatic location and mapping of critical areas with high efficiency, proving to be a useful tool for building inspection: differences were of approximately 5% for the facade with the worst and average state of conservation, however, the values deteriorate for the facade under good conditions, reaching the double of percentage. In terms of processing speed, StainView allows a facade mapping that is 8–12 times faster, and this difference tends to grow with the number of evaluated façades.

The development of automatic methods to recognize cracks in surfaces of concrete has been under focus in recent years, firstly through computer vision methods and more recently focusing on convolutional neural networks that are delivering promising results. Challenges are still persisting in crack recognition, namely due to the confusion added by the myriad of elements commonly found on concrete surfaces. The robustness of these methods would deal with these elements if access to correspondingly heterogeneous datasets was possible. Even so, this would be a cumbersome methodology, since training would be needed for each particular case and models would be case dependent. Thus, efforts from the scientific community are focusing on generalizing neural network models to achieve high performance in images from different domains, slightly different from those in which they were effectively trained. The generalization of networks can be achieved by domain adaptation techniques at the training stage. Domain adaptation enables finding a feature space in which features from both domains are invariant, and thus, classes become separable. The work presented here proposes the DA-Crack method, which is a domain adversarial training method, to generalize a neural network for recognizing cracks in images of concrete surfaces. The domain adversarial method uses a convolutional extractor followed by a classifier and a discriminator, and relies on two datasets: a source labeled dataset and a target unlabeled small dataset. The classifier is responsible for the classification of images randomly chosen, while the discriminator is dedicated to uncovering to which dataset each image belongs. Backpropagation from the discriminator reverses the gradient used to update the extractor. This enables fighting the convergence promoted by the updating backpropagated from the classifier, and thus generalizing the extractor enabling it for crack recognition of images from both source and target datasets. Results show that the DA-Crack training method improved accuracy in crack classification of images from the target dataset in 54 percentage points, while accuracy on the source dataset remains unaffected.

Despite all technological advances, mapping cracks on concrete structures mostly remains to be evaluated through sketches based on on-site observation and photographs. Methods based on image processing have been developed with clear advantages. However, most studies rely on perfectly identified areas or on single cracks without any other pathologies, being therefore unsuitable for on-site application. In addition, the accuracy is not usually quantified due to the absence of ground-truth. Thus, methods for automatic mapping of cracking patterns, sufficiently robust to deal with the surrounding pathologies, are of great interest. The Super Cluster-Crack method (SC-Crack method) is herein presented. It was developed for crack detection in concrete surfaces, with biological stains, by processing hyperspectral images. SC-Crack performs k-means clustering, followed by grouping clusters to composing a super cluster that stands for the cracks. The method was calibrated and validated by classifying hyperspectral images of concrete specimens, within bandwidths of 25nm in a wavelength range between 425nm and 950nm. Results are discussed by comparison with the ground-truth image. Finally, the super cluster composition is also validated. The SC-Crack method performs successfully both on clean and on surface with biological stains. In the latter case, hyperspectral images help to avoid mixing biological stains with crack pattern. Concerning the main goal of mapping the cracking pattern, the method performs perfectly on concrete clean surfaces, allowing to detect all the crack branches. In the case of surface with biological stains, the SC-Crack also detects the majority of cracking pattern, except for the thinner branches.

Purpose: To compare choroidal thickness (CT) between diabetic patients without diabetic retinopathy and a nondiabetic group. To explore how CT relates to disease duration, mean arterial pressure, glycemia, glycosylated hemoglobin, intraocular pressure, and ocular pulse amplitude. Methods: Choroidal thickness was assessed using a spectral-domain optical coherence tomography and enhanced depth mode at 13 locations (subfoveal and 3 measurements 500 μm apart in 4 directions—nasal, temporal, superior, and inferior). Linear regression models were used. Results: One hundred seventy-five patients were recruited (125 diabetic patients without diabetic retinopathy and 50 nondiabetic patients). In diabetic patients, although without statistical significance, CT showed a trend to be thicker in all locations (6.16–24.27 μm). Choroidal thickness was negatively associated with age (P < 0.001) in both groups, but only in the diabetic group, it was positively associated to ocular pulse amplitude (with a mean increase between 8.5 μm and 11.6 μm for each millimeter of mercury increase in ocular pulse amplitude). Diabetic patients' CT seems to stabilize after 150 months of diabetes, increase with higher glycemia levels (>160 mg/dL) while showing no fluctuation with glycosylated hemoglobin and mean arterial pressure. Conclusion: There seems to be a thickening of the choroid in diabetic patients without diabetic retinopathy. Moreover, this tissue may be functionally different in diabetes, as the pattern of associations seems to differ between groups.

Detection and mapping crack patterns are key issues for structural assessment of concrete structures. The use of image processing for identification of pathologies has undergone major developments, since it is a noninvasive technique providing the precision and reliability required for the task. The authors have developed a method, named SurfCrete, to materials and damages classification on concrete structures, including mapping cracks. This is based on analysis of multi-spectral images, including visible and near infra-red (NIR) regions of the electromagnetic spectrum. Latest improvements include the use of hyperspectral image analysis for crack detection, based on image clustering. The drawbacks of the developed methods are the difficulties usually shown when dealing with surfaces presenting several damages and materials besides cracks, namely due to the presence of biological colonization, repairing mortars, delamination and efflorescence, among other anomalies commonly found on concrete structures. Furthermore, when surfaces are subjected to different light conditions, this also influences the accurate classification of cracks. In this paper, an evolution of the method previously developed, herein named SurfCrete-HSV, is presented. The new method is completely focused on classifying biological colonization based on the classification of HSV false colour images, being therefore more robust and reliable. These HSV images are built from hyperspectral images (wavelengths from 450 nm to 950 nm and 25 nm of bandwidth) by selecting three channels, one from NIR region and two from the visible region of electromagnetic spectrum. The HSV space allows isolating the colour in a single data dimension to enable a brightness free clustering. An image of a concrete specimen with simulation of biological colonization over a smooth surface is used from a database of hyperspectral images, to evaluate SurfCrete-HSV method. Results show that the SurfCrete-HSV method is reliable for detection of biological colonization on concrete surfaces. The best set of channels to use results from combining one from Near Infra-Red with Red and Blue regions of the electromagnetic spectrum, which reveals high accuracy values with acceptable recall.

For concrete structures, the characterization of cracks, namely initiation, opening, and propagation, as well as of cracking pattern, plays an extremely important role in the scope of structural health monitoring. The authors have already developed a method, MCrack, based on computer vision and image processing with this goal. However, this is focused on a given concrete surface at a given instant and, thus, monitoring cracks through time, namely between periodic inspections, requires permanent and static camera stations. While this is most adequate for laboratorial tests, for on-site application it implies the need of work-intensive procedures (that are not error proof) to ensure the overlap of images acquired at different time instants. Recent research developments in image registration have a significant potential to be implemented in MCrack, allowing to monitor cracks with simple and light methods, more prone for on-site monitoring. In this paper, a first approach of MCrack-Propagation method is presented, as an evolution of MCrack, based on matching image features. The main improvement is the automatic feature matching between pairs of multi-temporal images of the same surface region. This enables to automatically track the evolution of cracks width on concrete surfaces. The MCrack-Propagation method improved the reliability of crack width measurement due to the higher number of measuring points than in previously presented methods. It is also light and easy to implement avoiding work intensive procedures, as those using image segmentation methods or traditional methods that require experts to contact with the cracks.

then sharpened into a tightly peaked distribution with a peak at around +1.0 D, with reductions in corneal and lens power and axial elongation. After 2–3 years, corneal power stabilizes, but the changes in lens power and axial length continue. Reductions in lens power reduce the myopic shift produced by axial elongation. There is some tightening of the distribution of refraction around approximately 1.0 D, with the less hyperopic eyes undergoing hyperopic shifts in refraction if the loss of lens power exceeds the rate of axial elongation. Under environmental conditions which produce myopia, these mechanisms are swamped by excessive axial elongation, and children then proceed to emmetropia and on to myopia.

Abstract Introduction The purpose of this study was to measure and to compare macular choroidal thickness (CT) between patients with mild Alzheimer's disease (AD), patients without AD, and elderly patients. Methods CT was measured manually in 13 locations at 500‐μm intervals of a horizontal and a vertical section from the fovea. Linear regression models were used to analyze the data. Results Fifty patients with a diagnosis of mild AD (73.1 years), 152 patients without AD (71.03 years), and 50 elderly without AD (82.14 years) were included. In the AD patients, CT was significantly thinner in all 13 locations (P < .001—comparing with age‐match group), and comparing with the elderly group, a more pronounced difference was found in two locations temporal to the fovea. Discussion Patients with AD showed a significant choroidal thinning even when compared with elderly subjects. The reduction of CT may aid in the diagnoses of AD, probably reflecting the importance of vascular factors in their pathogenesis.

PURPOSE: To identify changes in choroidal thickness (CT) and all retinal layers of diabetic patients without diabetic retinopathy (DR) after 1 year of follow-up.DESIGN: Prospective observational cohort study.METHODS: Overall, 125 diabetic patients without DR were included. Two visits were scheduled: the first visit (V1) and a second visit after 12 months (V2). At both visits, patients received a complete ophthalmologic evaluation that included OCT. Each retinal layer thickness was calculated for 9 ETDRS sectors, and CT was measured at 13 locations. Generalized linear mixed-effects models were used.RESULTS: Of the 125 patients, 103 completed the study, and 9 of the 103 developed DR (8.7%). CT was significantly higher at V2 than at V1, with an average value of 10-17 mu m at almost half the locations (500, 1000, and 1500 mu m temporal; 500 and 1000 mu m nasal; and 1000 mu m superior to the fovea) (P <.001.003). The thicknesses of the ganglion cell layer (13 and N6 sectors), inner plexiform layer (S6 and N6 sectors), inner nuclear layer (T6 and N6 sectors), and outer plexiform layer (S6 sector), as well as the overall retinal thickness (RT) (S3, N3, I3, S6, and T6 sectors), were decreased at V2 (P <.001). Visible retinopathy was negatively associated with overall RT (central, S3, T3, I3, and N3 sectors, P =.004.024) and the thickness of the ONL (T6 and 16 sectors, P =.007 and P =.009) and photoreceptor layer (N6 sector, P =.038). The presence of DR decreased the overall RT by 13.04-16.63 mu m.CONCLUSIONS: Diabetic patients without DR showed a thicker choroid and a thinner retina, particularly in inner layers, after 1 year of follow-up. These structural changes may correspond to the early neurodegenerative phase of DR. (C) 2016 Elsevier Inc. All rights reserved.

All large infrastructures worldwide must have a suitable monitoring and maintenance plan, aiming to evaluate their behaviour and predict timely interventions. In the particular case of concrete infrastructures, the detection and characterization of crack patterns is a major indicator of their structural response. In this scope, methods based on image processing have been applied and presented. Usually, methods focus on image binarization followed by applications of mathematical morphology to identify cracks on concrete surface. In most cases, publications are focused on restricted areas of concrete surfaces and in a single crack. On-site, the methods and algorithms have to deal with several factors that interfere with the results, namely dirt and biological colonization. Thus, the automation of a procedure for on-site characterization of crack patterns is of great interest. This advance may result in an effective tool to support maintenance strategies and interventions planning. This paper presents a research based on the analysis and processing of hyper-spectral images for detection and classification of cracks on concrete structures. The objective of the study is to evaluate the applicability of several wavelengths of the electromagnetic spectrum for classification of cracks in concrete surfaces. An image survey considering highly discretized wavelengths between 425 nm and 950 nm was performed on concrete specimens, with bandwidths of 25 nm. The concrete specimens were produced with a crack pattern induced by applying a load with displacement control. The tests were conducted to simulate usual on-site drawbacks. In this context, the surface of the specimen was subjected to biological colonization (leaves and moss). To evaluate the results and enhance crack patterns a clustering method, namely k-means algorithm, is being applied. The research conducted allows to define the suitability of using clustering k-means algorithm combined with hyper-spectral images highly discretized for crack detection on concrete surfaces, considering cracking combined with the most usual concrete anomalies, namely biological colonization.

Purpose To compare the thickness of all retinal layers between a nondiabetic group and diabetic patients without diabetic retinopathy (DR). Methods Cross-sectional study, in which all subjects underwent an ophthalmic examination including optical coherence tomography. After automatic retinal segmentation, each retinal layer thickness (eight separate layers and overall thickness) was calculated in all nine Early Treatment Diabetic Retinopathy Study (ETDRS) areas. The choroidal thickness (CT) also was measured at five locations. Generalized additive regression models were used to analyze the data. Results A total of 175 patients were recruited, 50 nondiabetic subjects and 125 diabetic patients without DR, stratified into three groups according to diabetes duration: group I (<5 years, n = 55), group II (5-10 years, n = 39), and group III (>10 years, n = 31). Overall, groups I and III of diabetic patients had a decrease in the photoreceptor layer (PR) thickness, when compared with the nondiabetic subjects in six ETDRS areas (P < 0.0007). Patients with more recent diagnosis (group I) had thinner PR than those with moderate duration (group II). Interestingly, patients with longer known disease (group III) had the thinnest PR values. There were no overall differences in the remaining retinal parameters. Conclusions Retinal thickness profile is not linear throughout disease duration. Even in the absence of funduscopic disease, PR layer in diabetic patients seems to differ from nondiabetic subjects, thus suggesting that some form of neurodegeneration may take place before clinical signs of vascular problems arise.

Excess sediment production in the upper parts of catchments may result in important impacts over morphodynamics of gravel bed-rivers. By changing morphodynamics, sediment overfeeding may induce important changes in the structure of near-bed flow, mainly in what concerns exchange of momentum and mass between flow within the roughness elements and flow in the upper regions. It is not well-known how turbulent statistics, including those characterizing the bursting cycle, are affected by bed load transport, for mobile but geometrically similar beds. This study addresses this issue. It is aimed at evaluating the impacts of sediment transport on flow hydrodynamics, namely on statistics of turbulent coherent structures. In order to accomplish the proposed objective, laboratory tests were undertaken. Two-dimensional instantaneous flow velocity fields in the stream-wise and vertical directions were measured with Particle Image Velocimetry.Two laboratory tests simulated a framework gravel bed with sand matrix and a framework gravel bed with sand matrix but with sediment transport imposed at near capacity conditions. The framework consists of coarse gravel whose diameters range between 0.5 cm and 7 cm and is kept immobile under the imposed flow conditions. The mobile sediments are sand with a mean diameter of 0.9 mm. For both tests, the quadrant threshold analysis technique was employed and shear stress distribution statistics were analyzed and discussed in what concerns their contribution and persistence.In the case of mobile conditions, sweep events tend to govern the flow in the near-bed region. Relevant differences between mobile and sub-threshold beds are found in the wake of roughness elements, mostly for sweep statistics. In the presence of bed-load, ejection events decrease their participation in the shear stress production processes. This decrease in the ejection events contribution is partially balanced with an increase in the frequency of inward events.