Crash severity is undoubtedly a fundamental aspect of a crash event. Although machine learning algorithms for predicting crash severity have recently gained interest by the academic community, there is a significant trend towards neglecting the fact that crash datasets are acutely imbalanced. Overlooking this fact generally leads to weak classifiers for predicting the minority class (crashes with higher severity). In this paper, in order to handle imbalanced accident datasets and provide a better prediction for the minority class, the random undersampling the majority class (RUMC) technique is used. By employing an imbalanced and a RUMC-based balanced training set, we propose the calibration, validation, and evaluation of four different crash severity predictive models, including random tree, k-nearest neighbor, logistic regression, and random forest. Accuracy, true positive rate (recall), false positive rate, true negative rate, precision, F₁-score, and the confusion matrix have been calculated to assess the performance. Outcomes show that RUMC-based models provide an enhancement in the reliability of the classifiers for detecting fatal crashes and those causing injury. Indeed, in imbalanced models, the true positive rate for predicting fatal crashes and those causing injury spans from 0% (logistic regression) to 18.3% (k-nearest neighbor), while for the RUMC-based models, it spans from 52.5% (RUMC-based logistic regression) to 57.2% (RUMC-based k-nearest neighbor). Organizations and decision-makers could make use of RUMC and machine learning algorithms in predicting the severity of a crash occurrence, managing the present, and planning the future of their works.

Tire rubber recycling for civil engineering applications and products is developing faster, achieving increasingly higher levels of maturation. The improvements in the material circle, where crumb rubber, generated as a by-product of the tire rubber making process, becomes the resource used for the construction of road asphalt pavement, is absolutely necessary for increasing the sustainability of the entire supply chain. The paper reports the results of an accurate data analysis derived from an extensive literature review of existing processes, technologies, and materials within construction of infrastructure. The current position, the direction, and rate of progress of the scientific efforts towards the reuse and recycling of tire rubber worldwide have been shown. Furthermore, an in-depth analysis of a set of important properties of Crumb Rubber Modified Asphalt has been carried out—fabrication parameters, standard properties, high and low-temperature performance, and rheological properties. Statistics over a sample of selected publications have been presented to understand the main processes adopted, rubber particle size, temperatures, and possible further modifications of crumb rubber modified binder.

This paper introduces a methodology for predicting and mapping surface motion beneath road pavement structures caused by environmental factors. Persistent Scatterer Interferometric Synthetic Aperture Radar (PS-InSAR) measurements, geospatial analyses, and Machine Learning Algorithms (MLAs) are employed for achieving the purpose. Two single learners, i.e., Regression Tree (RT) and Support Vector Machine (SVM), and two ensemble learners, i.e., Boosted Regression Trees (BRT) and Random Forest (RF) are utilized for estimating the surface motion ratio in terms of mm/year over the Province of Pistoia (Tuscany Region, central Italy, 964 km²), in which strong subsidence phenomena have occurred. The interferometric process of 210 Sentinel-1 images from 2014 to 2019 allows exploiting the average displacements of 52,257 Persistent Scatterers as output targets to predict. A set of 29 environmental-related factors are preprocessed by SAGA-GIS, version 2.3.2, and ESRI ArcGIS, version 10.5, and employed as input features. Once the dataset has been prepared, three wrapper feature selection approaches (backward, forward, and bi-directional) are used for recognizing the set of most relevant features to be used in the modeling. A random splitting of the dataset in 70% and 30% is implemented to identify the training and test set. Through a Bayesian Optimization Algorithm (BOA) and a 10-Fold Cross-Validation (CV), the algorithms are trained and validated. Therefore, the Predictive Performance of MLAs is evaluated and compared by plotting the Taylor Diagram. Outcomes show that SVM and BRT are the most suitable algorithms; in the test phase, BRT has the highest Correlation Coefficient (0.96) and the lowest Root Mean Square Error (0.44 mm/year), while the SVM has the lowest difference between the standard deviation of its predictions (2.05 mm/year) and that of the reference samples (2.09 mm/year). Finally, algorithms are used for mapping surface motion over the study area. We propose three case studies on critical stretches of two-lane rural roads for evaluating the reliability of the procedure. Road authorities could consider the proposed methodology for their monitoring, management, and planning activities.

In the present paper, we implemented the Bayesian regularization (BR) backpropagation algorithm for calibrating an artificial neural network (ANN) as an accident prediction model (APM) to be used on Italian four-lane divided roads. We chose the BR-ANN since it efficiently allows for dealing with small sample size and avoiding overfitting issues by adding a regularization term in the objective function to be minimized during training. Moreover, BR-ANNs are sparsely employed in road safety analyses, and their peculiarities deserve to be emphasized. In our work, the BR-ANN aims to predict the number of fatal and injury (FI) crashes across 236 road elements, for a total length of 78 km. The input features are road element length, horizontal and vertical alignment, cross-section geometry, operating speed, traffic flow, sight distance, and road area type (i.e., a categorical predictor accounting for the potential influence of merge and diverge influence areas). Training and test phases of the BR-ANN have been evaluated by determination coefficient ( R²), root mean square error (RMSE), overfitting ratio (OR), scatterplots, residuals analysis, and by the same ANN architecture trained with the gradient descent (GD) with momentum and adaptive learning rate backpropagation algorithm (GD-ANN). Results demonstrate that the BR-ANN markedly outperforms the GD-ANN, which suffers severe overfitting issues. Furthermore, BR-ANN does not overfit data (OR close to the unity), reports a satisfactory R² (0.726), and shows a Gaussian residual distribution with zero mean. Therefore, road authorities could consider regularized ANNs for performing appropriate safety analyses, especially when dealing with small road sample sizes.

This paper proposes a methodology for correlating products derived by Synthetic Aperture Radar (SAR) measurements and laser profilometric road roughness surveys. The procedure stems from two previous studies, in which several Machine Learning Algorithms (MLAs) have been calibrated for predicting the average vertical displacement (in terms of mm/year) of road pavements as a result of exogenous phenomena occurrence, such as subsidence. Such algorithms are based on surveys performed with Persistent Scatterer Interferometric SAR (PS-InSAR) over an area of 964 km² in the Tuscany Region, Central Italy. Starting from this basis, in this paper, we propose to integrate the information provided by these MLAs with 10 km of in situ profilometric measurements of the pavement surface roughness and relative calculation of the International Roughness Index (IRI). Accordingly, the aim is to appreciate whether and to what extent there is an association between displacements estimated by MLAs and IRI values. If a dependence exists, we may argue that road regularity is driven by exogenous phenomena and MLAs allow for the replacement of in situ surveys, saving considerable time and money. In this research framework, results reveal that there are several road sections that manifest a clear association among these two methods, while others denote that the relationship is weaker, and in situ activities cannot be bypassed to evaluate the real pavement conditions. We could wrap up that, in these stretches, the road regularity is driven by endogenous factors which MLAs did not integrate during their training. Once additional MLAs conditioned by endogenous factors have been developed (such as traffic flow, the structure of the pavement layers, and material characteristics), practitioners should be able to estimate the quality of pavement over extensive and complex road networks quickly, automatically, and with relatively low costs.

Cracks are fractures or breaks that occur in materials such as concrete, metals, rocks, and other solids. Various methods are used to detect and monitor cracks; among many of them, image-based methodologies allow fast identification of the distress and easy quantification of the percentage of cracks in the scene. Two main categories can be identified: classical and deep learning approaches. In the last decade, the tendency has moved towards the use of the latter. Even though they have proven their outstanding predicting performance, they suffer some drawbacks: a “black-box” nature leaves the user blind and without the possibility of modifying any parameters, a huge amount of labeled data is generally needed, a process that requires expert judgment is always required, and, finally, they tend to be time-consuming. Accordingly, the present study details the methodology for a new algorithm for crack segmentation based on the theory of minimal path selection combined with a region-based approach obtained through the segmentation of texture features extracted using Gabor filters. A pre-processing step is described, enabling the equalization of brightness and shadows, which results in better detection of local minima. These local minimal are constrained by a minimum distance between adjacent points, enabling a better coverage of the cracks. Afterward, a region-based segmentation technique is introduced to determine two areas that are used to determine threshold values used for rejection. This step is critical to generalize the algorithm to images presenting close-up scenes or wide cracks. Finally, a geometrical thresholding step is presented, allowing the exclusion of rounded areas and small isolated cracks. The results showed a very competitive F1-score (0.839), close to state-of-the-art values achieved with deep learning techniques. The main advantage of this approach is the transparency of the workflow, contrary to what happens with deep learning frameworks. In the proposed approach, no prior information is required; however, the statistical parameters may have to be adjusted to the particular case and requirements of the situation. The proposed algorithm results in a useful tool for researchers and practitioners needing to validate their results against some reference or needing labeled data for their models. Moreover, the current study could establish the grounds to standardize the procedure for crack segmentation with a lower human bias and faster results. The direct application of the methodology to images obtained with any low-cost sensor makes the proposed algorithm an operational support tool for authorities needing crack detection systems in order to monitor and evaluate the current state of the infrastructures, such as roads, tunnels, or bridges.