Nicla Maria Notarangelo

Flooding remains one of the most frequent and damaging natural disasters globally, exacerbated by climate changes and rapid urbanization. Understanding and mitigating urban flood risk requires near-real-time, fine-scale monitoring, including flood depth estimation. This study introduces a deep learning-based pipeline for estimating urban flood depth using device-independent street-level imagery, to complement existing remote, in-situ and hydrological approaches. By leveraging opportunistic sensing, this method exploits open-source tools to enhance spatial granularity and accessibility in flood monitoring.The dataset, derived from a publicly available source, consisted of 3,367 annotated images of submerged vehicles, categorized into five flood levels based on water height relative to vehicle features (e.g., tires, chassis, windows). Cars were selected as reference objects due to their standardized dimensions and prevalence in urban environments, enabling consistent and reliable flood depth estimation.The proposed pipeline processes images through an end-to-end workflow designed for real-time inference. It consists of four sequential stages: (1) vehicles are detected in street-level images using a pre-trained YOLO-World model; (2) detected vehicle regions are cropped and resized with a 20% bounding box enlargement to include flood visual indicators and additional context cues for the classification.; (3) images are super-resolved using pre-trained Enhanced Deep Super-Resolution (EDSR) networks to improve low-resolution imagery; (4) images are classified according the flood depth level using a ResNet50 model fine-tuned on the annotated dataset.The classifier demonstrated robust performance across the five flood levels. The confusion matrix revealed minor misclassifications between adjacent classes, particularly Levels 0 and Level 1. One-vs-all area under the receiver operating characteristic curves (AUC) values exceeded 0.85 for all classes, with the highest performance observed for Level 4 (AUC = 0.98) and Level 0 (AUC = 0.94). Real-world validation using crowdsourced images from the 2021 flood in Central Europe confirmed the pipeline's reliability, delivering accurate and consistent flood level predictions in near-real-time.This research advances urban flood monitoring by introducing a cost-effective and adaptable method for flood depth estimation that leverages existing devices without specialized hardware. The pipeline’s modular design ensures scalability and seamless integration into early warning systems and disaster response platforms. Future work will explore its application to aerial and drone imagery with oblique perspectives and develop cross-view geolocalization of flood depth measurements to improve spatial coverage and accuracy.

The COVID-19 pandemic has underscored the need for real-time, collaborative virtual tools to support remote activities across various domains, including education and cultural heritage. Virtual walkthroughs provide a potent means of exploring, learning about, and interacting with historical sites worldwide. Nonetheless, creating realistic and user-friendly applications poses a significant challenge. This study investigates the potential of collaborative virtual walkthroughs as an educational tool for cultural heritage sites, with a focus on the Sassi of Matera, a UNESCO World Heritage Site in Italy. The virtual walkthrough application, developed using RealityCapture and Unreal Engine, leveraged photogrammetric reconstruction and deep learning-based hand gesture recognition to offer an immersive and accessible experience, allowing users to interact with the virtual environment using intuitive gestures. A test with 36 participants resulted in positive feedback regarding the application’s effectiveness, intuitiveness, and user-friendliness. The findings suggest that virtual walkthroughs can provide precise representations of complex historical locations, promoting tangible and intangible aspects of heritage. Future work should focus on expanding the reconstructed site, enhancing the performance, and assessing the impact on learning outcomes. Overall, this study highlights the potential of virtual walkthrough applications as a valuable resource for architecture, cultural heritage, and environmental education.

Urban flood risk mitigation requires fine-scale near-real-time precipitation observations that are challenging to obtain from traditional monitoring networks. Novel data and computational techniques offer a valuable potential source of information. This study explores an unprecedented, device-independent, artificial intelligence-based system for opportunistic rainfall monitoring through deep learning models that detect rainfall presence and estimate quasi-instantaneous intensity from single pictures. Preliminary results demonstrate the models’ ability to detect a significant meteorological state corroborating the potential of non-dedicated sensors for hydrometeorological monitoring in urban areas and data-scarce regions. Future research will involve further experiments and crowdsourcing, to improve accuracy and promote public resilience.

Flood risk monitoring, alert and adaptation in urban areas require near-real-time fine-scale precipitation observations that are challenging to obtain from currently available measurement networks due to their costs and installation difficulties. In this sense, newly available data sources and computational techniques offer enormous potential, in particular, the exploiting of not-specific, widespread, and accessible devices.This study proposes an unprecedented system for rainfall monitoring based on artificial intelligence, using deep learning for computer vision, applied to cameras images. As opposed to literature, the method is not device-specific and exploits general-purpose cameras (e.g., smartphones, surveillance cameras, dashboard cameras, etc.), in particular, low-cost device, without requiring parameter setting, timeline shots, or videos. Rainfall is measured directly from single photographs through Deep Learning models based on transfer learning with Convolutional Neural Networks. A binary classification algorithm is developed to detect the presence of rain. Moreover, a multi-class classification algorithm is used to estimate a quasi-instantaneous rainfall intensity range. Open data, dash-cams in Japan coupled with high precision multi-parameter radar XRAIN, and experiments in the NIED Large Scale Rainfall Simulator combined to form heterogeneous and verisimilar datasets for training, validation, and test. Finally, a case study over the Matera urban area (Italy) was used to illustrate the potential and limitations of rainfall monitoring using camera-based detectors.The prototype was deployed in a real-world operational environment using a pre-existent 5G surveillance camera. The results of the binary classifier showed great robustness and portability: the accuracy and F1-score value were 85.28% and 85.13%, 0.86 and 0.85 for test and deployment, respectively, whereas the literature algorithms suffer from drastic accuracy drops changing the image source (e.g. from 91.92% to 18.82%). The 6-way classifier results reached test average accuracy and macro-averaged F1 values of 77.71% and 0.73, presenting the best performances with no-rain and heavy rainfall, which represents critical condition for flood risk. Thus, the results of the tests and the use-case demonstrate the model’s ability to detect a significant meteorological state for early warning systems. The classification can be performed on single pictures taken in disparate lighting conditions by common acquisition devices, i.e. by static or moving cameras without adjusted parameters. This system does not suit scenes that are also misleading for human visual perception. The proposed method features readiness level, cost-effectiveness, and limited operational requirements that allow an easy and quick implementation by exploiting pre-existent devices with a parsimonious use of economic and computational resources.Altogether, this study corroborates the potential of non-traditional and opportunistic sensing networks for the development of hydrometeorological monitoring systems in urban areas, where traditional measurement methods encounter limitations, and in data-scarce contexts, e.g. where remote-sensed rainfall information is unavailable or has broad resolution respect with the scale of the proposed study. Future research will involve incremental learning algorithms and further data collection via experiments and crowdsourcing, to improve accuracy and at the same time promote public resilience from a smart city perspective.

Automatic building extraction from high-resolution remotely sensed data is a major area of interest for an extensive range of fields (e.g., urban planning, environmental risk management) but challenging due to urban morphology complexity. Among the different methods proposed, the approaches based on supervised machine learning (ML) achieve the best results. This paper aims to investigate building footprint extraction using only high-resolution raster digital surface model (DSM) data by comparing the performance of three different popular supervised ML models on a benchmark dataset. The first two methods rely on a histogram of oriented gradients (HOG) feature descriptor and a classical ML (support vector machine (SVM)) or a shallow neural network (extreme learning machine (ELM)) classifier, and the third model is a fully convolutional network (FCN) based on deep learning with transfer learning. Used data were obtained from the International Society for Photogrammetry and Remote Sensing (ISPRS) and cover the urban areas of Vaihingen an der Enz, Potsdam, and Toronto. The results indicated that performances of models based on shallow ML (feature extraction and classifier training) are affected by the urban context investigated (F1 scores from 0.49 to 0.81), whereas the FCN-based model proved to be the most robust and best-performing method for building extraction from a high-resolution raster DSM (F1 scores from 0.80 to 0.86).

Near real-time rainfall monitoring at local scale is essential for urban flood risk mitigation. Previous research on precipitation visual effects supports the idea of vision-based rain sensors, but tends to be device-specific. We aimed to use different available photographing devices to develop a dense network of low-cost sensors. Using Transfer Learning with a Convolutional Neural Network, the rainfall detection was performed on single images taken in heterogeneous conditions by static or moving cameras without adjusted parameters. The chosen images encompass unconstrained verisimilar settings of the sources: Image2Weather dataset, dash-cams in the Tokyo Metropolitan area and experiments in the NIED Large-scale Rainfall Simulator. The model reached a test accuracy of 85.28% and an F1 score of 0.86. The applicability to real-world scenarios was proven with the experimentation with a pre-existing surveillance camera in Matera (Italy), obtaining an accuracy of 85.13% and an F1 score of 0.85. This model can be easily integrated into warning systems to automatically monitor the onset and end of rain-related events, exploiting pre-existing devices with a parsimonious use of economic and computational resources. The limitation is intrinsic to the outputs (detection without measurement). Future work concerns the development of a CNN based on the proposed methodology to quantify the precipitation intensity.