Increasing the Level of Safety and Security of Transport Infrastructure from Air Threats Based on Their Acoustic Identification

In recent years, innovative technologies for monitoring and managing civil infrastructure have been widely used by stakeholders and managing bodies, to ensure the correct and efficient maintenance of these critical elements. In this context, great attention is given to the environment surrounding these assets. Indeed, the environment plays a crucial role in altering the conditions of the civil infrastructures, either as a result of catastrophic natural events or due to the gradual morphological changes over time. Understanding the surrounding conditions of critical infrastructures is therefore essential to ensure accurate predictions regarding the changing conditions of the infrastructures over time, to prevent damages, and to intervene promptly and efficiently when necessary. In addition, comprehending the environmental conditions surrounding critical infrastructures is crucial for ensuring accurate long-term monitoring procedures, mitigating potential damages proactively, and facilitating efficient and timely interventions when necessary. Moreover, several international studies and research projects focused on these topics have been increasingly supported by industry and infrastructure managers in handling data from innovative monitoring technologies. Among these technologies, remote sensing techniques, including satellite analysis using MT-InSAR methods for assessing structural subsidence [1], and change detection techniques to evaluate temporal variations are gaining momentum. Furthermore, the use of Unmanned Aerial Vehicles (UAVs), to integrate information derived from other remote surveys, stands as a crucial topic to be more investigated. This research aims to identify a methodology for managing multi-sensor and multi-scale survey information integrating satellite remote sensing and ground-based Non-Destructive Testing for Digital Twin-based infrastructure monitoring. However, the interpretation of data derived from satellite remote sensing [1] and ground-based Non-Destructive Testing (NDT) [2] techniques remain an area awaiting comprehensive exploration within the realm of transport infrastructure monitoring. This approach is aimed at the definition of a Digital Twin of the analyzed infrastructure and the environment in which it is located. An experimental application was developed selecting a bridge, located in Italy, identified as a case study. Several data obtained from inspections performed by UAVs and satellite remote sensing were implemented, as well as a digital modeling process, specifically developed for integrating such database to create a Digital Twin of the bridge and the environment. This application stands as a starting point for defining a broader integrated monitoring methodology for the management of critical transport infrastructures. AcknowledgementsThis research is supported by the Project “SIMICOM”  accepted and funded by the Lazio Region, ItalyReferences[1] Gagliardi V., et Al. Digital twin implementation by multisensors data for smart evaluation of transport infrastructure. SPIE Optical Metrology. Multimodal Sensing and Artificial Intelligence: Technologies and Applications III, Munich, 2023. [2] Tosti F., et Al "Integration of Remote Sensing and Ground-Based Non-Destructive Methods in Transport Infrastructure Monitoring: Advances, Challenges and Perspectives," 2021 IEEE Asia-Pacific Conference on Geoscience, Electronics and Remote Sensing Technology (AGERS), Jakarta Pusat, Indonesia, 2021, pp. 1-7, 

With the increasing threat of unmanned aerial vehicles to critical railway infrastructure, the need for advanced detection technologies has become more urgent. This paper reviews existing railway monitoring solutions and outlines their limitations in identifying aerial threats. An acoustic analysis is conducted to extract distinctive unmanned aerial vehicle sound patterns using Mel-frequency cepstral coefficients, which serve as primary features for classification. Neural network models are applied to detect and differentiate aerial threats from environmental noise, achieving high recognition accuracy. The study also describes the development of an embedded artificial intelligence system based on STM32 microcontrollers, which combines real-time digital signal processing with efficient on-device neural inference. This solution offers a scalable and energy-efficient platform for decentralized audio-based drone detection in railway security applications.

ABSTRACTPrecise estimation of rice nitrogen (N) content is essential for optimizing fertilizer use. Traditional methods for estimating N content are time‐consuming, laborious, and costly. Unmanned aerial vehicles (UAVs) are time and money efficient substitutes allowing more accurate and flexible monitoring for larger rice areas. The objectives of this study were to: (i) develop random forest (RF) and artificial neural network (ANN) models for predicting and mapping the nitrogen content (%) in rice using seven vegetation indices derived from UAV multispectral sensors and; (ii) assess the key vegetation indices (VI) and their interrelationships with the predicted nitrogen content. Experiments were conducted at two locations in Cuttack district of Odisha, India, with different nitrogen levels. The UAV images were collected synchronizing with the maximum tillering stage of rice and seven indices were generated. The rice sampling was done on the date of flying UAV images and nitrogen content was estimated in the laboratory. RF and ANN models were developed using the N content as dependent and the VIs as independent variables. Both the models exhibited robust predictive capabilities, however, the RF model exhibited better performance, compared to the ANN model. Nitrogen content prediction using the developed RF and ANN models in testing site at farmer's field ranged from 0.78% to 1.95% (R2 of 0.67%) and from 0.5% to 1.78% (R2 of 0.55%), respectively. Normalized difference red edge (NDRE) and normalized difference vegetation index (NDVI) turned out as significant contributors in the development of both the models.

The main purpose of this work was to obtain background sound levels and speech intelligibility as well as to evaluate classification of speech commands in the presence of an unmanned aerial vehicle (UAV) equipped with four rotating propellers. Speech intelligibility was assessed using speech interference level (SIL) parameter according to ISO 9921. The UAV background sound levels were recorded in laboratory conditions using Norsonic140 sound analyzer in the absence of the UAV and in the presence of the UAV. The classification of speech commands/left, right, up, down, forward, backward, start, stop/recorded with Olympus LS-11 was evaluated in laboratory condition based on Mel-frequency cepstral coefficients and discriminant function analysis. The UAV was hovering at 1.5m during recordings. The A-weighted sound level obtained in the presence of the UAV was 70.5 dB(A). Speech intelligibility rating was poor in the presence of the UAV. Discriminant analysis based on Mel-frequency cepstral coefficients showed very successful classification of speech commands equal to 100%. Evaluated speech intelligibility did not exclude verbal communication with the UAV. The successful classification of speech commands in the presence of the UAV can enable the control of the UAV using voice commands and general communication with the UAV using speech.

This paper examines the defeat of various types of air attack weapons, such as fighters, helicopters, cruise missiles, reconnaissance unmanned aerial vehicles, armed unmanned aerial vehicles, drones and kamikazes by air defense systems, taking into account their low-altitude operation. For this purpose, efficiency coefficients are determined that characterize the defeat of low-altitude air attack weapons by anti-aircraft missile systems, taking into account their tactical and technical characteristics. To solve the problem, the first party was considered to be air attack weapons, such as fighters, helicopters, cruise missiles, reconnaissance unmanned aerial vehicles, armed unmanned aerial vehicles, drones and kamikazes, and the second party was considered to be the S-125 2TM, BUK-MB, Barak-8, S-300 PMU2, TOR-M2KM, Patriot MIM-104, Panisir S-1, Nasams-III, Igla-S, Strela-10, OSA-AK(M) anti-aircraft missile systems. Therefore, a mathematical model of the problem was developed using game theory. The obtained problem was solved using the simplex method. It was determined that the TOR-M2KM anti-aircraft missile system, which is an air defense system for engaging low-altitude air attack weapons, can be used primarily against fighters, helicopters, unmanned aerial vehicles, armed unmanned aerial vehicles and unmanned aerial targets during combat operations. Of the air attack weapons, it is proposed to use cruise missiles first of all during combat operations, and kamikazes at the next stage. It is advisable to use the Barak-8 air defense system against cruise missiles. The obtained results can be used in planning the effective implementation of an air defense system against low-altitude air attack weapons.

Performance prediction and analysis of a PEM fuel cell operating on pure oxygen using data-driven models: A comparison of artificial neural network and support vector machine

<p>Aiming to ensure the safety of the transport infrastructure during operation through the improvement of maintenance policies across Europe, the European Commission opened in 2019 the call for the Coordination and Support Action (CSA) “Monitoring and safety of transport infrastructure”. The main goal of this CSA is to support the preparation of a mandate for a CEN standard for the maintenance and control of the European transport infrastructure. In 2020, the CSA was granted to the IM-SAFE project consortium. Based on the analysis of standardisation needs, good practice and available knowledge and technology, the future perspectives of standardisation for the use of monitoring, inspection and testing in managing the safety risks to transport infrastructure has been studied.</p><p>This contribution presents the scope proposed by the IM-SAFE project for future harmonised European standards in the domain of monitoring, data-informed safety assessment and condition- based and risk-based predictive maintenance policies for bridges and tunnels, considering the integration of digital innovations as enabling technology.</p>

The article presents some signs of the full-scale armed aggression of the Russian Federation against Ukraine and analyzes the use of air and space attack means by the Russian Federation troops, in particular ballistic and cruise missiles, aircraft and unmanned aerial vehicles in the period from 24.02.22 to 24.02.24 at different stages, which differ in the tactics of their use and the performance of certain tasks. It is determined that the main direction of the use of air and space attack capabilities of the Russian Federation in the war against Ukraine remains the integrated use of cruise and ballistic missiles of various types and unmanned aerial vehicles for various purposes in the course of massive strikes on critical infrastructure of the country, as well as intensive air strikes to provide air support to enemy ground forces in their offensive operations. Between massive strikes, we should expect complex targeted (group) strikes by unmanned aerial vehicles and missiles of various types on the military and civilian infrastructure of the country in order to reduce the potential of the military-industrial complex, disrupt the functioning of the energy infrastructure and exert psychological influence on the minds of citizens and the military and political leadership of the country.This material is recommended for use by the operational personnel of the Air Forces of the Armed Forces of Ukraine during preparation and conduct of combat operations in order to plan and implement measures to ensure effective counteraction to the means of the Russian Federation's CBRN forces in order to protect important state facilities of the country and troops on the battlefield.

Unmanned aerial vehicles (UAVs) are widely applied in the civil domain. Meanwhile, the security issues concerning UAVs are emerging because the monitoring of small UAVs is difficult. In this paper, we designed a sound detection and alarm system for UAV monitoring. Firstly, the frequency domain feature extraction and Mel domain feature extraction are performed to obtain the features of UAV sound. The Mel frequency cepstral coefficient (MFCC) and support vector machine (SVM)-based sound detection method are designed. Besides, the real-time sound acquisition and detection are realized. The real test data has proved the efficiency of the designed UAV sound detection system. The advantages of the sound detection and alarm system in this paper are low cost and high accuracy, which may be widely applied in the area of UAV monitoring.

Applications of unmanned aerial vehicle (UAV) in road safety, traffic and highway infrastructure management: Recent advances and challenges

The problem of attending to the health of the aged who live alone has became an important issue in developed countries. One way of solving the problem is to check their health condition by a remote-monitoring technique and support them with well-timed treatment. The purpose of this study is to develop an automatic system that can monitor a health condition in real time using acoustical information and detect an abnormal symptom. In this study, cough sound was chosen as a representative acoustical symptom of abnormal health conditions. For the development of the system distinguishing a cough sound from other environmental sounds, a hybrid model was proposed that consists of an artificial neural network (ANN) model and a hidden Markov model (HMM). The ANN model used energy cepstral coefficients obtained by filter banks based on human auditory characteristics as input parameters representing a spectral feature of a sound signal. Subsequently, an output of this ANN model and a filtered envelope of the signal were used for making an input sequence for the HMM that deals with the temporal variation of the sound signal. Compared with the conventional HMM using Mel-frequency cepstral coefficients, the proposed hybrid model improved recognition rates on low SNR from 5 dB down to -10 dB. Finally, a preliminary prototype of the automatic detection system was simply illustrated.

Domestic violence between partners and family members is a worldwide problem increasing every day. As per academic studies and media articles, it escalated during the COVID-19 outbreak. Domestic violence can portray verbally and physically in several ways (for instance, between partners or against children and older people). Deep Learning (DL) combined with the Internet of Things (IoT) technology could support the detection of domestic violence, which is one of many societal issues. This paper describes a system that uses a Deep Learning model and smart microphones to identify domestic violence. The datasets used are from the Google AudioSet (GA) and from the Toronto Emotional Speech Set (TESS). For the training of the dataset, the system used spectrograms and MFCCs (Mel-Frequency Cepstral Coefficients). The system employs two approaches: (i) an ANN (Artificial Neural Network) model, and (ii) a ResNet model. The Resnet model obtained an accuracy of 71%. The ANN model, which brought an accuracy of 83%, was tested and loaded to a Raspberry Pi, i.e., connected to the microphone for audio recording. The recorded audio was fed to the trained model, classifying the audio, and alerting the domestic violence to relatives, friends, or volunteers registered with the system via e-mail. The designed system is compact and can be placed inside the home.KeywordsDomestic violenceSmart microphonesDeep learningSensorsANNResNetSpectrogramMFCC

Speech recognition technology makes human contact with the computer more accessible. There are two phases in the speaker recognition process: capturing or extracting voice features and identifying the speaker's voice pattern based on the voice characteristics of each speaker. Speakers consist of men and women. Their voices are recorded and stored in a computer database. Mel Frequency Cepstrum Coefficients (MFCC) are used at the voice extraction stage with a characteristic coefficient of 13. MFCC is based on variations in the response of the human ear's critical range to frequencies (linear and logarithmic). The sound frame is converted to Mel frequency and processed with several triangular filters to get the cepstrum coefficient. Meanwhile, at the speech pattern recognition stage, the speaker uses an artificial neural network (ANN) Madaline model (many Adaline/ which is the plural form of Adaline) to compare the test sound characteristics. The training voice's features have been inputted as training data. The Madaline Neural Network training is BFGS Quasi-Newton Backpropagation with a goal parameter of 0,0001. The results obtained from the study prove that the Madaline model of artificial neural networks is not recommended for identification research. The results showed that the database's speech recognition rate reached 61% for ten tests. The test outside the database was rejected by only 14%, and 84% refused testing outside the database with different words from the training data. The results of this model can be used as a reference for creating an Android-based real-time system.

Speech recognition technology makes human contact with the computer more accessible. There are two phases in the speaker recognition process: capturing or extracting voice features and identifying the speaker's voice pattern based on the voice characteristics of each speaker. Speakers consist of men and women. Their voices are recorded and stored in a computer database. Mel Frequency Cepstrum Coefficients (MFCC) are used at the voice extraction stage with a characteristic coefficient of 13. MFCC is based on variations in the response of the human ear's critical range to frequencies (linear and logarithmic). The sound frame is converted to Mel frequency and processed with several triangular filters to get the cepstrum coefficient. Meanwhile, at the speech pattern recognition stage, the speaker uses an artificial neural network (ANN) Madaline model (many Adaline/ which is the plural form of Adaline) to compare the test sound characteristics. The training voice's features have been inputted as training data. The Madaline Neural Network training is BFGS Quasi-Newton Backpropagation with a goal parameter of 0,0001. The results obtained from the study prove that the Madaline model of artificial neural networks is not recommended for identification research. The results showed that the database's speech recognition rate reached 61% for ten tests. The test outside the database was rejected by only 14%, and 84% refused testing outside the database with different words from the training data. The results of this model can be used as a reference for creating an Android-based real-time system.

It is important to establish protocols for dealing with suspicious Unmanned Aerial Vehicles (UAVs). To do this, it is necessary to develop efficient UAV recognition systems. This paper describes the process of planning and implementing a drone recognition system that focuses on the audio signals generated by UAVs. To do this, a small UAV audio data set was created, using Mel Frequency Cepstral Coefficients as the audio feature analyzed by both a Support Vector Machine model (SVM) and a Convolutional Neural Network (CNN) to create a system able of recognizing the audio generated by UAVs. The resulting system was evaluated and compared, and based on the results, the SVM models was chosen to implement the audio recognition system. Further advantages and disadvantages of UAV audio signal recognition systems, as well as recommendations for their implementation were noticed during this research.