ICT Applications for Smart Cities

This book is the result of four-year work in the framework of the Ibero-American Research Network TICs4CI funded by the CYTED program. In the following decades, 85% of the world's population is expected to live in cities; hence, urban centers should be prepared to provide smart solutions for problems ranging from video surveillance and intelligent mobility to the solid waste recycling processes, just to mention a few. More specifically, the book describes underlying technologies and practical implementations of several successful case studies of ICTs developed in the following smart city areas: • Urban environment monitoring • Intelligent mobility • Waste recycling processes • Video surveillance • Computer-aided diagnose in healthcare systems • Computer vision-based approaches for efficiency in production processes The book is intended for researchers and engineers in the field of ICTs for smart cities, as well as to anyone who wants to know about state-of-the-art approaches and challenges on this field. Preface Acknowledgements Contents 1 Art Graffiti Detection in Urban Images Using Deep Learning 1.1 Introduction 1.2 Detection of Art Graffiti 1.3 Object Detection Deep Architectures 1.3.1 YOLO Models 1.3.2 YOLOv4 1.3.3 YOLOv5 1.3.4 YOLOv4-tiny 1.4 Dataset 1.5 Experiments 1.5.1 Evaluation Metrics and Computing Resources 1.5.2 Experiments 1.6 Conclusion References 2 Deep Neural Networks for Passengers' Density Estimation and Face Mask Detection for COVID-19 in Public Transportation Services 2.1 Introduction 2.2 Proposed Architecture 2.3 Deep Learning for Object Detection 2.4 Experiments and Results 2.5 Conclusions and Future Work References 3 Epistemic Uncertainty Quantification in Human Trajectory Prediction 3.1 Introduction 3.2 Related Work and Problem Statement 3.3 Quantification of Aleatoric and Epistemic Uncertainties 3.3.1 Bayesian Deep Learning 3.3.2 Uncertainty Estimation Through Bayesian Deep Learning 3.4 Evaluating and Calibrating Uncertainties 3.4.1 Calibration of Uncertainties in 1D Regression 3.4.2 Highest Density Regions 3.4.3 HDR-Based Calibration 3.5 Experiments 3.5.1 Trajectory Prediction Base Model and Bayesian Variants 3.5.2 Implementation Details 3.5.3 Evaluation of the Uncertainties Quality 3.5.4 Evaluation of the Re-calibration Process 3.6 Conclusions References 4 Automatic Detection of Knives in Complex Scenes 4.1 Introduction 4.2 Related Work 4.3 YOLOv4 Architecture for Detection of Knives 4.3.1 Detection of Knives 4.3.2 YOLOv4 4.4 Datasets 4.4.1 DaSCI Dataset 4.4.2 MS COCO Dataset 4.4.3 Knife Classification Datasets 4.5 Pre-processings on Dataset 4.5.1 Dataset Preparation 4.5.2 Dataset Variabilities 4.5.3 Transfer Learning 4.6 Experimental Results 4.6.1 Description of Performance Metrics 4.6.2 Experimental Results 4.6.3 General Results 4.6.4 Results Considering Variabilities in Images 4.7 Conclusion References 5 Human Body Pose Estimation in Multi-view Environments 5.1 Introduction 5.2 Camera Pose Estimation 5.2.1 Siamese Network Architecture 5.2.2 Results from Real World Datasets 5.2.3 From Virtual Environments to Real World 5.3 Human Pose Estimation 5.3.1 Multi-view Scheme 5.3.2 Results from Multi-view Approach 5.4 Conclusions References 6 Video Analytics in Urban Environments: Challenges and Approaches 6.1 Introduction 6.2 Image Preprocessing 6.2.1 Camera Calibration 6.2.2 Background Subtraction 6.2.3 Image and Content Enhancement 6.3 Detection 6.4 Classification 6.5 Tracking 6.6 Applications 6.6.1 Traffic Scenarios 6.6.2 Pedestrian-Oriented Applications 6.7 Datasets 6.8 Conclusions References 7 Multimodal Sensor Calibration Approaches in the ATLASCAR Project 7.1 Introduction 7.2 Calibration of 3D and 2D LiDARs with Conical Targets 7.2.1 Proposed Calibration Process 7.2.2 Results 7.3 Spherical Target to Calibrate LiDARs and Cameras 7.3.1 Principles of the Calibration Approach 7.3.2 Geometric Transformations from Sets of Ball Centers 7.3.3 Results 7.4 Deep Learning to Detect a Sphere in Multimodal Images 7.4.1 Detection of the Calibration Target using the Same Neural Network 7.4.2 Conversion of the Coordinates of the Center of the Ball to Meters 7.4.3 Transformation Matrix Calculation 7.4.4 Results 7.5 Optimization Approach for Calibration 7.5.1 The Methodology 7.5.2 Results 7.6 Comparative Analysis of the Techniques 7.7 Conclusion and Future Perspectives References 8 Early Computer-Aided Diagnose in Medical Environments: A Deep Learning Based Lightweight Solution 8.1 Introduction 8.2 Methodology 8.2.1 Preprocessing 8.2.2 Attention Residual Learning 8.2.3 EfficientNets 8.3 Experiments and Results 8.3.1 Impact of the Preprocessing Methods in the Classification 8.3.2 Attention Residual Learning on EfficientNet 8.4 Conclusions References 9 Melamine Faced Panel Inspection, Towards an Efficient Use of Natural Resources 9.1 Introduction 9.2 Related Works 9.3 Theoretical Background 9.3.1 Local Binary Pattern (LBP) 9.3.2 Support Vector Machine (SVM) 9.4 Used Methodology for Surface Classification 9.4.1 Used Dataset 9.4.2 Dealing with Imbalanced Data 9.4.3 Used Programming Tools 9.4.4 Features Computation 9.4.5 Classification Training and Testing 9.5 Results 9.6 Comparisons with Previous Work 9.7 Conclusion References 10 Waste Classification with Small Datasets and Limited Resources 10.1 Introduction 10.2 Related Work 10.2.1 Waste Recycling 10.2.2 Network Distillation 10.3 Reviewing Network Distillation 10.4 Proposed Validation Protocol and Experimental Results 10.4.1 The TrashNet Dataset 10.4.2 Data Preprocessing 10.4.3 Evaluation 10.4.4 Implementation Details 10.4.5 Results 10.5 Conclusions and Future Work References