| Titre : | Identifying and tracking vehicles from a front facing camera using deep learning |
| Auteurs : | Lokmane Rezig, Auteur ; Mohamed Chaouki Babahenini, Directeur de thèse |
| Type de document : | Monographie imprimée |
| Editeur : | Biskra [Algérie] : Faculté des Sciences Exactes et des Sciences de la Nature et de la Vie, Université Mohamed Khider, 2023 |
| Format : | 1vol.(53p.) / ill.couv.ill.encoul / 30cm |
| Langues: | Anglais |
| Langues originales: | Anglais |
| Résumé : |
The identification and tracking of vehicles play a crucial role in various applications, includingb traffic management, surveillance systems, and autonomous driving. Traditional methods for vehicle detection and tracking often rely on handcrafted features and heuristics, which are limited in their ability to handle complex and diverse real-world scenarios. In recent years, deep learning approaches have emerged as a promising solution to overcome these limitations, leveraging the power of neural networks to learn complex representations directly from raw data. This research focuses on the development of an efficient and accurate system for identifying and tracking vehicles from a front-facing camera using deep learning techniques. The proposed system consists of two main components: vehicle detection and vehicle tracking. For vehicle detection, a deep convolutional neural network (CNN) is trained on a large dataset of annotated vehicle images to learn discriminative features. The trained CNN is capable of accurately localizing and classifying vehicles in real-time from the input video frames. This enables efficient detection of multiple vehicles in various environmental conditions, including challenging scenarios such as occlusion and varying lighting conditions. For vehicle tracking, a combination of object tracking algorithms and deep learningbased techniques is employed. The system maintains a set of vehicle tracks over consecutive frames, utilizing methods such as Kalman filtering, optical flow, and deep feature matching. Deep learning-based methods aid in handling challenging situations like abrupt motion changes, occlusion, and temporary object disappearance. The proposed system is evaluated on a benchmark dataset and compared against stateof- the-art methods. The results demonstrate the effectiveness of the deep learning-based approach in accurately identifying and tracking vehicles from a front-facing camera. The system achieves high detection and tracking accuracy, robustness to challenging scenarios, and real-time performance. In conclusion, this research presents an effective solution for identifying and trackingv vehicles from a front-facing camera using deep learning techniques. The proposed system offers improved accuracy and robustness compared to traditional methods, enabling its po- tential applications in traffic management, surveillance systems, and autonomous driving, ultimately contributing to safer and more efficient transportation systems. |
| Sommaire : |
1 General introduction 5 1.1 Problem statement . 5 1.2 Proposed method . . 6 1.3 Description of the dissertation .6 2 Tracking vehicles based on AI 8 2.1 Introduction of vehicle Tracking . 8 2.2 Machine learning for tracking vehicles .10 2.2.1 Supervised learning 11 2.2.2 Unsupervised learning . 12 2.2.3 Semi-supervised Learning . 12 2.3 Deep learning for tracking vehicles . 13 2.3.1 Terminology . . . 13 2.3.2 Convolutional Neural Networks : . 14 2.3.3 Issues and Solutions in CNN-based vehicle Tracking : . . 17 2.4 Vehicles tracking techniques classification . 19 2.4.1 Motivation . . 19 2.4.2 Geometries and Tradition techniques . 21 2.4.3 Tracking using SVM . . 26 2.4.4 Techniques based on CNN . 27 2.5 Conclusion . 29 3 CONTENTS 3 The Design of software pipeline to identify vehicles in a video from a front-facing camera on a car 30 3.1 Introduction . 30 3.2 General architecture 30 3.3 Architectural details . 32 3.4 Data preparation . 33 3.4.1 The dataset used . 33 3.4.2 General data center architecture . 33 3.4.3 dataset Division . 34 3.4.4 Training the model of learning 34 3.5 The model testing phase . 36 3.6 Conclusion . 37 4 Implementation of architecture 38 4.1 Introduction . 38 4.2 Development Environments and Tools . 38 4.3 Preparation of the data . 42 4.4 Realization of the CNN model . 44 4.5 Parameter initialization CNN . 45 4.6 Network learning settings . . 46 4.7 Results obtained . 47 4.8 Model evaluation on test data . 49 4.9 screenshot of result . .50 5 Conclusion and perspectives 53 |
Disponibilité (1)
| Cote | Support | Localisation | Statut |
|---|---|---|---|
| MINF/827 | Mémoire master | bibliothèque sciences exactes | Consultable |
