![]() | Only 14 pages are availabe for public view |
Abstract • Chapter one (introduction): This chapter discusses the motivation behind the thesis in addition to the problem statement. The chapter emphasis the main challenges that faces autonomous systems especially Visual Odometry algorithms. The thesis structure and contributions are also stated. • Chapter two (Literature survey and related work): This Chapter emphasizes the state of the art related to autonomous vehicle’s perception and localization in addition to the major algorithms used in modern Visual odometry techniques the chapter also discusses the challenges and gap of knowledge facing such algorithms. • Chapter three (Monocular Visual Odometry drift reduction using neural networks) : this chapter discusses the proposed Drift reduction neural network for monocular visual odometry algorithms. The chapter demonstrates the proposed pipeline, the methodology in addition to the experimental work and results of the work. • Chapter four (Stereo Visual Odometry Drift Reduction using Feed Forward Neural Networks): This chapter demonstrates a set of proposed drift reduction neural networks designed for a hybrid stereo Visual odometry algorithm that utilizes both monocular and stereo image information in order to reduce the drift in both orientation and translation motion estimates. The chapter shows the methodology, experimental work and the results. • Chapter five (Conclusion and future work): This chapter emphasizes the proposed work done conclusions in addition to the future work intended to be done. Keywords: Autonomous Vehicles, Perception, Localization, Drift Reduction, Machine Learning, Neural Networks |