Search In this Thesis
   Search In this Thesis  
العنوان
Improving SLAM Performance under Low Light Intensities /
المؤلف
Mohamed, Mohamed Hesham Mostafa Abdelaziz.
هيئة الاعداد
باحث / محمد هشام مصطفى عبدالعزيزمحمد
مشرف / محمود إبراهيم خليل
مناقش / حسن طاهر درة
مناقش / محمد واثق على كامل الخراشى
تاريخ النشر
2023.
عدد الصفحات
105 P. :
اللغة
الإنجليزية
الدرجة
ماجستير
التخصص
هندسة النظم والتحكم
تاريخ الإجازة
1/1/2023
مكان الإجازة
جامعة عين شمس - كلية الهندسة - قسم هندسة الحاسبات والنظم
الفهرس
Only 14 pages are availabe for public view

from 105

from 105

Abstract

Simultaneous Localization and Mapping (SLAM) is a computational method that creates a map while simultaneously localizing a moving item in an unknown environment, such as a car or perhaps a robot. State-of-the-art SLAM methods like the ORBSLAM3 algorithm achieve good performance in building maps of unknown environments; however, datasets generated by capturing underwater images suffer from low light conditions which in turn affects our ability to build a strong SLAM system for underwater activities such as pipeline and seabed inspection and many others. In this work, we study the effect of pre-processing on SLAM in low-light conditions. In this context, we compare the performance of the traditional contrast adaptive histogram equalisation (CLAHE) method with the more contemporary KinD deep learning approach and showed that the CLAHE algorithm improved the accuracy of the ORB-SLAM3 algorithm.
This thesis is broken into six chapters as follows:
Chapter 1: Introduction
Chapter one introduces the problem that this thesis aims to solve.
Chapter 2: Background
Chapter two is necessary to understand the rest of the thesis. This chapter gives a detailed overview of the SLAM concepts and their main workflow. The chapter is organized as follows: The first section starts by introducing SLAM and its important concepts. The second section discusses different types of cameras for SLAM. The third section gives an overview of visual SLAM workflow components. The last section gives an overview of the mathematical formulation of general SLAM Problems.
Chapter 3: Problem Formulation and Work Related
In chapter three we discuss the problem formulation, the related work, as well as the suggested approaches outlined. The chapter is organized as follows: in the first section of this chapter, we talk about SLAM systems that can work under challenging light conditions. In the second section, we talk about pre-processing techniques used for low-light image enhancement.
Chapter 4: Proposed Approach
In this chapter, we discuss the proposed approach to enhance the images as a pre-processing step before using it in the SLAM system.
Chapter 5: Evaluation of Proposed Approach and Dataset Preparation
In this chapter, we talk about the dataset we used in our work and the evaluation of our approach. This chapter is organized as follows: In the first part, we give more insights about our dataset and its preparation. In the second part, we evaluate our approach and the extracted results.
Chapter 6: Conclusion and Future Work
In this chapter, a summary of our work is outlined. Additionally, directions for future research are discussed.
Keywords: Computer vision, image processing, SLAM, Low light enhancement, Deep learning