(#57) Enhanced Object Localization via Neural Network Integration with Newton's Method
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
NSBM Green University
Abstract
Machine vision is crucial in autonomous vehicles and robotics, demanding precise object localization and environmental perception. Enhancing the baseline between cameras can improve accuracy and range but introduces challenges in stereo synchronization. This research presents a novel system that integrates neural networks with Newton's method to enhance object localization accuracy while reducing computational costs. Utilizing three cameras with known baselines, the system combines Convolutional Neural Network (CNN) models, such as YOLO, for critical component identification, with Newton's method for refining initial location estimates provided by machine learning algorithms. This integration ensures high accuracy by double-checking neural network outputs with mathematical models, avoiding local minima, and enhancing computational efficiency. Experimental results demonstrated successful camera calibration, accurate object detection, and significant improvements in point estimation accuracy using Newton's method. This approach promises advancements in 3D reconstruction and realtime object localization for autonomous systems.
Description
Citation
Balasooriya, T.N.; Dilpriya, H. (2024) (#57) Enhanced Object Localization via Neural Network Integration with Newton's Method, Technology Innovations & Digitalization Accompanying Creativity (TIDAC), NSBM Green University, Sri Lanka. P.114-119