Loading...

Using Optical Flow and Neural Network for Insect-Inspired Visual Navigation of an Autonomous Vehicle

Ghafelebashy, Atiyeh Sadat | 2023

30 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56322 (45)
  4. University: Sharif University of Technology
  5. Department: Aerospace Engineering
  6. Advisor(s): Banazadeh, Afshin
  7. Abstract:
  8. In this research, inspired by the navigation system of insects, a method for extracting the navigation information required for an autonomous device is presented, with an information input section in the form of image processing using optical flow and an information processing section with an artificial brain structure. In this method, the information obtained from the optical flow is processed by the neural network structure and the navigation information needed to control and guide the autonomous device is obtained. There are several methods to estimate the optical flow between two frames, in this research, the DEQ-Flow optical flow estimation network, which is a method based on deep learning, has been used due to its speed and better efficiency. The type of artificial neural network used for motion estimation has been selected by examining existing deep neural networks as a combination of convolutional network and auto-encoder network, and the famous KITTI dataset has been used for its training and evaluation, and its performance has been compared with previous methods. Finally, the designed navigation system will be able to calculate the position and condition of the device with acceptable accuracy, along with the existing navigation equipment. Such a comprehensive simulation of insect performance is one of the new issues in the field of bionics
  9. Keywords:
  10. Visual Navigation ; Artificial Neural Network ; Naturally Inspired Method ; Optical Flow ; Autoencoder Neural Networks ; Convolutional Neural Network

 Digital Object List

 Bookmark

No TOC