Enhanced user interface for the control of a deep tunnel robotic platform

• Main Author: Koh, Thean Chun
• Other Authors: Seet Gim Lee, Gerald
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/74531
• Institution: Nanyang Technological University

This Final Year Report describes the author’s work in developing image stitching method to enhance user interface for the control of a deep tunnel robotic platform, which will be utilized by Public Utilities Board, Singapore. The robotic platform is developed and be capable of being driven in the hazardous environment and operated to conduct the tasks that cannot be done by manual inspection. To provide necessary visual information inside deep tunnel, various cameras which can provide user a good observation to the tunnel environment would be implemented and programmed. Besides that, visual inspection can be improved through wide field of view of video frames using image stitching method. Before implementing the image stitching algorithm, camera calibration was needed to remove camera lens distortion and curvature distortion. Camera calibration was done by using chessboard and black and white fabric checker with suitable camera calibration algorithm. The distortion could be removed by comparing the expected distance and actual distance of chessboard corners and getting distortion coefficients and intrinsic matrix. Two image stitching methods were developed in this project: automatic image stitching and manual image stitching. Automatic image stitching involves features extraction, features matching and multi-band blending. Manual image stitching is a method that estimates homography transformation matrix manually to perform perspective transformation the images and combine these images side by side by setting a ROI (Region of Interest). The comparison between the results of two image stitching methods were carried out to find the best image stitching method for deep tunnel robotic platform. The environment constraint of deep tunnel was highlighted to make sure that the requirements could be achieved by the program design and get a better result for the user to inspect deep tunnel easily. After several tests were carried out, manual image stitching is chosen as the stitching method for deep tunnel images and videos as it requires less time to finish stitching process and achieves high stitching quality for the output deep tunnel images and videos.