Biography
Dr. Rostam Affendi Hamzah
Dr. Rostam Affendi Hamzah
Universiti Teknikal Malaysia Melaka, Malaysia
Title: 3D Terrain Surface Reconstruction from Stereo Image using Deep Learning
Abstract: 
This study presents a new framework of stereo matching algorithm that can generate three-dimensional from stereo camera. A stereo matching process produces a disparity map which provides the depth information required in 3D reconstruction. This map consists of disparity values of two corresponding points. Furthermore, the accuracy of 3D reconstruction depends on how precise the disparity being estimated on each pixel location. To get a good 3D reconstruction result, the propose stereo matching algorithm must be strong against the radiometric differences and edge distortions. Hence, this work proposes a new stereo matching algorithm with high accuracy for 3D surface reconstruction. First stage, hybrid Convolutional Neural Network (CNN) with Gradient Matching (SG) is proposed which uses deep learning and magnitude differences. The CNN and gradient matching are strong against the radiometric distortions due to different characteristics of the input stereo cameras. Second stage, the Adaptive Support Weight (ASW) with iterative Guided Filter (ASW iGF) is proposed to improve the edges of object matching. The last stage, Joint Weighted Guided Filter (JWGF) is suggested to reduce the remaining noise on the disparity map. This map is used for 3D surface reconstruction. Based on the standard quantitative benchmarking stereo dataset, the proposed work in this article produces good results and performs much better compared with before the proposed framework. This new algorithm is also competitive with some established methods in the literature. Then from this accurate framework, the 3D surface reconstruction is employed with the terrain or land images as an input using Point Cloud Library (PCL).
Biography: 
Rostam Affendi Hamzah graduated from Universiti Teknologi Malaysia where he received his B.Eng majoring in Electronic Engineering. Then he had his M. Sc. in Electronic System Design Engineering and Doctor of Philosophy majoring in Imaging from the Universiti Sains Malaysia. Currently he is a senior lecturer in the Universiti Teknikal Malaysia Melaka teaching digital electronics, digital image processing and embedded system. His areas of specialization are computer vision, pattern recognition and artificial intelligence. Besides, he was an invited speaker and presented papers at international conferences. As a result, he obtained several competitive research grants. Also, he refereed articles for, among others, international journals of high esteem, such as IEEE Access, IEEE Transactions on Mechatronics, IET Image Processing, Electronics Letters and IET Computer Vision. He can be contacted at email: rostamaffendi@utem.edu.my.