Title:	DEPTH VALUE DEDUCTION USING OPTICAL FLOW FOR REVERSE ENGINEERING
Author(s):	Rahmita Wirza Rahmat, Suhail Azmi
ISBN:	978-972-8939-48-9
Editors:	Yingcai Xiao
Year:	2011
Edition:	Single
Keywords:	Reverse engineering, optical flow
Type:	Short Paper
First Page:	245
Last Page:	248
Language:	English
Cover:
Full Contents:	click to dowload
Paper Abstract:	Optical flow is actually a tracking algorithm used for detecting moving features in images (at least two). With optical flow, we can get the motion magnitude and vector of the tracked features. In this paper, we proposed an algorithm to acquire 3D depth information from images using optical flow magnitude, while we filtered out the noise by using its vector. The idea; when rotating an object in front of sensors, nearer parts of the object will move more than further parts, which mean, the nearer part has a bigger magnitude than the latter. In this algorithm we are trying to achieve full automation in detecting 3D depth of the tracked features by exploiting the optical flow detected. To achieve this, we prepared an optimum environment for the implementation. Camera is placed at a specific location perpendicular to a black turntable with black background. With two images, one taken at zero degrees and one at 2.5 degree rotation, we apply a Pyramidal Implementation of Lucas Kanade optical flow to track detected features. Experiments are done to validate the 3D points constructed; three objects were tested quantitatively. The implementation of this framework succeeded in extracting features, estimating depth information and visualizing 3D result. Comparing with the real objects, we can conclude that the results are similar but not as accurate as conventional reverse engineering techniques. However, it will be a step forward to have a framework and a technique that greatly reduces cost and time, aside from being portable. Future research should be carried out in perfecting this framework because of its advantages and prospects.
	Go Back