SIFT Descriptor Robustness Analysis to Brightness Changes of Robowaiter Vision Sensor System
T N Nizar*, S Supatmi**, E P Putro***
Computer Engineering Department, Faculty of Engineering and Computer Science
Universitas Komputer Indonesia, Jl Dipatiukur 112-116 Bandung
*taufiq.nuzwir[at]email.unikom.ac.id, **sri.supatmi[at]email.unikom.ac.id, ***ekoprabowoputro[at]gmail.com
Abstract
This paper discussed about the feature detection problem in the computer vision affected by the brightness change. The presented descriptor is Scale Invariant Feature Transform (SIFT). The method is an algorithm in computer vision to detect and describe local feature in image which robustly identify object and invariant to uniform scaling, orientation, brightness changes, and partially invariant to affine distortion. We implement this algorithm to Robowaiters object detection system that must detect and recognize objects around its task like food, beverage, refrigerator and any kitchen objects. For this analysis case, we use beverage box image for sample image. The algorithm detects and recognize the image in normal brightness, and then the image brightness value increased and decreased. In the implementation, the algorithm success to detect and recognize the presented objects and distinguish it with success rate of 93.5% of image brightness increase and 95.5 % of image brightness decrease. Based the result, SIFT algorithm is robust to illumination changes for our case.
Keywords: SIFT, Brightness changes, computer vision, local image descriptor, feature.
Topic: Electrical and Computer Engineering