Multiple-object Detection Using Feature-Based Method in Manufacturing Line Inspection

Abstract

Abstract— This paper discuss on a combination of SURF and BRISK feature descriptor method to achieve multiple object detection in one image with shorter inspection time and using estimate geometric transformation. Tested inspected object used were three types of Arduino circuit board of Uno, Mega and Nano. In pre-processing stage, speed up robust feature and binary robust invariant scalable keypoint descriptor were employed for detecting and matching the tested object in cluttered image. By using estimate geometric transformation function, it enabling segmentation of orientated and distorted object. In processing stage, the background subtraction technique, called foreground detection was applied to recognize the abnormality of the tested object and label in status of that inspected object.