Abstract
Upon increasing number of computer usage in industries, the traditional methods and equipments of human subjective evaluations are gradually being replaced by the machine vision, which is a smart camera connected with computer based systems. It is called as Automated Visual Inspection System (AVIS). At current state, AVIS applications are mostly developed based on geometric and shape feature extraction methods. Unfortunately, that approach has some problems due to image scaling, illumination, noise and invariant issues. Therefore, the main objective of this research is to design spatial local feature transform for AVIS development. The AVIS methodology covers two parts: the first one is the hardware part which consists of webcam, source light and the conveyor belt while the software part comprising image processing based on the Bag-of-Words (BoW) model and object recognition based on K-Nearest Neighbor (K-NN) algorithm. Nine surgical instruments are used as the case study for the object recognition. This research is evaluated based on single and group classification technology. The obtained results show that the proposed AVIS work can achieve up to 92% and 89% of accuracy rate for individual and group classification in sequence.