Paper Title
Surf Based Fault Image Detection For Printed Circuit Board Inspection

Abstract
It is possible to collect, manage and analyze large amounts of data in real time because of development of fusion technology and spread of Internet of Things(IoT) in modern society. With the advanced technology, the high-tech product manufacturers began to offer differentiated services by producing fewer customized products. According to this, PCB (Printed Circuit Board, PCB), key component of the digital products, is also produced in small quantity batch production. Current PCB inspection systems require information about the normal image because it detects faults by comparison with the non-defective image. Therefore, this means that the test is not possible without the normal image and needs to have the all reference image. It is a major cause of reducing the efficiency of the PCB inspection system. This paper proposes a method for detecting the fault in PCB without normal image by learning the pattern of abnormal image. As a result of this methodology, it is expected to check more effectively the defects in the system to produce a variety of products and bring the time and cost savings in PCB inspection. Keywords� PCB Inspection, Speeded Up Robust Feature, Random Forest, Kernel Density Estimation.