Abstract
PCB inspection has been developed for many years, but it still needs some improvement such as, inspection time, incorrect, and fault miss rate etc. This paper will propose a more robust and faster PCB inspection system. The inspection system will be divided into two stages, namely, defects detection and defects classification. In order to find better inspection combination of indices, there are pattern matching method, regional gray level index (IL2), high gray level difference index (IT1), histogram subtraction index (IL1), run-length index, and projection to be analyzed and we will propose a new method to choose index. Because defects classification is more difficult, neural network may be adopted in inspection system. Furthermore, we will propose new indices in order to improve inspection system. After inspection system is finished, there will be lots of online images to verify the feasibility of system.