Performance Comparison of PCB Inspection using Machine Vision

Abstract

The human beings have very good power of vision as compared to many other species and very powerful brain to support when it comes to complete vision system of machine, there is very poor performance, in almost every aspect that man do in their daily life, the only reason is that human do not have enough computational power to process such amount of data in such a small time. Advantage that machines can provide in such situations where it takes data and quite some time to process it. One place like this is fault detection in PCBs, where a shot of PCB and check on the basis of that image whether the PCB is defective or not. This processing can be done at many levels and different image processing techniques can be used. The software implementation of vision processing algorithm is done in the language C# .NET and uses the latest development tools like Visual Studio 2003 and the implementation of the both the algorithms for image substitution and runlength encoding in this. The results are taken by test cases made in paintbrush they are all JPEG files of different resolution. The hardware implementation is done using VHDL and tools used are Model Sim 5.3 b and Leonardo spectrum for this purpose. The algorithm used is runlength encoding, which converts the image into gray scale then to black and white. The results of runlength algorithm are definitely better then image substitution algorithm which has been implemented in VHDL.