Research On PCB-AOI System Based On Stream Computation

Printed circuit board(PCB)as an electronic basic material,widely exists in a variety of electronic devices.Quality of PCB has a significant impact on the overall quality of the electronic product.Due to quantities of PCB layers,intensive wiring and some uncertain factors,PCB is likely to have some quality problems during the process of mass producing.In order to improve the reliability of electronic products,we need carry on the defect inspection to the PCB.Automatic Optical Inspection(AOI)is the latest technology and the most effective measure of PCB defect detection.It is a non-contact PCB defect detection method based on image processing.Currently,most AOI devices on the market use PC for image processing to determine defect points,but the PC processing method is inefficient due to the following reasons.1)The feature of low PC memory bandwidth is more obvious in AOI image processing program requiring intensive memory access,which limits the processing speed of the system.2)The small number of PC cores determines the low degree of parallelism.3)The PC is waiting while the camera is scanning and the image data is being transmitted.PC is not fully utilized during this time.Since the current method that takes PC as processing core exists problems,this paper proposes a stream computing method based on FPGA to increase the speed of PCB-AOI.The main research contents of this approach include: 1)Stream computation model and stream computation structure are deeply researched and relevant content of stream computation is formally decribed,including stream,stream function block and the combination of stream function blocks.2)Key technologies to implement stream computation on FPGA are studied,including multi-clock management,cross-clock domain data transmission and memory hierarchy model.This paper proposes a three-layer memory hierarchy model,whose core is local distributed stream register file,to improve the overall system bandwidth.3)Image processing module of defect detection of printed circuit board is implemented in the method of stream computation based on FPGA,which is divided into multi-stage pipelines to improve the parallelism of the system.Pipeline structure combined with three-layer memory model enables the module to take full advantage of the image capture and transmission time for computing,achieving processing while capturing and transmitting.4)Automatic optical defect inspection system of printed circuit board based on stream computation is implemented.Finally,based on the DE2-115 platform,this paper implements the PCB-AOI system in the method of stream computation.The experiments of image processing module are carried out,which verify the correctness and speed advatages of stream computation mode.and verifies the performance improvement of AOI system in stream computing mode.