Flexible Printed Circuit Board Defect Inspection System Research

In recent years, the Flexible Printed Circuit(FPC) boards have been widely used in consumer electronics. As this technology continues to develop and improve the manufacturing process, both the market for production and quality of FPC’s have increased higher demands. In actual production environment, the automatic optical inspection equipment is expected to identify defective products more quickly, more accurately and stability on the assembly line, improve productivity and production yield.To adapt to the development of modern computer trends, such as multi-core parallel computing, this paper has introduced a scheme of defect inspection system based on parallel comparison, the corresponding mechanical structure and working methods are also provided. It is expected to improve the real-time performance of the FPC defect inspection system, while simultaneously managing production costs.The defect inspection process is divided into three section; 1.image acquisition, 2.image pre-processing, 3.detection algorithms. Then using parallel computing methods to optimize the image pre-processing and detection algorithms processing speed under the two platform of a multi-core processor and a many-core graphics processor. Firstly, to solve the multi-channel visual input in parallel when resource contention and transmission bandwidth limitation by extending the USB to PCI-E board, to achieve image acquisition function independently and in parallel using multi thread. Following this, by analyzed the working methods and characteristics of image pre-processing and detection algorithm, two schemes called pipeline processing and data decomposition are designed to accelerate the program, with the comparison of the advantages and disadvantages of the two methods. Finally, set up the CPU+GPU heterogeneous computing platform. By analysis of the working mode and structure characteristics of GPU, programmed parallel running code to accelerate the image pre-processing and detection algorithm based on CUDA(Compute Unified Device Architecture). The experimental data show that the parallel processing optimization method had significant effect on improving the real-time performance in defect inspection system, with the increase of data quantity, the acceleration effect is better in CPU+GPU heterogeneous platform, can reach 1 to 2 orders of magnitude faster.A new idea of FPC’s defect inspection system implementation is provided by the combination of a scheme based on parallel comparison and CPU + GPU’s heterogeneous computing method, in order to improve the real-time performance. If further study will be continued, it is hoping to run in the actual production environment and greatly enhance the efficiency of existing defect inspection equipment.