Accurate Registration And Path Planning Of PCB Automated Optical Inspection System

Automated optical inspection (AOI) system for printed circuit board (PCB) can achieve automatic detection of the PCB surface mount components quickly and accurately. So it is an important equipment to improve the level of automation of electronics manufacturing and guarantee product quality. In order to enhance the detection accuracy and speed of AOI system, algorithms for registration of circle PCB fiducial marks and complete coverage path planning of camera are researched in this paper. The concretely works are listed as follow:(1)The research of improved Hough transform based fiducial marks registration algorithm:We take circle PCB fiducial mark as study object. Coordinate systems are established on mechanical home position, fiducial point and left upper point of FOV image respectively. Image-getting errors caused by mechanical platform positioning errors and PCB manufacturing errors are eliminated indirectly while using three coordinate systems to save position information. We use Hough transform to detect the center point of fiducial mark and improve traditional Hough transform to overcome the disadvantages of large computing and memory exhausting. Coarse circle parameters are estimated first using physical dimension information of fiducial marks and circular symmetry. Then accurate searching is implemented nearby coarse parameters. This method greatly reduces the Hough transform accumulator space and optimizes the efficiency of the algorithm. In order to obtain good edge information of fiducial mark, a sequence of image preprocessing is implemented to fiducial mark image, including gray processing, median filtering, binarization and Sobel edge detection.(2)The research of genetic-based camera complete coverage path planning algorithm:We build the problem’s mathematical model and analyze it. Traditionally, this problem is divided to into two parts. One is clustering problem, and the other is shortest path problem. But there are some drawbacks in two-step method. This paper proposes a new genetic-based comprehensive method to solve the problem. The chromosome, fitness function and genetic manipulation operators are defined solve our problem by genetic algorithm. While combing crossover operator, mutation operator and sequence operator, convergence speed is improved.The experimental results show that the precision of detected fiducial mark center reaches0.5pixel. And time consumption is millisecond level with real-time performance. The genetic algorithm is an effective solution to the camera path planning problem. It reduces AOI’s total working time. These two key engineering problems are successfully resolved with some practical significance.