Reference-free path-walking method for ball grid array inspection in surface mounting machines

In production processes that use surface mounting machines for assembly of printed circuit boards (PCBs), visual inspection is usually employed to recognize component locations and diagnose their faults. However, commonly used inspection algorithms fail to achieve both reliability and computational efficiency, especially when applied to components with high-density and large-scale packaging, such as ball grid array (BGA). In this paper, a scheme for BGA inspection is presented to address the challenges. First, we introduce an adaptive thresholding and a median-based Otsu method to separately determine the optimal threshold for segmenting each ball. Then, a reference-free path-walking mechanism is proposed to recognize locations of the balls in the grid. BGA position and orientation are computed based on the recognition. Finally, feature extraction and defect detection are performed by using the path-walking method. The proposed technique was implemented in the host computer of Samsung SMT482 machine, and comparisons were made with the Samsung's algorithm as well as with HALCON Software. Results indicate that our approach exhibits flexibility and robustness in a wide range of BGAs with different ball grids. Compared with the Samsung's algorithm, the proposed method demonstrates higher computational efficiency and promising inspection accuracy, even under unsatisfying illumination conditions.