A Hybrid Image Alignment System for Fast and Precise Pattern Localization

The need for accurate and efficient pattern localization prevails in many industrial applications, such as automated visual inspection and factory automation. The image reference approach is very popular in automatic visual inspection due to its general application to a variety of inspection tasks. However, it requires precise alignment of the inspection pattern in the image. To achieve precise pattern alignment, traditional template matching is extremely time-consuming when the search space is large. In this paper, we present a new hybrid image alignment system for fast and accurate pattern localization in a large search space. This pattern localization algorithm is very useful for applications in automated visual inspection and pick-and-place applications. The proposed hybrid alignment algorithm comprises a hierarchical nearest-neighbor search process and an optical-flow based energy minimization method. The hierarchical nearest-neighbor search, which is based on the learning-from-examples principle, produces rough estimates of the transformation parameters for the initial guesses of the optical-flow based energy minimization method. The energy minimization process iteratively refines the estimation results and provides associated confidence measures. Experimental results are shown to demonstrate the accuracy and efficiency of the proposed algorithm.