3D Measurements of Submillimeter-scale Micro-holes with Light-field Image Fusion under Multi-angle Illumination

Micro-holes, crucial components in various industries, pose challenges in accurate measurement due to their small dimensions and complex geometries. Traditional methods, such as coordinate measuring machines, profilometers, and 2D camera-based imaging systems, are effective but limited in scalability and efficiency. Light-field imaging offers promising solutions for addressing these challenges by providing spatial-angular information for depth reconstruction. However, uneven illumination and specular reflection on metal substrates hinder accurate depth estimation. To overcome these limitations, we propose a novel approach combining multi-angle illumination and exposure fusion. This method enhances image quality and consistency by capturing raw light-field images under varied illuminations and fusing them to mitigate intensity variations. Experimental verification demonstrates the effectiveness of our method in accurately characterizing micro-holes, with improvements concerning depth estimation accuracy observed up to 40% compared to normal cases. Multiple raw light-field images are recorded while the illumination is modulated for each image. Exposure fusion is performed for each sub-aperture image (SAI). The fusion process takes contrast and well-exposedness into account. The magnitude of error reduction can be affected by many factors other than applying multi-SAI fusion. The factors of illumination configurations, hole orientations, and surface properties will probably influence the performance. However, it can be estimated and concluded that by applying multi-SAI fusion, in most cases, the measurement error can be reduced by 15% to 40%. By addressing challenges related to uneven illumination and specular reflection commonly observed on metal surfaces, our method enhances depth reconstruction accuracy, enabling more precise characterization of micro-hole structures. The method takes advantage of the rapid recording capability of the light-field camera and fuses SAIs from multiple raw light-field images captured under different illumination. Experimental results validate the effectiveness of our approach, showcasing estimated improvements in depth estimation accuracy by up to 40% in most cases compared to normal illumination. The measuring efficiency can be above 10 000 effective points per second which qualifies in-situ inspections. Moving forward, further optimization and refinement of the proposed method could lead to broader applications in the measurement of other micro-structures under complicated surface conditions using light-field imaging.