Automated inspection of microlens arrays

Industrial inspection of micro-devices is often a very challenging task, especially when those devices are produced in large quantities using micro-fabrication techniques. In the case of microlenses, millions of lenses are produced on the same substrate, thus forming a dense array. In this article, we investigate a possible automation of the microlens array inspection process. First, two image processing methods are considered and compared: reference subtraction and blob analysis. The criteria chosen to compare them are the reliability of the defect detection, the processing time required per frame, as well as the sensitivity to image acquisition conditions, such as varying illumination and focus. Tests performed on a real-world database of microlens array images led to select the blob analysis method. Based on the selected method, an automated inspection software module was then successfully implemented. Its good performance allows to dramatically reduce the inspection time as well as the human intervention in the inspection process. 1. MICROLENS ARRAYS INSPECTION Inspection of microlens arrays produced through parallel microfabrication techniques borrowed from semiconduc- tor technology is a task for which a convenient solution has not yet been developed, as pointed out by researchers1 or industries2 active in this field. In this paper, we investigate the automation of the inspection process through image processing techniques. 1.1 Inspection task Microlens arrays are optical devices formed by a large number of small lenses, and are used in many applications including collimating, illuminating and imaging3. The work presented here concerns the inspection of devices with more than 2 millions lenses, with the specificity that gaps between lenses are coated with a reflective metal. In the inspection configuration considered here, the array is observed in reflection under a low magnification microscope, with an attached video camera, and the goal of the inspection is to spot defective lens or defects in the metal coating. Figure 1 shows a typical test image. The typical diameter of a lens is a few tens of µm, so that the microscope field of view covers more than 2000 lenses. Since a macroscopic device must be inspected, an xy platform is used to move the device under the microscope, and one image is acquired for each position. The total number of images acquired for each device is larger than 1800 (a certain amount of overlap between neighbor positions allow to ensure complete coverage). In the standard inspection procedure, a human operator examines each image, trying to identify and count the defects in the microlens array, in order to ascertain its quality. We investigated the automation of this inspection task. The motivations are first to relieve the human operator from the strain of watching the series of images, then to increase the reproducibility of the inspection procedure, and finally to reduce the inspection time. More specifically, we focused on developing an automated defect detection process. This process allows a semi-automated inspection, where the human operator needs only to examine images containing defects. 1.2 Overview of defective samples