Comparison of Autofocus Methods for Highly Reflective Surfaces : A Case Study on the Inspection of Optical Filters

Abstract: Visual inspection is a crucial final step when manufacturing products, as it determines if a product can be sold or rejected. The inspection is often performed manually by trained inspectors, which is time-consuming and often subjective. By automating the inspection, it can be performed faster and more accurately. Additionally, with faster and better results from inspection, faults in production can be noticed quickly and prevented, thus saving both money and resources. The automation often involves capturing images of the productand then analysing the images with a computer program. When capturing the images, the product must be in focus and under correct lighting. This master thesis set out to compare autofocus methods in the context of capturing images of optical filters for inspection. As the optical filters were highly reflective in the visual spectrum, the comparison was limited to autofocus methods proven to work on reflective surfaces. The literature proposes four autofocus methods for reflective surfaces. These four methods were compared regarding time, resolution, and working range. As these methods were designed for microscopes with high amplification and limited working range, they were overly complex and not applicable to the context of capturing images of optical filters. Therefore a more straightforward method with a lower resolution and larger working range that uses readily available components is proposed. The proposed method is a hybrid method using light triangulation and contrast detection, two standard autofocus methods for matte surfaces. The light triangulation quickly gives a rough focus, and then the final focus can iteratively be found with contrast detection. In the tests performed, the proposed method found focus and captured an image within 12 seconds. The proposed method achieved a resolution of just over 0.1 millimetres with a working range of 200 millimetres.