Image Quality Assessment on Multiple Cameras for Float-Zone Crystal Growth Production

Nowadays we find ourselves surrounded by electronic devices and it is highly likely they use integrated circuits and chips produced from monocrystalline silicon wafers. The process of producing silicon wafers starts with polycrystalline silicon that is first melted and then recrystallized in a monocrystalline structure. Usually, in an industrial setting, this process can be achieved with either the Czochralski or the Float Zone method. The growth process of the ingots is a time-consuming batch process and, in order to reduce the cost per unit of the final products, it is necessary to increase crystal yield. This can be achieved by properly regulating the parameters that affect the quality of the monocrystalline silicon obtained, avoiding anomalies that would lead to defective and low-quality ingots. Obtaining high-quality ingots also reduces material waste and energy consumption. Parameters regulation is based on the information acquired through different sensors and data from previous runs stored in an archive. The camera vision system is particularly important as it monitors the growth chamber throughout the whole process. The frames captured are constantly processed to extract important geometric parameters. This project is aimed at implementing an algorithm that ensures consistent high-quality images across the production plant, allowing a coherent feature extraction required for a proper process control. The high variation in the content of images captured across multiple cameras makes being able to properly tune camera focus a challenge. The focal length of the camera is manually adjusted with the aid of an algorithm that scores the blur content of the image. Properly focusing the camera ensures that the images acquired during the growth process are of high enough quality to allow feature extraction and coherent control of different process parameters.