Framework for Image Sensor Design Parameter Optimization for Pupil Detection

Machine vision systems (MVS) use image sensors to process and analyze image data. Depending on the application, the image sensor parameters are configured differently. However, some parameters are fixed for a specific product generation or product line. One of these parameters is the pixel pitch, the distance from one physical pixel to another. In this work, we introduce a framework, which allows to optimize design parameters of image sensors for pupil detection. We compare 2 different image sensor models with different pixel designs and generate images with different bit depths and resolutions. An evaluation of the design parameters is done with the generated images and a pupil detection algorithm. Furthermore, an existing pupil detection dataset is extended.