End-to-end simulation of a machine vision system for virtual task-based assessment

End-to-end imaging system simulation accelerates optical system time-to-market by reducing resources spent in designing, building, and testing real prototypes. In this work, we simulate a machine vision optical system including the lens system, detector, and Image Signal Processing with the objective of choosing the best lens and detector combination for the design. Machine vision is used in a broad range of applications, from circuit board assembly to produce sorting in the agricultural industry. The objective of the automated visual inspection system is to detect features that indicate variations from the desired product. The features could be characteristics such as size, texture, color, and shape. In this work, we simulate the optical design used in a system that uses machine vision to determine product quality. To simulate the system, we use the software ImSym– Imaging System Simulator. We use a utility in Python that connects to ImSym which lets us test multiple objects, lens systems, and detector combinations automatically. For each system combination, we run an Image Signal Processing routine to perform feature detection. Lastly, we show how to select the best lens and detector combination based on task performance.