Alinhamento automático no empacotamento de chips fotônicos baseado em visão computacional.

Abstract

This work deals with an automation process in the positioning of photonic components, based on image recognition, in an industrial positioning machine. The Pick & Place station was the target system of the application, which was granted by the company Idea! Electronic Systems as part of the training process in the Silicon Photonics project. In the Pick and Place station there is a programmable logic controller (PLC) operating in a mechanical stepper displacement system, limited by inductive motion sensors. The system is integrated with a digital platform through a serial communication channel linked to a client-server architecture via OPC protocol. A 10.55 MP monochrome CMOS camera is used for base-image acquisition. The camera calibration step of the image acquisition system was performed using the Zhang method, and the automatic recognition of the objects in the workbench was done by the image segmentation process, using the techniques of digital image processing such as Sobel operator, normalized correlation coefficient, smoothing linear filters, among others. The final results indicated a correct correlation of the dimensionless world of the pixels with the mechanical axis of the equipment stepper motors, mapped in millimeter scale. A 0.8mm error was computed at the first offset of each sequence. The time for a full standard "Pick" and "Place" operation for the image guided solution is 1min: 46s.