A multiscale calibration of a photon video microscope for visual servo control: Application to micromanipulation

Since many years the need of automated microassembly systems i.e. systems to assembly micrometric parts, MEMS (Micro-Electro-Mechanical Systems) , MOEMS (Micro-Opto-Electro-Mechanical Systems) etc., becomes more and more necessary because of the development of micro components based products. The corresponding tasks (supervision, control, quality inspection, ...) involve accurate metric measurements from images. The main image source is the photon videomicroscope whose features are the weakness of the depth-of-field and that of the field-of-view. The paper deals with the modelling and calibration of that kind of imaging system. A multiscale calibration paradigm is proposed and validated by means of a commercial video microscope. The model is derived from the usual perspective model. The latter is simplified by considering a single scale factor and then the link between the zoom and that scale factor is established. The calibration is performed through a virtual pattern made by tracking a silicon micro part in the images of the scene. The final error between the projection of a pixel and its model is about 1.45 pixel indicating the accuracy of the approach. At the end, the model is used to perform a visual servoing whose purpose is the aligning and centering of a micropart with a gripper.