Research On The Calibration Technique Of 3-D Microscopy Reconstruction System

Microscopes are now widely used in the filed of 3-D measurements for their large magnification and changeable physical focal length.There are two key steps before 3-D reconstruction with binocular imaging system: system calibration and feature matching,of which the system calibration is especially important.However,the small DOF(depth of focus)restricts camera from taking clear images with distinct tilts,and the calibration methods for normal lens requiring multiple views of the calibration board are not practicable.Concerning about these characteristics,some research carried out as follows.(1)A novel monocular imaging system calibration method,by which a digital camera coupled with a microscope is calibrated precisely,is proposed.To set the camera in focus,the calibration board should be near parallel to the imaging plane.Under the near parallel condition,a novel method,by which the magnification rate of the imaging system is calculated precisely and utilized to calculate the rotation angles,image distance and object distance,is proposed.For the invariance of the magnification rate,the calibration result is accurate and stable.To overcome the inadequacy that the physical focal length cannot be calibrated by one image,the measurement method based on the magnification of the microscope is adopted.The method is suit to the imaging system with the microscope of changeable physical focal length,and only one image is needed.The stability and the accuracy are evaluated by simulations and experiments.(2)A novel calibration method of the binocular fringe projection microscopic system,which is based on the BP neural network,is proposed.In the traditional methods,the binocular imaging system is decomposed into two monocular imaging system which are calibrated respectively.These methods are complicated and the calibration results cannot indicate the holistic characteristics.In our method,the BP neural network is combined with the binocular fringe projection system which is calibrated as a whole.The accuracy of this calibration method is proved by the simulation and experiences,and the size range of the object which need to be reconstruction by the neural network is detected.Besides,based on the 3-D reconstruction algorithm,an additional DMD is fixed to the binocular microscopic system to correct the reconstruction coordinate error caused by the irregular reflectivity.