Study On Technology Of Auto-focus Based On Image Processing

Auto-focus technology based on digital image processing, using the object imagingdefinition, is passive focusing technology. The image focusing systems don’t needdistance information to complete closed-loop focusing control, and are widely applied tomany fields, such as industry, medicine, laboratory research etc. The two basic problemsof image focusing system are evaluating image definition and searching in-focusposition. Based on the above two kinds of problem, this dissertation researches theauto-focus technology in the Cassegrain telescope, and the main contents and results areas follows.This dissertation studies the current image definition evaluation algorithms throughsimulation and experiments, and mainly studies the space-domain algorithms, andobtains the optimal algorithm for evaluating object which is Brenner algorithm. But theevaluation result accuracy of the traditional Brenner algorithm depends on the thresholdvalue. For this issue, an improved algorithm is proposed. The improved algorithm useshigh pass filter and bandpass filter to evaluate the image, and overcomes the limitationof traditional Brenner algorithm depending on the threshold value. The experiments andanalysis show that the improved algorithm can meet unimodality, accuracy, and improvesensitivity, and reduce calculating cost.In the optical system of point object imaging, the imaging of point object in thefocal plane actually is a circular light spot which has a certain area, and the evaluationalgorithm generally uses the energy distribution of point object imaging. The evaluationalgorithm has less calculation and clear physical meaning, but it doesn’t consider theimaging shape. Using the circularity to describe the imaging shape, an improved pointtarget evaluation algorithm is proposed. In the three-dimensional focusing system, theexperimental results show that the improved algorithm has better convergence speedand stability of convergence.Based on analyzing and comparing the current one-dimensional searchingalgorithms, this dissertation selects the hill-climbing algorithm to realize closed-loopfocusing control. But the hill-climbing algorithm is easy to fall into local extreme value,and the in-focus position searching is failure. This dissertation presents an improved hill-climbing algorithm, which uses the known information about the system. Theexperimental results show that the improved hill-climbing algorithm, which doesn’tchange the structure of system and doesn’t increase the sensors, has better searchingspeed and searching accuracy.The calculation of evaluating image definition is large, so the contradictionbetween the algorithm complexity and the system real-time is the prominent issue in theimage definition evaluation system. Based on comparing the current image processingplatforms, this dissertation selects the FPGA platform which can satisfy the demands.This dissertation uses the visual programming technology to realize evaluationalgorithm in FPGA.The divergence of primary and secondary optical axes reduces the imaging qualityin the Cassegrain telescope. The essence of adjusting primary and secondary mirrors toguarantee the imaging quality is adjusting the focus of primary and secondary mirrors tocoaxiality and superposition. This can be regard as generalized auto-focus technology,and requires a multidimensional auto-focus technology. The image focusing technologydoesn’t require specialized equipments and can simplify the structure of system.Furthermore, the evaluation result is intuitive.This dissertation designs the three-dimensional image focusing system platformwhich uses the Six Degrees of Freedom (6-DOF) platform. Based on analyzing andcomparing the current multidimensional searching algorithms, this dissertation selectsthe Stochastic Parallel Gradient Descent (SPGD) algorithm to realize three-dimensionalclosed-loop focusing control. According to the coupling relationship of the X-axis,Y-axis, and Z-axis, a modified SPGD control method is presented. The experimentalresults show that the closed-loop error is below3percent, which meets the precisionrequirements of the three-dimensional auto-focusing system. According to the results ofthe experiments, a detailed analysis of the precision of algorithms is conducted. Theexperiments of the three-dimensional image focusing system provide a solution for thepractical application, and have significant instruction meanings for engineering.