Study Of Smartphone Self-Calibration Method Without Control Information

In the field of photogrammetry,professional metric camera's price is too expensive although its precision is high.Non-metric camera is becoming more and more widely applied in photogrammetry and computer vision field with its low cost,convenience and increasingly mature performance.As a kind of measurement equipment,smartphone's imaging quality is improving continually.It is an important method to obtain image data for photogrammetry.However,compared with traditional measurement cameras,auto-focus smartphone's focal and principal point are difficult to acquire accurately and its optical distortion is large.These factors will directly affect the relationship between center of photography,object point and image point.Therefore,it is very important to choose suitable methods for smartphone's camera calibration.Traditional camera calibration methods are limited by high-precision three-dimensional inspection tinker or calibration template with known space structure,and the focal is always fixed compared with auto-focus smartphone.Aiming at this problem,this paper introduces the imaging model of photogrammetry and computer vision,calibration content and three classical camera calibration methods.The principle of SIFT algorithm,RANSAC algorithm is studied,then this paper put forward a kind of camera calibration method without control information.This method includes two steps:first,collect images according to 3 × 3 mode,select a reference image to take relative orientation with other images in order to build unit model,and build free net model through the proportional coefficient after model connection,then coordinates of projective centers and object space points will be all taken into one coordinate system of reference image;second,with reference image selected and additional parameter model,orientation elements of images relative to reference image and additional parameters will be gotten after self-calibration bundle block adjustment.Finally,this paper presents a strategy suited for smartphone's self-calibration on the basis of the analysis of principle of smartphone imaging and its auto-focus feature.When the object distance is less than 4m,use the method above to compute some times,then use the results to complete function fitting,which will establish the relationship between calibration parameters and object distance.When the distance is more than 4m,the calibration parameters could be used repeatedly after first calibration.Firstly,use proposed calibration method without control information and traditional calibration method with control information to conduct contrast experiments in indoor three-dimensional tinker,then use control points to verify experimental results.Secondly,use checkerboards to design five experiments to research changes of calibration results at different distance(0.2m,0.5m,1.0m,2.0m,3.0m),establish the relationship between calibration parameters and object distance and compare our method with Zhang Zhengyou's calibration method.Finally,use four smartphones such as iPhone6s,iPhone5,MI4,Nubia Z7 Mini to obtain data form the outside(object distance is more than 4m)to verify the applicability and stability of this method and compare the distortion of these cameras.