Research On Camera Self Calibration Method Based On Image Sequence

Camera calibration is the first and essential step of 3D reconstruction technology in computer vision.It has been applied in many fields,such as navigation,virtual games,medical detection,military and so on.Among them,the camera self calibration technology is more flexible than the traditional camera calibration technology,the calculation process is simpler,and more in line with the actual needs.Therefore,the camera self calibration technology has a better prospect.In view of the problems existing in the current camera self calibration method,an improvement is made to obtain more accurate camera related parameters.In this paper,the camera self calibration methods are divided into two categories: the self calibration with fixed camera internal parameters and the self calibration with different camera internal parameters.The main work is as follows:(1)In this paper,a camera self calibration method based on improved genetic algorithm is proposed on the premise that the internal parameters of the camera are fixed.First of all,the genetic algorithm is improved in three ways: first,by combining the elite retention strategy and the random League selection algorithm as the method of initializing the population;second,by improving the roulette selection method;third,by using the adaptive hybrid probability and mutation probability method to improve the genetic algorithm.Then,the simplified Kruppa equation defined by Hartley is transformed into the objective function,and the improved genetic algorithm is used to search the optimal value of the objective function.The experimental results show that this method can solve the problems of premature convergence,stagnation and falling into local optimum in the process of camera self calibration parameters optimization,and improve the accuracy of self calibration.(2)An improved self calibration method is proposed for different camera parameters.Aiming at the camera self calibration method based on isosceles triangle proposed by Bouchra Boudine et al.In order to obtain more accurate and effective calibration results,this paper improves its nonlinear cost function based on the relationship between two points in isosceles triangle and their projections.Firstly,the projection matrix is determined by the projection of two points on the image plane and the basic matrix.Then,the relationship between the two points of the scene and their projection on the image plane is determined,and the formula of nonlinear cost function is established.The relationship between the two points of the scene and its projection on the image plane is a matrix.When building the nonlinear cost function,this method not only considers the relationship between the twopoints of the matrix,but also considers the internal properties of the matrix itself.The experimental results show that the method is accurate.