Camera Calibration Based On Improved Teaching-learning-based Optimization Algorithm

With the rapid development of science and technology,computer stereo vision is widely used in electronic,medical,aerospace and other fields.The camera binocular vision system imitates human eyes,making the computer have visual cognitive function in order to get information from images.The camera calibration technique is the premise of all computer vision research,and the accurate calibration results prepare for the three-dimensional reconstruction of space object.So it has great practical significance to research the stereo vision calibration.In recent years,many researchers have used good optimization of swarm intelligence optimization algorithm in the camera calibration process to improve the accuracy of the calibrationresults.However,the optimization algorithm has some problems,such as population iterative process falls into local optimum,so calibration method still have room for improvement.In order to improve the accuracy of camera calibration,high convergence accuracy and strong robustness of teaching and learning optimization algorithm are used in the process of calibration optimization.But this algorithm also has the problem that swarm intelligence optimization algorithm falls into local extremum.So this article first puts forward an improved algorithm based on hybrid learning strategies,which is used in the calibration optimization model so that ensuring the stability of calibration at the same time improve the calibration accuracy.The concrete content is as follows:According to the basic teaching and learning algorithm easily falling into local optimization algorithm,put forward a kind of teaching and learning optimization algorithm based on hybrid learning strategy(Difference-disturbance Strategy Teaching-learning-based Optimization,DSTLBO).This algorithm mixes differential evolution algorithm and joins perturbation strategy,which improves the convergence of algorithm,at the same time,reduce the possibility of students in the late algorithm trapped in local optimum so that guaranteeing the population diversity and global optimality.The accuracy of existing calibration methods have a very large development space.And it has a large degree of dependence on the initial value of the camera.In this paper,the improved DSTLBO algorithm is applied to stereo vision camera calibration and an improved symmetric operator corner detection method(SS).This paper proposes binocular camera calibration based on improved teaching and learning optimization.The result of this experiment shows that this algorithm can reduce the calibration error effectively and its calibration accuracy is high without limiting the number of images and shooting angle.And it lays a good foundation for three-dimensional reconstruction of the object.