Research On Camera Calibration Algorithm Based On Monocular Vision Size Detection

Along with the rapidly development of computer vision theory and technology, there is a leap development in machine vision technology.In the modern world,Machine vision technology play an increasingly important role in social production and life, such as visual surveillance, automatic identification and measurement parts,3D reconstruction, terrain matching, medical imaging processing and so on.This paper firstly described various corner detection algorithm, on the basis of Harris corner algorithm,We proposed an improved sub-pixel corner extraction algorithm. This algorithm used a coarse-to-fine detection strategy. Firstly used Harris algorithm for coarse positioning. Then Forstner operator was used to precisely point positioning in sub-pixel. Next this paper analyzed several commonly used calibration methods.then based on Tsai’s two-step calibration method this paper put forward a three-step calibration method. Following in this paper we described some basic measurement theory in binocular vision detection and constructed two visual measurement model by monocular vision. The first method was based on the world coordinate system established by the template image,which was using image pixel coordinates to calculate the camera coordinate, then normalizing the camera coordinate and using outer parameter matrix to solve the world coordinate. The other method was using a single CCD camera to move to achieve the purpose of binocular vision detection.Focusing on the problem of image points matching,we proposed a match method which combined feature matching and area matching together based on plume constraints. At last, we used opencv computer vision library to implement corner detection and camera calibration.Through the study of the above size detection in monocular vision camera calibration methods, We achieved a three-dimensional reconstruction of the point in the object based on a single CCD, changed the traditional detection methods contact, slow detecting disadvantages.providing a new ideas for machine parts vision measuring,So it had advantages of efficient,stable, online, real-time and so on.