Feature Extraction And Pose Calculating Method Of Reference Object In Camera Calibration

In computer vision research, feature extraction of2D image is the first step, andcamera calibration is an essential step in3D reconstruction. Therefore, research of thefeature extraction and pose calculating method of reference object in cameracalibration has important practical application value.This paper focuses on the feature extraction and pose calculating method in theorthogonal constraint of reference object in camera calibration base on2D plane.It presents two different2D plane templates--templateⅠandⅡ. According to theprojective geometry knowledge, it uses cross ratio which is the most fundamentalprojective invariant to describe them, and has carried out actual experiments aboutthem. In the experiment, it collects and processes a series of different orientation andlevels of noise about the template image, in order to verify the effectiveness and noiseresisting ability of the template. By comparison, it selects the appropriate template,applied to the actual calibration.On the basis of previous studies, it proposes a method of calculating poseparameters in the orthogonal constraint based on2D plane, and has conducted theactual verification. It collects a series of images, calibrates them by Zhang’s classicalcamera calibration algorithm to obtain the internal parameters of the camera as theinitial parameters of the pose algorithms. Then, it calculates the camera externalparameters corresponding to each image, and compares with Zhang’s data, to verifythe effectiveness of the proposed algorithm.A lot of work has done in the preparation and writing of the paper, and achievedsome success, but there are still some deficiencies and areas for improvement. Forexample, explore and research more precise, fast and efficient extraction algorithm, inorder to improve the calibration accuracy; improve the pose algorithm procedures, and make it has better robustness and real-time performance.