Research On Binocular Vision Measurement Based On ORB Algorithm

Binocular vision measurement is bionics that mimicked human vision. The human uses their eyes to observe the world outside, obtains the information such as the position, the color, the shape and so on.The is the very process what binocular vision imitates. Binocular vision get the images through two cameras. By extracting the effective information from two images, the matching of the corresponding points is accomplished. The position of the target is acquired according to the spatial geometry. This thesis is based on this principle, and the position of target is measured.Firstly, this paper reviewed relevant literature and books, then studied the current status and developing trends of binocular vision measurement.The paper explored the principle of binocular vision in detail with the help of existing literature and books. Before the start of the measurement through binocular vision, we should establish the camera model according to the geometrical relationship of the binocular camera. The next step is to calib the inner and outer parameters of cameras. The parameters are calibed using the MATLAB calibration toolbox based on Zhang’s method.Next, the polar geometric constraints and stereo correction are studied.After the completion of the modeling, calibration and other basic work, the paper explored the feature extraction at length. During the stage of feature extraction, the ORB algorithm is used in this subject. Because of its good real-time property, the ORB algorithm is widely used.But it’s at a disadvantage in the case of scale transforming.So the structure of the Gauss scale space is added in the original ORB algorithm, which makes it performance well in feature detection under the circumstance of scale transformation.After the feature extraction is finished, the stereo matching of binocular vision measurement is studied. The constraint of matching method and stereo matching are introduced. Through the research of the original O RB algorithm, the matching search strategy is improved. The matching results are optimized by PROSAC algorithm, and the matching accuracy is improved.Finally, the binocular vision measurement platform based on Open CV and VC++ is completed, which integrates the algorithms introduced in this paper, and the binocular vision measurement experiment is finished. The experimental results show that the system can accurately accomplish the target pose measurement.