Three-dimensional Environment Reconstruction Based On Binocular Vision

With the continuous improvement of science and technology and the rapid development of society,3D reconstruction technology has been a popular research interest.The 3D environment reconstruction based on binocular vision has been applied in various areas because the binocular stereo vision system has many advantages,such as simplicity and low cost.However,in actual scene,factors such as the environment have a great influence on the key steps of 3D reconstruction,including stereo matching.In this paper,in order to reconstruct the indoor scene,we analyze the steps of binocular stereo vision and focus on researching and improving the stereo matching algorithm.The main work of this thesis is as follows:(1)This thesis firstly analyzes the camera imaging model and the binocular stereo vision model in detail.Then the Zhang Zhengyou calibration method,Bouguet stereo rectifying method and the fundamentals of stereo matching algorithm are discussed.(2)The adaptive weight stereo matching algorithm is analyzed and then the guided filtering stereo matching algorithm is introduced.In order to further reduce the complexity of the algorithm and improve the matching precision,the fast guidance filtering and cross-scale cost aggregation schemes are introduced respectively.The fast guided filtering stereo matching algorithm based on cross-scale cost aggregation and the normal guided filtering stereo matching algorithm based on cross-scale cost aggregation are proposed to achieve the improvement of the matching rate and the matching precision respectively.(3)The 3D environment reconstruction system based on binocular vision is built with binocular stereo camera.The system is designed and implemented according to the algorithm discussed in this thesis,and then the indoor scene is reconstructed.Finally,we test the performance of the proposed stereo matching algorithm based on the Middlebury standard test platform.The results show that the fast guided filtering stereo matching algorithm based on cross-scale cost aggregation with post-processing has obvious advantages on algorithm efficiency and the time of processing one image in the test set is around 1 second.The average mismatch rate of the normal guided filtering stereo matching algorithm based on cross-scale cost aggregation with post-processing on the four images in the test sets is 5.47%,which is almost 1%lower than that of other existing local stereo matching algorithm.The 3D environment reconstruction system based on binocular vision built in this paper can reconstruct indoor scenes well and achieve the expected goal.