Research Of The Key Algorithm In 3D Reconstruction Based On OpenCV

The research of 3D reconstruction algorithms in binocular computer system has always been a popular topic.By exploring the mechanism of human binocular vision,we are trying to equip the machines with the capability of 3D sensing in our world.Since the rapid and accurate acquisition of 3D information plays a key role in 3D display technology,automatic driving technology,virtual reality and augmented reality,and et al.The research and implementation of these algorithms are becoming very important in the field of computer vision.Stereo matching is the most difficult problem in the implementation of computer binocular vision system,and it aroused a great deal of interest in research and engineering.But stereo matching has proved to be a pathological problem,the existing algorithm cannot guarantee of producing accurate matching for every pixel.In most cases,we only need a sufficiently accurate approximation to the truth,which is the main research content of this paper.In this paper,we leveraged OpenCV(Open Source Computer Vision Library)as the platform of our stereo matching research,which has a great reputation of high performance and portability.We evaluated the performance of four stereo matching algorithms by scientific method,and discussed their practicalities.The fixed window local matching algorithms are obviously the best in time consuming,which makes them capable for highly fast stereo extraction applications.The dynamic programming algorithm based on a single scan line completed the calculation in a few seconds,but it is quite disappointed as to matching accuracy.The semi-global matching algorithm based on multi-scan lines yields good matching accuracy in my experiments within a relatively short time,makes it suitable for many applications that requires both high speed and precision.Finally,we proposed a fast global matching algorithm based on belief propagation.By introducing local matching methods into the iterative process of belief propagation and combining with other optimization techniques,we achieved a high precision result in a few number of iterations,so that makes this kind of global matching algorithm useful in a 3D reconstruction system which needs a fast and accurate result.