Research Of Underground Tunnel Reconstruction Based On Binocular Vision

Visualizing underground tunnel working environment is very important for unmanned mining.However,current technology can only visualize tunnels by collecting data after driving,which is no use for unmanned mining technology.In order to solve this problem,a visualizing system based on stereo vision is brought up to take pictures for working environment by stereo camera,obtain 3D scatters from pictures,build model for tunnels and implement underground tunnel visualization.Above all,the theory of stereo vision is introduced in this article,including stereo camera calibration,obtain of 3D scatters and the selecting and matching of key-points.The method of reconstruction based on binocular vision is introduced.A solid foundation is laid for researches and experiments later in the article.Second,a new visualization method based on sparse three-dimension scatters to do the visualizing reconstruction for the underground tunnels is proposed.The underground tunnel is described by its sections and center line.A model for trapezium-shaped and arch tunnel is built.The tunnel is broken into segments.A new method,“Best Center Line Principle”,is proposed to calculate parameters,like vertexes for sections,location and length of center line,for each segment.The tunnel can be reconstructed after every segment is properly calculated.Third,a simulation experiment is carried out to evaluate the algorithm.3D scatters are obtained using simulating algorithm in MATLAB.For big bend tunnels,a method of “Equal Width Step by Step” is proposed to avoid the change of width.For arch tunnels,a method of “Points Moving with Surface” is proposed to get even distributed 3D points.A parameter quantitatively evaluating accuracy of reconstruction is introduced to evaluate effects of algorithm in different tunnels.As the result turned out,methods proposed in this article can effectively reconstruct underground tunnels with certain noise immunity and stability.However,the result for ordinary tunnels is better than that for big bend tunnels.Last but not least,a system with both hardware and software is designed to verify the effectiveness of the algorithm.A simple trapezium-shaped tunnel model is built up.Photos are token from the tunnel by an FPGA with stereo camera.3D points are obtained from the photos in MATLAB.The tunnel is reconstructed by program implemented with C++ and finally repainted by MATLAB.The experiment result demonstrates that the algorithm in this article can effectively rebuild the tunnel model.