Intelligent Classification Of Tunnel Surrounding Rock And Optimization Of Support Parameters Based On Computer Vision

In bore-exploding tunnel construction,the joint information in excavation face is very important for the stability assessment of surrounding rock and the designing of tunnel support.The existing rock mass joint information identification methods mostly depend on laser scanning or 3-D image reconstruction.The former technology has small scope of application and it is lack of color information;and the researches of the latter method mostly focus on the Slope Engineering,so it is not suitable for the special state of tunnel excavation surface.To solve the above problems,a set of algorithms are designed based on Computer Vision,including tunnel surrounding rock joint extraction and multi-level feature fusion algorithms,based on which,intelligent stability assessment method of surrounding rock,and certain-random-coupling joint numerical model construction method are proposed.The methods are validated in Huxitai tunnel project on Lin-Jian Expressway in Zhejiang Province.The main research contents and achievements include:(1)3-D image segmentation algorithm based on the improved SLIC and joint extraction algorithm based on angle difference.According to the characteristics of 3D image data,the search zone and distance function of SLIC algorithm are improved to generate super-pixels with compact structure and strong homogeneity;Based on the results,a joint extraction algorithm based on angle difference is constructed to identify the joints in the tunnel excavation face image.(2)The joint feature fusion method based on multi-level clustering and the intelligent assessment of tunnel surrounding rock stability.In order to adapt to the multi-scale of the joints in excavation face,clustering operation is performed twice on the joint extraction results,thus small joints,medium-sized dominant joint groups and large-scale cross-mileage joint groups are selected by the multi-level clustering strategy.On this basis,the indexes such as volumetric joint count and structure plane occurrence correction coefficient are calculated automatically,and when combined with the geological prospecting data,the surrounding rock stability level exposed in tunnel excavation face in each mileage are quickly determined,so as to provide basis for the change of tunnel construction method and tunnel support parameters.(3)Construction method of certain-random-coupling joint model based on the 3-D image identification results.Based on the joint fusion result,the 3DEC conventional joints are set in corresponding zone for large and medium-sized joint groups;For small discrete joints,the modeling is completed with the help of 3DEC Discrete Fracture Network(DFN)module,and finally a certain-random-coupling joint model is formed;Based on the tunnel surrounding rock automatic grading results and tunnel numerical simulation results,the adaptability evaluation and optimization of the support parameters in the zones are carried out,guiding the tunnel construction.There are 139 pictures,19 tables and 91 references in this thesis.