Deformation Detection And 3d Modeling Of Vertical Rescue Well Wall Based On Machine Vision

Coal has always been the main energy in our country.The safety production of coal is very important to the economic development of our country,and the occurrence of mine disaster is making mine rescue more and more important.Timely and effective mine rescue can not only reduce casualties,but also reduce huge economic losses.In the process of mine rescue,it is particularly important to avoid making wrong rescue plan and decision due to the lack of rescue well information,resulting in secondary injury.In the current mine rescue process,the monitoring method for the rescue channel is relatively simple,the adaptability to the harsh environment of the mine is poor,and the deformation detection accuracy of the rescue shaft is low.There are many problems in using the detection results as the judgment basis for the rescue channel,which seriously affects the efficiency of mine rescue.In view of the current working environment of the rescue channel of the rescue well,this paper takes the depth camera as the detection equipment and the camera rotation scanning as the measurement scheme.The maximum relative error of the calibrated depth camera in the short distance is 1.4%,which can meet the rescue requirements of the rescue well.Aiming at the weak light environment in the rescue well,this paper uses the improved surf algorithm to increase the number of feature points extraction and the matching effect under the same time consumption.FLANN algorithm is used to roughly match the features of the adjacent images of the rescue well wall,RANSAC algorithm is used to calculate the H matrix,eliminate the image mismatch,and refine the feature matching points,so as to obtain the feature matching points suitable for low light working environment.The method of gradual in and gradual out is used to fuse the borehole wall image to obtain the panorama of the borehole wall where the deformation occurs,so as to observe the borehole wall deformation more intuitively.Through processing the scanning data of the rescue well wall,using the least square method to obtain the cross-section information of the rescue well in different depths,combined with the panorama of the wall at the deformation of the rescue well,the deformation of the well wall can be observed more quantitatively.Based on the experimental study of the rescue well platform,the maximum relative error of the deformation deformation of the rescue well measured by the least square method is 2.77%,and the maximum internal tangent circle of the section is calculated by the dichotomy method to determine the rescue channel.The maximum relative error of the maximum internal tangent circle measured is 0.75%,which meets the rescue requirements of the rescue well.Through the 3D software to deal with the cloud coordinates of the rescue well points,the 3D reconstruction of the rescue well wall is realized,and the 3D model accurately reflects the real information of the rescue well.