Digital Image Correlation Method And Its Application On Blasting Engineering

Explosion is a process of instantaneous release of huge chemical energy,which is often accompanied by the formation of high temperature and pressure gas products.Explosion in the medium will also be accompanied by the high-speed movement of media fragments.These complex processes make the traditional measurement methods unable to accurately study the blasting process.In recent years,the rapid development of image processing technology,stereo vision technology and artificial intelligence technology promotes the development of "intelligent blasting" technology.Digital image correlation method is an important method in mechanical measurement,which has the advantages of non-contact,high precision,full field measurement and so on.It has an important application in the research of blasting process and promotes the research blasting process towards "digitization","visualization" and "refinement".It provides an effective tool for the mechanism research of blasting process,thus provides a theoretical basis for the precise control of blasting process,which is an important technical direction for the blasting industry to realize the development of digitalization,networking,visualization,refinement and intellectualization.However,due to the instantaneity and complexity of blasting process,the conventional digital image correlation method has problems in measurement accuracy and calculation efficiency,which is difficult to meet the personalized requirements of blasting engineering.Therefore,aiming at the conventional digital image correlation method,this paper optimizes the accuracy,efficiency and stability of displacement search algorithms,and applies it to blasting engineering.The main contents of this paper are as follows:Firstly,in the integral pixel matching stage,aiming at the low efficiency caused by the complexity of traditional cross-correlation based matching criterion,the integral pixel matching algorithm based on image mutual information is proposed.Based on the entropy and combination entropy of reference subset and matching subset,this algorithm has the advantages of linear structure,single peak near the target point and with no need of iteration,which significantly reduces the amount of calculation.The feasibility of this algorithm is verified with simulation speckle experiment.In the contrast experiment with the traditional cross matching method,mountain climbing method and genetic algorithm based on cross-correlation criterion,it is proved that the elapsed time of this algorithm in the integral pixel matching is only 27% of the genetic algorithm,14% and 16% of the cross matching method and mountain climbing method.Secondly,in the sub-pixel matching stage,in view of the lack of accuracy and efficiency of the traditional surface fitting method,a Gauss surface fitting sub-pixel displacement matching algorithm based on moving least square is proposed.After determining the integral pixel displacement and correlation coefficient matrix,a sub window with the integral pixel target point as the center in the correlation value matrix is selected for surface fitting.Unlike the traditional surface fitting method,this algorithm gives different weights to the nodes in the fitting window,so that the nodes closer to the central point have a greater influence to the function value of the central point and the nodes further away from the central point have smaller influence to the function value of the central point.The weights of nodes on the boundary are “0” and these nodes will not influence the function value of the central point.In the moving least square method,the quadratic or higher-order polynomial basis function in which a 5×5pixel or larger window is needed for surface fitting is used as the basis function.Therefore,the Gauss basis function based moving least square method is proposed.There are 5 undermined coefficients in Gauss basis function,therefore,the3×3pixel fitting window could be used.A smaller fitting window greatly reduces calculation cost and the elapsed time could be reduced by 27.92% compared with quadratic basis function.Compared with the N-R method,gradient method,normal Gauss surface fitting method and moving least square surface fitting method using quadratic basis function,the results show that the maximum standard deviation of the displacement calculation is 0.00585 and 0.00672 pixels respectively at the level of 0.01 pixel and 0.1pixel.Compared with the traditional methods,it is 12.03?47.9% and 20.19?81.33% lower.In the real speckle translation test,the maximum relative error of displacement calculation is 5.0%,and the average error is only 2.08%.In addition,in the traditional digital image correlation method,invalid pixels out of the boundary of speckle region are often set to “0” which results in poor calculation accuracy of nodes on or near the boundary.Therefore,a new boundary processing method in which the pixels on the boundary are duplicated to the invalid pixels is proposed.The simulation speckle contrast experiment shows that the correlation coefficient values of different types of boundary nodes can be improved with this method,and then the relative error of displacement calculation can be reduced from 11.55% and 14.78% to 2.01% and1.99% respectively.The duration of a blasting process is a few milliseconds to seconds in general.The high-speed or ultra-high-speed cameras are often applied to obtain the images of blasting process.The frame rate is thousands to tens of thousands fps and the image processing workload is extremely huge.Therefore,it is of great importance to improve the correlation matching speed,therefore,the parallel computation technology based on GPU(graphic processing unit)is used in digital image correlation method.The processing mode of each pixel in the speckle region is similar and independent,therefore,the correlation processes of many nodes can be parallelly executed in GPU,which will greatly improve calculation efficiency.Using CUDA(compute unified device architecture)programming method,the displacement searching process in digital image correlation is accelerated in parallel.The simulation speckle experiment shows that with the increase of speckle region the efficiency can be improved by at least 20 times.Afterwards,the afore-mentioned digital image correlation algorithm is applied to three typical problems in blasting engineering and the relevant mechanism of blasting process is revealed:(1)in the Brazilian disk test of marble,the evolution process of displacement field on the surface of marble under static loading is observed and analyzed.The failure process of the specimen is the process of horizontal and vertical displacement continuously accumulating near the final failure crack.(2)the nonel tube is arranged on the blackboard in “zig-zag” type and an automatic detonation wave front searching algorithm based on piecewise vector is proposed,then the full detonation process is observed automatically.The detonation wave in nonel tube can be divided into the growth region,stable region,attenuation region and high temperature expansion region,with the length of each region is 3.42 cm,15.58 cm,12.92 cm and 1.05 m respectively.The growth length of nonel tube is 32?41cm.In addition to the explosive type inside the tube and temperature,bending or tensile deformation also affects the detonation rate of nonel tube.(3)by combining digital image correlation and digital image processing technology,the concrete surface crack propagation process under implosion loading are observed and analyze automatically.The propagation process of the surface crack can be divided into two stages,the length growth stage and the width expansion stage.In the first stage,the crack grows in length under the combination force of stress wave and explosion gas and the acting time is about 1.75 ms with a maximum velocity of 225.9m/s and an average velocity of 122.27m/s.In the second stage,the crack expands in width under the driven force of explosion gas and the acting time is up to 4.5ms.The dynamic fracture process zone under implosion loading is similar with that under static loading.The size of the fracture process zone is invariable if the concrete proportion and aggregate size are fixed,the length of fracture process zone is about 8?9 times of the maximum aggregate size.Finally,in view of the limitation of the conventional 2D digital image correlation method in the measurement of large out-of-plane displacement deformation,considering the spatial motion characteristics of blasting projectiles,combining the 2D digital image correlation algorithm and stereo vision technology,the 3D digital image correlation technology is studied.A 3D tracking and measuring system for blasting projectiles is built by means of a stereo measuring platform with dual high-speed cameras,and a concrete model blasting funnel test is carried out.By means of calibration technique,the internal and external parameters of the binocular are obtained,and the calculation model of spatial coordinates is established.The three-dimensional tracking and measurement of blasting projectiles is realized for the first time by means of stereo vision.According to the test results,the blasting projectile effect can be divided into two stages,the bulging motion stage and the throwing motion stage.In the bulging motion stages,the blasting projectile accelerates as an integral for about 9ms.At the end of the bulging motion stage,the projectile fractures into fragments due to the penetration of internal cracks.In the throwing motion stage,the movement process of fragments is more complicated as the trajectory is affected by several factors such as its location in the projectiles and its shape.This measurement system provides an effective new method for the follow-up research on blasting projectile effect.