Studies On Digital Image Correlation Method With Non-fixed Camera For Deformation Measurement

It is important to test deformations of structures with more and more artificial facilities are built.The fatigue conditions can be analyzed by displacement measurements.Then the risk of security can be detected timely to ensure the safety.Stress and strain measurements are the focus of experimental mechanics.Traditionally strain gauges are used to measure deformations of structures.The measurement results of strain gauges are reliable and stable.However,only deformations of several points can be measured simultaneously.It is difficult to get the stress and strain of the specimen surface by the method.What's more,it is difficult to paste gauges on large structures.The movements of the testing objects are inevitably affected by the attached gauges.Digital image correlation is one of the non-interference optical methods.Displacements of surfaces can be calculated by analyzing speckle images captured before and after deformation.Compared to interference optical method,the setups of digital image correlation is simple.A white light is sufficient instead of any leaser sources which are commonly used in other optical methods.The natural textures can be used or use artificial speckles made by spraying on the surfaces.Requirements of the testing environments are also lower than other optical methods.The researches of digital image correlation method are mainly proceeded in the laboratory.The camera is fixed to the specimen surface during the deformation which limit the application of the method outside of the laboratory.It is difficult to use strain guage or digital image correlation method with fixed setups to test lots of structures during a long time.A new digital image correlation method without fixed setups is developed.It can be used outside of the laboratory more conveniently than traditional digital image correlation method.The effects of measurement environments are analysed such as camera optical axis deviation,lens distortion and illumination changes.Artificial speckles are made by spraying to the specimen surface.Feature template is manufactured and attached tothe specimen surface.Several images of the speckles and feature template are captured from different orientations with constant camera focal length.The coordinates of feature points are calculated after image processing.Then the parameters of the camera are calibrated from them.The images are inevitably affected by lens distortion of the digital camera and corrected by distortion parameters.The coordinates are calculated from the corrected image and transformed to camera coordinates.The planar equation of the feature points is calculated by fitting of the coordinates.The plane of the specimen surface is parallel to the feature template and the distance of them is the thickness of it.The specimen surface is divided to several grids and the coordinates of the grid nodes are calculated.Then,the speckle patterns can be got from interpolation formula.At last,digital image correlation method is used to calculate the displacements of the specimen surface.In this method,the camera need not to be fixed with the specimen during the measurement.The reference and target images can be got from the camera calibration method.However,the speckles of the specimen surface may be rotated according to the feature template since they are not fixed together during the deformation.Then the orientation of the reference and target images may be not constant.The effects of image rotation to digital image correlation method are investigated and new techniques are developed to improve the calculation accuracy.The orientations of the target subsets are not always same to which of the reference subsets.Rotation search is proceeded to improve the calculation accuracy as well as displacement search in the method.Searching procedure is adjusted for the method to improve the calculation speed since too much subsets should be searched in the method.The specimen surface may be damaged during the processing such as observation of rock structure,residual stress measurement and etc.Gray levels of some pixels in deformed images may be significantly different to which in reference images.Digital image correlation is sensitive to damaged areas of images,which leads to difficulty of tracking subsets in reference and target images.The effects of damaged areas in specimen surfaces to digital image correlation are investigated.New digital image correlation techniques are developed to calculate displacement of specimen surfaces with damaged areas.In order to improve calculation accuracy and use all pixels sufficiently in subsets,coefficients are set to each pixel according to the qualities of them.The calculation results are mainly impacted by the high quality pixels rather than others.With the calculation path which is from pixels with higher correlation to lower ones,the displacement calculation accuracy of the pixels neardamaged areas is improved by the proposed method.Subset size selection is a necessary step of using digital image correlation method and has a large effect on the calculation accuracy.The subset size is selected manually by users according to the image pattern and deformation of specimen surface.It is a trade-off between using a large size and a small one so that sufficient pixels are included in a subset to distinguish itself and the calculation errors are relatively small.A new method for efficient using pixels is discussed and the relations between pixels and the central pixel are investigated.A relatively large subset size is chosen in the method and all the pixels in a subset are not treated equally.Coefficients are set to each pixel rely on the importance of them to identify a subset from target images which are calculated by the normal distribution based on the distances between them and the central pixel.Selection of subset size for traditional digital image correlation is replaced by selection of parameters for the equation in this method.The importance of pixels is changed with the parameters of the equation.The effect of subset size to correlation calculation is somewhat alleviated and the calculation accuracy is improved by the method.