An approach for synthetic images of sand deformation based on grain features and prescribed traces and its application

The image-based deformation analysis method (IDA) is currently an important deformation testing technology in the field of soil mechanics and geotechnical engineering. The speckle images are a common tool for the development, application and reliability evaluation of the IDA; whether they represent the characteristics of actual soil particles and are suited to assess the reliability of the IDA or not remains to be answered. By choosing the Fujian standard sand as a sample, a generation method for synthetic images describing the image and deformation characteristics of actual sands is addressed through the analysis of characteristics of sand images as well as their comparison to speckle images. Using the proposed method, four types of sequential deformation images with analytical formulas of the coordinates are generated to evaluate the reliability of two representative RG-DIC and PIVlab approaches. The results show that the speckle images have remarkable differences compared to the actual sand ones in texture, roundness and color values. The proposed generation method for synthetic images is capable of modifying the distribution parameters of particle size, roundness and color value, reflecting the characteristics of the actual soil images. By introducing time variable and arbitrary deformation functions, the generated sequential deformation images enable the reliability evaluation of IDA on dynamic and complex deformation. The speckle images apparently underestimate the analysis error of both the RG-DIC and the PIVlab methods, due to the black background and white circular spots with higher degree of recognition than the texture and particle features of the actual sand images. The deformation analysis in accuracy and stability from the RG-DIC exhibits more preferable than that from the PIVlab. The trends of analysis error under various deformation conditions using the two methods are in good agreement, and while the shear strain is smaller than 10^-3, the analysis error increases rapidly. The proposed method and results provide important supports and references for the development, application and reliability evaluation of soil image-based deformation analysis methods.