Digital Grain-size Analysis Based On Autocorrelation Algorithm

Grain size is one of the most important parameters that govern the movement of sediment particles, such as erosion, transport, and deposition. It has been a research focus in geology and coastal engineering for a long time. Recently, more and more researchers transfer to digital grain size analysis with the advances in optical technology and image processing technology. This thesis achieved a fast, in situ and convenient method to quantify grain size and its distribution based on autocorrelation algorithm.The estimation of well-sorted grains size is the first step to realize the digital grain size analysis on mixed-size sediments come from situ. According to the self-organization theory, similar texture patterned on well-sorted sediments lead to the occurrence of peak value in the autocorrelation curve. The corresponding offset could be used to infer grain size. Besides, this thesis extended the autocorrelation method from only one direction to eight directions and obtain the long and intermediate axes. The results were validated and proved compared well with them measured using a vernier caliper and sieving method.Regarding the mixed-size grains, this thesis introduced the image enhancement to eliminate the unusual patterns on the large-size grains and increase the contrast between the sediment particles and the pore, this step received an improvement of mean and median size estimation. A nonlinear programming method, which is different from the conventional’least-squares with non-negativity’method, the kernel density method and the maximum entropy method, is used to obtain the representative grain sizes and associated sediment inherent parameters, such as sorting, skewness, and kurtosis. Using the improved Gaussian function fitting method, the cumulative grain-size distribution curve and the probability density curve of the mixed-size sediments were obtained. The cumulative grain-size distribution curve obtained by image method was steeper than that resulted from sieving because of the shading of fine grains and the oblique laying of the surficial grains.A conversion step was taken to eliminate the difference between the two dimensional digital image method and the three-dimensional sieving method based on two main assumptions. It assumes that the sediment particles are balls with the same density. This convention overcomes the problem that traditional graphic method needs the 3D mass method to obtain the cumulative grain-size distribution curve and associated parameters.In addition, the present study needn’t to control the grain images under the same resolution after introducing a conversion step to get the autocorrelation curve under the desired image resolution by a conversion step. Subsequently, a transferred and available digital grain size analysis was realized because we could set up the calibration catalogue in the laboratory.Fourier transform and wavelet transform were proved to be available to infer the grain size as a time-frequency analysis. This study reviewed and discussed the grain size analysis based on the wavelet method and showed some progress about the selection of the mother wave and the scale.It demonstrated that digital grain size analysis could also been achieved based on wavelet method.