喷砂表面粗糙度在线测量仿真数据

该数据集主要面向基于数字全息技术的表面粗糙度检测研究，探讨采用无接触测量方式获取喷砂作业后的表面形貌信息，提升检测效率以及检测精度。采用混合聚类算法，将K-均值和高斯混合模型相结合的混合聚类算法应用于全息图的频谱分析，有效抑制零级频谱的干扰，实现两张全息图零级频谱的精准消除；将分数阶傅里叶变换理论与离轴数字全息重建理论进行了结合，通过分数阶傅里叶变换的旋转可加性结合粒子群优化算法精确估计全息图的最佳分数阶参数，在一定程度上解决分数阶傅里叶变换域滤波重建的视场损失问题，并实现高质量的振幅和相位重建；通过加窗傅里叶变换对差相位图进行降噪处理，结合解包裹算法获得连续相位分布；基于交替加权最小二乘法拟合的相位畸变补偿方法。该方法利用变量分割策略，将复杂的相位畸变补偿问题转化为更易于求解的交替优化问题，以物体相位和畸变相位为变量交替求解。此外，引入Tukey双权函数进行动态权值分配，使得算法在拟合过程中能够有效识别并抑制异常数据与噪声影响，从而得到高精度最终相位分布。基于相位与高度的量化关系，将相位分布转化为表面轮廓高度信息，进而计算得到算术平均粗糙度参数。数据于2024年8月采集于上海大学东区机自大楼，通过初始的两张全息图，经过Matlab算法处理后得到仿真数据，该仿真数据可用于后期的算法评估等用途，数据量为557M。

This dataset is primarily targeted at research on surface roughness detection based on digital holography, aiming to acquire surface topography information of sandblasted workpieces via non-contact measurement methods to improve detection efficiency and accuracy. A hybrid clustering algorithm combining K-means and Gaussian Mixture Model (GMM) is applied to the spectral analysis of holograms, which effectively suppresses the interference of zero-order spectra and achieves precise elimination of zero-order spectra in two holograms. The fractional Fourier transform (FrFT) theory is combined with the off-axis digital holography reconstruction theory. The rotational additivity of FrFT and the Particle Swarm Optimization (PSO) algorithm are utilized to accurately estimate the optimal fractional-order parameters of holograms, which partially addresses the field-of-view loss problem in filtered reconstruction within the FrFT domain and enables high-quality amplitude and phase reconstruction. Noise reduction is performed on the differential phase map via windowed Fourier transform, and a phase unwrapping algorithm is combined to obtain a continuous phase distribution. A phase distortion compensation method based on alternating weighted least squares fitting is proposed. This method adopts a variable segmentation strategy to transform the complex phase distortion compensation problem into an alternating optimization problem that is easier to solve, where the object phase and distortion phase are taken as variables for alternating solution. In addition, the Tukey's biweight function is introduced for dynamic weight assignment, enabling the algorithm to effectively identify and suppress the influence of outliers and noise during the fitting process, thereby obtaining a high-precision final phase distribution. Based on the quantitative relationship between phase and height, the phase distribution is converted into surface contour height information, and then the arithmetic average roughness parameter (Ra) is calculated. The dataset was collected in August 2024 at the College of Mechanical and Electrical Engineering, Eastern Campus of Shanghai University. The simulation data was obtained by processing the initial two holograms via Matlab algorithms, which can be used for subsequent algorithm evaluation and other purposes, with a total data size of 557 MB.