基于自适应阈值的齿轮干涉图像前景区域提取方法

齿面物体像灰度法是激光移相干涉测量中提取齿轮干涉图像前景区域的重要方法之一，针对该方法因人工设定阈值且忽略不同图像边缘特征从而导致的测量效率及精度受限问题，提出了一种基于自适应阈值的齿轮干涉图像前景区域提取方法。首先分析齿轮齿面形貌特征与各边缘顶点差异，对图像进行区域划分；然后根据边缘灰度变化规律通过邻域窗口筛选合格像素点并获取掩模结果，实现前景区域提取；最后根据5类图像评价指标分别对4组算法分割结果与传统方法分割结果进行数据对比。结果表明：算法在实现图像自动处理的基础上与参考结果匹配精确度提升约3.5%～4.5%,PRI(probabilistic rand index)提升约3%～4%,VOI(variation of information)提高约15%～25%,GCE(global consistency error)降低约2.5%～3.5%，最终相位信息准确度提升9μm～15μm。结果符合精度要求，该方法可广泛应用于齿轮干涉图像前景提取中。

Extraction method of gear interference image foreground region based on adaptive threshold

Gray method of tooth surface objects is one of the important methods to extract the gear interference image foreground region in laser phase-shifting interferometry. In view of the problem of limited measurement efficiency and limited accuracy caused by the manual threshold setting and neglect of different image edge features, a foreground region extraction method of gear interference image based on adaptive threshold was proposed. Firstly, the morphological characteristics of gear tooth surface and difference of each edge vertex were analyzed, and the image was divided into regions. Then, according to the changing rule of edge gray scale, the mask results were obtained by selecting qualified pixels through neighborhood window, so as to realize the foreground region extraction. Finally, the segmentation results of four groups of algorithms and traditional methods were compared according to five kinds of image evaluation indexes. The results show that the matching accuracy between the algorithm and reference results is improved by about 3.5%～4.5% based on the automatic image processing, the probabilistic rand index (PRI) is improved by about 3%～4%, the variation of information (VOI) is improved by 15%～25%, the global consistency error (GCE) is reduced by 2.5%～3.5%, and the final phase information accuracy is increased by 9 μm～15 μm. The results meet the requirements of accuracy, and the method can be widely used in foreground extraction of gear interference image.