为了实现复杂产品结构质量的准确快速检测，对于产品表面有多种结构体及空间尺寸的待检区域，采用变步长采样机制快速获取待测产品的周向方位图像序列，实现在有限方位下利用不完全数据对多个待识别区域的快速检测。首先采用垂直投影法确定各待检区域的旋转步长，为保证标准库中信息的完整性选一最小步长作为合格品的采样步长。其次采用尺度不变特征变换(SIFT)算法与折半查找法确定随机摆放的待检产品在标准库中的最优位置信息。最后通过相关度计算判别各区域有无缺陷。实验表明在保证检测准确率的前提下，采用变步长机制比固定步长检测平均可节省4.14 s。

In order to achieve the detection of the structural quality for complicated products accurately and rapidly, we adopt variable step-size sampling mechanism to quickly obtain the circumferential range image sequence for the product that has a variety of structural and spatial dimensional areas to be tested on the surface. With such an image sequence, rapid detection of multiple areas to be identified can be realized in the limited orientation by using incomplete data. Firstly, projection method is used to determine the rotational step and a minimum step size is selected as the qualified product sampling step in order to ensure the integrity of the information in the standard library. Secondly, the scale invariant feature transform (SIFT) algorithm and binary search are selected to find the optimal location information of product in the standard image library. Finally, the correlation degree is calculated to discriminate defects in some areas. The experimental result shows that the variable step-size method can save an average detection time of 4.14 s in comparison with the traditional fixed-step detection in the premise of ensuring the accuracy of detection.