复杂场景下的螺母快速识别算法研究∗

摘要

文章提出了一种综合应用哈夫变换实现螺母快速识别的算法。该算法在考虑螺母内外边界几何关系的基础上，应用哈夫变换实现目标边缘特征的提取，以内边界为圆、外边界邻边夹角120°为判据实现对螺母的快速识别。针对哈夫变换运算量大的问题，综合应用目标边缘信息，缩小搜索范围，有效减小了哈夫变换的调用次数和运算数据量，进而提高了识别效率。该方法具有哈夫变换这一传统经典处理方法的鲁棒性，对于存在不同对象间相互粘连的较为复杂场景下的图像，也能很好的实现螺母的快速识别。%A fast visual identification algorithm of nut is developed via combination of Hough transforms. The algorithm is based on the relationship of inner and outer boundary of nut;Hough transforms are applied to the feature extraction from the object boundary;and the fact that inner boundary is a circle and the adja-cent side of the outer boundary is 120 deg angle can be selected as the criteria for fast nut identification. The computational problem of Hough transform is carefully considered; the number of calls and the amount of data in computing can be reduced via comprehensive use of object boundary pixels and narrowing the search limit. The algorithm not only exhibits the robustness of Hough transforms as classic image processing meth-ods, but also shows good result for fast nut identification under complex scenes where different objects are partially inter-connected.