基于改进Hu矩算法的AGV字符识别研究

Research on AGV Character Recognition Based on Improved

摘要:

视觉导引AGV以识别标识符为基础实现工位转换。Hu矩算法可以提取字符的特征参数，结合机器学习分类算法，对特征值进行提取并分类，其运算速度快，但是识别准确率低。分析其原理公式，发现由于阶数高导致数值范围大，数据相近度差。基于此，对Hu矩公式进行改进，构造了新的算法。通过对UCI字符数据库进行检测，并在视觉导引AGV平台上进行在线测试，验证了改进后的Hu矩算法可以更快地提取特征值，并大大提高机器学习算法字符分类准确率，说明改进后的Hu矩算法在视觉导引AGV的字符识别上具有较好的实用性。

Abstract:

Visual guided AGV realizes position conversion based on identifiers recognition. The Hu moment algorithm is able to extract the feature parameters of characters and, combined with the machine learning classification algorithms, to classify them. The operation speed of Hu moment algorithm is fast, but its recognition accuracy is low. By analyzing its principle formulas, it was found that the large numerical range is due to the high order in formulas, which lead to a poor data closeness. Based on this, the Hu moment formulas were improved and new algorithm was constructed. Through detecting the UCI character database and taking an online testing on AGV, it is proved that the improved Hu moment algorithm can extract the feature faster and improve the accuracy of character classification of machine learning algorithm, indicating that the improved Hu moment algorithm has good practicability in character recognition of visual AGV.