基于机器视觉的图像形状特征提取方法研究进展

Research Progress in Shape Feature Extraction Methods Based on Machine Vision

基于机器视觉的图像形状特征提取应用的常见方法有阈值处理法、基于轮廓的形状特征提取法和基于区域的形状特征提取法3种。阈值处理法是常见的图像分割提取方法,具备操作简单、速度快等优势,但对于需精确提取图像形状和目标图像形状较为复杂的工况不适用;基于轮廓的形状特征提取方法,处理速度较快,但当处理复杂目标图像形状时,容易出现较大的偏差或错误;基于区域的形状特征提取方法,在提取形状特征时更加容易实现,且在处理复杂图像时更加准确,但需要的内部存储空间较大。由此可知,目前形状特征提取方法的应用局限性较大,而发展图像特征提取方法意义重大。

There are three common shape feature extraction methods based on machine vision including threshold processing method, coutour-based extraction method and region-based extraction method. Threshold processing method is one common method of shape segmentation and extraction and is easy to manipulate with high speed, but it is not suitable for complicated target shape features with high accuracy; Coutour-based extraction method could process with high speed but is more likely to bring about bias or error while dealing with complicated target shape features; Region-based extraction method requires enormous internal memory though it could be easily and accurately implemented to extract shape features. However, the limitation in extraction methods in application is still obvious and it is of significance in developing new extraction method.