基于轮廓转角特征的物体分类识别算法研究

Research on Contour-Based Turn-Angle Object Recognition and Classification Algorithm

图像物体分类识别是当今计算机视觉和模式识别领域中的研究热点。轮廓是物体的最显著的特征之一,因此利用物体的轮廓特征对物体进行分类具有十分重要的研究意义。研究一种基于轮廓特征的图像分类算法,包括基于转角特征的图像分类算法,并将其应用于对水产品(虾)的分类对于基于转角特征的分类识别算法,选取物体轮廓上特征点位置处的转角信息作为描述物体轮廓的特征,将所有训练样本轮廓特征对齐、平均之后得到该类物体的匹配模板。在识别阶段,通过计算待测物体与匹配模板之间的相似度,获得该待测物体的类相似性。在实验阶段,利用提出的基于转角特征的分类算法对水产品虾类进行分类,得到较好的物体形状分类效果。

Object recognition has already become an active research field in domains of computer vision and pattern recognition.As one of important characteristics of object,contour based feature plays an important role in the field of object recognition.This thesis investigates two kinds of contour-based image classification algorithms,including turn-angle based algorithm and shape context feature based algorithm.With re⁃gard to turn-angle based algorithm,this thesis describes object with the turn-angle attached to the selected points along the object contour.After normalizing the orientation of objects,a similarity measurement is performed to evaluate the distance between objects.This method is used to classify shrimp samples and obtains good result for the class of whole shrimp.