基于几何不变量的图像特征识别

图像的特征识别是图像处理和识别中的一个重要问题，几何不变量作为特征的特征值在很多领域已经得到了广泛的应用。实际中，普遍采用在仿射变换及射影变换下保持不变的仿射、射影不变量作为特征值。本文根据具体图像的特点，利用4类仿射和射影不变量构成特征的特征值空间，依据4步识别策略来识别图像中的特征点，从而完成识别任务。实验表明，这4类不变量能够较好地识别出实际图像中的特征。

RECOGNITION OF IMAGE FEATURE BASED ON GEOMETRY INVARIANTS

In image processing and recognition,the feature recognition of image is an important problem and as feature values of feature geometry invariants have been widely used in kinds of fields.In practice,we generally use affine and project invariants which keep invariant in affine and project transformation.Based on real images,this paper use four kinds of affine and project invariants to form the feature values space and recognize the feature points in images according as four steps of recognition tactic.So the mission is completed.The experiments show that it is good to recognize the features in real image used the four kinds of invariants.