基于菲涅耳变换的不变矩提取及目标识别方法

提出了一种基于菲涅耳变换的不变矩特征提取方法,并应用于图像目标识别.利用菲涅耳变换得到图像的衍射图样,将图像映射到菲涅耳衍射空间;在衍射空间提取几何矩或正交矩来获取图像全局信息的特征描述;利用k近邻法识别目标.实验结果表明基于菲涅耳变换的不变矩提取方法对于平移、尺度及旋转变化的图像目标具有更高的分类精度和抗噪声能力.

Fresnel transform based invariant moments extraction and object recognition method

An invariant moments feature extraction method based on Fresnel transform was proposed,and applied to image recognition.The original image was mapped into the Fresnel diffraction projection space by using the Fresnel transform.The geometric or orthogonal moment invariants were extracted in the diffraction space,from which the characterization of the global image information can be obtained.The k-nearest neighbor classifier was employed to implement objects classification.Experimental results show that the proposed moment invariants extraction method achieves high recognition rates for different combinations of translation,scaling and rotation and is very robust to noise.