摘要
在模式识别中,不变矩作为特征提取量被普遍使用。由于大多不变矩是在极坐标系下定义的,计算过程中需要多次进行坐标变换,这种变换既增加了计算复杂度,还产生了明显的量化误差。针对此问题,本文提出1种直接在笛卡尔坐标系中计算变形雅可比矩的算法,并将此方法分别应用于相似白细胞显微图像和8种家畜寄生虫卵显微图像的识别中,识别率分别为97.89%和92.18%。
Invariant moments were widely used in pattern recognition as a feature extractor,and the related computation for invariant moments usually adopted the coordinates transformation. Not only increase the computational burden greatly, but also create large quantized error. To solve this problem, an improved algorithm to compute Pseudo - Jacobi - Fourier moments directly in the Cartesian coordinate system is proposed in this paper. Then we apply this improved algorithm to classify similar leukocyte microscopic image and eight kinds of parasite egg images, and a satisfactory recognition rate of 97.89% and 92.18% achieved respectively.
出处
《内蒙古农业大学学报（自然科学版）》
CAS
2016年第5期120-126,共7页
Journal of Inner Mongolia Agricultural University(Natural Science Edition)
基金
国家自然科学基金项目(31060337)
国家自然科学基金项目(60967001)
内蒙古自治区自然科学基金项目(2011MS0904)
关键词
变形雅可比-傅里叶矩
笛卡尔坐标系
模式识别
Pseudo - Jacobi - Fourier moments
Cartesian coordinate system
pattern recognition
