改进的Fisher判别分析与折痕检测  

Improved Fisher Discriminant Analysis and Its Application to Crease Detection

作　　者：岳洪伟[1,2] 王克强[2] 廖伟[2] 郑永敏[2] 

机构地区：[1]五邑大学信息工程学院,江门529020 [2]仲恺农业工程学院自动化学院,广州510225

出　　处：《科学技术与工程》2015年第14期82-85,共4页Science Technology and Engineering

基　　金：广东省自然科学基金(S2013040014993);广东省大学生创新训练项目(1134713019);五邑大学青年科研基金(2014zk10)资助

摘　　要：针对毛杆折痕难以检测问题,首先将羽毛杆图像转化为一维信号,通过随机共振降低光照不均干扰,结合模极大值理论进行信号奇异点检测;利用奇异点位置完成子图像提取以减少对羽毛杆遍历检测带来的误判。然后采用协方差矩阵对目标子图像进行特征结构描述,引入仿射不变度量使得该空间满足黎曼流形的要求,并以此调整了Fisher判别分析的类间散度和类内散度计算。最后利用黎曼指数映射得到了样本的最佳映射空间,从而实现非线性空间的类别判别。实验结果验证了所提方法的有效性。Aiming at the detection difficult problem of feather quill crease, firstly, feather quill image is trans-formed into one dimensional signal; using the relationship between stochastic resonance and modulus maxima, the crease coordinate can be prejudged. Subimage can be extracted through the coordinate to reduce misjudgement caused by image traversal. Then covariance matrices are computed as the crease descriptors of feather quill, and an affine invariance metric which is adopted to make the space meet the requirement of Riemannian manifold is used to adjust class variance and within-class variance of. Finally, in order to implement the discrimination in nonlinear space, the best projection space of samples is gotten using the Riemannian index mapping. Experiment results on databases show the effectiveness of the nronosed method.

关 键 词：羽毛杆折痕 协方差矩阵 黎曼流形 FISHER判别分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]