基于扩展Shearlet变换、Krawtchouk矩和SVM的储粮害虫分类

为了进一步提高储粮害虫的识别精度,以便更有效地防治储粮害虫,提出了一种基于纹理和形状综合特征及全局混沌蜂群优化支持向量机(SVM)的储粮害虫分类方法。首先对储粮害虫图像进行扩展Shearlet变换,利用变换系数得到能量分布均值,加权后的能量分布均值构成纹理特征向量,用Krwtchouk矩不变量描述储粮害虫的形状特征;然后将纹理特征向量和形状特征向量分别归一化,两者结合构成储粮害虫的综合特征向量;最后用全局混沌蜂群算法优化SVM的核参数与惩罚因子,并应用参数优化的SVM进行分类。结果表明:与基于Gabor小波和支持向量机方法、基于Krawtchouk不变矩和支持向量机方法相比,本方法提取的储粮害虫特征信息更加完整,识别率更高。

A classification method of stored-grain pests based on the extended shearlet transform, Krawtchouk moment and SVM

To further improve the recognition accuracy of stored-grain pests and control stored-grain pests more effectively, a classification method of stored-grain pests is proposed based on synthetic features combining texture feature with shape feature and support vector machine (SVM) optimized by global chaotic bee colony algorithm. Firstly, an image with stored-grain pests is decomposed by the extended shearlet transform. The means of energy distribution are calculated by transform coefficients. The texture feature vector is made up of the weighted means of energy distribution. The shape feature vector of stored-grain pests is represented by Krawtchouk moment invariants. Then the normalized texture feature vector and the normalized shape feature vector are combined to form the synthetic feature vector of stored-grain pests. Finally, the kernel parameter and the penalty factor of support vector machine are optimized by global chaotic bee colony algorithm and the optimized support vector machine is used for classification of stored-grain pests. The experimental results show that, compared with the method based on Gabor wavelet and SVM, and the method based on Krawtchouk moment invariants and SVM, the proposed method extracts more complete characteristic information of stored-grain pests, and it has more higher recognition rate.