粮仓害虫的动态模糊聚类分析

Classification for Stored-Grain Pests based on Dynamic Fuzzy Clustering Analysis

综合利用计算机视觉技术和模糊聚类分析技术,实现了对粮仓害虫的动态模糊聚类分析。通过对粮仓害虫图像的CCD图像预处理,提取了9个几何特征参数和7个不变距,并通过优化选取其中6个几何特征参数和不变距1。最后利用所有样本的特征参数构造模糊等价矩阵,通过选择最佳阈值构造λ-截矩阵,从而实现对所有待识别样本的聚类,并从动态聚类图中可以分辨出各类样本的特征亲疏关系。

With full use of computer vision and fuzzy clustering analysis, dynamic clustering analysis for stored - grain pests was achieved. Through pretreatment to the CCD images of stored - grain pests, nine shape features and seven changeless moments were extracted. By the feature extraction, six features and changeless moment φ1 were obtained. By using fuzzy clustering analysis with front - parameters, an experiment for clustering twenty samples of four kinds of stored - grain pest was successfully performed. Based on dynamic clustering chart, some relations among the features of these pests can be detected.