摘要
基于纹理特征模型的檀香咖啡豹蠹蛾图像诊断方法,根据健康图像和虫害图像在纹理方面表现出的差异,提出海南省北部县市檀香受咖啡豹蠹蛾虫害"多纹理特征"的确定方法。针对每种图像类型,使用提取出的4维多纹理特征,组合得到6种数学模型,并对其进行评估。结果表明:模型1(自变量为熵值均值-相关性均值,因变量为熵值均值-能量均值)的模型精度与分类精度均为最佳,并且总体分类精度达到91.25%。与逐步聚类算法和K-means聚类算法、Logistic模型二分类法相比,该方法在保证分类精度的前提下减小了计算量,并为之后纹理图像分类提供了参考依据。
Abstract：According to differences in texture based on health images and pest images, a method of determining ＂multitexture features＂ of Zeuzera coffeae in Santalum album of northern counties in Hainan was put forward based on image diagnosis method of texture feature modeling. For each image type, 6 mathematical models were combined and evaluated by the extracted 4dimensional multitexture features. Results shows that model NO1 which X axis was the mean entropy correlation mean and Y axis was the mean entropy energy mean was the best in both fitting degree and classification accuracy, and the classification accuracy reached 9125%. Compared with stepwise clustering algorithm, Kmeans clustering algorithm and Logistic model twoclassifying method, this method could reduce the computational complexity under the premise of ensuring classification accuracy, and provide reference for texture image classification.
出处
《西南林业大学学报（自然科学）》
CAS
北大核心
2018年第1期117-125,共9页
Journal of Southwest Forestry University：Natural Sciences
基金
中央级科研院所基本科研业务费专项(CAFYBB2014MA006)资助
关键词
纹理特征
檀香
图像诊断
图像建模
灰度共生矩阵
texture feature
Santalum album
image diagnosis
image-based modeling
grayscale co-occur-rence matrix
