冬小麦田间图像的群体纹理性研究  

Study on the Canopy Image Texture of Winter Wheat in Field

作　　者：单成钢[1] 廖树华[1] 龚宇[1] 梁振兴[1] 王璞[1] 

机构地区：[1]中国农业大学农学与生物技术学院

出　　处：《麦类作物学报》2006年第5期88-91,共4页Journal of Triticeae Crops

基　　金：国家"863"项目(2002AA2Z4021-1)

摘　　要：为了满足快速实时测定小麦群体特征指标的需要，以水肥调控措施塑造了不同结构的冬小麦群体，研究了冬小麦图像的群体纹理性。结果表明，出苗、起身、拔节、孕穗期的图像纹理性因子为0．5～0．7，说明抽穗前的冬小麦各生育时期的图像均具有较强的纹理性。采用灰度其生矩阵方法进行进一步纹理分析，提取4个方向共生矩阵的纹理特征——能量、熵、对比度、逆差距、相关性的平均值和均方差，利用逐步判别的方法从中筛选分类变量，并对三种不同结构的冬小麦群体分类判别，分类正确率达92．6％，初步结果表明该方法在小麦群体的判别中具有一定的可行性。It was time-consuming and labor intensive to judge growth status of crop using traditional measurements. This paper focused on how to improve the traditional measurements using image processing technology. The texture reliability of canopy images was studied on different canopies of winter wheat which were established by the relevant managements of irrigation and fertilization. The preliminary results indicated： The texture reliability factor η in the growth stages of seedling, upstanding, shooting, booting were between 0.5 to 0.7, therefore, the image texture reliability was strong, and it was feasible to do the further texture analysis. The approach of gray level co-occurrence matrices was applied to extract the mean and mean square variance of energy, entropy, contrast, inverse different moment and correlation as texture features, then the classification variables were determined with the stepwise discriminant analysis, and the classification accuracy of 92.6% was achieved when the method was used to identify three kinds of canopies of winter wheat.

关 键 词：小麦 数字图像 纹理 共生矩阵 逐步判别分析 

分 类 号：S512.1[农业科学—作物学] S311