中文摘要：

郁闭度具有获取便捷的特点,是森林资源调查的基本因子之一,也是检测森林状态、检验营林效果的重要指标,并已成为许多重要特征遥感反演的基础数据。另一方面,纹理特征成为遥感信息挖掘的重要方向,并被用于判别单纯依靠光谱特征无法有效识别的信息,但纹理的应用范围与可行性仍缺乏验证。以福建省三明市、将乐县、沙县、延平区4县（市、区）HJ-1 CCD多光谱数据及实测松林郁闭度为主要数据,基于灰度共生矩阵法提取均值、方差、协同性、对比度、相异性、熵、角二阶矩及相关性等8类纹理特征（共计32个）,利用单因素方差分析法筛选24个对松林郁闭度变化有所响应的纹理,并分别以单个、单类纹理特征及多纹理特征联合为自变量,建立极弱度郁闭林、弱度郁闭林与中度郁闭林的Fisher判别函数。结果表明,单个纹理特征对松林郁闭度的判别能力普遍较差,仅COR_2、COR_4、CON_1、CON_2判别精度高于50%;单类纹理特征对郁闭度的判别能力有所提高,均值、方差、对比度的判别精度高于50%,而相关性的判别精度可达62.7%;多纹理特征联合对松林郁闭度有更高的判别能力,精度达74.6%。由此证明,纹理具备郁闭度的响应能力,在其他特征的遥感反演过程中可以应用,但需要深入挖掘其相关性与响应机制。

英文摘要：

Canopy density,which can be easily obtained,is one of the basic factors in forest resources survey,and also an important factor in forest state detection and forest management effect test,and has become a basic datum in remote sensing retrieval of many features.On the other hand,texture has become the significant direction of remote sensing data probing,but the application scope and feasibility of texture are still to be verified.Taking HJ-1 CCD multi-spectral image and measured pine forest canopy density materials in four counties（ cities,districts） of Sanming,Jiangle,Shaxian and Yanping of Fujian Province as the main data,eight classes of textual features（ total 32） such as mean,variance,homogeneity,contrast,dissimilarity,entropy,second moment,correlation were extracted with GLCM,24 textures affecting changes in canopy density were screened with ANOVA,and Fisher discriminant functions of very-weak canopy forest,weak canopy forest and moderate canopy forest independent variables respectively as single texture,single class textures,combination of multiple textures were constructed.The results showed that the discriminant abilities of single textual features on pine forests canopy density were generally poor,only the accuracies of COR_2,COR_4,CON_1,CON_2 were above 50%; the discriminant accuracies of single class textures increased,the accuracies of mean,variance,contrast were above 50%,the correlation reached 62. 7%; the combination of multiple textures had higher discriminant ability on pine forests canopy density with an accuracy up to 74. 6%. Therefore,texture has response ability to canopy density,and can be used in remote sensing retrieval of other features,but the relationships and response mechanisms should be deeply probed.