高级检索
    李蓉辉, 陈玲, 吴明晶, 余小龙, 赵秀海. 亚热带人工林高分影像GLCM纹理的应用策略研究[J]. 北京林业大学学报, 2021, 43(1): 1-9. DOI: 10.12171/j.1000-1522.20200139
    引用本文: 李蓉辉, 陈玲, 吴明晶, 余小龙, 赵秀海. 亚热带人工林高分影像GLCM纹理的应用策略研究[J]. 北京林业大学学报, 2021, 43(1): 1-9. DOI: 10.12171/j.1000-1522.20200139
    Li Ronghui, Chen Ling, Wu Mingjing, Yu Xiaolong, Zhao Xiuhai. Application strategies for GLCM textures from very high spatial resolution optical imagery over subtropical plantations[J]. Journal of Beijing Forestry University, 2021, 43(1): 1-9. DOI: 10.12171/j.1000-1522.20200139
    Citation: Li Ronghui, Chen Ling, Wu Mingjing, Yu Xiaolong, Zhao Xiuhai. Application strategies for GLCM textures from very high spatial resolution optical imagery over subtropical plantations[J]. Journal of Beijing Forestry University, 2021, 43(1): 1-9. DOI: 10.12171/j.1000-1522.20200139

    亚热带人工林高分影像GLCM纹理的应用策略研究

    Application strategies for GLCM textures from very high spatial resolution optical imagery over subtropical plantations

    • 摘要:
        目的  探讨纹理变量及其相应参数配置范围,阐明各纹理变量随输入参数的变化规律,以便指导高分光学影像纹理在林业上的应用。
        方法  以福建省将乐国有林场不同坡向不同龄组的杉木人工林为例,基于QuickBird影像的全色波段进行灰度共生矩阵(GLCM)纹理的计算与分析。
        结果  结果表明:(1)除均值外的所有GLCM纹理变量对阴坡的3个龄组的区分能力均强于阳坡,且纹理变量优选需同时考虑衡量指标和纹理变量之间的相关程度。(2)窗口大小是统计组和有序组纹理最关键的输入参数,合适的窗口大小与影像的分辨率以及研究对象的空间尺度有关,对比度组纹理与窗口大小无关,可随意设置。(3)应用统计组和有序组纹理,无需关注像元间距,而应用对比度组纹理,不可忽视像元间距。(4)应用统计组纹理,像元间距越大越需关注计算方向;而对比度组和有序组纹理则相反,即像元间距越小越需关注计算方向。(5)作为最不受研究人员重视的灰度量化等级,推荐采用32或者64。
        结论  高分光学影像的纹理信息对光谱重叠度较高的地物具有一定的区分能力,能部分“弥补”阴影导致的光谱信号损失,但在应用中需对纹理变量及其输入参数进行优化选择和配置。该文的研究结论能够为高分光学影像纹理信息的优化应用提供实用的参考借鉴。

       

      Abstract:
        Objective  This paper aims to discuss textural variables and their corresponding input parameters, and clarify a specific and simple strategy for the optimal utilization of textural information from very high spatial resolution (VHR) imagery in the field of forestry.
        Method  Based on a case of Chinese fir plantation with different age classes under varied topographic conditions from Jiangle State-Owned Forest Farm of Fujian Province of eastern China, seven gray level co-occurrence matrix (GLCM) textures and four input parameters were involved to demonstrate the optimal selection of textures and the setting of input parameters.
        Result  Almost all textures had higher age class separability over shady slope than sunny slope, and both the measurement index and the correlation between different textures should be used to select the optimal textures. Among the four input parameters, moving window size was the most important for textures in statistic and orderliness groups. Optimal window sizes should be jointly determined by the spatial resolution of images and spatial scale of research objects. Textures in contrast group were independent of window size, which means that any window size can be set for those textures. No attention need to be paid to the setting of displacement for textures in statistic and orderliness groups, but the situation was different for textures in contrast group. Moreover, orientation setting should be more concerned with the increase of displacement for textures in statistic group, but not for textures in the other two groups. As for the most neglected parameter, the grayscale quantization level can be set to 32 or 64.
        Conclusion  The textural information of VHR imagery can be used to distinguish objects with high spectral overlaps, which can “make up” the spectral information loss caused by shadows. However, textural variables and their input parameters should be optimized. The specific strategy and all the general rules from this study can provide practical suggestions for the optimal utilization of textural information from VHR imagery.

       

    /

    返回文章
    返回