【作者】 李卫正；

【导师】 彭世揆；

【作者基本信息】 南京林业大学 ， 森林经理学， 2009， 博士

【摘要】 使用改进的矢量中值滤波算法对广东省九曲水林场TM图像数据进行滤波消噪处理。处理结果与传统的滤波算法对比分析。结果表明,对林区TM图像而言,改进的矢量中值滤波器具有出色的平滑去噪能力,并且保持了很好的分类精度,可以替代传统的滤波方法。针对山区小班界线勾绘的特点—“对坡勾绘”,本文提出一种融合坡向信息的基于核的模糊C均值聚类算法(KFCM)算法,并研究了使用该方法提取小班界线的处理流程和技术,结果表明,对本次研究的TM数据而言,使用该方法提取效果较好。提取小班的符合程度可以达到76%。本文探索了利用DEM直接提取小班界线的具体实现方法。结果表明,由坡向图经图像处理后提取到的矢量多边形界线大部分与人工地面调查的小班界线基本相一致,重合程度可达70%。在对分水岭算法过分割问题分析基础上,提出一种结合沉浸分水岭和Normalized cut算法的图像分割方法,该方法具有分水岭变换速度快、边界定位准的特点,并利用Normalied Cut算法克服了图像过分割现象。直接使用该方法提取小班界线的符合程度可达到81%以上。如采用前期林班界线作为图像分割的控制线,对每个林班内对应的遥感影像分别使用结合沉浸分水岭算法和Normalized cut算法的图像分割方法,提取小班界线的效果最好,使用该方法提取的符合程度可达到84%以上。提取结果在进一步编辑修改后可以作为小班区划的工作底图。更多还原

【Abstract】 This research used TM image to extract the boundary line and the research site is in Jiuqushui forestry region , Wengyuan county, Guangdong province. First, the the vector median filtering was employed to dispose TM image data in order to removal the noise. Compared with other common filters, it has a good accuracy of classification and this method can be used instead of the other common used filtering methods.The KFCM (Kernel Fuzzy C-means cluster) algorithm was employed to classify the TM image data. and the classification results were used to extract the subcompartment line. The subcompartment line in mountain area is affected by the slope and the aspect. According to this, a new classification method combined aspect information to the Kernel Fuzzy C-means cluster method was used to extract subcompartment line. The extracting accuracy is up to 76%.In order to use DEM to draw the subcompartment boundary, the dissertation proposed a new method, which was used DEM to extract the slope map. During the extracting work, it used image processing technology on the slope map supported on GIS technology. The results show that the 70% of the extracted polygon lines can be superposed to the investigated subcompartment boundary.The immersion watershed method often led to oversegmentation and the normalized cut is low efficient in running time-cost. The new segmentation method used the watershed transform results as input parameter for the normalized cut. The extracting result indicated that the accuracy, which combined immersion watershed and normalized cut to extract subcompartment line, is up to 84%. This is a good result for drawing the subcompartment line by computer and can be instead of visual interpretation in the forest resource investigation.更多还原