【作者】 杨磊；

【导师】 李文彬；

【作者基本信息】 北京林业大学 ， 森林工程， 2008， 硕士

【摘要】 本论文的研究目的是为了提出一种基于单幅图象进行立木枝桠遮挡识别的方法,用于解决树枝的遮挡问题,并使其能应用于其他林业领域的计算机视觉研究。本论文的研究以立木图象处理为基础,提出一种基于遗传算法的图象分割方法对立木图象进行分割,并取得比较满意的结果;利用基于数学形态学的方法,完成图象中树枝信息的提取;在分析树枝生长特点的基础上,利用直线拟合技术,提出一种基于单幅图象的遮挡枝识别方法,实现了立木枝桠遮挡的识别。本论文取得的成果和创新点如下:1、在分析各种常用的图象分割技术的基础上,提出一种基于遗传算法的立木图象分割方法,解决了现有分割方法的局限性。实验结果表明,该方法不仅能较好的反映图象细节,而且能较好的消除背景噪声,适用于背景复杂的立木图象分割。2、提出一种基于数学形态学方法的树枝信息提取方法,先利用数学形态学基本运算—腐蚀和膨胀来实现树干信息的提取,然后将二值化的枝干图与提取得到的树干图进行差运算,从而得到树枝信息。3、在分析树枝生长特点的基础上,提出一种基于单幅图象的遮挡枝识别方法(已申报国家发明专利)。实验结果表明,本文提出的方法能较好地识别立木枝桠的遮挡,对计算机视觉技术在林业中的应用提供基础。更多还原

【Abstract】 In order to solve the occlusion of tree branches, this research puts forward a computer identifying method of standing tree branches’occlusion based on a single image, and makes it capable for other computer vision research in forestry.Based on standing tree image processing, this paper proposes a genetic algorithms-based method of image segmentation, and gets satisfying result of segmentation. Using a mathematical morphology-based method, the paper achieves to extract the information of branches. And based on analyzing the growth characteristic of branches, and using straight line fitting technique, the paper proposes an identifying method of standing tree branches’occlusion based on a single image, and achieves to identify the occlusion of standing tree branches.The achievement and innovation of this paper is as follows:1. Based on analyzing each kind of general image segmentation technique, the paper proposes a genetic algorithms-based method of image segmentation to solve the limitation of the segmentation method in existence. The result of the experiment shows that not only can this method reflect the detail of the image, but also can well eliminate the yawp of background,and it is fairly effective in image segmentation of standing tree in complex environment.2. This paper proposes a mathematical morphology-based method of branches information extracting, using the basic operation of mathematical morphology—erosion and dilation, achieves to extract the information of trunk, and subtracting the trunk image from the binary image of tree, achieves to extract the information of branches.3. Based on analyzing the growth characteristic of branches, the paper proposes an identifying method of standing tree branches’occlusion based on a single image, and applies for national invention patent.The result of the experiment shows that the method proposed by this paper can preferably identify the occlusion of standing tree branches, and provides elements for computer vision application in forestry.更多还原