基于立体视觉的郁闭林分主要测树因子的重建  

Reconstruction of Key Tree Factors in Shade Stand based on Stereo Vision

作　　者：王雪峰[1] 张青[2] 张超[3] 

机构地区：[1]中国林业科学研究院,北京100091 [2]北京林业大学 [3]中国农业大学

出　　处：《东北林业大学学报》2006年第4期24-26,30,共4页Journal of Northeast Forestry University

基　　金：国家自然科学基金(30271079)。

摘　　要：为解决林木图像由于树冠相互遮挡而难于匹配的问题,将立木分为枝下高与树冠高两部分,使用枝下高图像用立体视觉方法重建枝下高不同部位的直径与倾角,以此为基础估计树冠高部分,进而得到树木高度。同时,在不增加图像处理计算量的情况下,派生材积等间接测树因子。通过与干曲线的结合,解决了由于树冠遮挡难于用视觉方法重建主要测树因子的难题。试验地包括银杏、水杉两个树种。与实际树高相比,重建树高中最大误差是4.4%,最小误差为0.7%,在此基础上计算的材积最大、最小误差分别是4.7%和2.1%。与干曲线结合的立体视觉重建策略在恢复郁闭林分中的测树因子是可行的。A solution for measuring tree factors in computer vision is presented. Matching of tree images is very difficult because some trees are in the shelter. Stumpage was divided into two parts aiming to solve the matching problem, one is under branch and the other is crown. The stem diameters at different heights and angles between branches were reconstructed based on stereo vision methods. With the result of reconstruction, the crown height and tree height were evaluated, and also other factors such as volume were calculated. The crown shelter is an obstacle of reconstruction, which was overcome by combining with stem curve. Two tree species, gingkgo and metasequoia, were involved in experiment plots. Compared the evaluated height with the true tree height, the errors were within 0.7% to 4.4%, and consequently the errors of volume between 2.1% and 4.7%. It indicates that tree factors can be obtained by combining stem curve with stereo vision.

关 键 词：遮挡 树高测定 林木调查 立体视觉 三维重建 

分 类 号：S758.5[农业科学—森林经理学]