一种对年轮图像双边滤波增强的树龄测量方法

Improving the tree-age measurement by enhancing the tree-ring image with bilateral filter

树木圆盘年轮图像分析是一种具有较强实用价值的树龄测量方法,但树木横截面上的色斑、锯痕、木材组织中的粗大管孔等都会导致年轮图像发生灰度变化,降低树龄检测精度。为此,提出了一种新的树木年轮图像增强的树龄测量方法:首先采用双边滤波对树木年轮图像进行图像增强,在保留年轮边缘信息的同时抑制锯痕、色斑等导致的灰度变化;然后采用改进的Canny算法对年轮图像进行边缘检测,为了克服传统Canny算子对噪声敏感的缺点,在求取年轮图像的梯度时,计算水平、垂直、45°、135°等4个方向的梯度分量;最后根据多数投票原则,统计年轮图像中的边缘数量,实现树龄测量。实验结果表明,提出的方法能够有效抑制色斑、锯痕、粗大管孔等因素的干扰,得到的年轮边缘图像光滑、清晰,统计得到的树龄与真实情况一致。

Tree-ring image analysis has been proven to be a promising tool for predicting tree age and has attracted more and more attention. Many approaches were proposed to detect and count tree-ring boundaries in images of wood cross section. However, factors such as stain, saw mark and pore often lower the accuracy of tree-age prediction. In this paper, a practical method was proposed for tree-ring image enhancement and tree-age measurement. The images of wood cross sections were first enhanced by the bilateral filtering method due to its excellent noise reduction per- formance. The gray variation caused by stain, saw mark and pore was reduced, whereas the information of tree-ring edges was not affected. Then, an improved Canny algorithm was used to detect the edges of tree rings. During this process, gradients in directions of the horizon, the vertical, 45° and 135° were computed to overcome the drawback of the conventional Canny algorithm. Finally, a majority rule was proposed so that the tree-ring boundaries could be counted with high accuracy even though broken edges existed in the images. The tree-age measurement was thus a- chieved. In this study, four wood cross section images were used as the samples. One of them was polished and the other three remained untreated after saw cutting. The experimental results show that the proposed method can achieve good performance. Smooth and clear edges can be obtained even though stains, saw marks and pores exist. The tree ages obtained were consistent with the real data. The result of this study indicates that the proposed method has extensive application potentiality in extracting tree-age information from wood cross section images.