基于可见光谱通道的松树年轮参数获取方法研究

年轮参数是树木生长的重要指标。目前年轮的检测方法主要有人工测量法、扫描仪法和X射线法等。这些方法费时费力、检测成本高、操作难度大。为此，提出采用可见光谱进行年轮参数检测的新方法。首先针对活立木年轮检测需求，设计了木芯专用可见光谱分析装置。装置采用宽光谱对称式光源，封闭式暗箱设计，彩色CCD外置、固定距离采集方式。进而采集样品全彩色图像，分别提取不同光谱通道数据进行分析，最终获得年轮特征光谱参数。以从松树活立木钻取的木芯为例，经打磨后的木芯水平固定于载物台，光源均匀照射样品，CCD采集木芯漫反射RGB图像，基于对木芯全彩色图像的不同光谱通道的分析，分别提取RGB三帧灰度图像的强度曲线，区分早晚材分界线，得到树木年轮各项指标，实现树木年轮参数的快速提取。实验首先将CCD采集的RGB图像转换至NTSC色彩空间，以扩大色彩域。在设置滤波窗滤除背景，截取出木芯图像后，通过提取木芯图像的R，G和B三灰度分量图像，发现木芯B灰度图的早晚材区域差异最分明。基于该特征，可提取出早晚材分界线的位置信息。对木芯B灰度分量图微分，得到灰度沿水平方向的空间梯度。确定灰度变化率极大值对应的各点，鉴于木芯生长特性，灰度变化率极大值对应空间位置首先取为窄像素区域，进而在窄像素区域取中间值，此即木芯年轮各早晚材的中心点。其中，早材中心点对应光谱曲线的极大值，晚材中心点对应光谱曲线的极小值。结合专家经验，建立早晚材中心点与分界线的灰度关系，可得各分界线的位置。通过早晚材分界线与年轮各参数的关系，可进一步得出年轮各项指标。对比3位林木育种专家人工识别结果，本方法检测结果除了在木芯靠近端点处的位置外，都具有极高的准确度。采用可见光谱通道的数据采集与分析方法检测树木年轮参数，检测过程可实现全自动，高效、无损，精度可达到0.1 mm，结果精确。相比于人工测量法、扫描仪法，检测效率更高；相比于X射线法，检测过程更加安全、低耗，操作更方便。是一种具有较强应用性的方法。

Tree Ring Detection of Pine Tree Based on Visible Spectrum Channel

Annual ring parameter is an important indicator of tree growth. The current annual detection methods mainly include manual measurement, scanner method and X-ray method. These methods are time-consuming, labor-intensive, expensive to detect, and difficult to operate. For this reason, this paper proposed the method of using the visible spectrum to detect annual ring parameters. A special core analysis device was designed. The device is consist of a wide-spectrum symmetrical light source, a closed dark box, and a color CCD which is assembled external. Taking pine wood core as an example, The polished wood core is fixed horizontally on the stage, and capture images of the sample. Based on spectral analysis and extracting RGB grayscale intensity images, we can identify boundaries of early-wood and late-wood, and then a series of characteristic parameters are gained. Quick acquirement of tree-ring parameters by tree-ring picture processing can be realized. First, converts the RGB image acquired by the CCD to the NTSC color space to expand the color domain. Then, sets the filter window to filter out the background and cutting out the wood core image, by extracting the R, G, and B grayscale component images of the wood core image, it is found that the wood core B gray image has the most distinct difference in the early and late material regions. Based on this feature, the positional information of the boundary line of the early and late materials can be extracted. Differentiate the grayscale component map of the wood core B to obtain the spatial gradient of the gray level in the horizontal direction. The points corresponding to the maximal value of the gray change rate are determined. In view of the growth characteristics of the wood core, the maximum value of the gray change rate corresponding to the spatial position is first taken as a narrow pixel region, and then an intermediate value is taken in a narrow pixel region. Among them, the center point of the early material corresponds to the maximum value of the spectral curve, and the center point of the late material corresponds to the minimum value of the spectral curve. Combined with expert experience, establish the gray relationship between the center point of the early and late materials and the boundary line, and the position of each boundary line can be obtained. The indicators of annual rings can be further derived from the relationship between the boundaries of the wood and the annual rings. Comparing with the results of artificial identification of three forest tree breeding experts, the results of this method have extremely high accuracy, except for the position of the wood core near the end point. Using the data acquisition and analysis method of the visible spectrum channel to detect tree annual ring parameters, the detection process can be fully automated, highly efficient, and non-destructive. The accuracy can reach 0.1 mm and the result is accurate. Compared with the manual measurement method and the scanner method, the detection efficiency is higher; compared with the X-ray method, the detection process is safer, lower in cost, and more convenient to operate. It is a method with strong application.