树干直径自动识别与测量技术研究

在森林资源调查中为实现树干直径快速、精确测量,使用电子元件进行了树干影像自动识别与测量。通过摄像头和CMOS传感器获得图像,经灰度转换后,使用窗口大小为5行5列的高斯滤波器,利用Canny边缘检测算法提取边缘,包括平滑、非极大值剔除、双阈值的边缘连接等,在提取的边缘图像基础上,设计了树木直径自动提取算法。以4列的窗口为模板,自上而下,对经过Canny算法后的图像进行竖直线段的提取。在提取所有竖线后,将最大宽度的2条竖线当作树干轮廓线,通过焦距、物距、像距与像素宽度之间的关系计算出树干直径。最后选择了不同树种共计16棵树木进行了验证,实验结果表明树干直径识别的精度为96.9%,绝大部分测量数据符合森林资源调查要求。

Automatic Recognition and Measurement Technology of Tree Trunk Diameter

Tree trunk diameter is one of the most important tree measuring factors in forest inventory. To quickly and accurately measure the tree trunk, electronic components were applied for tree trunk image recognition and measurement. Application of image processing technology in forest mensuration proposes a solution for accurately measuring trees, which makes non professional technicians measure trees easily without experience. Images were taken by camera and processed by CMOS. The image was smoothed by Gaussian inverse filter after conversion from RGB to greyscale. Then edge was detected through non maximum suppression and double thresholds edge connection. Tree trunk automatic detection algorithm was developed on the base of the detecting image. The algorithm used a 4 column window that represented a vertical segment to extract the vertical segments from the images above. The algorithm got rid of the vertical segment with over two successive 0 values, including vertical, horizontal and diagonal directions. The detected 0 value was searched from top to bottom. In addition, individual points in the window were removed. When all the vertical lines were abstracted, the two vertical lines with the maximum width were represented as the tree trunk. Tree trunk diameter was computed according to the relationship among focal length, object distance, image distance and pixel size. The image recognition results were validated by selecting different trees. Results showed that the image recognition precision was 96.9% and most data were conformed to the requirement of forest inventory. The forestry intelligence was explored and the digitizing components was used to realize the forestry intelligence.