基于多目视觉的马尾松苗木形态学参数提取研究

生态文明建设、现代林业建设需要数量充足、质量优良的苗木,苗木质量评价是苗木质量调控的核心问题之一。为实现马尾松苗木形态特征快速精准评价,笔者设计了一套基于多目立体视觉的、高精度非接触式的马尾松苗木形态学参数提取系统。首先采用“张正友标定法”进行相机标定,然后软件系统对采集到的苗木图像序列进行预处理、图像校正、立体匹配等操作,匹配完成后获取到苗木的空间点云信息,最后运用泊松曲面重建算法对苗木的空间点云进行表面重建,得到马尾松苗木的三维模型。从重建出的模型中提取出苗木的根系体积,分别采用圆柱几何体、圆台几何体、圆柱与圆台几何体相结合、“体元模拟法”这4种方式进行拟合,为苗木质量的评价提供了更加精确的形态指标。
The construction of ecological civilization and the construction of modern forestry need the sufficient quantity and quality of seedlings. Seedling quality evaluation is one of the core problems of seedling quality control. A morphological characteristics extraction system of masson pine(Pinus massoniana Lamb.)seedlings is designed in this study. The system with multi-view stereo vision, high precision, and non-contact could reconstruct 3D model of masson pine seedlings by the multi-view image sequence, and quickly extracted the morphological characteristics of seedlings. Firstly, the camera calibration was carried out by using the calibration method from Zhang Zhengyou. Then the software system performed preprocesses, image correction and stereo matching on the image sequence, using SAD, SSD and NCC stereo matching algorithm to test the corrected image sequence in the process of matching, and the optimal matching algorithm was selected. And then the space point cloud information of the seedlings was obtained. Finally, using spatial Poisson surface reconstruction algorithm to carry out the surface reconstruction of seedlings space point cloud to obtain the 3D model of masson pine seedlings. After extracting the root volume of the seedlings from the reconstructed model, the volume was fitted with cylindrical geometry, round table geometry, geometry of the cylinder and the round table, and the “voxel simulation”, respectively, which embodied the unique advantages of “voxel simulation” when calculating the volume. Data extracted from the system software was compared with the traditional manual measurement data. It is proved that the 3D reconstruction method of masson pine seedlings has high precision and fast speed, and the reconstruction result is very close to the true morphology of the root of seedlings. The software can be used to replace the traditional manual measurement method, liberating people from the heavy and repeated measurement work, providing a more accurate morphological indicator for the evaluation of the seedling quality