基于立体视觉的玉米叶片形态测量与三维重建

Morphology measure and 3D reconstruction of corn leaf based on machine vision

提出一种基于立体视觉的玉米叶片形态测量和重建的方法。利用双目立体视觉系统获取玉米叶片的两幅图像,通过图像分割技术和边缘检测算法对每幅图像中的玉米叶片进行边缘提取;利用极线约束和彩色图像RGB值对图像进行匹配,计算出叶片边缘的三维坐标,从而恢复叶片的三维边缘;利用对叶片边缘的恢复技术,对叶片曲面进行三维重建;根据恢复的区域点云,可以测出任意两点的空间距离,实现了对叶片的三维测量。试验结果表明,此方法能够很好地恢复玉米叶片的三维信息,为玉米叶片三维形态的无损、快速检测监测提供了新的方法。

A method was proposed to gain the plant configuration information and 3D reconstruction. According to the principles of binocular stereo vision, two images were obtained. By image segmentation technology and edge detection algorithm, leaf edge of each image was extracted. Epipolar line and RGB of color image were applied to match stereo images, then 3D coordinates of the leaf edge were worked out, so the 3D edge of the leaf could be reconstructed. Besides, using the same technique, the surface of the leaf could be reconstructed. According to the point cloud, the distance between any two points could be got. Results of many experiments show that 3B information of corn leaf can easily be reconstructed by this method.