基于深度相机的玉米株型参数提取方法研究

提出了一种基于骨架提取的改进算法，可实现在大田环境下，使用PMD深度相机快速、无损测量玉米株型参数。首先利用深度图像RGB伪彩色和深度距离信息，提取深度图像的骨架，排除复杂背景干扰，得到单株玉米的二值骨架图像；然后利用基于角点检测的改进归类算法提取骨架图像特征点；最后建立骨架图像中特征点与深度图像的对应关系，利用空间几何数学方法，结合特征点计算出玉米的3种株型参数，即株高、茎粗、叶倾角。农田实验对比分析表明，所提方法的株高测量结果与人工测量结果的相关系数 r 为0.986，最大相对误差小于2 cm，农田作物育种抗逆性分析还表明玉米株型参数与抗倒伏性具有显著相关性。

Maize Plant Type Parameters Extraction Based on Depth Camera

During the whole crop growing period and high-precision breeding process, measuring crop plant type parameters to achieve its phenotypic analysis is one of the important link. In view of the problem that the maize plant type parameters are obtained mainly by artificial field measurement in China at present, and it is high labor intensity and time-consuming. Thus a rapid and nondestructive measurement method of maize plant type parameters was proposed based on the photonics mixer device (PMD) camera with improved skeleton extraction algorithm. Firstly, the RGB pseudo color depth and distance information of depth image were used, and the depth of the image skeleton extraction was obtained by the improved skeleton extraction algorithm without complex background interference. Secondly, the binary skeleton image per corn plant was got by taking advantage of the improved corner detection classification algorithm for extracting skeleton image feature points. Finally, the feature points in the skeleton image were corresponded with depth images. Three kinds of corn plant type parameters: plant height, stem diameter and leaf angle were calculated by using mathematics method combined with space geometry feature point. Field test of the method in the practical application environment showed that the plant height could be measured at the seedling stage of maize plant, the correlation coefficient ( r ) of maize plant type parameters between the measured results by using the proposed method and the results measured artificially was 0.986, and the maximum relative error was less than 2 cm. Farmland crops breeding resistance analysis also showed that maize plant type parameters and lodging resistance had significant correlation. The results provide technical support for the inversion of phenotypic analysis of crop breeding.