基于B样条曲线的水稻叶片几何参数测量系统

基于B样条曲线的水稻叶片几何参数测量系统能无损、精确获取水稻叶片几何参数,可为农学研究提供精准数据。该研究分析了B样条曲线拟合基本原理,采用1次B样条曲线获取水稻最大叶宽,基于3次B样条曲线交互绘制手势轨迹曲线逼近叶脉获取水稻叶长,控制点数量和位置可人工干预,实现轨迹曲线实时调整。运用变异系数和均方根误差评价测量结果,结果显示,叶长变异系数为2.4%,叶宽变异系数为5.4%,均小于15%,测量结果精度较高;建立系统预测值和实测值线性回归拟合模型,水稻叶长均方根误差为1.534 0 cm,叶宽均方根误差为0.101 9 cm,系统预测能力较好。利用B样条曲线测量叶片几何参数,前期图像获取简便,能够真正实现便捷、无损、准确获取水稻叶长、叶宽,为水稻叶片几何参数的测定提供了一种新的测量方法,可满足农业科学研究基本需要。

Rice leaf geometric parameter measurement system based on B-spline curve

In order to obtain accurate geometric parameters of rice leaves non-destructively and to provide accurate data for agronomic research, the rice leaf geometric parameter measurement system based on B-spline curve was developed. The basic principle of B-spline curve fitting was elaborated, and the maximum leaf width of rice was obtained by using first order B-spline curve and the leaf length was obtained with gesture trajectory curve drawn to approximate the leaf vein based on cubic B-spline curve. Moreover, the number and position of control points could be manually intervened, therefore to realize real-time adjustment of the trajectory curve. In the results analysis, coefficient of variation (CV) and root mean square error (RMSE) were applied to evaluate the precision and accuracy of results. The leaf length CV was 2.4%, the leaf width CV was 5.4%, both were less than 15%, which meant that the results were of high precision; Meanwhile, linear regression fitting model of system prediction and measured values was established, found that the RMSE of leaf length was 1.534 0 cm and the RMSE of leaf width was 0.101 9 cm, which demonstrated that the system had excellent prediction ability. The results showed that the B-spline curve based measurement system could truly realize accurate acquisition of rice leaf length and leaf width conveniently and non-destructively and without much restriction in photo shooting, which provided a new measurement method for the determination of rice leaf geometric parameters and could satisfy the basic needs of agricultural scientific research.