稻穗结构图像特征与籽粒数相关关系分析

针对人工测量低效,重复性差且无法同时获取多个参数的问题,提出稻穗结构图像特征测量方法。首先人工测量稻穗一次枝梗长度和每穗籽粒数,发现二者具有显著相关性;采用图像处理方法提取稻穗图像面积、一次枝梗长度特征,并分别分析二者与籽粒数的相关关系。试验结果表明,面积、一次枝梗长度特征与籽粒数的相关系数在0.90以上,预测籽粒数的最大误差平均值为7.90%,说明稻穗形态特征的图像提取方法可行有效,面积、一次枝梗长度均能用来表达或替代稻穗籽粒数特征。

Relative Analysis between Image Characteristics of Panicle Structure and Spikelet Number

Characteristics extraction method of rice panicle images was proposed to solve the traditional measurement problems, such as inefficiency, worse repeatability, obtaining multi-parameters difficultly. There was significant correlation between primary branch length and spikelet number obtaining by manual measurement. Consequently, rice panicle was spread out and images were captured. Image characteristics were extracted using some image processing operations, including area, primary branch length, and panicle skeleton. Experimental result showed that the correlation coefficients were up to 0.90 between image characteristics and spikelet number, and the average predicting error of the model was 7.90%. Consequently, characteristics extraction method of rice panicle images was effective and feasible. The area and primary branch length can be perfect expression and substitute for spikelet number.