基于随机森林的玉米发育程度自动测量方法

发布时间：2018-04-28 10:37

本文选题：玉米果穗 + 发育程度　；参考：《农业机械学报》2017年01期

【摘要】：为提高玉米果穗发育程度检测的自动化程度与精度,提出一种基于机器视觉技术的测量方法。在随机森林机器学习算法的基础上构造秃尖、干瘪和籽粒区域的识别模型。该模型由多个独立同分布的弱分类器构成,对输入的训练样本进行列和行两个方向上的随机采样。比较随机森林模型和决策树模型的分类效果可知随机森林模型有效避免了过拟合和局部收敛现象的产生,并具有良好的推广能力。为确定最优的弱分类器数目,选择弱分类器个数为训练样本数量的1/80、1/40、1/20、1/10、1/5、1/4时分别构建随机森林分类器。研究结果表明,当随机森林中弱分类器个数为训练样本数量的1/20时,模型的识别率与稳定性最好。然后,以最优的随机森林模型作为分类器构建玉米果穗不同发育程度自动检测方法。试验结果表明,各区域长度测量的准确性均在95%以上,测量速度可达30个/min以上。
[Abstract]:In order to improve the automation and precision of corn ear development detection, a method based on machine vision was proposed. Based on the stochastic forest machine learning algorithm, a recognition model of bald tip, dry area and grain area is constructed. The model is composed of several independent and distributed weak classifiers. The input training samples are sampled randomly in both columns and rows. Comparing the classification effect of stochastic forest model and decision tree model, we can see that stochastic forest model can avoid overfitting and local convergence effectively, and has good generalization ability. In order to determine the optimal number of weak classifiers, a random forest classifier is constructed when the number of weak classifiers is 1 / 80 / 40 / 1 / 40 / 20 / 10 / 10 / 5 / 4 of the number of training samples. The results show that the recognition rate and stability of the model are the best when the number of weak classifiers in the random forest is 1 / 20 of the number of training samples. Then, the optimal stochastic forest model was used as classifier to construct the automatic detection method for different development degree of corn ear. The experimental results show that the accuracy of each region length measurement is more than 95%, and the measuring speed can reach more than 30 / min.
【作者单位】：华中农业大学信息学院;华中农业大学理学院;
【基金】：国家自然科学基金项目(31301235)
【分类号】：S513;TP391.41

【相似文献】