高级检索

    烟苗整齐度评估分析模型研究

    An Analytical Model for Assessing Tobacco Seedling Uniformity

    • 摘要: 为实现集约化育苗工厂内烟苗整齐度的快速判断分析,本研究采用广义加性模型,对烟草苗床数据进行分析,筛选烟草苗床整齐度指标。通过随机森林算法、BP神经网络算法、支持向量机算法建立烟苗整齐度评估模型,并采用粒子群算法对模型分别进行优化。采用深度学习算法Alexnet、Resnet101和GoogleNet,2种优化器Adam和Nadam构建烟草苗床整齐度图像识别模型。研究结果表明,烟苗株高、茎围、有效叶数对烟苗整齐度有显著影响;粒子群优化随机森林算法模型性能最优,训练集准确率为96.67%,测试集准确率为88.00%,R2=0.69,MAE=0.13;Adam-GoogleNet模型识别性能最优,对烟苗整齐度测试数据识别平均准确率为93.89%。研究结果可为烟草苗床整齐度科学评价提供合理依据,为烟苗整齐度图像识别系统开发提供模型支撑。

       

      Abstract: To achieve rapid assessment and efficient analysis of the uniformity of tobacco seedlings in an intensive seedling factory, this study employs a generalized additive model (GAM) to analyze tobacco seed nursery data and screen for indicators of tobacco seedling uniformity. We evaluated Random Forest algorithm, BP Neural Network algorithm, and Support Vector Machine (SVM) algorithm. Particle Swarm Optimization (PSO) is then applied to optimize each of these models separately. This study constructs image recognition models for assessing the uniformity of tobacco seed nursery using deep learning algorithms, two optimizers, Adam and Nadam, specifically AlexNet, ResNet-101, and GoogLeNet. The research results indicate that the plant height, stem circumference, and number of effective leaves of tobacco seedlings have a significant impact on the uniformity of the tobacco seedlings. The Particle Swarm Optimized Random Forest model demonstrates the best performance, with an accuracy of 88.00%, an R2 value of 0.69, and a Mean Absolute Error (MAE) of 0.13. The Adam-GoogLeNet model shows the best recognition performance, achieving an averaged accuracy of 93.89%. Overall, findings of this study provide a reasonable basis for the scientific evaluation of tobacco nursery bed uniformity and offer support for the development of tobacco seedling uniformity image recognition systems.

       

    /

    返回文章
    返回