基于机器视觉的烤烟鲜烟成熟度判别模型优选

Optimization of fresh flue-cured tobacco maturity discrimination model based on machine vision

为解决目前靠人为主观经验判别鲜烟成熟度并不准确的问题,对不同成熟度鲜烟叶进行图像处理、特征提取,并建立鲜烟叶成熟度判别模型,以实现鲜烟成熟度的智能判别。通过采集云烟87品种不同成熟度上部烟叶图像的10种颜色特征和纹理特征,分别进行变量聚类分析以及相关性分析,筛选出每类特征与成熟度相关性最强的1个特征组成特征子集,利用基于遗传算法的支持向量机(GA-SVM)、基于粒子群算法的反向传播(PSO-BP)神经网络和极限学习机(ELM)进行鲜烟成熟度的判别研究。结果表明:以优选后5个烟叶图像特征作为模型输入时,所建立的GA-SVM、PSO-BP、ELM模型的判别准确率分别为92.00%、90.00%、84.00%。证明利用机器视觉技术判别鲜烟成熟度是可行的,为之后烟叶智能采收提供理论基础和技术支持。

In order to solve the problem of inaccurate judgment of fresh tobacco maturity based on subjective experience,image processing and feature extraction were carried out on fresh tobacco leaf with different maturity,and a fresh tobacco leaf maturity judgment model was established to realize intelligent judgment of fresh tobacco maturity.By collecting 10 color features and texture features of upper leaf images of Yunyan 87 varieties with different maturity,variable cluster analysis and correlation analysis were carried out respectively,and the one feature with the strongest correlation between each type of feature and maturity was selected to form a feature subset.In this paper,support vector machine based on genetic algorithm(GA-SVM),Back propagation neural network based on particle swarm algorithm(PSO-BP)and Extreme learning Machine(ELM)were used to identify the maturity of fresh tobacco.The results showed that the optimal five tobacco leaf image features were used as model inputs,the discrimination accuracy of the established GA-SVM、PSO-BP、ELM model was 92.00%、90.00%、84.00%.It is proved that it is feasible to discriminate the maturity of fresh tobacco by using machine vision technology,which provides theoretical basis and technical support for intelligent tobacco leaf harvesting.