摘要
针对武夷岩茶加工中杂质人工分拣效率低、成本高的问题,提出一种基于图像处理与BP(back propagation)神经网络的杂质识别方法。通过采集带杂质的武夷岩茶图像,采用小波变换去噪、直方图阈值分割等技术进行预处理,结合颜色空间转换、形状描述子(霍夫变换、轮廓检测)及纹理特征提取方法,构建包含轴比值、厚度值、卷曲程度的杂质特征向量。基于BP神经网络设计分类器,输入层接收特征向量,隐藏层采用Sigmoid激活函数,输出层通过Softmax分类,利用639幅杂质图像数据集进行训练和测试。实验结果表明:该模型对无规则类、球状类、棍棒类杂质的平均识别率达90%以上,显著优于支持向量机(support vector machine,SVM)的86.67%。该方法有效降低环境干扰对识别的影响,验证图像处理与BP神经网络在茶叶杂质自动化分拣中的可行性与高效性,为茶叶加工智能化提供技术支撑。
This study addresses the inefficiency and high costs of manual impurity sorting in Wuyi rock tea processing by proposing an impurity identification method integrating image processing and a BP neural network.Acquiring images of tea with impurities,preprocessing techniques including wavelet transform denoising and histogram threshold segmentation were applied,combined with color space conversion,shape descriptors(Hough transform,contour detection),and texture feature extraction to construct impurity feature vectors encompassing axis ratio,thickness value,and curvature degree.A BP neural network classifier was designed,where the input layer receives feature vectors,the hidden layer employs the Sigmoid activation function,and the output layer performs classification via Softmax.Training and testing were conducted using a dataset of 639 impurity images.Experimental results demonstrate that the model achieves an average recognition rate exceeding 90%for irregular,spherical,and rod-shaped impurities,significantly outperforming support vector machine(SVM)at 86.67%.The method effectively mitigates environmental interference,validating the feasibility and efficiency of image processing combined with BP neural networks for automated tea impurity sorting,thereby advancing intelligent solutions for tea processing industrialization.
作者
徐涛
闫彪
何曹尹宣
叶家亮
刘德生
XU Tao;YAN Biao;HE Caoyinxuan;YE Jialiang;LIU Desheng(Agricultural Machinery Intelligent Control and Manufacturing Technology Key Laboratory of Fujian Province,Wuyi University,Wuyishan,Fujian 354300;Wuyishan Yejiayan Tea industry Co.,Ltd.,Wuyishan,Fujian 354300)
出处
《武夷学院学报》
2025年第6期41-47,共7页
Journal of Wuyi University
基金
福建省自然科学基金(2024J01910)
福建省教育科学规划2024年常规课题(FJJKBK24-127)。
关键词
杂质识别
图像预处理
特征提取
BP神经网络
武夷岩茶
impurity identification
image preprocessing
feature extraction
BP neural network
Wuyi rock tea
