BP神经网络与HOG特征相结合的小花清风藤叶片智能鉴别模型建立

Establishment of Intelligent Identification Model of the Leaves of Sabia parviflora Combined with BP Neural Network and HOG Feature

为保障小花清风藤的药材质量,建立基于HOG特征与BP神经网络相结合的小花清风藤叶片智能鉴别模型。采用方向梯度直方图(HOG)特征提取算子提取小花清风藤与簇花清风藤叶片的HOG特征,结合BP神经网络理论构建基于HOG特征的小花清风藤叶片智能鉴别神经网络模型。结果表明:该模型对小花清风藤叶片的平均识别率为97.60%,簇花清风藤为98.61%,对任意样本为100%。建立的基于HOG特征与BP神经网络相结合的小花清风藤叶片智能鉴别模型适合于小花清风藤叶片鉴别。

In order to ensure the quality of the medicinal materials of Sabia parviflora,an intelligent identification model of S.parviflora was established based on integrated HOG characteristics and BP neural network.The HOG features of the leaves of S.parviflora and Sabia swinhoei were extracted by using the direction gradient histogram(HOG)feature extraction operator.Integrated with BP neural network theory,the intelligent neural network model of S.parviflora was constructed based on HOG features.Results:The average recognition rate of the model was 97.60%for the leaves of S.parviflora,98.61%for that of S.swinhoei,and 100%for any sample.The established intelligent identification model of S.parviflora based on the combination of HOG features and BP neural network is suitable for the identification of S.parviflora leaves.