基于X射线图像的核桃品种识别方法研究

Variety Classification of Walnut Based on X-ray Image

针对核桃品种混杂问题,研究基于X射线成像技术的核桃品种识别方法。采集不同核桃品种的X射线图像,并对图像进行预处理和背景分割;采用多种方法提取图像的纹理特征和形状特征,共计71个特征参数。分别利用极限学习机(ELM)和概率神经网络(PNN)建立核桃的品种判别模型。结果表明:在核桃品种判别的所有模型中,老树核桃被误判的个数最多,识别率有待提高,ELM判别模型整体识别率优于PNN判别模型;在3种核桃的判别过程中,ELM判别模型的总体判别率达到88.76%。因此基于X射线成像技术能有效判别核桃品种,为实现核桃品种自动分选提供了新的研究方向。

With the aim of solving the mixed problem of walnut varieties,the walnut variety identification method based on X-ray imaging technology was researched.The X-ray images of different walnut varieties were collected,then they were preprocessed and segmented by the background.The texture features and shape features of the image with a total of 71 feature parameters were extracted by various methods.Finally,Extreme Learning Machine(ELM)and Probabilistic Neural Network(PNN)were used to establish a two-species discriminant model and a multi-species discriminant model of walnut.The results showed that:among all the models for walnut variety discrimination,the old tree walnuts were misjudged the most,and the recognition rate needed to be improved.The overall recognition rate of the ELM discrimination model was better than the PNN discrimination model;in the process of discriminating walnut varieties,the overall discrimination rate of the ELM discriminant model reached 88.76%,indicating that the X-ray imaging technology can effectively discriminate walnut varieties and provide a new research direction for the automatic sorting of walnut varieties.