基于图像处理技术的储粮害虫检测过程中,需要解决多种害虫多特征的识别分类问题。提出运用BP神经网络来进行害虫特征分类的方法。通过对储粮害虫图像的预处理,获取并优化提取5个特征参数输入BP网络进行训练。仿真结果表明,训练网络对粮仓中4类常见害虫的识别率达到了95%,得到了较好的识别结果。 In the detection process of stored grain pests based on image processing technology, it is necessary to solve the identification and classification of multiple pests with many features. Proposed the use of BP neural network for pest feature classification methods. Through the pretreatment of stored pests images, five characteristic parameters were extracted and optimized and input into BP network for training. The simulation results show that the training network achieves a recognition rate of 95% for four kinds of common pests in silos, and obtains good recognition results.