一种探管式储粮害虫诱捕在线监测装置的设计研究

为实现对粮仓粮堆内害虫发生的早期监测，本研究基于探管式诱捕器和红外光电传感器研发了一种经济有效、可实时在线监测害虫发生的装置；基于该装置建立了9种主要储粮害虫成虫红外光电序列数据集；通过方差分析和数据可视化方法验证了其数据分布特性与害虫进入诱捕段的位置无关；对红外光电序列建立了高斯核支持向量机模型，通过实验室的测试，对蛀蚀性和粉食性两大害虫类别的判别达到了加权平均分类准确率87.7%、召回率87.5%和F1值87.6%。

Research on an Online Monitoring Device Based on Probe Trap for Stored-grain Insects in Grain Bulk

In order to detect the occurrence of insects inside the grain bulk, we developed an electronic grain probe trap used infrared photoelectric sensors, which can monitor the occurrence of the stored-grain insects in real time, and is cost-effective. An infrared photoelectric sequence dataset of 9 species of stored-grain insects was builded. Through variance analysis and data visualization, we verified that the infrared photoelectric sequence data of insects has no association with the position where the insect crawled into the trap section. A model based on Gaussian kernel support vector machine for the classification of infrared photoelectric sequence was given. The evaluation experiment was conducted in the laboratory. For the classification of internal-feeding insects and external-feeding insects, the weighted average accuracy of 87.7%, recall rate of 87.5% and F1 value of 87.6% were achieved.