Based on multi-dimensional models,such as AI deep learning and visual analysis,including Trans-former,Gaussian mixture model and ResNet18,in this paper,a digital intelligent application platform was construc-ted for population activity and control evaluation of grain storage pests,and online monitoring,prediction and control evaluation of pests were realized.Three core business models,i.e.pest monitoring,pest prediction and grain storage prevention and control evaluation,were involved:1)Two-stage monitoring based on video stream data in the ware-house,detection of pest moving targets by Gaussian mixture model and three-frame difference method,and con-struction of pest classification model based on ResNet18 to realize online real-time detection of pest number and spe-cies and determination of pest grain grade;2)In the pest prediction module,a prediction model based on Transform-er was proposed,and the probability of pest occurrence was predicted based on environmental data,such as grain temperature combined with pest population activity;3)The effect of pest control based on the pest monitoring model,the movement speed,activity posture and activity degree in the pest monitoring field were further analyzed to evaluate the effect of pest control.
关键词
AI多模型技术/害虫监测/害虫预测/防治评价
Key words
AI multi-model technology/pest monitoring/pest prediction/control evaluation
维普
万方数据