面向农业机器人应用的边缘人工智能和自适应嵌入式系统设计.pdf

1、An Edge AI and Adaptive Embedded System Design for Agricultural Robotics ApplicationsChun-Hsian Huang1,Zhi-Rui Chen2,and Huai-Shu Hsu21Dept.Electrical Engineering,National Changhua University of Education2Dept.Computer Science and Information Engineering,National Taitung UniversityOutlineIntroductio

2、nProposed methodData collection for AI modelsRecognition of target crops and their pest and disease severity(PDS)estimation using binarized neural networks(BNNs)PDS prediction using multimodal learningAgrBot designAgricultural cyber-physical system(CPS)System implementation and evaluationsConclusion

3、2OutlineIntroductionProposed methodData collection for AI modelsRecognition of target crops and their PDS estimation using BNNsPDS prediction using multimodal learningAgrBot designAgricultural CPSSystem implementation and evaluationsConclusion3IntroductionMonitoring crop pest and disease severity(PD

4、S)is crucial to ensure the healthy growth of cropsThe motivation of this work is to enable an agricultural robot to directly estimate and predict the crop PDS in the growth environment.Based on the PDS estimation and prediction,the agricultural robot can apply biological agents to protect the crops

5、from pests and diseases.4Agricultural Cyber-Physical System(CPS)Adaptive binarized neural network(BNN)hardware modulePrediction of PDS based on heterogeneous data5OutlineIntroductionProposed methodData collection for AI modelsRecognition of target crops and their PDS estimation using BNNsPDS predict

6、ion using multimodal learningAgrBot designAgricultural CPSSystem implementation and evaluationsConclusion6Data Collection for AI Models7Three Levels of PDS for Dragon Fruits8OutlineIntroductionProposed methodData collection for AI modelsRecognition of target crops and their PDS estimation using BNNs