基于深度学习的蔬菜田精准除草作业区域检测方法OA北大核心CSTPCD
Deep Learning Detection of Weeds in Vegetable Fields
[目的]蔬菜生长随机,杂草种类众多.传统杂草识别算法复杂,且仅识别出杂草,未能精准确定除草作业区域.本研究以蔬菜及其伴生杂草为研究对象,拟探索一种基于深度学习的杂草识别与精准除草作业区域检测方法.[方法]通过将原图切分网格图像,利用深度学习模型识别蔬菜、杂草及土壤,将包含杂草的网格图像标记为除草作业区域.选取ShuffleNet、DenseNet和ResNet模型开展识别试验,并采用精度、召回率、F1 值和总体准确率、平均准确率分别对验证集和测试集进行评价分析.[结果]所选的 3种网络模型均能较好地识别杂草和蔬菜,其中ShuffleNet为杂草识别最优模型,其对杂草的识别具有较为均衡的精度和召回率,分别为 95.5%、97%,且其识别速度也达最优,为 68.37 fps,能够应用于实时杂草识别.[结论]本研究提出的除草作业区域检测方法具有高度的可行性和极佳的识别效果,可用于蔬菜田间杂草的精准防除.
[Objective]Deep learning to accurately identify weeds for effective weeding in vegetable fields was investigated.[Method]Image of a vegetable field was cropped into grid cells as sub-images of vegetables,weeds,and bare ground.Deep learning networks using the ShuffleNet,DenseNet,and ResNet models were applied to distinguish the target sub-images,particularly the areas required weeding.Precision,recall rate,F1 score,and overall and average accuracy in identifying weeds of the models were evaluated.[Result]Although all applied models satisfactorily distinguished weeds from vegetables,ShuffleNet could simultaneously deliver a 95.5%precision with 97%recall and a highest detection speed of 68.37 fps suitable for real-time field operations.[Conclusion]The newly developed method using the ShuffleNet model was feasible for precision weed control in vegetable fields.
李卫丽;金小俊;于佳琳;陈勇
南京航空航天大学金城学院机电工程与自动化学院,江苏 南京 211156||南京林业大学机械电子工程学院,江苏 南京 210037南京林业大学机械电子工程学院,江苏 南京 210037||北京大学现代农业研究院,山东 潍坊 261325北京大学现代农业研究院,山东 潍坊 261325南京林业大学机械电子工程学院,江苏 南京 210037
计算机与自动化
蔬菜杂草图像处理深度学习作业区域检测
Vegetablesweedsimage treatmentdeep learningweeding area determination
《福建农业学报》 2024 (002)
199-205 / 7
国家自然科学基金项目(32072498);江苏省研究生科研与实践创新计划项目(KYCX22_1051)
评论