To address the challenges faced in real-world tomato ripeness detection,such as variable lighting conditions,complex backgrounds,and the trade-off between accuracy and the model being effectively lightweight,this stud...To address the challenges faced in real-world tomato ripeness detection,such as variable lighting conditions,complex backgrounds,and the trade-off between accuracy and the model being effectively lightweight,this study proposes a lightweight YOLOv11-MHS model.The improvements of the proposed model are reflected in three aspects:(1)the C3k2_MSCB module is designed,which integrates a multiscale convolutional block(MSCB)for multiscale feature extraction and fusion,thereby enhancing detection accuracy;(2)the neck of the model is redesigned as a high-level feature screening-fusion pyramid structure,which fuses key features to improve robustness in cluttered environments while reducing model size;and(3)the C2PSA module is enhanced by introducing the spatial and channel synergistic attention mechanism to improve the ability of the model to handle complex scenes.Experimental results on the same data set show that,compared to the baseline model YOLOv11n,YOLOv11-MHS achieves improvements of 1.7%in mAP0.5 and 2.9%in mAP0.5-0.95,while reducing parameters and model size by 35.2%and 32.7%,respectively.These results demonstrate that YOLOv11-MHS achieves both outstanding accuracy and lightweight performance in tomato ripeness detection,providing technical support for agricultural applications.展开更多
In order to address the challenge of non-destructive detection of tomato fruit ripeness in controlled environments,this study proposed a real-time instance segmentation method based on the edge device.This method comb...In order to address the challenge of non-destructive detection of tomato fruit ripeness in controlled environments,this study proposed a real-time instance segmentation method based on the edge device.This method combined the principles of phenotype robots and machine vision based on deep learning.A compact and remotely controllable phenotype detection robot was employed to acquire precise data on tomato ripeness.The video data were then processed by using an efficient backbone and the FeatFlowNet structure for feature extraction and analysis of key-frame to non-key-frame mapping from video data.To enhance the diversity of training datasets and the generalization of the model,an innovative approach was chosen by using random enhancement techniques.Besides,the PolyLoss optimization technique was applied to further improve the accuracy of the ripeness multi-class detection tasks.Through validation,the method of this study achieved real-time processing speeds of 90.1 fps(RTX 3070Ti)and 65.5 fps(RTX 2060 S),with an average detection accuracy of 97%compared to manually measured results.This is more accurate and efficient than other instance segmentation models according to actual testing in a greenhouse.Therefore,the results of this research can be deployed in edge devices and provide technical support for unmanned greenhouse monitoring devices or fruit-picking robots in facility environments.展开更多
基金financially supported by National Natural Science Foundation of China(12364011)Guangxi Science and Technology Plan,China(AD21220147,AD25069027)+1 种基金Liuzhou Science and Technology Program,China(2023PRJ0103,2024AA0204A001)Graduate Education Innovation Project,China(YCSW2024522).
文摘To address the challenges faced in real-world tomato ripeness detection,such as variable lighting conditions,complex backgrounds,and the trade-off between accuracy and the model being effectively lightweight,this study proposes a lightweight YOLOv11-MHS model.The improvements of the proposed model are reflected in three aspects:(1)the C3k2_MSCB module is designed,which integrates a multiscale convolutional block(MSCB)for multiscale feature extraction and fusion,thereby enhancing detection accuracy;(2)the neck of the model is redesigned as a high-level feature screening-fusion pyramid structure,which fuses key features to improve robustness in cluttered environments while reducing model size;and(3)the C2PSA module is enhanced by introducing the spatial and channel synergistic attention mechanism to improve the ability of the model to handle complex scenes.Experimental results on the same data set show that,compared to the baseline model YOLOv11n,YOLOv11-MHS achieves improvements of 1.7%in mAP0.5 and 2.9%in mAP0.5-0.95,while reducing parameters and model size by 35.2%and 32.7%,respectively.These results demonstrate that YOLOv11-MHS achieves both outstanding accuracy and lightweight performance in tomato ripeness detection,providing technical support for agricultural applications.
基金funded by the National Key R&D Program(Grant No.2022YFD2002305)Beijing Nova Program(Grant No.Z211100002121065,20220484202)Collaborative Innovation Center of Beijing Academy of Agricultural and Forestry Sciences(Grant No.KJCX201917).
文摘In order to address the challenge of non-destructive detection of tomato fruit ripeness in controlled environments,this study proposed a real-time instance segmentation method based on the edge device.This method combined the principles of phenotype robots and machine vision based on deep learning.A compact and remotely controllable phenotype detection robot was employed to acquire precise data on tomato ripeness.The video data were then processed by using an efficient backbone and the FeatFlowNet structure for feature extraction and analysis of key-frame to non-key-frame mapping from video data.To enhance the diversity of training datasets and the generalization of the model,an innovative approach was chosen by using random enhancement techniques.Besides,the PolyLoss optimization technique was applied to further improve the accuracy of the ripeness multi-class detection tasks.Through validation,the method of this study achieved real-time processing speeds of 90.1 fps(RTX 3070Ti)and 65.5 fps(RTX 2060 S),with an average detection accuracy of 97%compared to manually measured results.This is more accurate and efficient than other instance segmentation models according to actual testing in a greenhouse.Therefore,the results of this research can be deployed in edge devices and provide technical support for unmanned greenhouse monitoring devices or fruit-picking robots in facility environments.
