Опубликована: Дек. 20, 2024
Язык: Английский
Plant Methods, Год журнала: 2024, Номер 20(1)
Опубликована: Дек. 19, 2024
Verticillium wilt greatly hampers Chinese cabbage growth, causing significant yield limitations. Rapid and accurate detection of in the (Brassica rapa L. ssp. pekinensis) can provide agronomic benefits. Here, we propose a model, DSConv-GAN, which is based on images acquired by an unmanned aerial vehicle (UAV). Based YOLOv8, with addition dynamic snake convolution (DSConv) module improved loss function maximum possible distance intersection-over-union (MPDIoU), enhanced complex structures global characteristics under different growth conditions. To reduce difficulty acquiring diseased data, cycle-consistent generative adversarial network (CycleGAN) was used to simulate generate for multiple fields. The lightly infected plants achieved precision, recall, mean average precision (mAP), F1-score 81.3, 86.6, 87.7, 83.9%, respectively. DSConv-GAN outperforms other models terms speed, robustness, generalization. model combined software improve practicability proposed method. Our results demonstrate be effective intelligent farming tool that provides early, rapid, growing environments.
Язык: Английский
Процитировано
2
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, Год журнала: 2024, Номер 17, С. 13547 - 13557
Опубликована: Янв. 1, 2024
As a crucial economic crop, the health status of cotton directly impacts farmers' income and national economy. Therefore, timely accurate detection identification diseases pests are significant importance, aiding in reducing adverse effects on yield quality. Existing research struggles to address balance between resource consumption accuracy disease pest detection. Moreover, often occur beneath canopy, orthorectification drone imagery may result insufficient feature information prolonged processing time, among other issues. To aforementioned issues, this paper proposes precise method for Verticillium wilt based unmanned aerial vehicle multi-angle remote sensing guided by satellite time series monitoring model. Specifically, first, combining Sentinel-1 microwave Sentinel-2 optical images, we constructed model extreme gradient boosting algorithm identify areas affected invasion. Subsequently, after identifying blocks disease, collected multi-spectral data captured from multiple angles vehicles compared different combinations vegetation indices bands. Finally, classification support vector machine radial basis function method. Experimental results indicate that joint time-series achieved OA 81.73% Kappa coefficient 0.63, meeting requirements first stage. Based SVM with RBF optimal band combination, value comprehensive image at -58° angle reached 96.74%, 0.93, second
Язык: Английский
Процитировано
1
Frontiers in Plant Science, Год журнала: 2024, Номер 15
Опубликована: Авг. 27, 2024
Background Cotton pests have a major impact on cotton quality and yield during production cultivation. With the rapid development of agricultural intelligence, accurate classification is key factor in realizing precise application medicines by utilize unmanned aerial vehicles (UAVs), large devices other equipment. Methods In this study, insect pest model based improved Swin Transformer proposed. The introduces residual module, skip connection, into to improve problem that features are easily confused complex backgrounds leading poor accuracy, enhance recognition pests. 2705 leaf images (including three pests, aphids, mirids mites) were collected field, after image preprocessing data augmentation operations, training was performed. Results test results proved accuracy compared original increased from 94.6% 97.4%, prediction time for single 0.00434s. with seven kinds models (VGG11, VGG11-bn, Resnet18, MobilenetV2, VIT, small, base), respectively 0.5%, 4.7%, 2.2%, 2.5%, 6.3%, 7.9%, 8.0%. Discussion Therefore, study demonstrates significantly improves efficiency detection models, can be deployed edge such as thus providing an important technological support theoretical basis control precision drug application.
Язык: Английский
Процитировано
1
Plants, Год журнала: 2024, Номер 13(21), С. 2960 - 2960
Опубликована: Окт. 23, 2024
Cotton verticillium wilt significantly impacts both cotton quality and yield. Selecting disease-resistant varieties using their resistance genes in breeding is an effective economical control measure. Accurate severity estimation of this disease crucial for resistant varieties. However, current methods fall short, slowing the process. To address these challenges, paper introduces CVW-Etr, a high-precision method estimating wilt. CVW-Etr classifies into six levels (L0 to L5) based on proportion segmented diseased leaves lesions. Upon integrating YOLOv8-Seg with MobileSAM, demonstrates excellent performance efficiency limited samples complex field conditions. It incorporates RFCBAMConv, C2f-RFCBAMConv, AWDownSample-Lite, GSegment modules handle blurry transitions between healthy regions variations angle distance during image collection, optimize model's parameter size computational complexity. Our experimental results show that effectively segments lesions, achieving mean average precision (mAP) 92.90% accuracy 92.92% only 2.6M parameters 10.1G FLOPS. Through experiments, proves robust severity, offering valuable insights applications.
Язык: Английский
Процитировано
0
Food Chemistry X, Год журнала: 2024, Номер 24, С. 102031 - 102031
Опубликована: Ноя. 22, 2024
Язык: Английский
Процитировано
0
Опубликована: Дек. 20, 2024
Язык: Английский
Процитировано
0