PWDViTNet: A lightweight early pine wilt disease detection model based on the fusion of ViT and CNN

Computers and Electronics in Agriculture, Journal Year: 2025, Volume and Issue: 230, P. 109910 - 109910

Published: Jan. 10, 2025

Language: Английский

---

A Pine Wilt Disease Detection Model Integrated with Mamba Model and Attention Mechanisms Using UAV Imagery

Remote Sensing, Journal Year: 2025, Volume and Issue: 17(2), P. 255 - 255

Published: Jan. 13, 2025

Pine wilt disease (PWD) is a highly destructive worldwide forest quarantine that has the potential to destroy entire pine forests in relatively brief period, resulting significant economic losses and environmental damage. Manual monitoring, biochemical detection satellite remote sensing are frequently inadequate for timely control of disease. This paper presents fusion model, which integrates Mamba model attention mechanism, deployment on unmanned aerial vehicles (UAVs) detect infected trees. The experimental dataset presented this comprises images trees captured by UAVs mixed forests. were gathered primarily during spring 2023, spanning months February May. subjected preprocessing phase, they transformed into research dataset. comprised three principal components. initial component backbone network with State Space Model (SSM) at its core, capable extracting features high degree efficacy. second network, enables our center PWD greater optimal configuration was determined through an evaluation various mechanism modules, including four modules. third component, Path Aggregation Feature Pyramid Network (PAFPN), facilitates refinement data varying scales, thereby enhancing model’s capacity multi-scale objects. Furthermore, convolutional layers within have been replaced depth separable (DSconv), additional benefit reducing number parameters improving speed. final validated test set, achieving accuracy 90.0%, recall 81.8%, map 86.5%, parameter counts 5.9 Mega, speed 40.16 FPS. In comparison Yolov8, enhanced 7.1%, 5.4%, 3.1%. These outcomes demonstrate appropriate implementation edge devices, such as UAVs, effective PWD.

Language: Английский

---

Algorithm for Detecting Trees Affected by Pine Wilt Disease in Complex Scenes Based on CNN-Transformer

Forests, Journal Year: 2025, Volume and Issue: 16(4), P. 596 - 596

Published: March 28, 2025

Pine wilt disease, a highly destructive forest disease with rapid spread, currently has no effective treatments. Infected pine trees usually die within few months, causing severe damage to ecosystems. A and accurate detection algorithm for diseased is crucial curbing the spread of this disease. In recent years, combination drone remote sensing deep learning become main methods detecting locating trees. Previous studies have shown that increasing network depth cannot improve accuracy in task. Therefore, lightweight semantic segmentation model based on CNN-Transformer hybrid architecture was designed study, named EVitNet. This reduces parameters while improving recognition accuracy, outperforming mainstream models. The IoU discolored reached 0.713, only 1.195 M parameters. Furthermore, considering diverse complex terrain where are distributed, fine-tuning approach adopted. After small amount training, new samples increased from 0.321 0.735, greatly enhancing practicality algorithm. model’s speed task identification meets requirements real-time performance, its exceeds future, it expected be deployed drones recognition, accelerating entire process discovering infected

Language: Английский

---

Deep learning models and methods for solving the problems of remote monitoring of forest resources

Bulletin of the Tomsk Polytechnic University Geo Assets Engineering, Journal Year: 2024, Volume and Issue: 335(6), P. 55 - 74

Published: June 27, 2024

Relevance. The need for precise data analysis in remote monitoring of Earth's forest resources through satellites and unmanned aerial vehicles. Aim. Analysis the current research status via vehicles, formulation directions prospective development this area; implementation investigation new deep learning models analyzing high very high-resolution images coniferous forests. Objects. Hardware, models, methods, information systems, technologies real-time resources, obtained form images. Methods. Deep methods classifying trees images; methodology conducting monitoring; training, validation, convolutional neural networks. Results conclusions. Analytical review data; list formulated tools efficient two Mo-U-Net Mo-Res-U-Net, based on classical U-Net model. Two datasets imagery from an vehicle were created these models. results solving multiclass classification tasks Siberian fir (A. sibirica) pine (P. infested by insect pests. studies showed that unlike model, provide a higher accuracy all classes A. sibirica P. trees, including intermediate classes, with IoU mIoU metrics above threshold value 0.5, indicating practical such forestry industry.

Language: Английский

---

Identification of Pine Wilt-Diseased Trees Using UAV Remote Sensing Imagery and Improved PWD-YOLOv8n Algorithm

Drones, Journal Year: 2024, Volume and Issue: 8(8), P. 404 - 404

Published: Aug. 18, 2024

Pine wilt disease (PWD) is one of the most destructive diseases for pine trees, causing a significant effect on ecological resources. The identification PWD-infected trees an effective approach control. However, effects complex environments and multi-scale features PWD hinder detection performance. To address these issues, this study proposes model based PWD-YOLOv8 by utilizing aerial images. In particular, coordinate attention (CA) convolutional block module (CBAM) mechanisms are combined with YOLOv8 to enhance feature extraction. bidirectional pyramid network (BiFPN) structure used strengthen fusion recognition capability small-scale diseased trees. Meanwhile, lightweight FasterBlock efficient (EMA) mechanism employed optimize C2f module. addition, Inner-SIoU loss function introduced seamlessly improve accuracy reduce missing rates. experiment showed that proposed PWD-YOLOv8n algorithm outperformed conventional target-detection models validation set ([email protected] = 94.3%, precision 87.9%, recall 87.0%, rate 6.6%; size 4.8 MB). Therefore, demonstrates superiority in diseased-tree detection. It not only enhances efficiency but also provides important technical support forest control prevention.

Language: Английский

---