Development of a Lightweight Model for Rice Plant Counting and Localization Using UAV-Captured RGB Imagery

Agriculture, Journal Year: 2025, Volume and Issue: 15(2), P. 122 - 122

Published: Jan. 8, 2025

Accurately obtaining both the number and location of rice plants plays a critical role in agricultural applications, such as precision fertilization yield prediction. With rapid development deep learning, numerous models for plant counting have been proposed. However, many these contain large parameters, making them unsuitable deployment settings with limited computational resources. To address this challenge, we propose novel pruning method, Cosine Norm Fusion (CNF), lightweight feature fusion technique, Depth Attention Module (DAFM). Based on innovations, modify existing P2PNet network to create P2P-CNF, model counting. The process begins trained using CNF, followed by integration our module, DAFM. validate effectiveness conducted experiments datasets, including RSC-UAV dataset, captured UAV. results demonstrate that method achieves MAE 3.12 an RMSE 4.12 while utilizing only 33% original parameters. We also evaluated other show high accuracy maintaining architecture.

Language: Английский

---

Weighted Feature Fusion Network Based on Multi-Level Supervision for Migratory Bird Counting in East Dongting Lake

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(5), P. 2317 - 2317

Published: Feb. 21, 2025

East Dongting Lake is an important habitat for migratory birds. Accurately counting the number of birds crucial to assessing health wetland ecological environment. Traditional manual observation and low-precision methods make it difficult meet this demand. To end, paper proposes a weighted feature fusion network based on multi-level supervision (MS-WFFNet) count MS-WFFNet consists three parts: EEMA-VGG16 sub-network, multi-source aggregation (MSFA) module, density map regression (DMR) module. Among them, sub-network cross-injects enhanced efficient multi-scale attention (EEMA) into truncated VGG16 structure. It uses multi-head nonlinearly learn relative importance different positions in same direction. With only few parameters added, EEMA effectively suppresses noise interference caused by cluttered background. The MSFA module integrates mechanism fully preserve low-level detail information high-level semantic information. achieves aggregating features enhancing expression key features. DMR applies output each path ensures local consistency spatial correlation among multiple results using distributed supervision. In addition, presents bird dataset DTH, collected monitoring equipment Lake. combined with other object datasets extensive experiments, showcasing proposed method’s excellent performance generalization capability.

Language: Английский

---

Vision foundation model for agricultural applications with efficient layer aggregation network

Expert Systems with Applications, Journal Year: 2024, Volume and Issue: 257, P. 124972 - 124972

Published: Aug. 10, 2024

Language: Английский

---

Pattern Classification of an Onion Crop (Allium Cepa) Field Using Convolutional Neural Network Models

Agronomy, Journal Year: 2024, Volume and Issue: 14(6), P. 1206 - 1206

Published: June 3, 2024

Agriculture is an area that currently benefits from the use of new technologies and techniques, such as artificial intelligence, to improve production in crop fields. Zacatecas one states producing most onions northeast region Mexico. Identifying determining vegetation, soil, humidity zones could help solve problems irrigation demands or excesses, identify spaces with different levels soil homogeneity, estimate yield health crop. This study examines application intelligence through deep learning, specifically convolutional neural networks, patterns can be found a field, this case, zones. To extract mentioned patterns, K-nearest neighbor algorithm was used pre-process images taken using unmanned aerial vehicles form dataset composed 3672 (1224 for each class). A total six network models were classify namely Alexnet, DenseNet, VGG16, SqueezeNet, MobileNetV2, Res-Net18. Each model evaluated following validation metrics: accuracy, F1-score, precision, recall. The results showed variation performance between 90% almost 100%. Alexnet obtained highest metrics accuracy 99.92%, while MobileNetV2 had lowest 90.85%. Other models, ResNet18, 92.02% 98.78%. Furthermore, our highlights importance adopting agriculture, particularly management onion fields Zacatecas, findings farmers agronomists make more informed efficient decisions, which lead greater sustainability local agriculture.

Language: Английский

---

Automatic Counting and Location of Rice Seedlings in Low Altitude UAV Images Based on Point Supervision

Agriculture, Journal Year: 2024, Volume and Issue: 14(12), P. 2169 - 2169

Published: Nov. 28, 2024

The number of rice seedlings and their spatial distribution are the main agronomic components for determining yield. However, above information is manually obtained through visual inspection, which not only labor-intensive time-consuming but also low in accuracy. To address these issues, this paper proposes RS-P2PNet, automatically counts locates point supervision. Specifically, RS-P2PNet first adopts Resnet as its backbone introduces mixed local channel attention (MLCA) each stage. This allows model to pay task-related feature dimensions avoid interference from background. In addition, a multi-scale fusion module (MSFF) proposed by adding different levels features backbone. It combines shallow details high-order semantic seedlings, can improve positioning accuracy model. Finally, two seedling datasets, UERD15 UERD25, with resolutions, constructed verify performance RS-P2PNet. experimental results show that MAE values reach 1.60 2.43 counting task, compared P2PNet, they reduced 30.43% 9.32%, respectively. localization Recall rates 97.50% 96.67%, exceeding those P2PNet 1.55% 1.17%, Therefore, has effectively accomplished seedlings. RMSE on public dataset DRPD 1.7 2.2, respectively, demonstrating good generalization.

Language: Английский

---