FID-Net: A Feature-Enhanced Deep Learning Network for Forest Infestation Detection

Forest pests threaten ecosystem stability, requiring efficient monitoring. To overcome the limitations of traditional methods in large-scale, fine-grained detection, this study focuses on accurately identifying infected trees and analyzing infestation patterns. We propose FID-Net, a deep learning model that detects pest-affected trees from UAV visible-light imagery and enables infestation analysis via three spatial metrics. Based on YOLOv8n, FID-Net introduces a lightweight Feature Enhancement Module (FEM) to extract disease-sensitive cues, an Adaptive Multi-scale Feature Fusion Module (AMFM) to align and fuse dual-branch features (RGB and FEM-enhanced), and an Efficient Channel Attention (ECA) mechanism to enhance discriminative information efficiently. From detection results, we construct a pest situation analysis framework using: (1) Kernel Density Estimation to locate infection hotspots; (2) neighborhood evaluation to assess healthy trees’ infection risk; (3) DBSCAN clustering to identify high-density healthy clusters as priority protection zones. Experiments on UAV imagery from 32 forest plots in eastern Tianshan, China, show that FID-Net achieves 86.10% precision, 75.44% recall, 82.29% mAP@0.5, and 64.30% mAP@0.5:0.95, outperforming mainstream YOLO models. Analysis confirms infected trees exhibit clear clustering, supporting targeted forest protection. FID-Net enables accurate tree health discrimination and, combined with spatial metrics, provides reliable data for intelligent pest monitoring, early warning, and precise management.

💡 Research Summary

This paper addresses the pressing need for accurate, large‑scale monitoring of forest pests and diseases by developing a UAV‑based detection and analysis system called FID‑Net. The authors first highlight the shortcomings of traditional ground surveys, aerial visual inspections, and satellite remote sensing—namely limited spatial/temporal resolution, high labor costs, and poor sensitivity to early‑stage symptoms. To overcome these issues, they collect high‑resolution visible‑light imagery using a DJI Mavic 3 Pro equipped with a 20 MP Hasselblad camera over 32 plots in the eastern Tianshan Mountains (China). After orthorectification and quality control, 312 high‑quality 1024 × 1024 tiles containing 40,241 annotated tree instances (infected vs. healthy) form the East Tianshan Tree Detection Dataset (ET‑TDD).

The detection backbone is YOLOv8n, a lightweight one‑stage detector. The authors augment it with three novel modules:

1. Feature Enhancement Module (FEM) – a parallel branch that processes the RGB input with 1 × 1 and 3 × 3 convolutions, explicitly learning pest‑sensitive color and texture cues while adding less than 4 % extra parameters.

2. Adaptive Multi‑scale Feature Fusion Module (AMFM) – aligns and fuses multi‑scale features from the original backbone and the FEM‑enhanced branch using learned scale‑wise attention weights, preserving information for both small and large trees.

3. Efficient Channel Attention (ECA) – a 1‑D convolution‑based channel attention that highlights discriminative channels with negligible computational overhead.

Together, these components increase the model’s capacity without sacrificing real‑time performance (≈30 FPS). On the test split, FID‑Net achieves 86.10 % precision, 75.44 % recall, 82.29 % mAP@0.5, and 64.30 % mAP@0.5:0.95, outperforming YOLOv8s/m/l by 4–7 % absolute mAP. Ablation studies confirm that each module contributes 1.5–2.3 % mAP improvement, and the full combination yields the best results.

Beyond detection, the authors propose a comprehensive spatial‑analysis pipeline built on the detection outputs:

• Kernel Density Estimation (KDE) maps the density of infected trees, revealing high‑density hotspots that indicate active outbreak zones.
• Neighborhood risk assessment computes a risk score for each healthy tree based on the number of infected neighbors within a 30 m radius, providing an early‑warning layer for potential spread.
• DBSCAN clustering groups healthy trees into high‑density clusters; these clusters are designated as priority protection areas where preventive measures should be concentrated.

These three metrics together transform raw detection counts into actionable forest‑management intelligence, enabling targeted surveillance, resource allocation, and early intervention.

The study acknowledges limitations: reliance solely on visible‑light imagery restricts spectral information; the dataset is confined to a specific geographic area and two conifer species (Siberian larch and Schrenk spruce), which may affect generalization; and detection of extremely subtle early symptoms remains challenging. Future work is suggested to incorporate multispectral/hyperspectral and LiDAR data for richer feature representation, and to develop temporal models using sequential UAV flights for disease progression forecasting.

In summary, FID‑Net delivers a lightweight yet highly accurate solution for UAV‑based forest pest detection, and its integrated spatial analysis framework provides quantitative guidance for precision forest protection. The approach sets a new benchmark for real‑time, fine‑grained monitoring of forest health at landscape scales.