Detecting small forest fire targets in unmanned aerial vehicle (UAV) images is difficult, as flames typically cover only a very limited portion of the visual scene. This study proposes Context-guided Compact Lightweight Network (CCLNet), an end-to-end lightweight model designed to detect small forest fire targets while ensuring efficient inference on devices with constrained computational resources. CCLNet employs a three-stage network architecture. Its key components include three modules. C3F-Convolutional Gated Linear Unit (C3F-CGLU) performs selective local feature extraction while preserving fine-grained high-frequency flame details. Context-Guided Feature Fusion Module (CGFM) replaces plain concatenation with triplet-attention interactions to emphasize subtle flame patterns. Lightweight Shared Convolution with Separated Batch Normalization Detection (LSCSBD) reduces parameters through separated batch normalization while maintaining scale-specific statistics. We build TF-11K, an 11,139-image dataset combining 9139 self-collected UAV images from subtropical forests and 2000 re-annotated frames from the FLAME dataset. On TF-11K, CCLNet attains 85.8% mAP@0.5, 45.5% mean Average Precision (mAP)@[0.5:0.95], 87.4% precision, and 79.1% recall with 2.21 M parameters and 5.7 Giga Floating-point Operations Per Second (GFLOPs). The ablation study confirms that each module contributes to both accuracy and efficiency. Cross-dataset evaluation on DFS yields 77.5% mAP@0.5 and 42.3% mAP@[0.5:0.95], indicating good generalization to unseen scenes. These results suggest that CCLNet offers a practical balance between accuracy and speed for small-target forest fire monitoring with UAVs.
- Article type
- Year
- Co-author
Open Access
Article
Issue
As forests are an indispensable component of terrestrial ecosystems, accurate and effective estimation of forest carbon stocks is an important basis for effectively responding to climate change and achieving the goal of carbon neutrality.
Taking the arboreal forests in Changsha City, Hunan Province, as the research object, the arboreal forests in Changsha City were categorized according to the dominant species and species groups into fir group, masson pine group, camphor group, foreign pine group, oak group and maple. Landsat-8 remote sensing data were used to extract band information, texture characteristics, vegetation index and topographic factors to obtain 80 modeling factors. Combined with the data from the Third National Land Survey of Changsha City, Multiple linear regression (MLR), Support vector regression (SVR) and Random forest (RF) algorithms were used to construct forest carbon stock inversion models for different dominant tree species and species groups.
The coefficients of determination of the RF model (R2=0.933 9-0.967 9) were significantly higher than those of the MLR model (R2=0.011 8-0.584 5) and the SVR model (R2=0.229 7-0.904 1), and the RMSE and MAE were significantly lower. Comparing the inversion results, the spatial distribution values of forest carbon stock without inversion of dominant tree species and species group classification ranged from 13.68-40.98 t·hm-2, while the spatial distribution values of forest carbon stock after inversion by dominant tree species and species group classification ranged from 6.03-57.98 t·hm-2 and forest carbon stock in Changsha City in 2020 was 4.946 8 Tg.
The forest carbon stock inversion model constructed by the RF algorithm after the classification of dominant tree species and species groups eliminated the problems of over-fitting and underestimation of the peak value when estimating under unclassified conditions, and provided a reference for the remote sensing inversion of forest carbon stock on a large scale.
京公网安备11010802044758号