Problems: Tomato Spotted Wilt Virus (TSWV) severely affects tobacco yield and quality, creating an urgent need for accurate, rapid, non-destructive monitoring to support disease management. While existing TSWV detection methods perform well at the leaf scale, their field-scale application remains challenging. Due to complex crop canopy structures, spectral characteristics at the field level differ significantly from leaf-level observations, and TSWV-sensitive spectral features are still unclear. This study therefore aims to develop a field-scale TSWV identification model using UAV-based hyperspectral imaging to enable targeted disease control. Methodology: A UAV-mounted hyperspectral camera (400–1000 nm) was deployed to capture imagery of tobacco plants at the rosette stage, enabling comparative spectral analysis between healthy and infected specimens. To identify sensitive features associated with tobacco plants infected with TSWV, six distinct feature extraction methodologies encompassing traditional statistical approaches (spectral ratio, correlation analysis, and principal component analysis [PCA]), machine learning-based techniques (relevant features [Relief], successive projections algorithm) and vegetation indices were utilized. Subsequently, we conducted a systematic evaluation of 18 classification models developed using three machine learning algorithms—support vector machine (SVM), k-nearest neighbors, and extreme gradient boosting —with the derived feature variables. Results: This study demonstrates that while all integrated models combining Relief- and Correlation- selected feature bands with three machine learning algorithms delivered excellent performance, the SVM-Relief model achieved the most outstanding results (OA = 97.3%, AUC = 0.994, Kappa=0.947). Based on the SVM-Relief combination, a proposed method called RPR —which integrates PCA with recursive feature elimination— was further employed to reduce the number of feature indicators from 15 to 4 (775.6/772.9/781.1/756.4 nm). The resulting SVM-RPR combination model achieved performance (OA = 97.3%, AUC = 0.990, Kappa=0.947) comparable to that of the SVM-Relief model. Contribution: This indicated that red-edge bands were of significant value in distinguishing healthy and TSWV-infected tobacco plants. Our study indicates the significant potential of integrating UAV-based hyperspectral imaging with machine learning techniques for rapid, non-destructive detection of tobacco TSWV at the field scale. The proposed approach offers a novel and efficient pathway for remote sensing-based monitoring of viral diseases in crops, with implications for precision agriculture and plant disease management.

Birdsong is a valuable indicator of rich biodiversity and ecological significance. Although feature extraction has demonstrated satisfactory performance in classification, single-scale feature extraction methods may not fully capture the complexity of birdsong, potentially leading to suboptimal classification outcomes. The integration of multi-scale feature extraction and fusion enables the model to better handle scale variations, thereby enhancing its adaptability across different scales. To address this issue, we propose a multi-scale hybrid convolutional attention mechanism model (MUSCA). This method combines depthwise separable convolution and traditional convolution for feature extraction and incorporates self-attention and spatial attention mechanisms to refine spatial and channel features, thereby improving the effectiveness of multi-scale feature extraction. To further enhance multi-scale feature fusion, a layer-by-layer alignment feature fusion method is developed to establish a deeper correlation, thereby improving classification accuracy and robustness. Using the above method, we identified 20 bird species on three spectrograms, wavelet spectrogram, log-Mel spectrogram and log-spectrogram, with recognition rates of 93.79%, 96.97% and 95.44%, respectively. Compared with the resnet18 model, it increased by 3.26%, 1.88% and 3.09%, respectively. The results indicate that the MUSCA method proposed in this paper is competitive compared to recent and state-of-the-art methods.

Revealing difference in bird vocalization changes from the perspectives of song recognition and acoustic indices has become a hot topic and challenge in recent ecological landscape research. This paper proposes a fine-grained (Dawn, noon, night) bird vocalization recognition framework based on a two-layer deep network to identify the same species’ bird vocalization at different times of the day. Additionally, a new acoustic index method, the Log-Mel Acoustic Complexity Index (Log-Mel ACI), is introduced to explore the differences in bird vocalization of the same species throughout the day. The results of two-layer deep network showed significant separability of the bird vocalization of the same species at dawn, noon, and night based on Log-Mel spectrum. Furthermore, it was found that the improved ACI based on Log-Mel exhibits better circadian rhythmic performance than the traditional ACI, being highest at dawn, followed by night, and lowest at noon. These findings demonstrate that Log-Mel is effective in both deep network recognition and ACI calculation.

Forest fires cause severe ecological and economic losses, so timely and accurate detection becomes crucial for effective prevention and control. Edge devices with intelligent algorithms can detect forest fires in real time. Current deep learning algorithms can achieve high accuracy, but they are not suitable for edge devices because they require substantial computing resources. To address this issue, this study proposes a real-time lightweight forest fire detection network (RLFNet) improved from YOLOv11n, with three key enhancements to the backbone, neck, and head. (1) A Parallel Multi-Scale Extraction Block (PMEB) improves C3k2 with a dual-branch parallel strategy to enhance multi-scale feature extraction efficiency; (2) a Bidirectional Cross Fusion Module (BCFM) replaces simple Concat with a context-aware cross-gating mechanism to suppress background noise and reduce false alarms; and (3) a Faster Inference Detection Head (FIDH) leverages structural re-parameterization and group normalization to boost inference efficiency while reducing parameters. In addition, a Layer-Adaptive Magnitude-based Pruning (LAMP) strategy is applied to further improve model’s computational efficiency. Experimental results on the self-constructed Diverse Fire Scenario (DFS) dataset demonstrate that RLFNet reduces parameters and GFLOPs by 25.2% and 20.6%, boosts mAP50 by 5.3%, and achieves an inference speed of 225 FPS, attaining the best accuracy and speed among the compared models. Validation on a public remote sensing dataset further confirms its strong generalization. These results indicate that RLFNet provides a high efficiency and lightweight solution for edge devices to real-time detect forest fires.