Pine wilt disease (PWD) is a forest disease characterized by rapid spread and extremely high lethality, posing a serious threat to the ecological security of China’s forests and causing significant economic losses in forestry. Given the extensive forestry area, limited personnel for inspection and monitoring, and high costs, utilizing UAV-based remote sensing monitoring for diseased trees represents an effective approach for controlling the spread of PWD. However, due to the small target size and uneven scale of pine wilt disease, as well as the limitations of real-time detection by drones, traditional disease tree detection algorithms based on RGB remote sensing images do not achieve an optimal balance among accuracy, detection speed, and model complexity due to real-time detection limitations. Consequently, this paper proposes Light-ViTeYOLO, a lightweight pine wilt disease detection method based on Vision Transformer-enhanced YOLO (You Only Look Once). A novel lightweight multi-scale attention module is introduced to construct an EfficientViT feature extraction network for global receptive field and multi-scale learning. A novel neck network, CACSNet(Content-Aware Cross-Scale bidirectional fusion neck network), is designed to enhance the detection of diseased trees at single granularity, and the loss function is optimized to improve localization accuracy. The algorithm effectively reduces the number of parameters and giga floating-point operations per second (GFLOPs) of the detection model while enhancing overall detection performance. Experimental results demonstrate that compared with other baseline algorithms, Light-ViTeYOLO proposed in this paper has the least parameter and computational complexity among related algorithms, with 3.89 MFLOPs and 7.4 GFLOPs, respectively. The FPS rate is 57.9 (frames/s), which is better than the original YOLOv5. Meanwhile, its mAP@0.5:0.95 is the best among the baseline algorithms, and the recall and mAP@0.5 slightly decrease. Our Light-ViTeYOLO is the first lightweight method specifically designed for detecting pine wilt disease. It not only meets the requirements for real-time detection of pine wilt disease outbreaks but also provides strong technical support for automated forestry work.

Crop yield estimation plays a crucial role in agricultural production planning and risk management. Utilizing simultaneous localization and mapping (SLAM) technology for the three-dimensional reconstruction of crops allows for an intuitive understanding of their growth status and facilitates yield estimation. Therefore, this paper proposes a VINS-RGBD system incorporating a semantic segmentation module to enrich the information representation of a 3D reconstruction map. Additionally, image matching using L_SuperPoint feature points is employed to achieve higher localization accuracy and obtain better map quality. Moreover, Voxblox is proposed for storing and representing the maps, which facilitates the storage of large-scale maps. Furthermore, yield estimation is conducted using conditional filtering and RANSAC spherical fitting. The results show that the proposed system achieves an average relative error of 10.87% in yield estimation. The semantic segmentation accuracy of the system reaches 73.2% mIoU, and it can save an average of 96.91% memory for point cloud map storage. Localization accuracy tests on public datasets demonstrate that, compared to Shi–Tomasi corner points, using L_SuperPoint feature points reduces the average ATE by 1.933 and the average RPE by 0.042. Through field experiments and evaluations in a strawberry field, the proposed system demonstrates reliability in yield estimation, providing guidance and support for agricultural production planning and risk management.

Self-supervised learning, due to its label-independent nature, has been widely applied in the field of deep learning. However, existing methods still face limitations in semantic information extraction and model compression within the convolutional neural network channels. This article proposes a method combining self-supervised learning distillation and pruning to optimize channel importance identification and model performance. By leveraging self-supervised learning to capture channel features, the method identifies critical channels and deactivates redundant ones, effectively decoupling semantic information and significantly improving accuracy and efficiency in subclassification tasks. Specifically, the proposed approach evaluates channel importance through a gated mechanism, constructing task-specific efficient subnetworks while employing structured pruning to reduce computational resource consumption. Experiments conducted on the CIFAR-10 dataset validate the proposed method, demonstrating its ability to significantly reduce model parameters and improve subclassification accuracy while maintaining high overall accuracy.