An Improved TransMVSNet Algorithm for Three-Dimensional Reconstruction in the Unmanned Aerial Vehicle Remote Sensing Domain.

It is important to achieve the 3D reconstruction of UAV remote sensing images in deep learning-based multi-view stereo (MVS) vision. The lack of obvious texture features and detailed edges in UAV remote sensing images leads to inaccurate feature point matching or depth estimation. To address this problem, this study improves the TransMVSNet algorithm in the field of 3D reconstruction by optimizing its feature extraction network and costumed body depth prediction network. The improvement is mainly achieved by extracting features with the Asymptotic Pyramidal Network (AFPN) and assigning weights to different levels of features through the ASFF module to increase the importance of key levels and also using the UNet structured network combined with an attention mechanism to predict the depth information, which also extracts the key area information. It aims to improve the performance and accuracy of the TransMVSNet algorithm's 3D reconstruction of UAV remote sensing images. In this work, we have performed comparative experiments and quantitative evaluation with other algorithms on the DTU dataset as well as on a large UAV remote sensing image dataset. After a large number of experimental studies, it is shown that our improved TransMVSNet algorithm has better performance and robustness, providing a valuable reference for research and application in the field of 3D reconstruction of UAV remote sensing images.

Resistance of termite mounds to variation in long-term fire regimes across semi-arid African savannas.

Fire regimes are expected to change with climate change, resulting in a crucial need to understand the specific ways in which variable fire regimes impact important contributors to ecosystem functioning, such as mound-building termites. Termite mounds and fire are both important agents of savanna ecosystem heterogeneity and functioning, but there is little understanding of how they interact across savanna types. We used very high-resolution LiDAR remote sensing to measure the size and distribution of termite mounds across approximately 1300 ha of experimental burn plots in four South African savanna landscapes representing a wide range of fire treatments differing in seasonality and frequency of burning. In nutrient-poor granitic savannas, fire had no impact on termite mound size, densities and spatial distributions. In nutrient-rich basaltic savannas with high mammalian herbivore abundance and intermediate rainfall, very frequent fires caused a decrease in termite mound size, whereas in arid nutrient-rich basaltic savannas, fires that occurred at intermediate frequencies and in transitional seasons (i.e. late dry season and late wet season) decreased the degree of spatial overdispersal exhibited by mounds. Overall, our results suggest that termite mounds are resistant to variation in fire seasonality and frequency, likely indicating that ecosystem services provided by mound-building termites will be unaffected by changing fire regimes. However, consideration of changes to termite mound size and distribution could be necessary for land managers in specific savanna types, such as nutrient-rich soils with high mammalian herbivore abundance.

[Research on remote sensing recognition of wild planted Lonicera japonica based on deep convolutional neural network].

Identification of Chinese medicinal materials is a fundamental part and an important premise of the modern Chinese medicinal materials industry. As for the traditional Chinese medicinal materials that imitate wild cultivation, due to their scattered, irregular, and fine-grained planting characteristics, the fine classification using traditional classification methods is not accurate. Therefore, a deep convolution neural network model is used for imitating wild planting. Identification of Chinese herbal medicines. This study takes Lonicera japonica remote sensing recognition as an example, and proposes a method for fine classification of L. japonica based on a deep convolutional neural network model. The GoogLeNet network model is used to learn a large number of training samples to extract L. japonica characteristics from drone remote sensing images. Parameters, further optimize the network structure, and obtain a L. japonica recognition model. The research results show that the deep convolutional neural network based on GoogLeNet can effectively extract the L. japonica information that is relatively fragmented in the image, and realize the fine classification of L. japonica. After training and optimization, the overall classification accuracy of L. japonica can reach 97.5%, and total area accuracy is 94.6%, which can provide a reference for the application of deep convolutional neural network method in remote sensing classification of Chinese medicinal materials.

Research on remote sensing recognition of wild planted Lonicera japonica based on deep convolutional neural network / 中国中药杂志

Identification of Chinese medicinal materials is a fundamental part and an important premise of the modern Chinese medicinal materials industry. As for the traditional Chinese medicinal materials that imitate wild cultivation, due to their scattered, irregular, and fine-grained planting characteristics, the fine classification using traditional classification methods is not accurate. Therefore, a deep convolution neural network model is used for imitating wild planting. Identification of Chinese herbal medicines. This study takes Lonicera japonica remote sensing recognition as an example, and proposes a method for fine classification of L. japonica based on a deep convolutional neural network model. The GoogLeNet network model is used to learn a large number of training samples to extract L. japonica characteristics from drone remote sensing images. Parameters, further optimize the network structure, and obtain a L. japonica recognition model. The research results show that the deep convolutional neural network based on GoogLeNet can effectively extract the L. japonica information that is relatively fragmented in the image, and realize the fine classification of L. japonica. After training and optimization, the overall classification accuracy of L. japonica can reach 97.5%, and total area accuracy is 94.6%, which can provide a reference for the application of deep convolutional neural network method in remote sensing classification of Chinese medicinal materials.