Live fuel moisture content (LFMC) is an important index used to evaluate the wildfire risk and fire spread rate. In order further improve retrieval accuracy, two ensemble models combining deep learning were proposed. One a stacking model based on LSTM, TCN LSTM-TCN models, other Adaboost model. Measured LFMC data, MODIS, Landsat-8, Sentinel-1 remote sensing data auxiliary such as canopy height land cover of forest-fire-prone areas in Western United States, selected for our study, results...

As one of the more difficult problems in field computer vision, utilizing object image detection technology a complex environment includes other key technologies, such as pattern recognition, artificial intelligence, and digital processing. However, because an can be complex, changeable, highly different, easily confused with target, target is affected by factors, insufficient light, partial occlusion, background interference, etc., making multiple targets extremely robustness algorithm low....

The Internet of Things (IoT)-based passive acoustic monitoring (PAM) has shown great potential in large-scale remote bird monitoring. However, field recordings often contain overlapping signals, making precise information extraction challenging. To solve this challenge, first, the inter-channel spatial feature is chosen as complementary to spectral obtain additional correlations between sources. Then, an end-to-end model named BACPPNet built based on Deeplabv3plus and enhanced with polarized...

Deep convolutional neural networks (DCNNs) have achieved breakthrough performance on bird species identification using a spectrogram of vocalization. Aiming at the imbalance vocalization dataset, single feature model (SFIM) with residual blocks and modified, weighted, cross-entropy function was proposed. To further improve accuracy, two multi-channel fusion methods were built three SFIMs. One these fused outputs extraction parts SFIMs (feature mode), other classifiers (result mode). The...

Recognizing wildlife based on camera trap images is challenging due to the complexity of wild environment. Deep learning an optional approach solve this problem. However, backgrounds captured from same infrared are rather similar, and shortcut recognition models occurs, resulting in reduced generality poor model performance. Therefore, paper proposes a data augmentation strategy that integrates image synthesis (IS) regional background suppression (RBS) enrich scene suppress existing...

In the wild, bird vocalizations of same species across different populations may be (e. g., so called dialect). Besides, number is unknown in advance. These two facts make task recognition based on vocalization a challenging one. This study treats this as an open set (OSR) cross-corpus scenario. We propose Instance Frequency Normalization (IFN) to remove instance-specific differences corpora. Furthermore, x-vector feature extraction model integrated Time Delay Neural Network (TDNN) and Long...

Infrared camera trapping, which helps capture large volumes of wildlife images, is a widely-used, non-intrusive monitoring method in surveillance. This can greatly reduce the workload zoologists through automatic image identification. To achieve higher accuracy recognition, integrated model based on multi-branch aggregation and Squeeze-and-Excitation network introduced. adopts transformation to extract features, uses block adaptively recalibrate channel-wise feature responses explicit...

Abstract Bird-caused damages for overhead transmission lines become one of the main fault types power grid, which will affect safety and reliability grid operation usually bring huge economic losses. The effectiveness existing bird repeller would deteriorate obviously due to habituation birds. Bird bioacoustic vocalization can elicit avoidance behaviour conspecific, delay process birds repeller. An intelligent based on species variations was proposed. Based birds’ appearance respectively,...

Based on the transfer learning, we design a bird species identification model that uses VGG-16 (pretrained ImageNet) for feature extraction, then classifier consisting of two fully-connected hidden layers and Softmax layer is attached. We compare performance proposed with original VGG16 model. The results show former has higher train efficiency, but lower mean average precisions(MAP). To improve MAP model, investigate result fusion mode to form multi-channel best reaches 0.9998. number...

As the population and distribution of Crested Ibis (Nipponia nippon) become larger, it is necessary to propose a highly efficient census method estimate size Ibis. Passive acoustic monitoring (PAM) has very good prospect for monitoring. To realize automatic with PAM, individual identification based on vocalization key technology. A novel model was proposed in this paper, which built autoencoder LSTM obtain meaningful latent representation from raw recording directly, further, embedded...

We propose a novel Dynamic Restrained Uncertainty Weighting Loss to experimentally handle the problem of balancing contributions multiple tasks on ICML ExVo 2022 Challenge. The multitask aims recognize expressed emotions and demographic traits from vocal bursts jointly. Our strategy combines advantages Weight Average, by extending weights with restraint term make learning process more explainable. use lightweight multi-exit CNN architecture implement our proposed loss approach. experimental...

Currently, wireless acoustic sensor networks (WASN) are commonly used for wild bird monitoring. To better realize the automatic identification of birds during monitoring, enhancement audio is essential in nature. distributed beamformer most suitable method WASN. However, there still several disadvantages this method, such as large noise residue and slow convergence rate. overcome these shortcomings, an improved minimum variance distortionless response (IDMVDR) beamforming WASN proposed...