Due to the various shapes, textures, and colors of fires, forest fire detection is a challenging task. The traditional image processing method relies heavily on manmade features, which is not universally applicable to all forest scenarios. In order to solve this problem, the deep learning technology is applied to learn and extract features of forest fires adaptively. However, the limited learning and perception ability of individual learners is not sufficient to make them perform well in complex tasks. Furthermore, learners tend to focus too much on local information, namely ground truth, but ignore global information, which may lead to false positives. In this paper, a novel ensemble learning method is proposed to detect forest fires in different scenarios. Firstly, two individual learners Yolov5 and EfficientDet are integrated to accomplish fire detection process. Secondly, another individual learner EfficientNet is responsible for learning global information to avoid false positives. Finally, detection results are made based on the decisions of three learners. Experiments on our dataset show that the proposed method improves detection performance by 2.5% to 10.9%, and decreases false positives by 51.3%, without any extra latency.

Diseases and insect pests of tea leaves cause huge economic losses to the tea industry every year, so the accurate identification of them is significant. Convolutional neural networks (CNNs) can automatically extract features from images of tea leaves suffering from insect and disease infestation. However, photographs of tea tree leaves taken in a natural environment have problems such as leaf shading, illumination, and small-sized objects. Affected by these problems, traditional CNNs cannot have a satisfactory recognition performance. To address this challenge, we propose YOLO-Tea, an improved model based on You Only Look Once version 5 (YOLOv5). Firstly, we integrated self-attention and convolution (ACmix), and convolutional block attention module (CBAM) to YOLOv5 to allow our proposed model to better focus on tea tree leaf diseases and insect pests. Secondly, to enhance the feature extraction capability of our model, we replaced the spatial pyramid pooling fast (SPPF) module in the original YOLOv5 with the receptive field block (RFB) module. Finally, we reduced the resource consumption of our model by incorporating a global context network (GCNet). This is essential especially when the model operates on resource-constrained edge devices. When compared to YOLOv5s, our proposed YOLO-Tea improved by 0.3%–15.0% over all test data. YOLO-Tea’s ${AP}_{0.5}$, ${AP}_{TLB}$, and ${AP}_{GMB}$ outperformed Faster R-CNN and SSD by 5.5%, 1.8%, 7.0% and 7.7%, 7.8%, 5.2%. YOLO-Tea has shown its promising potential to be applied in real-world tree disease detection systems.

In the previous years, vision transformer has demonstrated a global information extraction capability in the field of computer vision that convolutional neural network (CNN) lacks. Due to the lack of inductive bias in vision transformer, it requires a large amount of data to support its training. In the field of remote sensing, it costs a lot to obtain a significant number of high-resolution remote sensing images. Most existing change detection networks based on deep learning rely heavily on the CNN, which cannot effectively utilize the long-distance dependence between pixels for difference discrimination. Therefore, this work aims to use a high-performance vision transformer to conduct change detection research with limited data. A bibranch fusion network based on axial cross attention (ACABFNet) is proposed. The network extracts local and global information of images through the CNN branch and transformer branch, respectively, and then, fuses local and global features by the bidirectional fusion approach. In the upsampling stage, similar feature information and difference feature information of the two branches are explicitly generated by feature addition and feature subtraction. Considering that the self-attention mechanism is not efficient enough for global attention over small datasets, we propose the axial cross attention. First, global attention along the height and width dimensions of images is performed respectively, and then cross attention is used to fuse the global feature information along two dimensions. Compared with the original self-attention, the structure is more graphics processing unit friendly and efficient. Experimental results on three datasets reveal that the ACABFNet outperforms existing change detection algorithms.

Forest fires are highly unpredictable and extremely destructive. Traditional methods of manual inspection, sensor-based detection, satellite remote sensing and computer vision detection all have their obvious limitations. Deep learning techniques can learn and adaptively extract features of forest fires. However, the small size of the forest fire target in the long-range-captured forest fire images causes the model to fail to learn effective information. To solve this problem, we propose an improved forest fire small-target detection model based on YOLOv5. This model requires cameras as sensors for detecting forest fires in practical applications. First, we improved the Backbone layer of YOLOv5 and adjust the original Spatial Pyramid Pooling-Fast (SPPF) module of YOLOv5 to the Spatial Pyramid Pooling-Fast-Plus (SPPFP) module for a better focus on the global information of small forest fire targets. Then, we added the Convolutional Block Attention Module (CBAM) attention module to improve the identifiability of small forest fire targets. Second, the Neck layer of YOLOv5 was improved by adding a very-small-target detection layer and adjusting the Path Aggregation Network (PANet) to the Bi-directional Feature Pyramid Network (BiFPN). Finally, since the initial small-target forest fire dataset is a small sample dataset, a migration learning strategy was used for training. Experimental results on an initial small-target forest fire dataset produced by us show that the improved structure in this paper improves mAP@0.5 by 10.1%. This demonstrates that the performance of our proposed model has been effectively improved and has some application prospects.

Remote sensing image change detection plays an important role in urban planning and environmental monitoring. However, the existing change detection algorithms have limited ability in feature extraction, feature relationship understanding, and capture of small target features and edge detail features, which leads to the loss of some edge detail information and small target features. To this end, a new dual attention-guided multiscale feature aggregation network is proposed. In encoding stage, the fully convolutional dual-branch structure is used to extract the semantic features of different scales, and then the multi-scale adjacent semantic information aggregation module is used to aggregate the adjacent semantic features at different scales, which can better capture and fuse the features of different scales, thereby improving the accuracy and robustness of change detection. In decoding stage, the dual attention fusion module is proposed to guide and fuse the features extracted from different scales along the spatial and channel directions, and reduce the background noise interference. In addition, this paper also proposes a three-branch feature fusion module and a global semantic information enhancement module to make the network better integrate global semantics and differential semantics and further integrate high-level semantic features. We also introduce an auxiliary classifier in decoding stage to provide additional supervision signals, and fuse the output of the three auxiliary classifiers with the output of the main decoder to further achieve multi-scale feature fusion. The comparative experiments on three remote sensing data sets show that the proposed method is superior to the existing change detection methods.