Tea diseases have a significant impact on the yield and quality of tea during the growth of tea trees. The shape and scale of tea diseases are variable, and the tea disease targets are usually small, with the intelligent detection processes of tea diseases also easily disturbed by the complex background of the growing region. In addition, some tea diseases are concentrated in the entire area of the leaves, needing to be inferred from global information. Common target detection models are difficult to solve these problems. Therefore, we proposed an improved tea disease detection model called TSBA-YOLO. We use the dataset of tea diseases collected at the Maoshan Tea Factory in China. The self-attention mechanism was used to enhance the ability of the model to obtain global information on tea diseases. The BiFPN feature fusion network and adaptively spatial feature fusion (ASFF) technology were used to improve the multiscale feature fusion of tea diseases and enhance the ability of the model to resist complex background interference. We integrated the Shuffle Attention mechanism to solve the problem of difficult identifications of small-target tea diseases. In addition, we used data-enhancement methods and transfer learning to expand the dataset and relocate the parameters learned from other plant disease datasets to enhance tea diseases detection. Finally, SIoU was used to further improve the accuracy of the regression. The experimental results show that the proposed model is good at solving a series of problems encountered in the intelligent recognition of tea diseases. The detection accuracy is ahead of the mainstream target detection models, and the detection speed reaches the real-time level.

Forest fires occur frequently around the world, causing serious economic losses and human casualties. Deep learning techniques based on convolutional neural networks (CNN) are widely used in the intelligent detection of forest fires. However, CNN-based forest fire target detection models lack global modeling capabilities and cannot fully extract global and contextual information about forest fire targets. CNNs also pay insufficient attention to forest fires and are vulnerable to the interference of invalid features similar to forest fires, resulting in low accuracy of fire detection. In addition, CNN-based forest fire target detection models require a large number of labeled datasets. Manual annotation is often used to annotate the huge amount of forest fire datasets; however, this takes a lot of time. To address these problems, this paper proposes a forest fire detection model, TCA-YOLO, with YOLOv5 as the basic framework. Firstly, we combine the Transformer encoder with its powerful global modeling capability and self-attention mechanism with CNN as a feature extraction network to enhance the extraction of global information on forest fire targets. Secondly, in order to enhance the model’s focus on forest fire targets, we integrate the Coordinate Attention (CA) mechanism. CA not only acquires inter-channel information but also considers direction-related location information, which helps the model to better locate and identify forest fire targets. Integrated adaptively spatial feature fusion (ASFF) technology allows the model to automatically filter out useless information from other layers and efficiently fuse features to suppress the interference of complex backgrounds in the forest area for detection. Finally, semi-supervised learning is used to save a large amount of manual labeling effort. The experimental results show that the average accuracy of TCA-YOLO improves by 5.3 compared with the unimproved YOLOv5. TCA-YOLO also outperformed in detecting forest fire targets in different scenarios. The ability of TCA-YOLO to extract global information on forest fire targets was much improved. Additionally, it could locate forest fire targets more accurately. TCA-YOLO misses fewer forest fire targets and is less likely to be interfered with by forest fire-like targets. TCA-YOLO is also more focused on forest fire targets and better at small-target forest fire detection. FPS reaches 53.7, which means that the detection speed meets the requirements of real-time forest fire detection.

The frequent occurrence of forest fires in recent years has not only seriously damaged the forests’ ecological environments but also threatened the safety of public life and property. Smoke, as the main manifestation of the flame before it is produced, has the advantage of a wide diffusion range that is not easily obscured. Therefore, timely detection of forest fire smoke with better real-time detection for early warnings of forest fires wins valuable time for timely firefighting and also has great significance and applications for the development of forest fire detection systems. However, existing forest fire smoke detection methods still have problems, such as low detection accuracy, slow detection speed, and difficulty detecting smoke from small targets. In order to solve the aforementioned problems and further achieve higher accuracy in detection, this paper proposes an improved, new, high-accuracy forest fire detection model, the OBDS. Firstly, to address the problem of insufficient extraction of effective features of forest fire smoke in complex forest environments, this paper introduces the SimAM attention mechanism, which makes the model pay more attention to the feature information of forest fire smoke and suppresses the interference of non-targeted background information. Moreover, this paper introduces Omni-Dimensional Dynamic Convolution instead of static convolution and adaptively and dynamically adjusts the weights of the convolution kernel, which enables the network to better extract the key features of forest fire smoke of different shapes and sizes. In addition, to address the problem that traditional convolutional neural networks are not capable of capturing global forest fire smoke feature information, this paper introduces the Bottleneck Transformer Net (BoTNet) to fully extract global feature information and local feature information of forest fire smoke images while improving the accuracy of small target forest fire target detection of smoke, effectively reducing the model’s computation, and improving the detection speed of model forest fire smoke. Finally, this paper introduces the decoupling head to further improve the detection accuracy of forest fire smoke and speed up the convergence of the model. Our experimental results show that the model OBDS for forest fire smoke detection proposed in this paper is significantly better than the mainstream model, with a computational complexity of 21.5 GFLOPs (giga floating-point operations per second), an improvement of 4.31% compared with the YOLOv5 (YOLO, you only look once) model mAP@0.5, reaching 92.10%, and an FPS (frames per second) of 54, which is conducive to the realization of early warning of forest fires.

Forest fires seriously destroy the world’s forest resources and endanger biodiversity. The traditional forest fire target detection models based on convolutional neural networks (CNNs) lack the ability to deal with the relationship between visual elements and objects. They also have low detection accuracy for small-target forest fires. Therefore, this paper proposes an improved small-target forest fire detection model, STPM_SAHI. We use the latest technology in the field of computer vision, the Swin Transformer backbone network, to extract the features of forest fires. Its self-attention mechanism can capture the global information of forest fires to obtain larger receptive fields and contextual information. We integrated the Swin Transformer backbone network into the Mask R-CNN detection framework, and PAFPN was used to replace the original FPN as the feature fusion network, which can reduce the propagation path of the main feature layer and eliminate the impact of down-sampling fusion. After the improved model was trained, the average precision (AP0.5) of forest fire target detection at different scales reached 89.4. Then, Slicing Aided Hyper Inference technology was integrated into the improved forest fire detection model, which solved the problem that small-target forest fires pixels only account for a small proportion and lack sufficient details, which are difficult to be detected by the traditional target detection models. The detection accuracy of small-target forest fires was significantly improved. The average precision (AP0.5) increased by 8.1. Through an ablation experiment, we have proved the effectiveness of each module of the improved forest fire detection model. Furthermore, the forest fire detection accuracy is significantly better than that of the mainstream models. Our model can also detect forest fire targets with very small pixels. Our model is very suitable for small-target forest fire detection. The detection accuracy of forest fire targets at different scales is also very high and meets the needs of real-time forest fire detection.

No abstract