Abstract:Target detection technology based on unmanned aerial vehicle (UAV)-derived aerial imagery has been widely applied in the field of forest fire patrol and rescue. However, due to the specificity of UAV platforms, there are still significant issues to be resolved such as severe omission, low detection accuracy, and poor early warning effectiveness. In light of these issues, this paper proposes an improved YOLOX network for the rapid detection of forest fires in images captured by UAVs. Firstly, to enhance the network's feature-extraction capability in complex fire environments, a multi-level-feature-extraction structure, CSP-ML, is designed to improve the algorithm's detection accuracy for small-target fire areas. Additionally, a CBAM attention mechanism is embedded in the neck network to reduce interference caused by background noise and irrelevant information. Secondly, an adaptive-feature-extraction module is introduced in the YOLOX network's feature fusion part to prevent the loss of important feature information during the fusion process, thus enhancing the network's feature-learning capability. Lastly, the CIoU loss function is used to replace the original loss function, to address issues such as excessive optimization of negative samples and poor gradient-descent direction, thereby strengthening the network's effective recognition of positive samples. Experimental results show that the improved YOLOX network has better detection performance, with mAP@50 and mAP@50_95 increasing by 6.4% and 2.17%, respectively, compared to the traditional YOLOX network. In multi-target flame and small-target flame scenarios, the improved YOLO model achieved a mAP of 96.3%, outperforming deep learning algorithms such as FasterRCNN, SSD, and YOLOv5 by 33.5%, 7.7%, and 7%, respectively. It has a lower omission rate and higher detection accuracy, and it is capable of handling small-target detection tasks in complex fire environments. This can provide support for UAV patrol and rescue applications from a high-altitude perspective.

What problem does this paper attempt to address?

The paper primarily addresses the issue of small object detection in forest fire images captured by Unmanned Aerial Vehicles (UAVs). Specifically, the paper identifies several key problems encountered when using UAVs for forest fire patrol and rescue: 1. **Severe omissions**: In complex environments, especially from high-altitude perspectives, small objects like flames and smoke are difficult to accurately identify. 2. **Low detection accuracy**: Traditional methods have low accuracy in detecting small objects such as forest fires. 3. **Poor early warning effectiveness**: Existing technologies are not effective in early warning. To address these issues, the authors propose an improved version of the YOLOX network (YOLOX stands for "You Only Look Once", a real-time object detection algorithm), named Enhanced Lightweight YOLOX. This network enhances the detection performance of small objects in forest fires through the following methods: 1. **Multi-level feature extraction structure (CSP-ML)**: This structure is designed to enhance the network's feature extraction capability in complex fire environments, thereby improving the detection accuracy of small fire areas. 2. **CBAM attention mechanism**: Embedded in the neck structure of the network, it is used to reduce the interference of background noise and irrelevant information. 3. **Adaptive feature extraction module**: Introduced into the feature fusion part of the YOLOX network to prevent the loss of important feature information during fusion, further enhancing the network's feature learning ability. 4. **CIoU loss function**: Used to replace the original loss function, addressing issues of over-optimization of negative samples and poor gradient descent direction, enhancing the network's effective recognition of positive samples. Experimental results show that the improved YOLOX network significantly outperforms the traditional YOLOX network and some other deep learning algorithms (such as Faster R-CNN, SSD, and YOLOv5) in the task of small object detection in forest fires. Especially in multi-target flame and small target flame scenarios, the improved model's mean Average Precision (mAP) reached 96.3%, which is 33.5% to 7% higher than other algorithms, and it has a lower omission rate and higher detection accuracy. In summary, this paper aims to solve the problem of rapid detection of small forest fire targets in UAV images by proposing an improved lightweight YOLOX network, aiming to provide technical support for UAV patrol and rescue applications.

Enhanced Lightweight YOLOX for Small Object Wildfire Detection in UAV Imagery

FL-YOLOv7: A Lightweight Small Object Detection Algorithm in Forest Fire Detection

LUFFD-YOLO: A Lightweight Model for UAV Remote Sensing Forest Fire Detection Based on Attention Mechanism and Multi-Level Feature Fusion

Tiny-Object Detection Based on Optimized YOLO-CSQ for Accurate Drone Detection in Wildfire Scenarios

UAV forest fire detection based on lightweight YOLOv5 model

Intelligent UAV fire monitoring method based on improved YOLOv5 algorithm

A novel small object detection algorithm for UAVs based on YOLOv5

YOLOv7-FIRE: A tiny-fire identification and detection method applied on UAV

A Wildfire Smoke Detection System Using Unmanned Aerial Vehicle Images Based on the Optimized YOLOv5

Efficient Detection of Forest Fire Smoke in UAV Aerial Imagery Based on an Improved Yolov5 Model and Transfer Learning

Forest Fire Detection and Recognition Using YOLOv8 Algorithms from UAVs Images

UN-YOLOv5s: A UAV-Based Aerial Photography Detection Algorithm

A Small Target Forest Fire Detection Model Based on YOLOv5 Improvement

YOLOv7-P: a lighter and more effective UAV aerial photography object detection algorithm

An Improved Lightweight YOLOv8 Network for Early Small Flame Target Detection

Fire-YOLO: A Small Target Object Detection Method for Fire Inspection

YOLO-Drone: An Optimized YOLOv8 Network for Tiny UAV Object Detection

Small Target Detection Algorithm for UAV Based on Improved YOLOv5

Forest Fire Smoke Detection Based on Deep Learning Approaches and Unmanned Aerial Vehicle Images

Lightweight Small Target Detection Algorithm with Multi-Feature Fusion