Novel Recursive BiFPN Combining with Swin Transformer for Wildland Fire Smoke Detection

Abstract

The technologies and models based on machine vision are widely used for early wildfire detection. Due to the broadness of wild scene and the occlusion of the vegetation, smoke is more easily detected than flame. However, the shapes of the smoke blown by the wind change constantly and the smoke colors from different combustors vary greatly. Therefore, the existing target detection networks have limitations in detecting wildland fire smoke, such as low detection accuracy and high false alarm rate. This paper designs the attention model Recursive Bidirectional Feature Pyramid Network (RBiFPN for short) for the fusion and enhancement of smoke features. We introduce RBiFPN into the backbone network of YOLOV5 frame to better distinguish the subtle difference between clouds and smoke. In addition, we replace the classification head of YOLOV5 with Swin Transformer, which helps to change the receptive fields of the network with the size of smoke regions and enhance the capability of modeling local features and global features. We tested the proposed model on the dataset containing a large number of interference objects such as clouds and fog. The experimental results show that our model can detect wildfire smoke with a higher performance than the state-of-the-art methods.

MMST: A Multi-Modal Ground-Based Cloud Image Classification Method

“…In this section, we used local binary patterns (LBP) [23] and completed LBP (CLBP) [24], bag-of-visual-words (BoVW) [25], and pyramid BoVW (PBoVW) [26] to conduct experiments on the MGCD dataset to explore the performance of the hand-crafted methods. In the experiments of this subsection, the values of (P, R) of LBP were set to (8,1), (16,2), and (24, 3), respectively. Table 2 illustrates the classification results of these methods on the MGCD dataset.…”

“…The Swin Transformer is proposed to build hierarchical feature maps by merging image patches [16], as shown in Figure 5. Compared with the method of keeping the feature map size invariant in Vit, hierarchical feature maps not only use multi-scale features for modeling, but they can also greatly reduce the complexity of self-attention operations [14].…”

An enhanced YOLOv8 for flame and smoke detection with dilated convolution and image dehazing

“…[8] . More recently, Li designed an attention model, RBiFPN, for the fusion of smoke features [9] . [10] implemented transfer learning on pre-trained models like VGG16, for fire and smoke detection.…”

Improving the Accuracy of Robot Collecting Organisms in Marine Environment Based on Yolov5 Improvement

“…To address this issue, this research replaces the PANet structure with the BiFPN [ 36 ] architecture. Unlike PANet's singular path in both top‐down and bottom‐up directions, BiFPN treats each bi‐directional path as a feature network layer and duplicates the same layer multiple times to achieve a deeper level of feature fusion.…”