Comparative Research on Forest Fire Image Segmentation Algorithms Based on Fully Convolutional Neural Networks

Abstract

In recent years, frequent forest fires have plagued countries all over the world, causing serious economic damage and human casualties. Faster and more accurate detection of forest fires and timely interventions have become a research priority. With the advancement in deep learning, fully convolutional network architectures have achieved excellent results in the field of image segmentation. More researchers adopt these models to segment flames for fire monitoring, but most of the works are aimed at fires in buildings and industrial scenarios. However, there are few studies on the application of various fully convolutional models to forest fire scenarios, and comparative experiments are inadequate. In view of the above problems, on the basis of constructing the dataset with remote-sensing images of forest fires captured by unmanned aerial vehicles (UAVs) and the targeted optimization of the data enhancement process, four classical semantic segmentation models and two backbone networks are selected for modeling and testing analysis. By comparing inference results and the evaluation indicators of models such as mPA and mIoU, we can find out the models that are more suitable for forest fire segmentation scenarios. The results show that the U-Net model with Resnet50 as a backbone network has the highest segmentation accuracy of forest fires with the best comprehensive performance, and is more suitable for scenarios with high-accuracy requirements; the DeepLabV3+ model with Resnet50 is slightly less accurate than U-Net, but it can still ensure a satisfying segmentation performance with a faster running speed, which is suitable for scenarios with high real-time requirements. In contrast, FCN and PSPNet have poorer segmentation performance and, hence, are not suitable for forest fire detection scenarios.

AF-Net: An Active Fire Detection Model Using Improved Object-Contextual Representations on Unbalanced UAV Datasets

“…In [10], the published FLAME data set facilitates the development of fire analysis tools for UAV-based fire monitoring systems. This work provides a baseline model of U-Net to extract small fire segments from UAV imagery [30].…”

Semantic segmentation of UAV remote sensing images based on edge feature fusing and multi-level upsampling integrated with Deeplabv3+

“…Based on Dee-plabv1 [33], the researchers have proposed Deeplabv2 [9], Deeplabv3 [34], and Deeplabv3+ [8], which gradually improve the algorithm segmentation performance by optimizing the network structure. Wang et al [10] performed remote sensing of forest fires based on Deeplabv3 + and achieved quite well segmentation performance; Zhang et al [11] achieved promising results in urban land use classification based on Deeplabv3+ and UAV remote sensing technology; Wang et al [12] added a class feature attention mechanism to Deeplabv3+ and achieved high overall segmentation accuracy; Du et al [35] incorporated Deeplabv3+ and object-based image analysis strategy to label remote sensing image, which achieves impressive accuracy.…”

“…The Deeplab semantic segmentation network was improved from FCN and has been developed to Deeplabv3+ [8], which combines the advantages of encoder-decoder structure and spatial pyramid pooling (ASPP) [9] module and has shown an excellent comprehensive performance in semantic segmentation recently. Wang et al [10] investigated the application of Dee-plabv3+ in remote sensing of forest fires and achieved satisfying segmentation performance and running speed; Zhang et al [11] performed urban land use classification based on Dee-plabv3+ and optimized the classification results using the fully connected conditional random field (CRF); Wang et al [12] integrated class feature attention mechanism into Deeplabv3 + and improved the segmentation accuracy, but it still has problems of not being able to accurately segment small targets and having numerous model parameters. The above studies show that Deeplabv3+ performs quite well in semantic segmentation of remote sensing images, but its network structure is complex and requires a lot of computational resources and time to converge during training.…”

Edge AI-Based Tree Trunk Detection for Forestry Monitoring Robotics

“…The early detection of wildfires can help prevent their progress over forest lands, hence reducing their ecological and societal impact. Other ways of detecting forest fires are about segmentation of burned areas in UAV images [8], flame detection [9][10][11], flame segmentation [12][13][14][15] or smoke detection [16] with DL in terrestrial and aerial images. These works can enable an early alert to firefighters about the appearance of a possible forest fire, thus increasing forest protection and, consequently, forest monitoring.…”