Research On Forest Fire Risk Monitoring System

Forests are not only a vital strategic resource for social development but also have a significant impact on the conservation of species diversity.Forest fire has always been critical issue forest managers face,and fire risk monitoring is of great significance for fire prevention and resource allocation.Nanjing Laoshan National Forest Park has many precious natural resources and tourist attractions.Its fire prevention methods mainly rely on manual patrols and watchtower monitoring.There is no suitable fire risk assessment,fire detection model or fire risk monitoring system.In this paper,the fire risk monitoring system of Nanjing Laoshan National Forest Park is constructed through three modules: fire risk prediction and assessment,forest fire target detection,and forest road network fire protection design,to improve the management ability of Laoshan forest fire.The main work is as follows:(1)The forest fire risk assessment model of Nanjing Laoshan National Forest Park was constructed.The model consists of factors(elevation,slope aspect,terrain moisture index,slope,distance from roads and residential areas,normalized vegetation index and temperature)that have a more significant impact on the probability of fire in Laoshan.The importance of each factor is inconsistent,and the importance weight of each factor is calculated by the multi-level analysis method,and the consistency is verified.According to the weights,the forest fire risk assessment model is obtained,and then the Laoshan forest fire risk map is drawn.Compared with the MODIS fire anomaly point data,it is found that the accuracy rate of the Laoshan forest fire risk assessment model reaches 76.65%.(2)A ViT-ARGNet model is proposed for object detection of forest fires in medium and high-risk areas.The model uses ALCN-Vi TYOLOX as the backbone network to extract forest fire image features.Recursive Feature Pyramid(RDDFPN)with Deconvolution and Dilated Convolution is connected after ALCN-Vi TYOLOX for feature fusion,and Global Optimal Non-Maximum Suppression(GO-NMS)is used in the post-processing step in object detection to achieve the final detection result.The experimental results show that the number of parameters of Vi T-ARGNet is as low as 57.8M,the m AP reaches 78.3%,the m AP50 reaches 89.45%and,the m AP75 reaches 81.13%,the FPS reaches 134.7,and the GFLOPs reaches 57.34.Compared with other mainstream methods,this method has the advantages of good real-time detection and high accuracy.(3)A new fire-fighting distance criterion,FFDC,is proposed to analyze forest road networks to increase fire-fighting coverage.In this study,multi-criteria evaluation(intensity evaluation standard and absorption evaluation standard)was adopted to evaluate further the forest road network’s intensity and forest absorptive capacity.Do A relatively low-cost is created that covers the maximum effective fire protection area and continues to develop environmental versatility.The results showed that the newly planned forest road network coverage rate increased from37.2% to 57.95%,the strength evaluation was 88.05%,and the absorption evaluation was68.22%.(4)A forest fire risk monitoring system is built,which strictly follows the three-tier structure,which is convenient for managing the services provided by the system.Designed and realized the configuration of solar wireless monitoring,the connection and operation of the system and solar monitoring,fire risk assessment function,forest fire target detection and alarm function,forest road network management,system configuration and other functions,and passed the system functional test and non-functional test.