Research On Remote Sensing Monitoring And Rescue Path Planning Of Forest Fire In Beijing

Forest fires are known to be one of the world’s eight most impactful natural disasters;they are highly unpredictable,highly destructive,cause severe damage to the atmosphere,and have a profound impact on human habitat.Therefore,in the face of harsh reality,monitoring forest fires,determining the relevant weight factors for potential fire points,and planning fire rescue paths are of practical and urgent significance.However,traditional forest fire monitoring methods have shortcomings,such as relying on manual operations,low efficiency,high cost,and limited coverage.Moreover,quickly reaching the fire scene is the top priority of disaster relief work in the complex forest environment.Based on the above practical problems,this study proposes a remote sensing monitoring and rescue path planning method for forest fires,which is realized by using multi-scale remote sensing images,deep learning technology,and knowledge in the field of path planning.Firstly,use medium-resolution MODIS images for high-frequency remote sensing monitoring of large-scale fire conditions,and then use medium-resolution and high-resolution satellite remote sensing images combined with relevant research areas in Beijing to use DEM slope and NDVI as weighting factors for fire rescue path planning for the occurrence area.The main contributions of this paper are as follows:1)This study proposes a remote sensing forest fire recognition neural network model named VTRSN,which mainly uses the residual convolutional neural network as the backbone network and combines the visual conversion module of the multi-head attention mechanism,aiming to make full use of the general and local information of the image to improve the recognition performance of the model.Experimental results show that the accuracy,Kappa coefficient,OE,and CE of the VTRSN model are 80.51,0.8254,30.58,and 11.61,respectively.By comparison,the method used in this paper can have a better classification effect than ResNet and ViT alone.2)According to the characteristics of mountain forest fires,NDVI and slope weight factors are introduced,and the optimized path is obtained by improving the A~*path planning algorithm.This algorithm adopts the simplest space-for-time method: for a given pixel(x,y)in the area,calculate the distance between it and the nearest road pixel and build an estimated grid to cache all the distance between the pixel and the road.After clarifying the relevant data in the OPEN table,it is arranged and stored in the form of a small root heap,and after the node at the top of the heap is obtained,the node with the minimum cost value has been found.The experiment uses the OpenCV open-source computer vision library,the Gaussian operator to smooth and filter the image,and the Laplacian operator for convolution filtering.When the path planning program is started,the program automatically calls the load method to load the previously saved grid,coordinate set data,and search for the pathfinding starting point.The cost function in the experiment is jointly determined by the NDVI factor Factor＿NDVI and the slope size factor Factor＿G.In order to explore a suitable cost function,it is necessary to adjust the mixing ratio of Factor＿NDVI and Factor＿G and conduct comparative experiments.Experiments show that the planning of forest fire rescue paths is highly feasible.The path calculated by different weights is displayed in the production function image.The change in the path length shows that when the weight is gradually increased from 0 to 0.3,it reaches the maximum value and then begins to decline.Therefore,when using the improved A~* algorithm for path planning and superimposing DEM data and NDVI data for path calculation,when the weight of NDVI exceeds 0.3,the higher the weight,the shorter the calculated path will be.Finally,the best rescue path based on mountain fire rescue can be obtained.