UAV-borne LiDAR Route Optimization Method For Densely Vegetated Areas

The Southwest region is prone to frequent landslides,which have sudden,multiple,and group characteristics.The occurrence of landslides poses a serious threat to people’s lives and property,and the prevention and monitoring of landslides face enormous challenges.Terrain data is an important foundation for disaster prevention and reduction.Drones have been widely used in terrain mapping due to their high mobility,flexible and convenient use,and low cost.In particular,unmanned aerial vehicle(UAV)threedimensional laser scanning technology,which can operate 24/7 and penetrate vegetation,has become an important means of obtaining accurate terrain information in disaster prevention and control work.However,due to the complexity of China’s terrain,landslides often occur in complex areas with high vegetation cover,such as high-altitude mountains and gorges.Conventional flight route planning methods that do not consider surface features result in excessive redundant data and sparse ground point data,which have an adverse effect on later detailed terrain extraction.Therefore,this thesis explores and studies the UAV three-dimensional laser scanning(Light Detection and Ranging,Li DAR)system flight route planning method in vegetation-covered areas.The aim is to improve the quality of surface data acquisition,provide powerful technical support for obtaining more effective surface points in landslide areas and constructing accurate terrain.This thesis takes the landslide in Guang’an Village,Dahe Township,Wuxi County,Chongqing City as the research area.Based on the current situation where UAV flight path planning usually only considers factors such as operating time,plane and elevation accuracy,it is difficult to solve the problem of sparse ground laser points in vegetationcovered areas and the subsequent refined terrain extraction.The study focuses on the UAV Li DAR adaptive route planning method in landslide areas with dense vegetation.Firstly,based on high-resolution remote sensing images and the need for high-precision terrain acquisition in vegetation-covered areas,the "fuzzy" calculation method of vegetation coverage based on visible light band is explored on the basis of random forest algorithm.Secondly,the classic ant colony algorithm is optimized based on the "fuzzy" calculation results of vegetation coverage,and the adaptive route planning method for UAV Li DAR in vegetation-dense areas is studied based on three contents of the classic ant colony algorithm.Then,considering the flight speed of UAVs,flight height,data acquisition mode of three-dimensional laser scanners,and the size of remote sensing images in the research area,a route planning vector environment with different node numbers at different locations is established.Finally,the adaptive route planning method for UAV Li DAR in vegetation-dense areas is verified in different route planning vector environments.The main research results are as follows::(1)"Fuzzy" calculation method of vegetation coverageGiven the vegetation-penetrating ability of unmanned aerial vehicle(UAV)-borne3 D laser scanners,dense vegetation with strong ground-occlusion capability is accurately identified as vegetation objects,while non-vegetation and sparse vegetation with weak ground-occlusion capability are "fuzzily" classified as non-vegetation objects,to achieve "fuzzy" calculation of vegetation coverage in the study area and meet the requirements of route planning optimization.First,a single-band image of the remote sensing image of the study area is extracted,and band operations are performed to obtain the corresponding vegetation index.Secondly,classification samples,training data sets,and the training of the random forest classifier are prepared based on all vegetation indices,screening vegetation indices corresponding to classification models with classification accuracy greater than 95%.Then,the vegetation and non-vegetation are classified based on the screened vegetation index,using the corresponding classification model.During the classification process,the excess green index(EXG),modified green-red vegetation index(MGRVI),normalized green-red difference index(NGRDI),and red-green ratio index(RGRI)accurately identify dense vegetation with strong ground-occlusion capability as vegetation objects,and non-vegetation,low-lying vegetation,and sparse vegetation in landslides are "fuzzily" classified as non-vegetation objects.Finally,the vegetation coverage of the classification results for the excess green index,modified green-red vegetation index,normalized green-red difference index,and red-green ratio index are calculated separately,and the average value is taken as the vegetation coverage of the study area.(2)Vector environmental impact analysis of route planningThe number of location nodes in the vector environment of route planning directly affects the speed and convergence of the route planning method proposed in this study.In the different categories of vector environments established in this study,the fewer the number of location nodes,the faster the speed and convergence of the route planning method.After comparing the route planning results obtained by this method in different categories of route planning vector environments,when facing urgent route planning tasks in vegetation-dense areas,a vector environment with relatively fewer location nodes can be selected to efficiently complete the route planning task.When the primary purpose of the route planning task is to obtain more ground point data,and to provide data basis for establishing high-precision surface data,a vector environment with relatively more location nodes can be selected to complete the route planning task.(3)Optimization of route planning algorithmBased on the classic ant colony algorithm,the optimization of the information pheromone initialization,state transition rules,information pheromone calculation and update strategies is implemented by introducing the "fuzzy" calculation results of vegetation coverage.The adaptive route planning method for UAV-borne Li DAR in vegetation-dense areas is obtained,and the optimization of the three contents mentioned above is verified separately.The route planning results of the proposed method are compared with those of the classic ant colony algorithm before optimization in different categories of route planning vector environments.The experimental results show that the proposed method has a faster iteration speed than the classic ant colony algorithm,i.e.,the time cost of completing one route planning task is less.