Precise and rapid identification of greenhouses and encroachment risk assessment in the Xichang section of the Chengdu-Kunming railway based on cloud platform
Railways constitute an integral part of contemporary transportation infrastructures,with their safety and seamless operation being intrinsically linked to economic stability and growth.In recent years,the expansion of agricultural activities has led to a notable rise in the presence of lightweight objects,such as agricultural films and plastic greenhouse covers,along railway corridors.Under extreme weather conditions,particularly strong winds,these objects may be lifted and encroach upon the railway's safety perimeter—an event termed"encroachment".Such encroachment can obscure overhead power lines(catenaries),causing power outages and posing significant risks to railway safety.Historically,the early detection of these encroachment relied heavily on manual inspections,a method that proved inefficient,susceptible to oversight,and ill-equipped to satisfy the demands for extensive,real-time surveillance.Consequently,there is an exigent need for accurate,rapid identification of floating debris along railways and comprehensive risk assessments of large-scale intrusion potential.This research harnesses the capabilities of the Google Earth Engine(GEE)cloud platform,amalgamating Sentinel-1 synthetic aperture radar(SAR)data with Sentinel-2 optical imagery to construct a multidimensional feature space comprising 17 variables.These encompass dual radar backscatter coefficients(VV and VH),ten spectral reflectance bands(B2 through B8A and Bll,B12),and five spectral vegetation indices(NDVI,SAVI,MNDWI,NDBI,EBSI).Employing the random forest(RF)algorithm,the study successfully delineated greenhouse regions along the Xichang segment of the Chengdu-Kunming railway spanning from 2019 to 2023.Furthermore,by incorporating datasets on greenhouse density kernels,proximity to rail tracks,and local wind speeds/directions,a sophisticated risk assessment model was formulated to quantify the likelihood of intrusion events across different zones adjacent to the railway.Key findings include:(1)High accuracy in greenhouse detection.Utilizing the random forest algorithm yielded exceptional accuracy in identifying greenhouse structures,with results closely aligning with actual remote sensing observations.Over a five-year period,the model achieved an average precision rate of 93.90％and a recall rate of 94.95％.(2)Geographical distribution & growth trends.Greenhouses are predominantly clustered in the northwestern and southwestern sectors of the study area.Notably,there was a substantial surge in greenhouse proliferation,peaking at an annual increment of 20.27 square kilometers between 2021 and 2022.(3)Risk assessment by season.High-risk zones for wind-induced encroachment during spring and winter were pervasive in the northwest and southwest,whereas fewer high-risk areas were identified in summer and autumn,which were mostly confined to select locations in the northeast.This study provides technical support for the precise and rapid identification of floating objects and intrusion risk warning along railway lines,contributing to enhanced safety and stability of railway operations.
NETL
NSTL
万方数据
