Research on satellite remote sensing-based intelligent monitoring technologies for new construction land
The accurate and frequent extraction of information regarding new construction land is essential in natural resource monitoring and land law enforcement supervision.To fulfill these practical demands,this study constructed a technical framework for the intelligent monitoring of new construction land via satellite remote sensing.The framework includes a"spatial-temporal-spectral-classified"monitoring hypercube,a base map generation for monitoring,a sample annotation iteration,component-based artificial intelligence(AI)change detection model establishment,parcel information filtering,and post-processing.To meet the demand for accurate applications in large areas and complex scenes,this study fully combined different AI algorithms and network structures,such as attention mechanism,domain adaptation,and visual transformers,to develop a component-based AI change detection model for improving the accuracy and reliability of the algorithm.Meanwhile,to address issues,such as misidentification during the automatic extraction of new construction land parcels,parcel fragmentation,and edge inaccuracy,geomorphological principles were comprehensively utilized to set constraints and investigate post-processing parcel refinement methods.Experiments by region and time were conducted on large-scale remote sensing monitoring of new construction land to verify the feasibility of the proposed concept of the"spatial-temporal-spectral-classified"monitoring hypercube.Moreover,through ablation analysis of the component-based AI change detection model,the advantages and disadvantages of the algorithms were compared and analyzed.In particular,the visual transformer module exhibits evident advantages in terms of the feature completeness,edge accuracy,and recall rate of new construction land extraction.On the basis of certain operational data of satellite image-based law enforcement and supervision,cloud cover filtering was conducted.Wrongly extracted parcels accounted for about 0.84％.In addition,after the post-processing parcel refinement method proposed in this study was adopted,the accuracy and practicability of the monitoring results were further enhanced.The satellite remote sensing-based technologies and methods for the intelligent monitoring of new construction land proposed in this study have been applied to natural resource monitoring,such as land law enforcement and supervision.
万方数据
维普
