Road Extraction With Multiple Features And Energy Optimization Model

Remote sensing has become the most effective method to acquire information from earth surface.And extracting roads from remote sensing images will provide support for industries,so there is an urgent need for it.Both extracting new roads and updating information of roads from remote sensing images will greatly cut down the cost of human resources and time.However,current road extracting technology still cannot meet up with need for real production.Technologies of road extraction from remote sensing images have developed for decades,the scenes move from rural areas to complex urban scene,resolutions of images have been higher and higher,and the task difficulty is also higher and higher.In the meantime,more robust technologies have been brought out,but road object extraction methods still remain to be broken through to adapt to different situations.This thesis explores the method of road extraction from high resolution remote sensing images,the main content is as follows:1.Bring out a new fast method of road regions of interest extraction by exploiting multiple road features and topology feature of roads.2.Bring out a new road extraction technology based on existing road data or simple input from user.The incremental markov random field method in combination with specific color modeling,texture modeling and region length constraint modeling is bought out.To use existing road data for guidance,simple road vector rectifying method has been brought out.