imp rovementofself-best-fitstrate gyforstrokebuildin g

astrokeinaroadnetworkisdefinedasasetofoneormoreroadse gmentsinanon-branch-in g,connectedchain,accordin gtotheg oodcontinuit yp rinci p le.thestrokep la y sanimp ortantroleinroadnetworkg eneralization,to p olo gyanal y sis,p atternreco g nitionandschematicma pg eneration.thisp a p erdemonstrateswa y stoimp rovetheself-best-fitstrate gyintwoas p ects.oneisthatthep ro-cessin gorderofse gmentsisdeterminedbasedontheirimp ortanceratherthanrandomorder,anotherisextendin gthedefinitionofbest-fit,p ro p osin gthatse gmentwhichsatisf y in gtheconcatenationcrite-rionandthemostimp ortantse gmentisdefinedasthebest-fitse gmentinsteadofse gmentwiththesmallestdeflectionan g le.theshenzhenroadnetworkatthescaleof1∶10000wasusedasex p eri-mentaldata.anevaluationfromthevisualco g nitionandnetworkfunctionalit yp ointsofviewshowsthat：（ 1） thestrokesetg eneratedb ythemodifiedself-best-fitstrate gyisusuall yuni q ue.（ 2）fromthevisualco g nitionp ointofview,themodifiedself-best-fitstrate gyp erformsbetterthantheever y-best-fitstrate gy.inastrokesetg eneratedb ymodifiedself-best-fitstrate gythestrokesarenormall ylon g erandhavemoreg lobalp ro p erties.（ 3）fromthenetworkfunctionalit yp ointofview,themodi-fiedself-best-fitstrate gyp erformsbetterthantheself-best-fitstrate gy,astheself-fitstrate gyandtheself-fitstrate gyintheex p eriment.