資源描述:
《基于jseg算法的彩色遙感圖像分割碩士論文》由會員上傳分享,免費在線閱讀,更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、基于JSEG算法的彩色遙感圖像分割摘要遙感圖像分割是遙感圖像分析和解譯中的關(guān)鍵技術(shù)之一,對有效利用遙感數(shù)據(jù),進(jìn)一步進(jìn)行信息提取與目標(biāo)識別等工作具有重要意義。高分辨率遙感圖像包含豐富的顏色、紋理等信息,圖像本身也含有大量噪聲,因此如何較好地移除噪聲,并合理利用圖像所包含的各類信息對圖像進(jìn)行準(zhǔn)確分割是高分辨率遙感圖像分割研究中的一個經(jīng)典問題。算法是一種可依次實現(xiàn)圖像濾波、顏色量化和空間分割的彩色圖像分割方法,但該方法直接用于遙感圖像分割時,往往由于遙感圖像中區(qū)域邊界較模糊而導(dǎo)致對區(qū)域邊界分割不準(zhǔn)確,或由于區(qū)域
2、內(nèi)不同陰影而出現(xiàn)過分割現(xiàn)象。為了有效實現(xiàn)對區(qū)域邊界的準(zhǔn)確分割,本文利用能更好描述區(qū)域內(nèi)顏色的同質(zhì)性的局部同質(zhì)矩陣校正傳統(tǒng)JSEG算法中的局部值,以實現(xiàn)對區(qū)域邊界的準(zhǔn)確反映,提高區(qū)域邊界分割的準(zhǔn)確性。為了減弱或消除傳統(tǒng)算法的過分割現(xiàn)象,本文利用能穩(wěn)定描述圖像紋理特征的算子進(jìn)行具有相似紋理信息的顏色類圖的合并。最后,采用分析法、優(yōu)度實驗法中的評價測度,在算法中值和均值計算的基礎(chǔ)上給出了值和均值評價方法,對本文方法的分割效果進(jìn)行評價。仿真實驗結(jié)論表明本文提出的改進(jìn)算法可有效克服傳統(tǒng)的算法在高分辨遙感圖像分割時存
3、在的邊界分割不準(zhǔn)確及過分割現(xiàn)象。關(guān)鍵詞:高分辨率遙感圖像,算法,局部同質(zhì),算子Abstract(重新翻譯)Inordertoeffectivelyanalyzeandinterprettheremotesensingdata,theremotesensingimageprocessingtechnologyhasbecomehotspot.Theremotesensingimagesegmentationisoneofkeyfactors,whichdecidethesuccessofremotesens
4、ingimageanalyzeandcalculation.Becauseonlyifthebettersegmentationeffectisobtained,thefollow-upworkscanbegetbettereffects,suchasinformationextractionandtargetrecognitionetal.Thebackgroundofthispaperishighresolutionremotesensingimage.Astheseremotesensingimag
5、escontainawealthofcolor,textureandalotofnoiseetal,thekeypointoftheresearchishowtoeffectivelyremovenoise,smooththeimageandreasonablyusevariousinformationtosegmenttheimage.Firstly,theimagefiltering,colorspacequantizationandspatialsegmentationcanbecompletely
6、simultaneouslybyusetheJSEGalgorithm,whichdirectlyutilizecolorinformation.Then,InordertosolvethelimitationcaseoftheJSEG,suchasoversegmentation,theimprovedJSEGalgorithmisproposedinthispaper.Forobtainingthebettereffectofsegmentation,thelocalJvalueiscorrected
7、bylocalhomogeneousmatrix,whichgivesreasonablydescriptionforcolorhomogeneityinregionandtheboundaryofdifferentregion.Inaddition,forweakeningoreliminatingtheoversegmentation,theclassmapismergedbyLBP/Coperator.Itgivesstabledescriptionforthetextureofimage.Thef
8、inally,theevaluationofimagesegmentationisanalyzedandsummarized.Inaddition,theevaluationmethodofJvalueandJaveragevalueisproposedbasedoncalculatingtheJvalueandJaveragevalueinJSEGalgorithm.Themeritisreflectedbyobjectiv