Automatic Recognition of Linear features, Symbols and Textured Areas within Maps

1'111s 1)aper describes advances in ~netliods li)r tlie segliielitation of linear features, sucli as grids. loads. contours and rivers, symbols atid textured areas witl~in binary tnap images. Due to the l~ipli level of cotnplexity and occlusion within binary nlap data. tlie automatic classification of small non-linear co~iiponetits discussed in the paper, sl~ows retriarkable progress towards a co~nl)!etely autotiiatic map recogtiition system .Flie need for total automation it1 capturing digitised 111nl3 dntn 118s primarily been driven by tlie liectls of t~rbnn 1)lannitig and public utility colnpaliles. sr~cli as gas. electricity and water. Autotnatic recogt~itior~ of Iilnll cotiil~onents olTei-s dramatic be~~elits it1 ~)la~irl~~ig and maintaining the data 11setl I>v sucll conil)a~iies since their databases ale colttext tleoeritlerit ul)ori tnap ir~lbrtnatiori A lecent i~iteresting application at Nottiligham Ilniversitv lias I)ecti the capture of terrain dates for virtual ~eality systenis Our approach to the probletii is to scan in a paper-based map. thus producing a raster ittiage. convert the image to a vectorised format, arid tlien allelnpt to extract or segment sitnilar structuretl cotnpotients witliin the niap iniage into various layers or strata. However, maps are a source of extremely coniplex binary images due to tlie intense occlusion incurred by the constituent co~iil)o~icrits For this reason the task of autotnatic recogtlitiotl of all classes of pattern withill sucli documents retilaitis a diflicult one to solve. l'lle layout of the paper includes an overall c;~tegorisatioti of tlie tilap data used. Followed by a ~lescr~l)t~ori of tlie tiiodel by wliicli tlie data is extracted, along with results showing lines arid areas srrccesslillly extracted arid classified. aritl coliclritles wltli a suliililary of sollie of tlie benefits this syste~ri oll'ets in relation to previous work carried out it1 tlie liclti 2. Overall Catcgorisatio~~ or Rial) Data The higlily co~nplcx hiliary itnages. produced as a result of digitisation of geographical maps. reflects tlie abundant amount of intricate detail these tlocuriierits cotitaitl. UK Ordanance Survey Landranger Series tnap with a scale of 1.50 000 niay typically illr~strate eleven tiiain categories of constituents witliin its key. However, a category may iticlude ul1to twenty five direretit sub-classes. Soti~etinles there is only a subtle distinction in structure between these With the aid of colour, tliese subtle differences are easily distinguishable, brrt when presented it1 binary form. a hierarchical model rnust be cot~structed to identify eacli class. 111 our system. liiap data from tlie Ordatiance Survey 1,atldranger Series (1.50000 scale) lias been taken. The …