Digit Recognition in Handwritten Weather Records

Abstract. This paper addresses the automatic recognition of handwritten tem-peraturevaluesinweatherrecords. Thelocalizationoftablecellsisbasedon line detection using projection profiles. Further, a stroke-preservinglineremovalmethodwhichisbasedongradientimagesisproposed. Thepresented digit recognition utilizes features which are extracted using aset of filters and a Support Vector Machine classifier. It was evaluatedontheMNISTandtheUSPSdatasetandourowndatabasewithabout17,000RGBdigitimages. Anaccuracyof99.36%perdigitisachievedfortheentiresystemusingasetof84weatherrecords. 1 Introduction Inadditiontohandwrittendocumentsstoredinhistoricarchivestherearepro-cesses, like manually filling out forms, which still depend on pen and paper.However, accessing the information stored in those documents requires timeand manpower [9]. By digitizing those documents, the advantages of digitallystored information, such as ease of access, can be exploited for handwrittendocuments[9].Thedocumentscoveredinthispaper,i.e. weatherrecords,consistofknownprintedformswithhandwritteninformation. Eventhoughthestructureoftheform is known beforehand, due to prior handling and the scanning process,global and local deformations are introduced. Further, the structure of formssuggestscertainareaswhichareapplicableforhandwriting,yetwritersarenotbound to this constraint. Additionally, handwriting varies significantly dueto different writing styles. Even the handwriting of a single writer exhibitsvariations. In Figure 1 an example of a weather record is shown. The originsoftheweatherrecordsregardedinthispaperareseveralmeasurementstationslocatedinAustria(LowerAustria).Figure1: WeatherrecordfromFebruaryoftheyear2000fromameasurementstation in Lower Austria. The temperature values are located in the column"Lufttemperatur