Forgery Detection by Local Correspondence Forgery Detection by Local Correspondence Forgery Detection by Local Correspondence

Signatures may be stylish or unconventional and have many personal characteristics that are challenging to reproduce by anyone other then the original author For this reason signatures are used and accepted as proof of authorship or consent on personal checks credit purchases and legal documents Currently signatures are veri ed only informally in many environments but the rapid development of computer technology has stimulated great interest in research on automated signature veri cation and forgery detection In this thesis we focus on forgery detection of o line signatures Although a great deal of work has been done on o line signature veri cation over the past two decades the eld is not as mature as on line veri cation Temporal information used in on line veri cation is not available o line and the subtle details necessary for o line veri cation are embedded at the stroke level and are hard to recover robustly We approach the o line problem by establishing a local correspondence between a model and a questioned signature The questioned signature is segmented into consecutive stroke segments that are matched to the stroke segments of the model The cost of the match is determined by comparing a set of geometric properties of the corresponding sub strokes and computing a weighted sum of the property value di erences The least invariant features of the least invariant sub strokes are given the biggest weight thus emphasizing features that are highly writer dependent Random forgeries are detected when a good correspondence cannot be found i e the process of making the correspondence yields a high cost Many simple forgeries can also be identi ed in this way The threshold for making these decisions is determined by a Gaussian statistical model Using the local correspondence between the model and a questioned signature we perform skilled forgery detection by examining the writer dependent information embedded at the sub stroke level and trying to capture unballistic motion and tremor information in each stroke segment rather than as global statistics Experiments on random simple and skilled forgery detection are presented This research was funded in part by the Department of Defense and the Army Research Laboratory under Contract MDA C LAMP TR CAR TR CS TR MDA C April Forgery Detection by Local Correspondence Jinhong Katherine Guo Forgery Detection by Local Correspondence Jinhong Katherine Guo Center for Automation Research University of Maryland College Park MD Abstract Signatures may be stylish or unconventional and have many personal characteristics that are challenging to reproduce by anyone other then the original author For this reason signatures are used and accepted as proof of authorship or consent on personal checks credit purchases and legal documents Currently signatures are veri ed only informally in many environments but the rapid development of computer technology has stimulated great interest in research on automated signature veri cation and forgery detection In this thesis we focus on forgery detection of o line signatures Although a great deal of work has been done on o line signature veri cation over the past two decades the eld is not as mature as on line veri cation Temporal information used in on line veri cation is not available o line and the subtle details necessary for o line veri cation are embedded at the stroke level and are hard to recover robustly We approach the o line problem by establishing a local correspondence between a model and a questioned signature The questioned signature is segmented into consecutive stroke segments that are matched to the stroke segments of the model The cost of the match is determined by comparing a set of geometric properties of the corresponding sub strokes and computing a weighted sum of the property value di erences The least invariant features of the least invariant sub strokes are given the biggest weight thus emphasizing features that are highly writer dependent Random forgeries are detected when a good correspondence cannot be found i e the process of making the correspondence yields a high cost Many simple forgeries can also be identi ed in this way The threshold for making these decisions is determined by a Gaussian statistical model Using the local correspondence between the model and a questioned signature we perform skilled forgery detection by examining the writer dependent information embedded at the sub stroke level and trying to capture unballistic motion and tremor information in each stroke segment rather than as global statistics Experiments on random simple and skilled forgery detection are presentedSignatures may be stylish or unconventional and have many personal characteristics that are challenging to reproduce by anyone other then the original author For this reason signatures are used and accepted as proof of authorship or consent on personal checks credit purchases and legal documents Currently signatures are veri ed only informally in many environments but the rapid development of computer technology has stimulated great interest in research on automated signature veri cation and forgery detection In this thesis we focus on forgery detection of o line signatures Although a great deal of work has been done on o line signature veri cation over the past two decades the eld is not as mature as on line veri cation Temporal information used in on line veri cation is not available o line and the subtle details necessary for o line veri cation are embedded at the stroke level and are hard to recover robustly We approach the o line problem by establishing a local correspondence between a model and a questioned signature The questioned signature is segmented into consecutive stroke segments that are matched to the stroke segments of the model The cost of the match is determined by comparing a set of geometric properties of the corresponding sub strokes and computing a weighted sum of the property value di erences The least invariant features of the least invariant sub strokes are given the biggest weight thus emphasizing features that are highly writer dependent Random forgeries are detected when a good correspondence cannot be found i e the process of making the correspondence yields a high cost Many simple forgeries can also be identi ed in this way The threshold for making these decisions is determined by a Gaussian statistical model Using the local correspondence between the model and a questioned signature we perform skilled forgery detection by examining the writer dependent information embedded at the sub stroke level and trying to capture unballistic motion and tremor information in each stroke segment rather than as global statistics Experiments on random simple and skilled forgery detection are presented This research was funded in part by the Department of Defense and the Army Research Laboratory under Contract MDA C