Inclusive Lambda sample are used to determine the Proton Identification efficiency and related systematic uncertainty for the Belle experiment. The number of tagged Lambda's can be estimated from the reconstructed mass spectrum of proton and pi- from Lambda decays. We compare momentum-dependent efficiency between Data and MC with various proton identification criteria. Background events from Ks and gamma's are rejected properly using kinematical information. In the study, we used on-resonance data from EXP7 to EXP49, corresponding to an integrated luminosity of 492 fb-1. The results indicate that the performance of proton identification at Belle is quite stable and the associated systematic uncertainty is within 2%.