Entity resolution on camera records without machine learning