Partial discharge (PD) detection and recognition are of great significance to the condition monitoring of gas-insulated switchgear (GIS). Metallic particles defects are a type of defect that are easy to occur in GIS. The study focuses on the development process of PD creeping discharge induced by metallic particles in GIS. A GIS simulation experiment platform which is set up according to the actual operating conditions of GIS is employed for this. Step up method is applied to study the PD of metallic particles. The discharge pattern of various stages is measured by pulse current method, meanwhile the characteristic parameters which can characterize the severity of PD effectively are extracted. Intensified CCD (ICCD) is also set up to record the evolution process of PD creeping discharge of the metallic particles, meanwhile the development process is divided with K-Means clustering algorithm combined with characteristic parameters. The results indicate that the evolution process of PD creeping discharge induced by metallic particles in GIS is divided into three stages: corona discharge domination, coexistence of corona discharge and surface streamer discharge and surface streamer discharge domination.