GUIs(Graphical　User　Interfaces)　are　an　important　part　of　many　software　applications,　especially　for　mobile　gaming　applications　whose　GUI　screens　contain　a　large　number　of　widgetized　control　elements　for　triggering　state　transitions.　Traditional　testing　tools　such　as　Monkey　Test　generate　random　event　sequences　that　fail　to　generate　valid　events　for　these　manipulated　controls,　which　can　easily　lead　the　test　into　a　loop　that　cannot　be　recovered　from　a　particular　interface　state.　For　this　reason,　this　paper　proposes　a　new　automated　testing　framework　named　YOLOv5-ABLN-Test,　which　utilizes　the　model　YOLOv5ABLN　proposed　in　previous　studies　for　control　identification.　Meanwhile,　a　new　interface　state　determination　strategy　is　proposed　based　on　the　improved　ORB　algorithm.　The　framework　confirms　the　actual　operable　GUI　controls　by　initiating　GUI　events　to　the　recognized　controls.　We　have　implemented　the　framework　as　an　Android　application　testing　tool　and　compared　it　with　the　existing　techniques　in　the　industry　through　case　studies　of　real-world　applications.　©　2024　IEEE.