This　paper　introduces　a　novel　alternated　two-particle　discrete-time　quantum　walk　model　on　arbitrary　graphs,　overcoming　the　limitation　that　conventional　shared　coin　schemes　are　confined　to　regular　graphs.　By　employing　a　fixed-dimension　coin　operator　and　encoding　the　graph　structural　information　into　the　shift　operator,　the　new　model　becomes　applicable　to　arbitrary　graphs.　Further,　by　adding　interaction　between　the　two　particles,　an　extension　is　presented,　which　can　enhance　the　model＇s　dynamic　properties　and　facilitate　more　intricate　quantum　interference　phenomena.　Based　on　the　new　model,　a　kind　of　quantum　graph　isomorphism　algorithm　framework　is　proposed.　It　just　need　O(N0.5*divided　by　E　divided　by)　steps　of　quantum　walk　and　O(N2)　dimensions　of　Hilbert　space,　which　offers　a　significant　reduction　in　complexity　compared　to　other　quantum　walk　based　algorithms.　The　graph　isomorphism　testing　is　performed　in　the　framework　using　non-interacting　and　interacting　quantum　walks　respectively.　Experimental　results　illustrate　that　the　algorithm　based　on　interacting　particles　achieves　a　100%　success　rate　in　discriminating　all　test　graphs.　While　the　algorithm　with　non-interacting　particles　performs　a　100%　distinction　success　rate　on　general　non-regular　graphs,　albeit　being　ineffective　on　strongly　regular　graphs.　Due　to　the　effectiveness　of　the　algorithm　on　arbitrary　graphs,　the　algorithm　has　broad　application　prospects　in　the　identification　and　comparison　of　chemical　molecular　structures,　the　analysis　of　social　networks　and　so　on.