A　distributed　dynamic　clustering　and　collaborative　tracking　algorithm　is　proposed　for　maneuvering　target　tracking　problems　in　sensor　networks.　The　sensor　node　is　selected　adaptively　and　a　sensor　cluster　is　activated　online　by　optimizing　the　performance　measure　of　tracking　and　cost　of　communication.　Accuracy　of　tracking　is　improved　by　dynamic　collaboration　and　information　fusion　of　the　sensor　nodes.　The　particle　filtering　is　employed　to　predict　and　estimate　the　probability　distribution　of　target　states　due　to　nonlinear　problems　and　randomness　of　the　sensor　nodes.　The　Gaussian　mixture　particle　filtering　and　the　shortest　routing　algorithm　are　utilized　for　information　exchange　between　the　sensor　nodes　to　save　energy　of　communication.　An　efficient　particle　method　is　proposed　for　approximating　expected　posterior　mean　square　error　to　optimize　sensor　selection.　The　simulation　shows　significant　improvement　of　the　proposed　algorithm　over　existing　IDSQ　methods　in　tracking　accuracy　for　maneuvering　target.