The　hierarchical　processing　capabilities　of　the　human　visual　system　can　significantly　enhance　the　efficiency　of　data　processing　in　the　central　nervous　system.　Volatile　and　non-volatile　devices　are　key　components　in　simulating　the　central　nervous　system.　Realizing　both　volatile　and　non-volatile　functionalities　on　a　single　device　is　ideal;　however,　challenges　such　as　complex　preparation　and　cumbersome　switching　persist.　In　this　study,　a　tunable　synaptic　transistor　with　volatile　and　non-volatile　switching　capabilities　is　developed,　offering　ease　of　fabrication　and　convenient　switching.　It　can　simulate　various　forms　of　synaptic　plasticity　and　exhibits　excellent　storage　performance　in　non-volatile　mode.　Finally,　we　design　an　image　preprocessing　and　classification　system　based　on　visual　selective　attention,　which　enables　efficient　neuromorphic　computation　through　hierarchical　data　processing.