The aim of this paper is to develop the detection system for present fast-developing LED (Light Emitted Diode) photoelectric Industry. To cope with the processing size of next generation dwindles, the gear of the inspection equipment in ultra precision and fast inspection specifications will become one of the important research keys. This paper presents a machine vision to inspect LED micro-structure by piezo actuation stage, while implemented vision system is used for feedback control of the planar motor. Ultra precision and fast positioning for searching the LED Pad/Die is achieved by using the features of piezoelectric actuator (PEA) that is embedded In the probing stage. Such piezo-actuated probing stage is controlled by using feed-forward neural network (FFNN) learning algorithm to compensate for the nonlinear and hysteresis properties of piezoelectricity. Experimental results show each LED Pad/Die can be finished lighting within the rate of 4 pieces per second of one probe. This work validates a new inspection machine structure and demonstrates the feasibility on future multi probes for next manufacturing in lower dimensions.