PURPOSE: To easily identify a self position of a moving body at any place with high accuracy by a method wherein a land mark image is abstracted from an image obtained by a TV camera and the recognition result of the land mark is outputted.

CONSTITUTION: A land mark recognition section 1 recognizes a land mark in an image of a TV camera at any time and transmits information of the kind and position thereof to an observation map making section 2. The making section 2 adds the newly recognized land mark information to an observation map information memory section 3. At that time, a moving quantity calculation section 4 obtains a rough moving quantity required for a moving vehicle by using an internal sensor or the like. A rough position calculation section 6 calculates a rough present position of the moving vehicle and a map data managing section 7 cuts out data of a region of a circumference of the present position of the moving vehicle obtained by the calculation section 6 from a map data memory section 8 to transmit to a checking section 5. The checking section 5 checks a chain of land marks described in the cut-out map data with a chain of land marks stored in the memory section 3 to identify the self position.