The invention relates to a three-dimensional (3D) visual navigation method based on multi-sensor information fusion. The multi-sensor data registration and fusion are finished based on acquiring multi-sensor data by means of a photoelectric sensor (infrared and visible lights), an airport radar monitor system (air traffic control radar and ground surveillance radar) and navigation positioning systems (radio navigation positioning system and global positioning system); and the detection, identification, track and behavior prediction are carried out on the targets (obstacles, humans, vehicles and planes) in the air or on the ground; the plane safety take-off and straight-in approach optimization information is automatically generated according to the target information in the air or on the ground; the target information and the plane safety take-off and straight-in approach optimization information are fused and displayed in an airport 3D scene to form a 3D visual navigation; and the system can automatically generate the 3D visual navigation information and transmits the navigation information to the aircraft via the GPS to realize the 3D visual navigation on the aircraft. The realization of the method comprises the following specific steps of: firstly, acquiring the multi-sensor data; secondly, registering and fusing the multi-sensor data; thirdly, detecting, identifying and tracking the targets in the air or on the ground by utilizing the fused data, and predicting the behavior of the targets; fourthly, establishing a 3D model of the airport, and fusing the targets and the behavior prediction data thereof and displaying the data in the airport 3D scene to form a 3D visual navigation system; and finally, carrying out the 3D visual navigation on the aircraft by means of the 3D visual navigation system to realize the safe take-off or approach of the aircraft.