PURPOSE: To achieve a higher image recognition processing accuracy by enabling comparision of an actual image data with a forecast image data without misrecognition.

CONSTITUTION: For example, an actual image data A having feature points and feature segments is obtained as a result of processings for the differentiation or binary coding of an image with an actual image computing unit 3 after taken in with an image input unit 1. On the other hand, an extractor 5 extracts a forecast image data B having feature parts from a memory 4 provided with a map based on positional information of a moving body. This forecast image data B is entirely provided with a first shade, namely, a first weighted part B1 with a first forecast image computing unit 6 while done with a second shade, namely, a second weighted part B2 at feature points C with a second forecast image computing apparatus 8. The forecast image data B thus weighted is inputted into a comparator 9 once memorized into a forecast image memory 7 to be compared with the actual image data A.