For detection of large structure surface cracks, a positioning method based on image processing technology and coordinate mapping is presented. First, the highest resolution frame is extracted from the sequence images, and the Retinex algorithm is used to compensate the uneven brightness. Second, the landmark image coordinates are mapped to observation coordinates. The mapping relationship between the world coordinate system and the observation coordinate system is calculated. Third, the morphological image processing algorithm is used, such as convex hull, pixels and thinning, to extract crack image coordinates. If there is a new crack, its observation coordinate is mapped to the world coordinate. If there is an old crack, its world coordinate is mapped to the observation coordinate. The observation angle in current observation system is calcualted. The result shows that this method is efficient, accurate and convenient, and the position error is less than 0.07°in 16 seconds.