Abstract: Considering the traditional bundle block adjustment of rational function model is highly constrained by small error of attitude and orbit, narrow field camera and good image intersection angle, a block adjustment method of satellite imagery based on line-of-sight vector rectification was presented. Firstly, the light ray of an image point is calculated by using the rational polynomial coefficients attached to the image, and then the pseudo orbit and attitude of sensor is restored when the point was acquired, then an error compensation model is constructed for the virtual orbit and attitude, and finally the model parameters and object coordinates of tie points are solved simultaneously by the least square method. As the compensation model is modelled from the reason of system errors, it can avoid approximation assumptions and constraints of traditional strategy. Several comparative experiments of simulation data, mapping satellite and non-mapping satellite data are designed. The results show that this algorithm can achieve higher accuracy than the traditional method under various severe conditions such as images with large attitude angle error, large field angle and weak intersection angle.

Key words: rational function model, block adjustment, line-of-sight vector, mapping satellite, non-mapping satellite