In order to analyze the characteristic of large attitude angle errors of CBERS-02B satellite imagery

this paper proposes a specific strategy and algorithm to improve the point determination accuracy by using block adjustment. First

the errors of attitude angles are calibrated according to four Ground Control Points （GCPs） around the four corners of each image. Then the Rational Polynomial Coefficients （RPCs） of each imagery are solved using terrain-independent approach. The Rational Function Model （RFM） with the affine transformation parameters is applied for the bundle block adjustment of CBERS-02B satellite images. Two CBERS-02B stereo pairs in level 0 acquired by a single CCD are selected in our test. The experimental results have shown that an accuracy of ±3 pixels can be attained both in plane and elevation. Moreover

the elevation accuracy is better than the planimetric accuracy. Comparing to the bundle block adjustment without pre-correction

the method proposed in this paper can improve the accuracy both in plane and elevation to a similar accuracy level of foreign High-resolution Satellite Images （HRSIs）

which indicates that Chinese satellite images have good potential for geometric positioning. These experiments further show

the pre-correction before bundle block adjustment is necessary and feasible for the images in geometric processing.