An attitude measurement method based on multi-sensor combination was proposed to solve the problem of target feature points occlusion due to spatial obstruction in large-scale spatial attitude measurement. The single-coordinate reference measurement of the measured feature point was realized by the digital level and the attitude probe, and the initial attitude of the target was obtained by the level difference of the feature point and its known geometric constraint relationship. On this basis, the attitude rotation matrix between high-precision tilt sensor and measured target was calibrated, and then the target attitude was calculated in real time from the sensor output based on the coordinate transformation theory. The experimental results show that the relative accuracy of attitude measurement is better than 0.001 5° in the range of 10 m, and the repeatability measurement error is less than 0.000 4°, which can be used for the precise real-time measurement of large-scale obstructed spatial attitude.