To measure accurately the attitude of a spatial object that was fast-changing in a suspension state, an attitude measurement method was proposed based on three linear CCD cameras for measuring multiple point cooperation targets. By proposed method, the nonlinear systemic errors and more calibration parameters caused by more overconstraints for camera positions were overcome in the traditional linear CCD attitude measurement. Three 1D cameras composed of linear CCDs and cylindrical lenses were used to measure simultaneously the multiple point cooperation targets on the measured object. Aiming at the restriction of measuring principle on the attitude angle, the simulated calculation method was used to calculate the attitude angle range and the BP neural network was taken to calibrate the CCD cameras and give the spatial 3D coordinates of point cooperation targets. Furthermore, an attitude calculation model based on Rodrigues parameters was established to solve the attitude angles of measured objects. Experimental results indicate that the measurement accuracy of spatial point position and the calculating precision of spatial attitude angle in this system are higher than those of traditional methods, which verifies that this attitude measurement method is feasible and available to measure the attitude.