ing at the fact that the pose parameter measurement of moving targets in a large field of view is susceptible to factors such as model cumulative error, imaging distortion and insufficient feature information, a new vision measurement method is proposed. First, an efficient multi-source feature data fusion model is established, and it suitable for the visual measurement process, which can solve the problem of single feature point. Then, a bidirectional closed measurement mode based on feature point cloud information is build, which changes the one-way transfer process from image data to spatial feature information in the traditional method, and returns the confirmed spatial data as control information to the measurement process, which can effectively avoid the larger the measurement space, the larger the cumulative error of the measurement model. Finally, the experimental results show that the proposed method achieves a target attitude measurement accuracy better than ±1.5° in a large field of view space of 10 m×8 m×3 m, and the position accuracy is better than 2 mm. The obtained measurement results verify that the target pose parameters obtained by the bidirectional closed cloud control measurement mode have high accuracy and strong stability, and can meet the actual engineering application requirements.