For the requirements of interior space fire early automatic positioning，the traditional positioning method is improved based on machine vision technology. A simple and practical plane calibration board is designed for camera calibration，and the problem of fastly and automatically extracting the pixel coordinates of the feature points on the calibration board is solved by using Halcon. For the limitation that the traditional method only considers radial distortion，a more comprehensive model of camera distortions is established. The homography matrix and distortion coefficient are got and optimized with imaging model and distortion model by iteration and space positioning is realized by three-dimensional reconstruction. The method has been applied in a fire gum project and the results suggest that the improved method is effective，fast and accurate.