Quality of MMS forward-photographed image intersection

This paper studies the quality of object point positioning by means of image intersection by using close-range images forward taken on a land-based MMS van driving along a street. Due to the weak geometry of intersection of multiple corresponding image rays, photo triangulation often diverges or provides worse results. Nevertheless, the problems can be solved so that the positioning accuracy still suits for updating 1/1000 maps. Tests are done by using different sets of images with diverse baseline lengths, constraint data, image point distributions, and different number of images. All corresponding image points are measured either by SIFT or manually. In each photogrammetric intersection network, the geometric strength figures are analyzed such as inner reliability, lengths and directions of semi-axes of error ellipses, as well as the standard deviations of ground coordinates. Also, a number of ground check points are used to evaluate the quality of the results. They demonstrate that, under a proper constraint configuration, the positioning technique is available for cartography of 1/1000 maps. For example, by using three full control points, the average accuracy of ground coordinates reaches 9∼15cm.