Abstract:
The 3D reconstruction of moving object is an important underlying technology for target statistics, target search, abnormal alarms, and in various data mining applications and other improvements in spatio-temporal analysis on target geographic data. In this paper, we make use of computer vision technology to restore the three-dimensional object space coordinates of the target from its two-dimensional image space coordinates in a real-time camera frame. By applying a pattern classification method based on the object side feature fields and taking the object size and location into account, more accurate target screening results are obtained, this improves the accuracy of object tracking and reduction of false-alarms. Through a monocular stereotactic algorithm based ground planar constraint, video data composed of frames are interpreted to generate the three dimensional geographic information of object consisting of trajectories. The obtained spatio-temporal trajectory as a GIS layer, is further integrated into the GIS system to implement spatio-temporal video-based data management.