Recently, 3D measurement of soil and sand surfaces has been required for a disaster observation and a construction field. We can generally acquire 3D coordinate values with stereo camera or laser scanner. Although current laser scanners can measure objects directly, a real-time data acquisition for moving objects is difficult due to a line scanning. On the other hand, stereo camera can measure moving objects in real time. Therefore, we have proposed 'cross-matching methodology' consisted of a temporal matching with optical flows and a stereo measurement with a template matching. Firstly, feature points are tracked in temporal with optical flow estimation after feature point acquisition from a image. Secondary, the corresponding feature points are acquired from temporally synchronous images with stereo matching. Finally, 3D tracking for moving objects is applied with a combination of temporal and stereo matching without markers.