Diplomarbeit, Jan-Hendrik Palic
Betreuer: Ingo Schiller
The aim of this work is to derive accurate global positions of a multi-Sensor 3D marine seismic acquisition system by means of visual detection. High accuracy relative coordinates are needed to stack datasets of the multiple receivers. Thus the usage of high spatial resolution seismic systems demands a sufficiently accurate positioning system.
We use a triangulation approach to obtain relative positions of seismic receivers towed in a perpendicular line behind a research vessel. The receivers a mounted under colored buoyages which are to be triangulated and tracked during acquisition by using color histograms. we use two cameras that are mounted on a crane at the stern of the ship for triangulation. Because of roll and pitch of the ship, an inertial sensor is needed to adjust the camera positions. Finally the positions must be related to global GPS coordinates, obtained from a ship-based GPS.
The system consists of two firewire cameras in water resistant housings. The rig is controlled by an embedded PC, Pentium M with 2GHz, 1GB RAM and a 750GB hard disk. For a stable positioning of the seismic receivers, it is necessary to get a triggered picture of the receiver array at each time the seismic acquisition starts with the ignition of a seismic source. The system works at a frame-rate of 3Hz. The second part of the acquisition system is a computer for recording the inertial sensor data, triggered by the embedded computer. The GPS data is recorded by the system triggering the seismic source.
Calibration of the stereo camera rig
The calibration of the camera system is difficult in this case. The cameras have a base distance of 5,80m. To calibrate the camera system, a pair of pictures with distinctive objects have been used. Calibrating the stereo rig on the ocean does not seem sufficient, because there are only indistinguishable waves and, which is more important, the waves are on a plane. So it is difficult to choose corresponding pixels and the result might be insufficient for an accurate positioning.
Detection of the hydrophones
The detection works with color histograms that are created for every receiver device. These reference histograms are compared to histograms of a chosen Region-Of-Interest in each picture including the device array. The color histograms are created in HSL color space to be robust for strong changing of luminance. The comparison algorithm uses the Battacharyya-Distance. It can be shown that a successful detection needs more criteria than compared histograms when .i.e. the tracker looses the objects because of daylight changes. Therefor we try to add more constraints consisting the behavior of movement of the towed devices.
The global positioning uses the GPS data of the onboard ship device and the coordinates of the detected hydrophones.
The 3D position of the hydrophones are calculated using the stereo or the mono camera system, either using triangulation or by intersecting the viewing ray from the camera with the calculated water plane.
The relative position of the cameras towards the GPS receiver is known, and with this knowledge the calculated 3D positions of the balls can be calculated in global GPS coordinates.