GPR scanning through the use of external scene knowledge
A method for scanning an area using a ground penetrating radar (GPR) by moving the GPR along at least one scanning trajectory. The method includes determining at least one landmark object image feature by applying object detection techniques to images of the area to be scanned, determining a position and/or type of at least one landmark object corresponding to the at least one landmark object image feature in a scanning-area coordinate frame, determining a candidate position or candidate type of at least one candidate underground asset in the area to be scanned by using the determined position or type of the at least one landmark object determining the at least one scanning trajectory using the candidate position and/or candidate type of the at least one candidate underground asset.
1. A method for planning at least one scanning trajectory for scanning an area using a ground penetrating radar (GPR) by moving the GPR along at least one planned scanning trajectory, the method comprising:
determining at least one landmark object image feature by applying object detection techniques to images of the area to be scanned capturing an overground scene;
determining a position or type of at least one landmark object corresponding to the at least one landmark object image feature in a scanning-area coordinate frame;
determining a candidate position or candidate type of at least one candidate underground asset in the area to be scanned by using the determined position or type of the at least one landmark object; and
determining the at least one planned scanning trajectory using the candidate position or candidate type of the at least one candidate underground asset,
wherein the landmark object image feature is determined by the object detection techniques applied to the images that is carried out using machine learning real-time object detection techniques.
2. The method for scanning an area according to claim 1 , wherein the candidate position or candidate type of the at least one candidate underground asset are determined from at least the determined position or type of the at least one landmark object using a rule-based algorithm, wherein the rule-based algorithm encodes known craft-specific information on relationships between landmark objects and corresponding underground assets, wherein the candidate position or candidate type of the at least one candidate underground asset are provided in geometrical and classification form, the geometrical form summarizing the spatial location and extent of the at least one candidate underground asset and the classification form summarizing the type of the at least one candidate underground asset.
3. The method for scanning an area according to claim 1 , wherein the determined candidate type of the at least one candidate underground asset comprises information about at least one material of the at least one candidate underground asset.
4. The method for scanning an area according to claim 1 , wherein the at least one scanning trajectory is determined in such a way as to optimize detectability of at least a part of the at least one candidate underground asset.
5. The method for scanning an area according to claim 4 , wherein the at least one scanning trajectory is orthogonal to a main direction of at least the part of the at least one candidate underground asset and the at least one scanning trajectory is provided with a start and end point, or
the at least one scanning trajectory is at an oblique angle to a main direction of at least the part of the at least one candidate underground asset and the at least one scanning trajectory is provided with a start and end point.
6. The method for scanning an area according to claim 1 , wherein a plurality of scanning trajectories is determined, wherein the scanning trajectories possess different directions and have different lengths.
7. The method for scanning an area according to claim 1 , further comprising:
determining a current position and orientation of the GPR in the scanning-area coordinate frame during movement along the at least one scanning trajectory using information provided by at least one sensor integrated into or attached to the GPR;
providing the images of the area to be scanned by using at least one image sensor;
determining the current position and orientation of the at least one image sensor using a known geometrical relationship between the at least one image sensor and the GPR;
fusing the so obtained current position and orientation of the at least one image sensor with the position and orientation of the at least one image sensor provided by a simultaneous localization and mapping (SLAM) algorithm to obtain an updated position and orientation of the at least one image sensor in the scanning-area coordinate frame; and
determining an updated position and orientation of the GPR by way of the known geometrical relationship between the GPR and the at least one image sensor using the updated position and orientation of the at least one image sensor.
8. The method for scanning an area according to claim 1 , further comprising: projecting a next determined scanning trajectory along which the GPR is to be moved next onto the area to be scanned.
9. The method for scanning an area according to claim 1 , further comprising: projecting a signal onto the area to be scanned to indicate the candidate position or candidate type of a candidate underground asset comprised by the at least one candidate underground asset.
10. The method for scanning an area according to claim 1 , further comprising: providing to a user of the GPR an indication signal to indicate that a start position of a scanning trajectory comprised by the at least one scanning trajectory is reached.
11. The method for scanning an area according to claim 1 , further comprising: providing indication to the user of the GPR about potential deviations of the moving GPR away from a current scanning trajectory along which the GPR is supposed to move.
12. The method for scanning an area according to claim 1 , further comprising: using markers to indicate the location and spatial extent of the area to be scanned.
13. A computer program product with program code being stored on a non-transitory machine readable medium,
the program code being configured to provide the candidate location or candidate type of the at least one candidate underground asset and the at least one scanning trajectory in the area to be scanned using at least the images of the area to be scanned as in the method according to claim 1 .
14. A GPR system comprising a GPR, a computing unit, a computer program product with program code according to claim 13 being stored on the computing unit, and at least one image sensor providing images to the computing unit.