Systems and methods for persistent surveillance and large volume data streaming
View Patent ↗In general, the present disclosure relates to persistent surveillance (PS), wide medium and small in area, and large volume data streaming (LVSD), e.g., from orbiting aircraft or spacecraft, and PS/LVSD specific data compression. In certain aspects, the PS specific data or LVSD compression and image alignment may utilize registration of images via accurate knowledge of camera position, pointing, calibration, and a 3D model of the surveillance area. Photogrammetric systems and methods to compress PS specific data or LVSD while minimizing loss are provided. In certain embodiments, to achieve data compression while minimizing loss, a depth model is generated, and imagery is registered to the depth model.
1. A photogrammetric method for compressing and transmitting persistent aerial surveillance data of a location of interest, the method comprising:
generating a textured depth model of the location of interest for persistent aerial surveillance based on initial aerial imagery of the location of interest captured from different angles circling around the location of interest;
predicting aerial imagery of the location of interest based on the textured depth model of the location of interest;
capturing aerial imagery of the location of interest using at least one camera, the at least one camera having a known model and known ephemeris parameters;
registering at least one frame of the captured aerial imagery to the textured depth model of the location of interest;
generating a difference image between a captured frame of the captured aerial imagery and a predicted frame of the predicted aerial imagery, wherein the difference image represents pixels of an independent object that are absent from the textured depth model corresponding to a changed region in time or in space;
spatially compressing the difference image to generate a compressed aerial surveillance data stream; and
transmitting the compressed aerial surveillance data stream.
2. The method of claim 1 , further comprising:
decompressing the compressed aerial surveillance data stream to generate a decompressed difference image;
rendering a depth model of the location of interest to an image plane to create a predicted image; and
adding the decompressed difference image to the predicted image to generate a reconstructed, decompressed image.
3. The method of claim 1 , wherein registering the at least one frame of the captured aerial imagery and generating the difference image are carried out substantially contemporaneously with capturing the aerial imagery of the location of interest.
4. The method of claim 1 , wherein generating the textured depth model is carried out substantially contemporaneously with capturing the aerial imagery of the location of interest.
5. The method of claim 1 , wherein the textured depth model is obtained from a previous persistent surveillance collection.
6. The method of claim 1 , wherein the textured depth model is obtained or refined through an additional, independent sensing modality.
7. The method of claim 6 , wherein the additional, independent sensing modality comprises a parallel track of Light Detection and Ranging (LIDAR).
8. The method of claim 1 , wherein the captured aerial imagery is registered to the textured depth model with reduced parallax induced motion effects so as to improve compression ratios.