Line rate visual analytics on edge devices

1. A method, comprising:

receiving, at an edge device in a network, a stream of data packets originated by a visual data source, the data packets including at least one of video data, image data, and geospatial data, wherein the edge device acts as a local proxy device within its respective local area network (LAN) between the network and a core network to forward data packets into and out of the network;

extracting, at the edge device, a visual data attribute from the stream of data packets;

vectorizing, at the edge device, the extracted visual data attribute via quantization vectors;

indexing, at the edge device, the vectorized visual data attribute in a schema-less database that stores indexed visual data attributes;

receiving, at the edge device, a search query including a vectorized visual data attribute to be searched; and

searching, at the edge device, the indexed visual data attributes for at least one candidate visual data attribute comparable to the vectorized visual data attribute to be searched.

2. The method as in claim 1 , further comprising:

defining a region of interest in the stream of data packets, wherein the extracted visual data attribute is extracted from the region of interest.

3. The method as in claim 1 , further comprising:

determining whether the stream of data packets includes video data, image data, or geospatial data, wherein the extracting at the edge device is based on the determination.

4. The method as in claim 1 , further comprising:

determining the visual data source by which the stream of data packets is originated, wherein the extracting at the edge device is based on the determination.

5. The method as in claim 1 , further comprising:

compressing the stream of data packets; and

placing the compressed stream of data packets in a visual data container.

6. The method as in claim 1 , wherein the extracting of the visual data attribute comprises:

projecting an object in Q dimensions onto an object plane with Q- 1 dimensions;

calculating a bounding region on the object plane in which the projected object is located, wherein the bounding region is defined by at least one bounding parameter.

7. The method as in claim 6 , wherein the at least one bounding parameter includes at least one of a first and second area value, a separation amount of the first and second area value, a center point of a line segment, and an angular orientation of the line segment.

8. The method as in claim 1 , wherein the extracting of the visual data attribute is performed in a transform space.

9. The method as in claim 1 , further comprising:

creating the schema-less database using at least one bounding parameter as a field.

10. The method as in claim 1 , further comprising:

providing access to the schema-less database to at least one other edge device in the network.

11. The method as in claim 1 , further comprising:

identifying visual data trends based on the extracted visual data attribute; and

storing the extracted visual data attribute at a storage location.

12. An apparatus, comprising:

one or more network interfaces that communicate with a network;

a processor coupled to the one or more network interfaces and configured to execute a process; and

a memory configured to store program instructions which contain the process executable by the processor, the process comprising:

receiving, as an edge device in the network, a stream of data packets originated by a visual data source, the data packets including at least one of video data, image data, and geospatial data, wherein the edge device acts as a local proxy device within its respective local area network (LAN) between the network and a core network to forward data packets into and out of the network:

extracting, as the edge device, a visual data attribute from the stream of data packets;

vectorizing, as the edge device, the extracted visual data attribute via quantization vectors;

indexing, as the edge device, the vectorized visual data attribute in a schema-less database that stores indexed visual data attributes;

receiving, as the edge device, a search query including a vectorized visual data attribute to be searched; and

searching, as the edge device, the indexed visual data attributes for at least one candidate visual data attribute comparable to the vectorized visual data attribute to be searched.

13. The apparatus as in claim 12 , wherein the process further comprises:

defining a region of interest in the stream of data packets, wherein the extracted visual data attribute is extracted from the region of interest.

14. The apparatus as in claim 12 , wherein the process further comprises:

determining whether the stream of data packets includes video data, image data, or geospatial data, wherein the extracting at the edge device is based on the determination.

15. A tangible, non-transitory, computer-readable media having software encoded thereon, the software when executed by a processor operable to:

receive, as an edge device in the network, a stream of data packets originated by a visual data source, the data packets including at least one of video data, image data, and geospatial data, wherein the edge device acts as a local proxy device within its respective local area network (LAN) between the network and a core network to forward data packets into and out of the network;

extract, as the edge device, a visual data attribute from the stream of data packets;

vectorize, as the edge device, the extracted visual data attribute via quantization vectors;

index, as the edge device, the vectorized visual data attribute in a schema-less database that stores indexed visual data attributes;

receive, as the edge device, a search query including a vectorized visual data attribute to be searched; and

searching, as the edge device, the indexed visual data attributes for at least one candidate visual data attribute comparable to the vectorized visual data attribute to be searched.