Methods and systems for improving DVS features for computer vision applications
The disclosure relates to methods and systems for non-linear mapping of motion-compensated DVS events to DVS images. In the non-linear mapping of motion-compensated DVS events to DVS images, the current pixel increments depend on the existing number of accumulated events at that location. Further, the initial events are given a larger weightage to preserve the tracked features which are used for bundle adjustment. Further, disclosure relates to methods and systems for representing polarity in single channel DVS Frame. As the polarity adds additional constraints on feature matching and therefore accurate optical flow is seen in the image. Moreover, disclosure relates to methods and systems for using event-density as a measure of input contrast after DVS event accumulation and this is used to determine the target range for contrast stretching.
1. A method of imaging an event captured by a dynamic vision sensor (DVS) camera, said method comprising:
receiving one or more events captured by the DVS camera over a period of time; and
non-linearly incrementing a motion-compensated pixel location of each of the one or more events by:
segregating the motion-compensated pixel location of each of the one or more events into one or more groups, each of the one more groups corresponding to a range of number of events occurring at the motion-compensated pixel locations; and
incrementing pixel-intensity for the one or more groups by different amounts to non-linearly increment pixel intensities.
2. A system for imaging event captured from dynamic vision camera, the system comprising:
an event-camera configured to capture one or more events over a period of time; and
an image processing system configured to non-linearly incrementing ( 604 ) a motion-compensated pixel location of each of the one or more events by:
segregating the motion-compensated pixel location of each of the one or more events into one or more groups, each of the one more groups corresponding to a range of number of events occurring at the motion-compensated pixel locations; and
incrementing pixel-intensity for the one or more groups by different amounts to non-linearly increment pixel intensities.
3. A dynamic vision sensor (DVS) device comprising:
a memory storing one or more instructions; and
a processor configured to execute the one or more instructions to:
receive one or more events captured by a DVS camera;
process the one or more events based on at least one of:
a non-linear mapping of the one or more events;
single channel polar DVS representation of the one or more events; and
event-density based adaptive normalization of the one or more events; and
perform extraction and tracking on the processed one or more events; and
apply computer vision algorithms based on the tracked one or more events.
4. The DVS device according to claim 3 , wherein the processor is further configured to perform the non-linear mapping of the one or more events by non-linearly incrementing a motion-compensated pixel location of each of the one or more events by:
segregating the motion-compensated pixel location of each of the one or more events into one or more groups, each of the one more groups corresponding to a range of number of events occurring at the motion-compensated pixel locations; and
incrementing pixel-intensity for the one or more groups by different amounts to non-linearly increment pixel intensities.
5. A dynamic vision sensor (DVS) method comprising:
receiving one or more events captured by a DVS camera;
processing the one or more events based on at least one of:
a non-linear mapping the one or more events;
single channel polar DVS representation of the one or more events; and
event-density based adaptive normalization of the one or more events; and
performing extraction and tracking on the processed one or more events; and
applying computer vision algorithms based on the tracked one or more events.
6. The DVS method according to claim 5 , wherein the non-linear mapping of the one or more events comprises:
non-linearly incrementing a motion-compensated pixel location of each of the one or more events by:
segregating the motion-compensated pixel location of each of the one or more events into one or more groups, each of the one more groups corresponding to a range of number of events occurring at the motion-compensated pixel locations; and
incrementing pixel-intensity for the one or more groups by different amounts to non-linearly increment pixel intensities.
7. A non-transitory computer readable storage medium having stored thereon a program for performing a dynamic vision sensor (DVS) method comprising:
receiving one or more events captured by a DVS camera;
processing the one or more events based on at least one of:
a non-linear mapping the one or more events;
single channel polar DVS representation of the one or more events; and
event-density based adaptive normalization of the one or more events; and
performing extraction and tracking on the processed one or more events; and
applying computer vision algorithms based on the tracked one or more events.
8. The non-transitory computer readable storage medium according to claim 7 , wherein the non-linear mapping of the one or more events comprises:
non-linearly incrementing a motion-compensated pixel location of each of the one or more events by:
segregating the motion-compensated pixel location of each of the one or more events into one or more groups, each of the one more groups corresponding to a range of number of events occurring at the motion-compensated pixel locations; and
incrementing pixel-intensity for the one or more groups by different amounts to non-linearly increment pixel intensities.