People detection method for auto-framing and tracking in a video conference
View Patent ↗A videoconference apparatus and method coordinates a stationary view obtained with a stationary camera to an adjustable view obtained with an adjustable camera. The stationary camera can be a web camera, while the adjustable camera can be a pan-tilt-zoom camera. As the stationary camera obtains video, participants are detected and localized by establishing a static perimeter around a participant in which no motion is detected. Thereafter, if no motion is detected in the perimeter, any personage objects such as head, face, or shoulders which are detected in the region bounded by the perimeter are determined to correspond to the participant.
1. A people detection method for auto-framing and tracking in a video conference, the method comprising:
capturing a first image of a first view of an environment with a first camera;
detecting, using a processor, at least one participant in the first image;
comparing, using the processor, the first image to a previous image of the environment to identify one or more motion pixels;
determining, using the processor, a boundary in the first view of the environment by iteratively searching for a static region in the first view based on the one or more motion pixels, the boundary at least partially surrounding the at least one participant;
detecting, using the processor, a personage object within the boundary in a second image of the first view;
assigning, using the processor, an identifier to the personage object indicating that the personage object corresponds to the at least one participant based on determining that the boundary was static between capture of the first image and capture of the second image; and
generating, using the processor, a time-of-last-motion array, wherein generating the time-of-last-motion array comprises comparing the first image to the previous image, and wherein the one or more motion pixels are identified based on the time-of-last-motion array.
2. The method of claim 1 , wherein the personage object comprises shoulders.
3. The method of claim 1 , wherein the personage object comprises a torso.
4. The method of claim 1 , wherein the personage object comprises a head, shoulders, torso, face or facial feature, or some combination of any of these.
5. The method of claim 1 , further comprising:
adjusting, using the processor, a second view of a second camera to a framed view of the environment based on the determined boundary; and
capturing second video in the framed view with the second camera.
6. The method of claim 1 , further comprising tracking, using the processor, a location of the at least one participant using the identifier.
7. The method of claim 1 , wherein detecting the at least one participant in the first image includes:
detecting, using the processor, a face in the first image;
detecting, using the processor, a torso in the first image; an
determining, using the processor, that the face corresponds to the torso based on a position of the face relative to the torso.
8. A non-transitory computer readable medium storing instructions, the instructions comprising instructions executable by at least one processor to cause the processor to:
capture a first image of a first view of an environment with a first camera;
detect at least one participant in the first image;
compare the first image to a previous image of the environment to identify one or more motion pixels;
determine a boundary in the first view of the environment by iteratively searching for a static region in the first view based on the one or more motion pixels, the boundary at least partially surrounding the at least one participant;
detect a personage object within the boundary in a second image of the first view;
assign an identifier to the personage object indicating that the personage object corresponds to the at least one participant based on determining that the boundary was static between capture of the first image and capture of the second image; and
generate a time-of-last-motion array, wherein generating the time-of-last-motion array comprises comparing the first image to the previous image, and wherein the one or more motion pixels are identified based on the time-of-last-motion array.
9. The non-transitory computer readable medium of claim 8 , wherein the personage object comprises shoulders.
10. The non-transitory computer readable medium of claim 8 , wherein the personage object comprises a torso.
11. The non-transitory computer readable medium of claim 8 , wherein the personage object comprises a head, shoulders, torso, face or facial feature, or some combination of any of these.
12. The non-transitory computer readable medium of claim 8 , wherein the instructions further comprise instructions executable to cause the processor to:
adjust a second view of a second camera to a framed view of the environment based on the determined boundary; and
capture second video in the framed view with the second camera.
13. A videoconferencing apparatus, the videoconferencing apparatus comprising:
at least one processor;
at least one camera coupled to the processor; and
a memory coupled to the processor, and storing instructions executable by the processor to cause the processor to:
capture a first image of a first view of an environment using the at least one camera;
detect at least one participant in the first image;
compare the first image to a previous image of the environment to identify one or more motion pixels;
determine a boundary in the first view of the environment by iteratively searching for a static region in the first view based on the one or more motion pixels, the boundary at least partially surrounding the at least one participant;
detect a personage object within the boundary in a second image of the first view; and
assign an identifier to the personage object indicating that the personage object corresponds to the at least one participant based on determining that the boundary was static between capture of the first image and capture of the second image; and
generate a time-of-last-motion array, wherein generating the time-of-last-motion array comprises comparing the first image to the previous image, and wherein the one or more motion pixels are identified based on the time-of-last-motion array.
14. The videoconferencing apparatus of claim 13 , further comprising at least one second camera, and wherein the instructions are further executable to cause the processor to:
adjust a view of the second camera to a framed view based on the determined boundary; and
capture second video in the framed view with the second camera.
15. The videoconferencing apparatus of claim 14 , wherein the at least one camera comprises a web camera and the at least one second camera comprises a pan-tilt-zoom camera.