Method for detecting and tracking objects in image sequences of scenes acquired by a stationary camera
In a sequence of images of a scene acquired by a stationary camera, objects are detected and tracked by determining a first set of candidate foreground regions according to a background model. A second set of candidate foreground regions is determined according to a set of foreground models. Then, candidate foreground regions in the first set and the second set are validated to produce a final set of foreground regions in the image that include the objects.
1. A method for detecting and tracking objects in a sequence of images of a scene acquired by a stationary camera, comprising for each image in the sequence the steps of:
determining a first set of candidate foreground regions according to a background model;
determining a second set of candidate foreground regions according to a set of foreground models; and
validating the candidate foreground regions in the first set and the second set to produce a final set of foreground regions in the image that include the objects present in the image, wherein the validating uses amounts of overlap between regions from the first set and the second set and uses all of the following cues to produce the final set of foreground regions: object identity, object size, object motion energy, and correlation with a representative background image, such as a mean background image, wherein object motion energy is determined using a difference image determined from recent images from the sequence, and wherein the steps are performed in a processor.
2. The method of claim 1 , further comprising:
updating the background model according to the final set of foreground regions.
3. The method of claim 1 , further comprising:
updating the foreground models according to the final set of foreground regions.
4. The method of claim 1 , wherein the image is subject to one or more of the following, changing lighting conditions, shadows, occlusion, small camera motions such as caused by vibration, moved background objects, and left-behind objects.
5. The method of claim 1 , wherein recent images from the sequence are stored in a buffer.
6. The method of claim 1 , wherein the background model includes mean color values and a covariance of color values for each pixel in the image.
7. The method of claim 1 , wherein the background model includes a Gaussian distribution of color values for each pixel in the image.
8. The method of claim 1 , wherein according to the background model, each pixel in the image has an associated probability of being part of the foreground, and further comprising:
aggregating and filtering the probabilities to determine the first set of candidate foreground regions.
9. The method of claim 8 , wherein the filtering uses a Markov random field.
10. The method of claim 1 , wherein an object identity is represented using a histogram of color values of pixels associated with the object.
11. The method of claim 1 , wherein the scene is indoors.
12. The method of claim 1 , wherein the camera and the processor are located in a climate control device.
13. The method of claim 12 , wherein the final set of foreground regions is used to control the climate control device.
14. The method of claim 1 , wherein the camera is located in an appliance.
15. The method of claim 14 , wherein the final set of foreground regions is used to control the appliance.
16. The method of claim 1 , wherein the camera is located in an entertainment device.
17. The method of claim 16 , wherein the final set of foreground regions is used to control the entertainment device.
18. The method of claim 1 , wherein the camera is located in a device or appliance and the final set of foreground regions is used to control a different device or appliance.