Methods and apparatus for detecting pets
Methods and apparatus for pet detection in security systems. In one example, a method includes detecting a motion event in a scene using a motion detector, based on detecting the motion event, activating an image capture device to acquire a plurality of images of the scene, applying a motion detection process to the plurality of images to detect motion in the scene, applying an object detection process to at least one of the plurality of images to detect an object in the scene, pairing the motion with the object to locate a moving object, identifying the moving object as a pet, and based at least in part on identifying the moving object as a pet, deactivating the image capture device.
1. A method comprising:
detecting a motion event in a scene using a motion detector;
based on detecting the motion event, activating an image capture device to acquire a plurality of images of the scene;
applying a motion detection process to the plurality of images to detect motion in the scene, wherein applying the motion detection process includes producing a first bounding box identifying a location of the motion in at least one image of the plurality of images;
applying an object detection process to the at least one image to identify a pet in the scene, wherein applying the object detection process includes producing a second bounding box identifying a location of the pet in the at least one image;
pairing the motion with the pet to identify a moving pet by determining that the first bounding box and the second bounding box at least partially overlap;
based on identifying the moving pet, acquiring a plurality of additional images of the scene;
processing the plurality of additional images to confirm (i) detection of the motion in the scene and (ii) identification of the pet in the scene; and
based at least in part on confirming identification of the pet, deactivating the image capture device.
2. The method of claim 1 , wherein applying the motion detection process to the plurality of images to detect the motion includes processing multiple consecutive image frames.
3. The method of claim 2 , wherein processing the multiple consecutive image frames includes applying a Kalman filter to the multiple consecutive image frames.
4. The method of claim 2 , wherein processing the multiple consecutive image frames includes:
converting each image frame of the multiple consecutive image frames to greyscale to produce multiple consecutive greyscale image frames;
comparing pixel intensities between the multiple consecutive greyscale image frames; and
detecting the motion based on differences in at least some of the pixel intensities between the multiple consecutive greyscale image frames exceeding a predetermined threshold value.
5. The method of claim 1 , wherein detecting the motion event includes detecting the motion event using a passive infrared sensor.
6. The method of claim 1 , wherein processing the plurality of additional images further comprises:
applying the motion detection process to the plurality of additional images to confirm detection of the motion in the scene; and
applying the object detection process to at least one of the plurality of additional images to confirm identification of the pet in the scene.
7. The method of claim 1 , wherein applying the object detection process includes applying an adaptive neural network.
8. A method comprising:
detecting a motion event using a motion detector;
based on detecting the motion event, activating an image capture device;
acquiring a first set of images of a scene with the image capture device;
processing the first set of images to identify a moving pet in the scene, wherein processing the first set of images includes
applying a motion detection process to the first set of images to produce, in at least one image of the first set of images, a first bounding box identifying a location of motion in the scene,
applying an object detection process to the at least one image to produce a second bounding box identifying a location of a pet in the scene, and
pairing the motion with the pet to identify the moving pet by determining that the first and second bounding boxes at least partially overlap;
based on identifying the moving pet, acquiring a second set of images of the scene with the image capture device;
processing the second set of images to confirm identification of the moving pet; and
based on confirming the identification of the moving pet, deactivating the image capture device.
9. The method of claim 8 , wherein applying the motion detection process includes applying a Kalman filter to the first set of images.
10. The method of claim 8 , wherein processing the second set of images includes:
applying the motion detection process to the second set of images to confirm detection of the motion in the scene; and
applying the object detection process to at least one image of the second set of images to confirm detection of the pet in the scene.
11. A sensor comprising:
a motion detector configured to detect motion events;
an image capture device;
at least one processor; and
a data storage device storing instructions that when executed by the at least one processor configure the sensor to
detect a motion event using the motion detector,
based on detecting the motion event with the motion detector, activate the image capture device,
acquire a plurality of image frames of a scene using the image capture device,
apply a motion detection process to the plurality of image frames to produce, in at least one image frame of the plurality of image frames, a first bounding box identifying an instance of motion in the scene,
apply an object detection process to the at least one image frame to produce a second bounding box identifying a location of an object in the scene,
pair the motion with the object to locate a moving object by determining that the first and second bounding boxes at least partially overlap,
determine a classification of the moving object, and
take a response action based on the classification of the moving object;
wherein to take the response action, the sensor is configured to
trigger an alarm based on the classification of the moving object corresponding to a person,
control the image capture device to record a video sequence based on the classification of the moving object corresponding to an unrecognized object, and
deactivate the image capture device based on the classification of the moving object corresponding to a pet.
12. The sensor of claim 11 , wherein the data storage device further includes instructions that when executed by the at least one processor cause the sensor to:
operate in a low power mode of operation in which the image capture device is deactivated;
detect the motion event while operating in the low power mode of operation; and
based on detecting the motion event, operate in a normal mode of operation in which the image capture device is activated.
13. The sensor of claim 11 , further comprising:
a battery coupled to the motion detector, the image capture device, the data storage device, and the at least one processor.
14. The sensor of claim 11 , wherein the motion detector is a passive infrared sensor.
15. The sensor of claim 11 , wherein the data storage device further includes instructions that when executed by the at least one processor cause the sensor to:
based on the classification of the moving object corresponding to a pet or an unrecognized object and prior to taking the response action,
acquire one or more additional image frames of the scene using the image capture device, and
process the one or more additional image frames to confirm the classification of the moving object; and
based on confirmation of the classification of the moving object, take the response action.
16. The sensor of claim 11 , wherein to apply the motion detection process, the data storage device further includes instructions that when executed by the at least one processor cause the sensor to:
convert the plurality of image frames to greyscale to produce a plurality of greyscale image frames;
compare pixel intensities between the plurality of greyscale image frames; and
detect the motion based on differences in at least some of the pixel intensities between the plurality of greyscale image frames exceeding a threshold value.
17. The sensor of claim 11 , wherein to apply the object detection process, the data storage device further includes instructions that when executed by the at least one processor cause the sensor to apply an adaptive neural network.
18. The method of claim 9 , wherein applying the motion detection process to the first set of images includes:
converting the first set of images to greyscale to produce a set of greyscale images;
comparing pixel intensities between the set of greyscale images; and
detecting the motion based on differences in at least some of the pixel intensities between the set of greyscale images exceeding a threshold value.