Nikon Working on Improvements to Face Detection and Tracking Technology
We can exclusively reveal today that Nikon are actively working on improvements to their already powerful face detection and tracking technology. A patent published towards the end of last year reveals details of the upcoming improvements, which include changes in the way the system handles multiple faces within the same scene. Our analysis of the new information is that the system is able to ‘remember’ faces based on measurements of the position of facial features, and then use the stored proportions to track where subjects are within a scene. Currently, Nikon’s Face Tracking AF allows you to use the camera’s joystick to select between up to 35 different faces in a single image to set them as the point of focus. This patent appears to suggest improvements that would allow the camera to more accurately detect whether a face is a new face within the scene, or a pre-existing one that has simply moved closer or further away to the camera.
Face detection is widely used across Nikon’s range of digital cameras and their ongoing development of the feature could lead to exciting new capabilities in the near future!
PROBLEM TO BE SOLVED: To enable setting of a region of a feature which is most suitable as a region used for AF adjustment out of plural features detected from an imaging range.
SOLUTION: An imaging device includes an imaging element for repeatedly imaging a light flux passed through an optical system to generate an image signal, detection means for detecting a first feature within an image based on the image signal, setting means for setting the position of the first feature in the image as a target position, comparison means for comparing the first feature and a second feature different from the first feature when the second feature is detected from an image based on an image signal which is an image signal out of repetitively captured image signals and obtained by the imaging element after the position of the first feature is set as the target position, and control means for making the setting means reset the position of the second feature in the image as the target position based on the comparison result of the comparison means.
For example, in performing face detection using through image data for the first time, characteristic quantity of the face detected from through image data in this case is carried out with characteristic quantity information from characteristic quantity information not being memorized at the internal memory 34, and it memorizes to the internal memory 34.
On the other hand, when face detection using through image data is performed in the past, characteristic quantity information is memorized by the internal memory 34. The face primary detecting element 42 compares the characteristic quantity of the face detected from through image data with the characteristic quantity information memorized by the internal memory 34. When the characteristic quantity of the face detected from through image data by this comparison corresponds with characteristic quantity information, it is judged with that face being a face detected in the past. In this case, even if it is a case where characteristic quantity is in agreement, while the size and position of a face area are changing, characteristic quantity information is rewritten. Even if it is a case where the face of the object included in a photographic range is in a photographic range, and moves by this, it becomes possible to track the face. On the other hand, when the characteristic quantity and the characteristic quantity information on a face which are detected from through image data are not in agreement, it judges with it being the newly detected face. In this case, the characteristic quantity of the face judged that is a newly detected face, and the position in the size and photographic range of a face area are summarized, characteristic quantity information is created, and it writes in the internal memory 34.
- Publication number: 2016-197893
- Date of publication of application: 24.11.2016
- Application number: 2016-133754
- Date of filing: 05.07.2016
- Applicant: NIKON CORP