Primary user selection for head tracking
摘要
說(shuō)明書(shū)
Once a face pair is created, a subsequent image captured by each camera is analyzed to detect the primary face (e.g., tracking-by-detection). Some tracking algorithms are designed to track a face only when it is detected in a previous image. Placing such a limitation, however, could render such a method inoperable when a face is not recognized in, for example, a second image by the face detection algorithm, but is detected in the first and third image in a respective sequence. In such an instance, a second tracking algorithm, such as a Median Flow Tracker (MFT), can be used when a face is detected in a previous image captured by each camera, but is not detected in any of the current images from the cameras. Thus, in this example, the MFT algorithm detects the face in at least one of the current images and, then, in a subsequent frame, the process of attempting to detect the face using the face detection algorithm resumes. Therefore, if the face pair is detected in a previous frame, tracking-by-detection is performed. If the face pair is not detected in the previous frame, the MFT algorithm is used.
