| 摘要 |
Disclosed is a multi-target segmentation and tracking method based on a depth image, belonging to the field of image processing, mode recognition, computer application, etc. The method includes the following steps: collecting a depth image; updating background modeling; discovering a background; judging whether a foreground is found for the first time, and if the foreground is found for the first time, then performing target segmentation; if the foreground is not found for the first time, then performing dynamic target tracking; and judging whether there is a data input, and if there is a data input, then taking the new data to re-update the background modeling, otherwise, ending same. The present invention effectively improves the reliability of target recognition and tracking by way of highly efficient target segmentation processing and dynamic target tracking processing, which provides preparation information to further subsequent processing such as image analysis, dynamic modeling, 3D man/machine interaction, etc. |
