Target tracking method based on kernel correlation filtering
The invention discloses a target tracking method based on kernel correlation filtering. The method comprises the steps that a first frame image is input, a reference sample of target SIFT features iscalculated, all target samples are obtained through a cyclic matrix, and then classifiers are trained...
Gespeichert in:
Hauptverfasser: | , , , , , , , , , , , , , , , |
---|---|
Format: | Patent |
Sprache: | chi ; eng |
Schlagworte: | |
Online-Zugang: | Volltext bestellen |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Zusammenfassung: | The invention discloses a target tracking method based on kernel correlation filtering. The method comprises the steps that a first frame image is input, a reference sample of target SIFT features iscalculated, all target samples are obtained through a cyclic matrix, and then classifiers are trained according to the target samples; a next frame image is input, a new target region is collected, and all training samples are calculated; a kernel correlation matrix of all the target samples and all the training samples and a classifier regression coefficient are calculated; and the kernel correlation matrix and the classifier regression coefficient are used to calculate function response values of all positions, a maximum response position is detected, and then a target position of a currentframe is calculated. According to the method, cyclic shift is performed by use of the target region of the cyclic matrix, dense sampling is realized, and sufficient samples are created for the classifiers; and then a kernel pr |
---|