The invention provides a correlation filter tracking method based on multi-feature fusion. The method comprises the following steps: the target position and the size of a target block in an image are read at a t-1 moment; in the image, a target area with the target position as a center is extracted, and the HOG feature, the depth color feature and the depth motion feature of the target area are calculated; according to the HOG feature, the depth color feature and the depth motion feature, a filter is built; data of the image at a t moment are read, a candidate area is extracted with the target position at the t-1 moment as the center in the image, and the HOG feature, the depth color feature and the depth motion feature of the candidate area are calculated; in the image at the t moment, the candidate area is extracted with the target position at the t moment as the center, and the HOG feature, the depth color feature and the depth motion feature of the candidate area are calculated; and according to the HOG feature, the depth color feature and the depth motion feature at the t moment, the filter is updated. The method is a robust target tracking method.