Long-time correlation filtering target tracking method based on adaptive feature fusion

The invention relates to a long-time correlation filtering target tracking method based on adaptive feature fusion, and belongs to the field of computer vision. The method comprises the following steps: in a target position prediction stage, respectively extracting an HOG feature and a color histogram feature of each frame of image, respectively obtaining a response diagram of each feature by using a kernel correlation filter and a color histogram model, and normalizing a peak value of the response diagram of each feature so as to dynamically adjust weights of different features; the adaptive fusion of the features is realized, and the position of the tracking target is estimated according to the fused feature response graph; two confidence coefficient detection indexes are set at the same time, and a detection filter and an SVM classifier are additionally trained and are respectively used for detecting whether a current tracking target is reliable or not, namely, whether the target is lost or not and repositioning the position of the lost target, so that long-time tracking of the target is realized.