This paper proposes a method of detecting moving objects against changing background based on correlative feature matching. The result of the mean of pixel value minus the variance of pixel value is used as the adaptive threshold to separate the object from the background; the features are extracted and parameterized from the image; using the Euclidean distance between the feature to detect objects; Filtering the static object through three continuous frame. This method reduce the calculation greatly compare to the method using correlative parameter

and the method is not subject to the same background between the neighbouring frames. The validity of the method is proved by experiment.