A Kalman filter based image tracking algorithm is developed. When target is locked, a Kalman filter model is utilized to track the target under a partially shaded manoeuvring path and a set of invariant features of homomorphic skew-ness and kurtosis, which are derived from the spectrum histogram of the target image, is used to check whether the target has breaklocked or not in every image frame during the tracking phase. Another set of invariant features derived from the geometrical shape of the target image is used as checking indices. After simulations, we find that the tracking algorithm combining the Kalman filter model with the invariant features of skewness and kurtosis has a good performance even at a signal to noise ratio as low as 0·4 dB.