The performance of feature-aided tracking (FAT) relies largely on the quality of features extracted from signature sensors. In our previous work, we introduced a new FAT algorithm for tracking with ground moving target indicator (GMTI) and high resolutional range (HRR) measurements, where features were extracted from an HRR sensor using the technique of mixture density estimation. Although satisfactory results were achieved, an additional improvement is expected if a target rigidity constraint is incorporated. In this paper, we exploit the rigidity property of targets to alleviate the inherent local convergence problem of the expectation-maximization algorithm used in mixture density estimation. Simulation results show a performance improvement over the previous FAT algorithm.