To solve the tracking drift problem caused by object appearance change in complex environments, the paper proposes an object tracking algorithm on the basis of multi-feature fusion and classifier online learning. The algorithm trains the sub-classifier with different apparent features, and calculates the reliability of each classifier by building the loss function, and then the optimum target state estimation by means of the weighted fusion prediction results of each sub-classifier is obtained. Meanwhile, it updates the training sample set coarsely according to the nearest-farthest boundary principle as well as the co-training theory, and gets more representative ones with the refined selection criterion, which further updates the sub-classifier adaptively. Experimental evaluations demonstrate that the proposed algorithm achieves favorable tracking performance against state-of-the-art methods on various typical testing scenarios.