FCT is a simple yet effective and efficient tracking algorithm, despite much success has been demonstrated, numerous issues remain to be addressed. In FCT, because of the sparsity of the compression measurement matrix, the spatial information of the sample is neglected, so the feature cannot represent the tracking target correctly and there is no remedy when tracking error. In this paper, we propose an improved fast compressive tracking algorithm considering the sample space information and extracts generalized Haar-like features randomly in block; target motion estimation method is used to correct target location, as the classifier is wrong. Adjusting the sparse degree of vector in compression measurement matrix and threshold of naive Bayes classifier can realize accurate target tracking. The experimental results show that compared with FCT, the improved algorithm achieves much better results in terms of both similarity and success rate and subjective visual perception.