A compressive template tracking algorithm based on online learning is proposed by using the compressive sensing (CS) matrix in this paper.Firstly, the compressive template of the object appearance feature is set up with a very sparse random CS matrix that satisfies the restricted isometry property (RIP).Se condly,the target area is determined by template matching,and the appearance change of the target is learned in the tracking by w hich the template is update.The local template update strategy is applied in which the foreground features and t he background ones in the template are updated,respectively.The strategy produces the more accurate target templat e,which can reflect the changes of the target in real time.Our algorithm adopts the deterministic tracking method,w hich greatly reduces the sampling of the positive and negative samples,and reduces the complexity of algorithm.At t he same time,the local template update strategy of online learning ensures the accuracy of target feature.Compar ed with compressive tracking (CT) algorithm,the time performance of our algorithm is raised obviously,and the time performance o f our algorithm adopted by Kalman filter is improved by more than 6times.The tr acking success rate of our algorithm is also improved greatly which prevents the drift of target in changing scenarios.