Correlation filter based object tracking has recently gained popularity due to continuous improvements in the tracking accuracy and robustness. However, these trackers are limited by the model drift problem due to wrong target appearances learned from an incorrectly tracked frame. The model drift increases as more frames are processed and restricts the ability of long term tracking in correlation filter trackers. The proposed method introduces tracking resumption in correlation trackers using a detector mechanism that re-initializes the tracker upon a target loss identified using an adaptive threshold. Online training of both the tracker and detector modules incorporates temporal information into the proposed framework, making it robust to appearance changes of the object. The tracker and detector stages complement each other in correcting the false appearances learned from any frame, thereby mitigating the model-drift problem. A similarity matching technique estimates the final target location. Extensive experimental analysis on benchmark datasets indicate that the proposed tracker is well suited for robust long-term tracking and is superior to other state of the art methods both qualitatively and quantitatively. © 2019 Elsevier B.V.