This project showcases an implementation of the RCNNshift，which is a kernel-based tracker that uses new features generated by the random ignition mechanism of Random-coupled Neural Network,significantly improves the stability and accuracy of tracking.