Multi-target Real-time Tracking Technology Based On Android Platform

In the past, owing to the visual tracking technology for hardware and real-time requirements, which is only applied to PC, and some embedded platforms. However, mobile devices in recent years has undergone enormous changes, and its hardware performance has been greatly improved, which will provide hardware technical support for us to apply visual tracking technology to mobile devices. Now Android platform is the most user-owned mobile device platform, visual tracking algorithm applied to the Android platform will have great significance.In this thesis, a lot of research work on visual tracking algorithm, and compared from today’s popular visual tracking algorithm, we found Kernelized Correlation Filter(KCF) tracking algorithm in many scenarios have good robustness and high real-time performance, which is very suitable for transplant to the Android platform. This thesis makes a deep research on the algorithm, for it can not track the target with multiple scales, an improved multi-scale KCF tracking algorithm was proposed. we transplanted the improved algorithm to the Android platform, designed and completed a multi object real-time tracking system. A detailed study of the expanded work as follows:(1) On the KCF tracking algorithm in-depth study, a detailed analysis of the principles and processes of the algorithm, moreover, a large number of experiments are made to prove KCF tracking algorithm is robustness and high real time.(2) Since original KCF algorithm can not track the target with multiple scales. Based on the original KCF algorithm combines the bilinear interpolation algorithm, a multi-scale KCF tracking algorithm improved. A large number of experiments are made to verify the improvement of the new algorithm.(3) With improved multi-scale KCF tracking algorithm as the foundation, based on the Android platform design and completed a multi target real time tracking system. The robustness and real-time of the system are tested, and it proves that the system can effectively track multiple targets in real time.