物件追蹤廣泛應用在民間及軍事上，在此應用領域有許多各式的演算法，其目標在於如何能夠更精準且穩定的追蹤目標物件，在實際應用上，除了考慮演算法的精準度及穩定度之外，也包括了在運算上的處理速度。本次研究目的在於改善物件追蹤演算法最小平均誤差和(Minimum Output Sum of Squared Error, MOSSE)，其特點在於處理速度快且容易實現，不過在追蹤過程中容易發生目標丟失的問題。本研究透過加入YOLO物件追蹤導入目標丟失後再追蹤的機制以及多模板的方式，改善MOSSE在執行物件追蹤時，由於環境因素影響例如目標丟失以及不同視角下的場景因素，造成無法順利追蹤的窘境。經由在Object Tracking Benchmark (OTB)上的27支影片進行追蹤性能測試，以時序上穩健性評估(Temporal Robustness Evaluation, TRE)曲線下面積(Area under curve, AUC)作為評估指標，透過加入再追蹤的機制，將原先MOSSE的性能從0.25提升至0.308；此外以多模板的方式，也讓MOSSE的表現從0.25提升至0.302；最後將多模板以及再追蹤的機制合在一起更將AUC指標提升至0.346。驗證本研究以YOLO及多模板的方式，可有效地提升MOSSE演算法之性能。

The application of object tracking is widely used in civil and military. The goal of the object tracking algorithms is how to track the target precisely and robustly. Practically, it also considers computing cost and hardware capability in the real-time application. This research aims to improve the Minimum Output Sum of Squared Error (MOSSE) in tracking performance. The characteristics of MOSSE are the fast processing speed and easy to implement, but it is easy to lose the target during the tracking because it is based on 2D image and requires continuous updating the position of the object. We add a re-track mode with YOLO, which is an algorithm of object detection, to provide MOSSE the potential positions to track the target again. Also, there are several challenges such as appearance distortion and rotation that cause MOSSE to lose the target. We use the different views of target appearance to get the multiple templates to improve the tracking performance of MOSSE. The tracking performance was tested by 27 videos on the Object Tracking Benchmark (OTB) and used Temporal Robustness Evaluation (TRE) and Area under curve (AUC) as the evaluation indices. The indices show that the performance compared to MOSSE is improved from 0.25 to 0.308 after including the re-track mode. The performance of adaptive multiple templates is improved from 0.25 to 0.302. The combination of the adaptive multiple templates and the re-track mechanism can improve the AUC to 0.346. The results verify that we can use YOLO and multiple templates to improve the performance of MOSSE effectively.

[1] Wu, Y. Lim, J. and Yang, M. H., 2013, “Online Object Tracking: A Benchmark,” Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 2411-2418.
[2] Verma, R., 2017, “A Review of Object Detection and Tracking Methods,” International Journal of Advance Engineering and Research Development, 4(10), pp. 569-578.
[3] Soleimanitaleb, Z. Keyvanrad, M. A. and Jafari, A., 2019, “Object Tracking Methods: A Review,” 2019 9th International Conference on Computer and Knowledge Engineering, ICCKE, pp. 282-288.
[4] Rinosha, J. S. M. Augasta, G., 2021, “Review of recent advances in visual tracking techniques,” Multimedia Tools and Applications, 80.
[5] Danelljan, M. Häger, G. Khan, F. S. and Felsberg, M., 2017, “Discriminative Scale Space Tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 39(8), pp. 1561-1575.
[6] Hare, S. Golodetz, S. and Saffari, A., 2016, “Struck: Structured Output Tracking with Kernels,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 38(10), pp. 2096- 2109.
[7] Babenko, B. Yang, M.-H. and Belongie, S., 2009, “Visual Tracking with Online Multiple Instance Learning,” 2009 IEEE Conference on Computer Vision and Pattern Recognition, pp. 983–990.
[8] Fiaz, M. Mahmood, A. Javed, S. and Jung, S. K., 2019, “Handcrafted and deep trackers: Recent visual object tracking approaches and trends,” ACM Computing Surveys, 52(2), Paper Number a43.
[9] Kumar, A. Walia, G. S. and Sharma, K., 2020, “Recent trends in multicue based visual tracking: A review,” Expert Systems with Applications, 162(30), Paper Number 113711.
[10] Bolme, D. S. Beveridge, J. R. Draper, B. A. and Lui, Y. M., 2010, “Visual Object Tracking using Adaptive Correlation Filters,” 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, CA, USA, pp. 2544-2550.
[11] Henriques, J. F. Caseiro, R. Martins, P. and Batista, J., 2015, “High-Speed Tracking with Kernelized Correlation Filters,” IEEE Transactions on Pattern Analysis and Machine Intelligence, 37(3), pp. 583-596.
[12] Shin, J. Kim, H. Jeong, D. and Palk, J., 2019, “Automatic Failure Detection and Correction for Real-Time Object Tracking with Kernelized Correlation Filter,” 2019 IEEE International Conference on Consumer Electronics (ICCE).
[13] Yuan, D. Lu, X. and Li , D., 2019, “Particle filter re-detection for visual tracking via correlation filters,” Multimedia Tools and Applications, 78(11), pp. 14277-14301.
[14] Kim, B. H. Lukezic, A. Lee, J. H. Jung, H. M and Kim, M. Y., 2020, “Global Motion-Aware Robust Visual Object Tracking for Electro Optical Targeting Systems,” Sensors, 20(2), Paper Number 566.
[15] Aziz, N. N. A. Mustafah, Y. M. and Azman, A. W., 2018, “Features-Based Moving Objects Tracking for Smart Video Surveillances: A Review,” International Journal on Artificial Intelligence Tools, 27(2), Paper Number 1830001.
[16] Jiang, M. Li, R. Liu, Q. Shi, Y. and T, E., 2021, “High Speed Long-Term Visual Object Tracking Algorithm for Real Robot Systems,” Neurocomputing, 434, pp. 268-284.
[17] Jie, Z. Lei, L. Xiaqing, G. and Shanmei, L., 2020, “Tracking and Landing of QUAV under Complex Conditions based on Ground-Air Information Fusion,” 2020 IEEE 2nd International Conference on Civil Aviation Safety and Information Technology (ICCASIT), Paper Number 9368815.
[18] He, S. Wang, H. Zhang, S. and Tian, B., 2019, “Vision Based Autonomous Tracking and Landing on an Arbitrary Object of Quadrotor,” 2019 IEEE 9th Annual International Conference on CYBER Technology in Automation, Control, and Intelligent Systems (CYBER), Paper Number 9066614.
[19] Feng, Q. Ablavsky, V. Bai, Q. Li, G. and Sclaroff, S, 2020, “Real-Time Visual Object Tracking with Natural Language Description,” Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), pp. 700-709.
[20] Jacobs, D., 2005, “Correlation and Convolution,” Journal CMSC, pp. 426.
[21] Redmon, J. Divvala, S. Girshick, R. and Farhadi, A., 2016, “You Only Look Once: Unified, Real-Time Object Detection,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 779-788.
[22] Redmon, J. and Farhadi, A., 2017 “YOLO9000: Better, Faster, Stronger,” Proceedings of the IEEE conference on computer vision and pattern recognition.
[23] Everingham, M. Gool, L. V. Williams, C. K. Winn, J. and Zisserman, A., 2010, “The Pascal Visual Object Classes (voc) Challenge,” International journal of computer vision, 88(2), pp. 303- 338.
[24] Lin, T. Y. Maire, M. Belongie, S. Hays, J. Perona, P. Ramanan, D. Dollar, P. and Zitnick, C. L., 2014, “Microsoft coco: Common Objects in Context,” In European Conference on Computer Vision, pp. 740–755.
[25] Bochkovskiy, A. Wang, C.-Y and Liao, H.-Y. M., 2020, “Yolov4: Optimal Speed and Accuracy of Object Detection,” arXiv preprint arXiv:2004.10934.