隨著科技的進步，多軸飛行器的性能越來越好，價格越來越低廉，因
此發展也越發蓬勃。多軸飛行器可以應用在軍事、運輸、娛樂、救援、環 保、科學，工程，…等，其中應用在娛樂的空拍最為常見，空拍多用於運 動、攝影、環保、探勘，監視…等。通常多軸機是用搖控器控制其飛行路 徑，所以當要追蹤目標物時，就可以用遙控器手動操控多軸機追蹤目標物。
為了追蹤方便，多軸機後來發展出被動式追蹤，也就是在被追蹤目標上放
一個被追蹤裝置，多軸飛行器透過該裝置的訊號來跟蹤，但追蹤非特定目 標時就無法事先放置被追蹤裝置在目標上。
在本論文中，我們提出多軸機電腦視覺追蹤系統，在多軸機上裝設相
機，透過電腦視覺達到追蹤任意目標而不需額外被追蹤裝置的目的。系統
分為兩個部分，一個是電腦視覺追蹤、二是多軸飛行器控制，電腦視覺追 蹤是利用改良後可適應尺度變化的 Kernel correlation filter (KCF) 進行追 蹤，當 KCF 追蹤失敗後，即刻使用基於特徵匹配的再偵測演算法重新找 回目標。電腦視覺追蹤演算法執行於 NVIDIA TK1 嵌入式板子上， Arduino 接收 NVIDIA TK1 的控制指令，接著轉成脈衝寬度調變 (pulse width modulation, PWM) 訊號給多軸飛行器，讓飛行器能持續跟著目標。
在實驗分析中，我們有縮放和遮蔽兩種特殊影片，用來測試演算法對
於縮放與遮蔽的追蹤能力；除此之外，我們利用追蹤率、重疊準確率和執
行速度做為評估依據，先用人工方式框出目標在影像中的位置做為成功
追蹤的判定依據，接著利用這些框出的位置計算演算法的追蹤率、重疊準 確率和執行速度。我們的演算法執行速度26 fps、在縮放的案例中追蹤率 88%、在遮蔽的案例中追蹤率 98%、重疊準確度 87%，在速度與準確率 的取捨中，我們犧牲了演算法執行速度來換取更穩健的追蹤。

With the progress of science and technology, the guiding of multicopters is getting better, cheaper, and more affordable. The applications of multicopters are also progressively prosperous. Multicopters can be used in military, transportation, entertainment, rescue, environmental protection, science, engineering, etc. In which, the applications for aerial photography is most popular. Frequently, aerial photography is used in sports, photography, environmental protection, exploration, monitoring, etc. In general, the flying path of a multicopters is controlled by a remote controller. When we want to track a target, we can use a remote controller manually controlling the multicopters to follow the moving target. In order to conveniently track targets, a multicopter is equipped by a passive tracking device to follow a specified target. However, if we want to track a non-controlled target, the passive tracking device is failed.
In this paper, we propose a vision-based tracking system for multicopters. We install a camera on a multicopter, through computer vision method to track any target without additional tracking devices. The system is divided into two parts, one is the computer vision tracking module and the other is the multicopters control module. Computer vision tracking module is implemented by an adaptable scaled KCF algorithm, when the KCF tracking is failed, a feature-based matching detector is then used to re-detect the target. The computer vision tracking algorithm is executed on an NVIDIA TK1 embedded board. An Arduino MCU receives the NVIDIA TK1 control instruction, then generates a pulse width modulation (PWM) signal to the multicopter to control the multicopter to continuously track the target.
In experiments, there are two cases of scaling and occluded targets in the videos to be evaluated. The experimental results reveal that the proposal method can adapt the change of scale and occlusion of targets. In addition to the above ability, the tracking rate, accuracy, and execution speed are also evaluated. The proposed algorithm performs speed to 26 fps, tracking rate in the scaled cases can reach 88%, tracking rate in occluded cases can reach 98%, and overlap rate can reach 87%. With trade-off between speed and accuracy, we sacrifice the execution speed of the algorithm to exchange for more robust tracking.

[1] K. Zhang, L. Zhang, and M.-H. Yang, "Real-time compressive tracking," in Proc. European Conf. on Computer Vision, Firenze, Italy, Oct.7-13, 2012, vol.7574, no.3, pp.864-877.
[2] Z. Kalal, K. Mikolajczyk, and J. Matas, "Tracking-learning-detection," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.34, no.7, pp.1409-1422, 2012.
[3] B. D. Lucas and T. Kanade, "An iterative image registration technique with an application to stereo vision," in Proc. Int. Joint Conf. on Artificial Intelligence, Vancouver, British Columbia, Canada, Aug.24-28, 1981, pp.674-679.
[4] R. E. Kalman, "A new approach to linear filtering and prediction problems," Journal of Basic Engineering, vol.82, no.1, pp.35-45, 1960.
[5] D. Comaniciu and P. Meer, "Mean shift: a robust approach toward feature space analysis," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.24, no.5, pp.603-619, 2002.
[6] K. Kim, T. H. Chalidabhongse, D. Harwood, and L. Davis, "Real-time foreground-background segmentation using codebook model," Real-Time Imaging, vol.11, no.3, pp.172-185, 2005.
[7] C. Stauffer and W. E. L. Grimson, "Adaptive background mixture models for real-time tracking," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, Fort Collins, CO, Jun.23-25, 1999, vol.2, pp.246-252.
[8] E. Rosten and T. Drummond, "Machine learning for high-speed corner detection," in Proc. European Conf. on Computer Vision, Graz, Austria, May 7-13, 2006, vol.3951, pp.430-443.
[9] C. Harris and M. Stephens, "A combined corner and edge detector," in Proc. Alvey Vision Conf., Oxford, UK, Aug.31-Sep.2, 1988, pp.147-151.
[10] M. Calonder, V. Lepetit, C. Strecha, and P. Fua, "BRIEF: binary robust independent elementary features," in Proc. European Conf. on Computer Vision, Hersonissos, Greece, Sep.5-11, 2010, vol.6314, no.4, pp.778-792.
[11] D. G. Lowe, "Distinctive image features from scale invariant keypoints," Int. Journal of Computer Vision, vol.60, no.2, pp.91-110, 2004.
[12] H. Bay, T. Tuytelaars, and L. VanGool, "SURF: Speeded up robust features," in Proc. European Conf. on Computer Vision, Graz, Austria, May 7-13, 2006,vol. 3951, pp.404-417.
[13] X. Cheng, N. Li, S. Zhang, and Z. Wu, "Robust visual tracking with SIFT features and fragments based on particle swarm optimization," Circuits, Systems, and Signal Processing, vol.33, no.5, pp.1507-1526, 2014.
[14] J. Kennedy and R. Eberhart, "Particle swarm optimization," in Proc. IEEE Conf. on Particle swarm optimization, Perth, Australia, Nov.27-Dec.1, 1995, vol.4. pp.1942-1948.
[15] Q. Miao, G. Wang, C. Shi, X. Lin, and Z. Ruan, "A new framework for on-line object tracking based on SURF," Pattern Recognition Letters, vol.32, no.13, pp.1564-1571, 2011.
[16] H. Grabner and H. Bischof, "On-line boosting and vision," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, New York, NY, Jun.17-22, 2006, vol.1, pp.260-267.
[17] I. Leichter, M. Lindenbaum, and E. Rivlin, "Mean shift tracking with multiple reference color histograms," Computer Vision and Image Understanding, vol.114, no.3, pp.400-408, 2010.
[18] Z. Kalal, K. Mikolajczyk, and J. Matas, "Forward-backward error: automatic detection of tracking failures," in Proc. IEEE Conf. on Pattern Recognition, Istanbul, Turkey, Aug.23-26, 2010, pp.2756-2759.
[19] T. Ojala, M. Pietikainen, and T. Maenpaa, "Multiresolution gray-scale and rotation invariant texture classification with local binary patterns," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.24, no.7, pp.971-987, 2002.
[20] D. Bolme, J. R. Beveridge, B. A. Draper, and Y. M. Lui, "Visual object tracking using adaptive correlation filters," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, San Francisco, CA, Jun.13-18, 2010, pp.2544-2550.
[21] C. Cortes and V. Vapnik, "Support-vector networks," Machine Learning, vol.20, no.3, pp.273-297, 1995.
[22] Y. Freund and R. Schapire, "A desicion-theoretic generalization of on-line learning and an application to boosting," Computer and System Sciences, vol.55, no.1, pp.119-139, 1995.
[23] H. Lu, W. Zhang, F. Yang, and X. Wang, "Robust tracking based on PSO and on-line AdaBoost," in Proc. IEEE Conf. on Intelligent Information Hiding and Multimedia Signal Processing, Kyoto, Japan, Sep.12-14, 2009, pp.690-693.
[24] W. Zhu, S. Wang, R.-S. Lin, and S. Levinson, "Tracking of object with SVM regression," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, Kauai, Hawaii, Dec.8-14, 2001, vol.2, pp.240-245.
[25] B. Babenko, M.-H. Yang, and S. Belongie, "Visual tracking with online multiple instance learning," in Proc. IEEE Conf. on Computer Vision and Pattern Recognition, Miami, FL, Jun.20-25, 2009, pp.983-990.
[26] J. F. Henriques, R. Caseiro, P. Martins, and J. Batista, "High-speed tracking with kernelized correlation filters," IEEE Trans. on Pattern Analysis and Machine Intelligence, vol.37, no.3, pp.583-596, 2015.
[27] E. Rublee, V. Rabaud, K. Konolige, and G. Bradski, "ORB: An efficient alternative to SIFT or SURF," in Proc. IEEE Conf. on Computer Vision, Barcelona, Spain, Nov.6-13, 2011, pp.2564-2571.