用於攝像機監控系統的多配置物件偵測框架

Multi Configuration Object Detection Framework for Camera Surveillance System

Memen Hamzah Obaid Salihi(Memen Hamzah Obaid Salihi)

指導教授 ： Hung-Yu Wei ； Michele Zorzi

國立臺灣大學/電機資訊學院/電信工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202201030

摘要

兩階段物件偵測方法所產生的高延遲限制了影像偵測的使用情境。因此，現有的研究僅將低負載物件偵測方法應用於物聯網裝置，但這些探測器的問題在於低準確性。在延遲性和準確性的權衡下，我們提出了多配置架構，將每組攝像機都設置在一個配置下。解決最佳化問題後，系統基於最低成本的配置下即能決定在哪處理任務；運算決策可以在本地（邊緣）或遠端（雲端）執行。實驗結果顯示。本研究所提出的多配置物件偵測架構在低延遲性和高準確性的表現優於現有的單一配置系統。

關鍵字

監視系統 ； 邊緣雲端運算 ； 物件偵測 ； 物聯網

並列摘要

The high delay produced by the two-stage Object detection methods is limiting their use in video surveillance scenario. Therefore, most of the existing works use only light object detection method to be applied on IoT devices.
The problem of these detectors is the low accuracy. To find a trade off between thelatency and the accuracy we proposed multi configuration framework where each group of cameras are set under one configuration. After solving the optimization problem, the system will be able to decide where to process the tasks based on the minimum cost for all the configurations.The computation decision can be done locally (Edge) or remotely (cloud).
The experiment result demonstrates that the proposed multi-configuration object detection framework outperforms the existing uni configuration system in terms of lower latency and higher accuracy.

並列關鍵字

Surveillance system ； Edge-Cloud computing ； Object detection ； IoT

參考文獻

References

[1] Anaconda software distribution, 2020.

[2] M. Abadi, P. Barham, J. Chen, Z. Chen, A. Davis, J. Dean, M. Devin, S. Ghemawat, G. Irving, M. Isard, M. Kudlur, J. Levenberg, R. Monga, S. Moore, D. G.Murray, B. Steiner, P. Tucker, V. Vasudevan, P. Warden, M. Wicke, Y. Yu, and X. Zheng. Tensorflow: A system for large-scale machine learning. In 12th USENIX Symposium on Operating Systems Design and Implementation (OSDI 16), pages 265–283, 2016.

[3] S. M. Beitzel, E. C. Jensen, and O. Frieder. MAP, pages 1691–1692. Springer US, Boston, MA, 2009.

[4] S. Bi and Y. Zhang. Computation rate maximization for wireless powered mobile- edge computing with binary computation offloading. IEEE Transactions on Wireless Communications, 17(6):4177–4190, 2018.