運用動作分類之快速動作比對系統應用於門禁管制

A Fast Action Recognizing System Using Action Classification with Kinect Camera for Access Control System

施宇倫(Yu-Lun Shih)

指導教授 ： 張厥煒

國立臺北科技大學/電資學院/資訊工程系研究所/碩士(2013年)

https://doi.org/10.6841/NTUT.2013.00542

摘要

傳統的喇叭鎖、密碼電子鎖、指紋鎖以及手機的滑動解鎖等，皆是用來防範竊盜、未經允許使用，或是誤觸開啟等情況發生。本論文藉由具備深度資訊以及骨架資訊的體感器，實作出不同以往傳統特性的鎖，本研究稱為體感鎖。本次體感鎖的設計是透過體感器提供的深度資訊經由影像處理所得到之骨架座標資訊，再由使用者預先錄製數個連續動作，並設定動作先後順序形成獨特的體感鑰匙，最後系統會將動作單元以及體感鑰匙存入資料庫。解鎖時，只要依據先前設定動作單元先後順序做出動作即可解鎖。
而為提升動作比對的效率，本研究採用動作分類過濾器來分解動作單元的前後動作關係；結尾動作過濾器嘗試對齊動作單元的結尾，使用兩個過濾器先篩選出候選動作單元，最後再利用動態時間校正演算法計算出動作單元相似程度進而達到快速動作比對的目的。

關鍵字

體感攝影機 ； 動作比對 ； 動態時間校正

並列摘要

Nowadays people use every kind of locks in real world, such as cylindrical lock, iPhone’s “slide to unlock”, Android phone’s screen lock, fingerprint lock etc. Lock has been transformed to different ways with new technologies but with the same goal like: access control, anti-theft, accidently touch something or use without permission. According to the Kinect-like camera is more widely be used in living, human skeleton capture is becoming such easier than before. With human skeleton data, human action match can be recognized by calculating the skeleton’s data which provided by Kinect-like camera. In this paper, we present a fast action recognizing system for access control system trying to create a human action lock. Users can record their actions and setup the unlock sequence. To unlock, users just do the right actions and sequence.
To sum up, we proposed action classification filter and ending pose filter to enhance the efficiency of action matching. In the end, we calculate similarity score by full action matching using dynamic time warping algorithm.

並列關鍵字

Dynamic Time Warping ； Action Match ； Kinect Camera

參考文獻

[1] L. A. Schwarz, A. Mkhitaryan, D. Mateus, and N. Navab, "Human Skeleton Tracking from Depth Data Using Geodesic Distances and Optical Flow," Image and Vision Computing, vol. 30, no. 3, 2012, pp. 217-226.

[3] L. Teofilo, P. Nogueira, and P. Silva, "GEMINI: A Generic Multi-Modal Natural Interface Framework for Videogames," Advances in Information Systems and Technologies, Advances in Intelligent Systems and Computing A. Rocha, A. M. Correia, T. Wilson and K. A. Stroetmann, eds., Springer Berlin Heidelberg, 2013, pp. 873-884.

[5] PrimeSense. [Online]. Available: http://www.primesense.com/

[7] OpenNI. [Online]. Available: http://www.openni.org/

[11] R. Poppe, "A Survey on Vision-Based Human Action Recognition," Image and Vision Computing, vol. 28, no. 6, 2010, pp. 976-990.