使用深度點雲技術手勢手形比對門禁系統

 

A Hand Gesture and Shape Matching Access Control System Using Depth Point Cloud Technology

 

厥煒

國立北科技大學 資訊工程系

忠孝東路31

台北市106大安區

cwchang@ntut.edu.tw

 

林鈺皓

國立北科技大學 資訊工程研究所

忠孝東路31

台北市106大安區

peter1205@gmail.com

 

 

摘要

本論文運用3D「手勢與手形」結合「鎖」的概念,設計並實作出一種新的以互動方式解鎖的「手勢鎖」。藉由每個人手勢手形特徵及手部習慣動作的不同,所產生出來的個別獨特生物特徵,做為解鎖時的依據。

此「手勢鎖」系統透過RGBD感測器,追蹤其手部區域的影像資訊,利用手部遮罩圖疊合點雲影像,取得完整的手部點雲資訊。再利用點雲技術3D物件辨識演算法,進行手勢點雲的特徵抽取與比對。使用者在解鎖時,依序比劃出先前所設定儲存的手勢手形鑰匙,透過手勢點雲的3D形狀與順序排列,取得其對應的密碼序列,系統即可成功解鎖。經實驗及實測結果,此方法有其安全性精確性與可行性。

 

關鍵詞: 深度感測器、手勢辨識、點雲技術、門禁系統、3D形狀比對。

 

Abstract

In this paper, we design and implement a new kind of interactive lock-open method for the access control, called “Hand Gesture Lock”, by using the idea of 3D hand gesture and hand shape matching approach. This lock-open method is based on the familiar behavior of hand shape change or the hand moving trajectory, or even the hand size of each person. And these familiar behavior of hand can be treated as a unique biometric feature to solve the lock-open problem.

This “Hand Gesture Lock” system is designed by four basic steps: (1) acquiring the RGB and depth images from an RGBD sensor, (2) tracking the moving hand position, (3) extracting the 3D hand shape depth image from the hand position mask, (4) matching the 3D hand shape object with the target hand shapes in the pre-trained database by the point cloud ICP technology. When a user wants to open the lock, he needs to show both the correct hand gestures and he’s correct pre-set corresponding gesture code sequence. After several experiments and testing, this system shows very secure, accuracy and feasibility.

 

Keywords: Depth Sensor, Hand Gesture Recognition, Point Cloud Technology, Access Control System, 3D Shape Matching