智慧安控系統

Abstract

摘要 本專題研究主要目的為辨識出人體的各處關節點與姿勢，並判斷偵測中的人是否做出翻牆動作進而觸發警報，以達到安控的目的。 使用者可以利用外接鏡頭並擺放至需要偵測的場所位置，並依照需求調整觸發警報之牆位高度以及斜率，此系統會藉由偵測腰部節點是否超過偵測線以判斷畫面中人物是否正在進行翻牆的動作，如此一來可以避免經過牆前的行人以及偶然在牆下蹲下撿東西或綁鞋帶的人誤觸警報。 另外，此系統將會即時偵測並計算出現在畫面中的人數以及觸發警報之人數，並顯示在畫面上告知使用者，使用者可以藉由人物身上的紅框快速辨別觸發警報的人物。系統在每次有人觸發危險警報時，會在電腦中建檔將觸發警報時之螢幕畫面儲存為照片，並以當時之時間日期命名資料夾，讓使用者後續得以方便快速的查找偵測紀錄。在本專題研究中，此系統已能準確偵測人體並達到以上功能。

Abstract The main purpose of this independent study is to identify the joints and postures of the human body, and determine whether the detected person makes a movement over the wall to trigger an alarm to achieve the purpose of security control. Users can use external lenses and place them in the place that needs to be detected, and adjust the height and slope of the wall that triggers the alarm according to their needs, the system can detect whether the waist node exceeds the detection line to determine whether the character in the picture is moving over the wall, so as to avoid pedestrians passing in front of the wall and people who accidentally squat under the wall to pick up things or tie their shoes by mistake. In addition, the system will detect and calculate the number of people appearing in the screen and the number of people who trigger the alarm in real time, and display it on the screen to inform the user, and the user can quickly identify the person who triggered the alarm through the red frame on the character. Every time someone triggers a danger alarm, the system will create a file in the computer to save the screen when the alarm is triggered as a photo, and name the folder with the time and date at that time, so that users can easily and quickly find the detection record in the future. In this independent study, this system has been able to accurately detect the human body and achieve the above functions.