近年來,校園霸凌事件層出不窮,學校也藉由裝設大量攝影監視器試圖預防、減少及發現覇凌事件。但受限於人力,無法隨時監看大量的錄影內容,而無法即時防止霸凌事件的發生。因此,本研究利用影像辨識技術建置一套警示系統,藉由偵測錄影監視畫面之異常現象,並即時發出警示,以提醒校安人員前往處理,防止霸凌產生嚴重傷害。 本論文的系統架構主要包含三大模組:(1)前景影像偵測模組-運用背景相減法將感興趣的影像分離並轉換為黑白影像,利用侵蝕與擴散運算去除雜訊並強化影像;(2)物件辨識模組-以相鄰元素編號演算法將畫面中影像分割成個別物件,並使用Otsu 演算法原理判斷物件是否為接觸物件;(3)動作偵測模組-依據物件長寬比例,配合系統設定參數,判別物件目前動作是否異常,並進行加權計算,以利觸發警示模組。最後,透過真實影片的實驗測試,驗證本系統確實可達到覇凌行為之偵測效果。
Recently, bullying seems to emerge in an endless stream. Although schools attempt to prevent, reduce and reveal the bullying by equipping with a great deal of photography monitors, bullying still can’t be prevented in real time for lacking of supervising labor. Therefore, a bullying detection system based on image recognition technology is introduced in this research. With detecting the abnormal scene and alarming in real time, the guard will be alerted and go for preventing the severely harm caused by bullying. In this research, the system consists of three main module: (1) the foreground image detection module; (2) object recognition module; (3) motion detection module. To conclude, the system will be verified with a real video test to prove the accomplishment of this research.