基於異質攝影機整合之智慧型監控系統

Abstract

在智慧型監控應用中，不同的攝影機有不同的優缺點，一般的可見光攝影機會受到環境的影響，在光源不足的地方無法清楚地辨識出環境中是否有可疑人物。經常使用於夜間的近紅外線攝影機由於是主動式打光，可偵測的距離有限。中紅外線攝影機雖然可不受環境影響清楚辨識場景中的高溫人物，但背景較模糊且價格昂貴。本論文將以整合兩種性質不同的攝影機為目的，提出兩個分別在日間與夜間使用之監控系統。前者利用可見光與中紅外線攝影機偵測可見光影像中是否有無法辨識的可疑人物，若有則發出警訊，並且會在可見光影像中畫出此人物之近似輪廓。而後者則使用近紅外線與中紅外線攝影機動態拍攝，並利用夜間環境中預先架設好的近紅外線LED當作地面定位參考點，對環境中高溫物體進行定位。

Different types of camera have different advantages and disadvantages for intelligent surveillance applications. Visible cameras cannot correctly recognize suspicious humans when the illumination is insufficient. Near-infrared cameras, which are often used the nighttime and illuminate the environment by LEDs, usually have limited depth of view. Mid-infrared cameras can correctly recognize objects of high temperature in different environments, but the background of the image can be blurred, and the price of camera is high. In this thesis, by integrating heterogeneous cameras, we propose two systems for daytime and nighttime surveillance, respectively. In the daytime system, a visible camera and a mid-infrared camera are used. When the visible camera cannot recognize an object correctly due to low illumination, an approximate outline will be drawn using the data from another camera, and a warning message will be issued. In the nighttime system a near-infrared camera and a mid-infrared camera are used. The system will locate the objects of high temperature using near-infrared LEDs with known positions on the floor.