在紅外線監控系統中,有別於一般可見光影像的特性,能夠發展出不同於可見光影像分析的應用。但紅外線影像品質不佳,低對比度、少特徵點等等缺點,所以先利用中值濾波器進行影像復原與重建,來提升紅外光影像的品質,此濾波器是基於排序統計理論的一種,能有效地去除雜訊,是一種非線性信號處理的技術,運算簡單、速度快。前置處理後,先建立參考背景影像,儲存在SDRAM,同時擷取連續影像與參考背景影像進行相減動作,計算出影像差異值。背景相減法後會因光源變化或移動快慢等因素造成雜訊及物件破碎的問題,以形態學處理之閉合運算來修補破碎不完整的區塊,斷開運算來去除不必要的雜訊或小區塊。為了達到移動物件追蹤,先將不同移動物件進行影像分割,採取方法為影像標籤化,並利用重心法分別求出重心坐標與邊界座標,最後將移動物件提取出來。 大多數監控系統多以PC作為平台,不僅效率低且佔據空間。本論文以Altera Cyclone II FPGA DE2-70為系統核心,紅外線攝影機透過ADV7180類比轉數位,ITU-R 656解碼電路,影像透過SDRAM當作資料緩衝並同時進行交錯式掃描,再經由許多模組進行影像演算法運算,接著把YCbCr訊號轉換為RGB色彩空間,最後將處理後的結果顯示於LCD上。系統皆以Verilog硬體電路完成,所以能快速地運算演算法,達到即時物件追蹤的功能。
Unlike optical image analysis characteristics, various applications can be developed for infrared monitoring systems. Because infrared images are characterized by poor quality, low contrast, and few feature points, median filters are employed to recover and reconstruct images to improve the quality of infrared images. Based on the theory of order statistics, median filters are an easy-calculating and fast non-linear signal processing technology that can denoise signals effectively. After pre-processing, the reference background images are established and saved on SDRAM. We then retrieve the sequence images and the reference background images to conduct background subtraction to calculate image differences. Following background subtraction, changes in light sources and the speed of moving objects can cause noise and object appears shattered. Thus, we use morphological closing operation to mend shattered, incomplete segments and morphological opening operation to remove unnecessary noises and small segments. To realize moving object tracking, we employ image labeling to segment various moving objects and the barycentric method to calculate the barycentric and boundary coordinates. Finally, the moving objects were retrieved. Most monitoring systems are PC platform based and have the disadvantages of low efficiency and space-occupation. The system core of this study was an Altera Cyclone II FPGA DE2-70. Using the ADV7180 analog-to-digital conversion and the ITU-R 656 decoding circuit, interlaced scans were conducted on the images using SDRAM as the data buffer and numerous modules to calculate video algorithms. YCbCr signals were then converted to RGB color spaces. Finally, the processed results were displayed on an LCD. The system was completed by the Verilog hardware circuit that enables the rapid application of algorithms on images, thereby realizing real-time recognition.