使用大範圍高解析視訊監控系統從事目標物之偵測與影子去除

Abstract

隨著國際間恐怖攻擊事件頻傳，世界各國對於反恐意識提升，於近年來相繼於境內各大城市或重要地點架設視訊安全監控系統。在視訊安全監控系統中，大範圍、高解析度的監控畫面以及智慧化的監控系統是不可或缺的。構建此系統必需同時擁有高品質的輸入影像和顯示裝置以及自動化偵測物體的功能。 鑑於高品質的顯示裝置發展迅速，而高解析度監控攝影機發展緩慢，與顯示器相比並沒有廣泛的被使用。在本文章的第一部分中，我們設計了一個創新的攝影機架構，包含固定式廣角攝影機以及高解析度快速球型攝影機，來建構大範圍、多倍數和多重解析度的視訊監控系統，其提供多重解析度的移動物體資訊。 首先，我們發展一套新的攝影機校正方式，計算固定式廣角攝影機以及高解析度快速球型攝影機之間的對應關係、快速球型攝影機自身旋轉校正以及快速球型攝影機多倍數校正。快速球型攝影機多倍數校正，是基於不同放大倍數在不同角度上具有一致性的特性，來加速校正的過程且不影響正確性; 此校正方式為一新穎的校正方式。完成雙攝影機的校正後，我們使用快速球型攝影機合成一個大範圍高解析的背景影像。當前景物在固定式廣角攝影機中被偵測到後，快速球型攝影機就會被驅使並對使用者選擇的物體進行連續追蹤。最後，我們整合了預先建構好的大範圍高解析背景以及分別由固定式廣角及快速球型攝影機所取的低高解析度前景影像，產生大範圍、多倍數以及高解析度監控畫面。 對於智慧化視訊安全監控系統，自動偵測動物體是一個重要的議題，使用背景相減法來偵測前景物，是研究多年卻仍然很重要的部分。一個好的背景相減演算法可以忍受環境的變化，例如：動態背景和光照的突然變化。在本文章的第二部分中，我們設計了一個空間背景模型(spatial background model, SBM)的新架構。包含兩個主要成分，背景模型(background model, BM)和背景梯度提取器(background gradient extractor, BGE)，來提取前景物體。對於每一張影像，我們都透過傳遞鄰居的資訊來建構背景模型，用於處理動態背景和突然的光線變化。背景梯度提取器與背景模型為同時建構和更新。為保持前景物形狀的完整性，我們利用背景梯度資訊設計禁止傳遞的策略。該方法可以有效地擷取前景和消除背景噪音。 此外，在視訊安全監控的應用中，物體的影子偵測和移除是物體偵測不可或缺的部分。在本文章的第三部分中，我們提出一個新的物體影子去除架構。整合兩個主要偵測器，移動物體偵測器和影子偵測器。對於移動物體，我們利用空間背景模型來偵測移動物體。對於影子去除，我們首先抽取出影子的特徵，包含色度、物理以及紋理性質。然後，使用隨機森林演算法學習影子特徵並產生隨機森林影子偵測模組。接著，我們對時空背景模型產生的結果使用隨機森林影子偵測模組進行影子去除。我們所提出的方法可以有效地檢測出移動物體並移除陰影的影響。此外，透過與其他技術做比較來展示我們所提出的方法，物體偵測與影子去除，的性能。 採用上述方法，使用雙攝影機架構建構出大範圍、高解析度影像，並利用時空背景模型有效的偵測出前景物資訊，進一步利用隨機森林演算法去除移動物體影子的影像，更精準的擷取出移動物體範圍，有效提高智慧化視訊監控的正確性。

Due to the terrorist attacks occur frequently, the anti-terrorism awareness of each country is raising. Therefore, the visual surveillance monitoring systems are setting up at important sites or in major cities in recent years. In visual surveillance monitoring system, the large-area high-resolution visual monitoring view and intelligence monitoring systems are indispensable in surveillance applications. To construct such systems, high-quality image capture, high-resolution display devices and automated detection of objects are required. Whereas high-quality displays have rapidly developed, the high-resolution surveillance cameras have progressed slowly and remain not widely used compared with displays. In the first part of this study, we designed an innovative framework, using a dual-camera system comprising a wide-angle fixed camera and a high-resolution pan-tilt-zoom (PTZ) camera to construct a large-area high-resolution visual-monitoring system that features multiresolution monitoring of moving objects. First, we developed a novel calibration approach to estimate the relationship between the two cameras and calibrate the PTZ camera. The PTZ camera was calibrated based on the consistent property of distinct pan-tilt angle at various zooming factors, accelerating the calibration process without affecting accuracy; this calibration process has not been reported previously. After calibrating the dual-camera system, we used the PTZ camera and synthesized a large-area high-resolution background image. When foreground targets were detected in the images captured by the wide-angle camera, the PTZ camera was controlled to continuously track the user-selected target. Last, we integrated preconstructed high-resolution background and low-resolution foreground images captured using the wide-angle camera and the high-resolution foreground image captured using the PTZ camera to generate a large-area high-resolution view of the scene. For intelligence visual surveillance monitoring system, the background subtraction is a crucial component, which has been studied over years. However, an efficient algorithm that can tolerate the environment changes such as dynamic backgrounds and sudden changes of illumination is still demanding. In the second part of this study, we design an innovative framework called the spatial background model (SBM) from a single-layer codebook model. Two main components, the background model (BM) and the background gradient extractor (BGE), are constructed to extract the foreground objects. The background model is built for each single frame with spatial information propagated from the neighbor locations, which is useful for handling dynamic background and sudden lighting changes. The background gradient extractor is also constructed and updated, and we design a propagation forbidden policy for background updating, so as to keep the completeness of foreground shape via the background gradient information. The proposed method can efficiently capture the foreground and eliminates the noise of background. Moreover, Cast shadows detection and removal is indispensable in the object detection to many surveillance applications. In the third part of this study, we present a novel framework for removing cast shadow of moving objects. Two main components, moving foreground detector and shadow detector, are integrated. For moving objects, we utilize the spatial background model (SBM) to detect the moving objects. For shadow removal, we first extract the shadow features which are chromaticity, physical property, and texture. Then, using the classifier, Random Forest, to learn the shadow model with shadow features. After that, removing the shadow from the result of SBM with the Random Forest shadow detector (RFSD). The proposed method can effectively detect the moving objects and remove the effect of shadow. Furthermore, we demonstrate the performance of our method compared with some techniques of object detection and shadow removal. Using these methodologies, constructing a large-area and high-resolution view using a dual-camera system, detecting the moving objects with spatial background model, and using random forest shadow detector to remove the shadow effect, to improve the accuracy of intelligence of monitoring surveillance.