傅立成臺灣大學:資訊工程學研究所莊振勛Chuang, Cheng-HsiungCheng-HsiungChuang2010-06-022018-07-052010-06-022018-07-052007U0001-2207200818083700http://ntur.lib.ntu.edu.tw//handle/246246/184783本篇論文提出利用增強性方向梯度直方圖(Augmented Histograms of Oriented Gradients (AHOG))於移動式平台上進行多人偵測，在本篇研究中，我們利用人體的幾何特徵來加強方向梯度直方圖(Histograms of Oriented Gradients (HOG))描述人型外觀的能力，其中我們把直立人型中存在的對稱性，每個身體部位的相對距離，以及人型在梯度特徵中的密度分佈加入HOG特徵中，來提升HOG特徵的描述能力，包含了上述人型特徵的HOG在此篇研究稱為AHOG，接著利用串接式AdaBoost演算法建立ㄧ個人型串接式分類器，用來對輸入影像中的可能區域進行偵測，由此人型分類器所決定之區域，則被考慮為人型可能區域，除此之外，利用串接式分類器的架構，可以減少偵測人型的時間；最後人型可能區域會再經由人型輪廓驗證，來確信此區域確實有人型存在，並且減少因為由複雜背景所引發的錯誤訊息，藉此降低錯誤偵測的發生。在此研究實驗中，於多種不同的實驗環境中，都可以提供可靠的人型偵測準確率。In this thesis we introduce an Augmented Histograms of Oriented Gradients (AHOG) feature for human detection from a non-static camera. This research tries to increase the discriminating power of original Histograms of Oriented Gradients (HOG) feature by adding human shape properties, such as contour distances, symmetry, gradient density, and shape approximation. The relations among AHOG features are characterized by the contour distances to the centroid of human. By observing on the biological structure of a human shape, we impose the symmetry property on every HOG feature and compute the similarity between feature itself and its symmetric pair so as to weigh HOG features. After that, the capability of describing human features is greatly improved when being compared with that of traditional one, especially when the moving humans are under consideration. Besides, we also augment the gradient density into AHOG to mitigate the influences caused by repetitive backgrounds. Moreover, we reject the false detections via an elliptical verifier learned when one tries to approximate a human shape. In the experiments, our proposed human detection method demonstrates highly reliable accuracy and provides the comparable performance to the state-of-the-art human detector on different databases.口試委員會審定書 ＃謝 i文摘要 iiBSTRACT iiiONTENTS ivIST OF FIGURES viiIST OF TABLES ixhapter 1 Introduction 1.1 Motivation 2.2 Challenges of Human Detection 3.3 Related Work 8.4 Objective 11.5 Organization 11hapter 2 Preliminaries 13.1 Problem Definition 13.2 Support Vector Machine (SVM) 14.2.1 Objective of SVM 15.2.2 Preliminary Knowledge of SVM 16.3 AdaBoost Algorithm 18.3.1 Objective of AdaBoost Algorithm 19.3.2 Preliminary Knowledge of AdaBoost Algorithm 19.3.3 Preliminary Knowledge of Cascaded AdaBoost Algorithm 21.4 Approach Overview 25.5 Summary of Contributions 25hapter 3 Human Candidate Detection 27.1 Augmented Histograms of Oriented Gradients 27.1.1 Feature Type 28.1.2 Gradient Computation 31.1.3 Symmetry 33.1.4 Gradient Density 36.1.5 Contour Distance 37.1.6 Dominant Orientation Rotation 39.1.7 Orientation Histogram Construction 41.2 Training 42.3 Detection 43.3.1 Human Potential Location 43.3.2 Classification 44hapter 4 Human Candidate Verification 46.1 Ellipse Approximation 47.1.1 Connected Components 47.1.2 Moments 48.2 Training & Verification 50hapter 5 Experiment 52.1 Environment Description 52.2 Database 52.3 Training 53.3.1 Discussion of Training Process 55.4 Experiment Results 57.4.1 Performance of MIT Database 57.4.2 Performance of Our Database 58hapter 6 Conclusion 63eferences 64application/pdf3340683 bytesapplication/pdfen-US人型偵測方向性梯度直方圖行人偵測human detectionhistograms of oriented gradientsAdaBoostthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/184783/1/ntu-96-R95922141-1.pdf