In this master thesis, the problem of tracking groups using an image sequence dataset is examined. Target tracking can be defined as the problem of estimating a target's state given prior knowledge about its motion and some sensor measurements related to the target's state. A popular method for target tracking is e.g. the Kalman filter. However, the Kalman filter is insufficient when there are multiple targets in the scene. Consequently, alternative multitarget tracking methods must be applied along with methods for estimating the number of targets in the scene. Multitarget tracking can however be difficult when there are many unresolved targets, e.g. associating observations with targets in dense crowds. A viable simplification is group target tracking, keeping track of groups rather than individual targets. Furthermore, group target tracking is preferred when the user wants to know the motion and extension of a group in e.g. evacuation scenarios. To solve the problem of group target tracking in video surveillance, a combination of GM-PHD filtering and mean shift clustering is proposed. The GM-PHD filter is an approximation of Bayes multitarget filter. Pedestrian detections converted into flat world coordinates from the image dataset are used as input to the filter. The output of the GM-PHD filter consists of Gaussian mixture components with corresponding mean state vectors. The components are divided into groups by using mean shift clustering. An estimate of the number of members and group shape is presented for each group. The method is evaluated using both single camera measurements and two cameras partly surveilling the same area. The results are promising and present a nice visual representation of the groups' characteristics. However, using two cameras gives no improvement in performance, probably due to differences in detections between the two cameras, e.g. a single pedestrian can be observed being at two positions several meters apart making it difficult to determine if it is a single pedestrian or multiple pedestrians.

三維城市模型之建置需求漸趨繁多，可提供都市規劃、城市導航及虛擬實境等相關應用，過去研究多以建置LOD2城市模型為主，且較著重於屋頂結構。近年來，逐漸利用垂直影像及傾斜影像作為原始資料，提供建物牆面之建置，並且，隨著無人機系統(Unmanned Aircraft System, UAS）發展，可利用其蒐集高解析度且高重疊垂直及傾斜拍攝之建物影像，並採影像密匹配技術產製高密度點雲，進而快速取得建物包含屋頂及牆面之三維資訊，而這些資訊可進一步提供後續建置LOD3建置層級之模型，而在建置前，首先須對資料進行特徵分析，萃取特徵點、線、面，進而提供建置模型所需之資訊。 因此，本研究期望利用密匹配點雲，計算其點雲特徵，並採用Mean Shift分群法（Comaniciu and Meer, 2002）萃取建物點雲資訊，並提供一最佳分群策略。首先，本研究將以UAS為載具，設計一野外率定場率定相機，並蒐集建物高重疊UAS影像密匹配產製高密度點雲，針對單棟建物高密度點雲，實驗測試點雲疏化程度後，依據疏化成果計算點雲特徵，並以此批點雲資料實驗測試Mean shift分群法（Cheng, 1995）中之參數，後設計分群流程以分離平面點群及曲面點群，探討分群成果以決定最佳分群策略。實驗結果顯示本研究提出之分群策略，可自動區分平面點群及曲面點群，並單獨將平面點群分群至各牆面。 / Unmanned Aerial System (UAS) offer several new possibilities in a wide range of applications. One example is the 3D reconstruction of buildings. In former times this was either restricted by earthbound vehicles to the reconstruction of facades or by air-borne sensors to generate only very coarse building models. UAS are able to observe the whole 3D scene and to capture images of the object of interest from completely different perspectives. Therefore, this study will use UAS to collected images of buildings and to generate point cloud from dense image matching for modeling buildings. In the proposed approach, this method computes principal orientations by PCA and identifies clusters by Mean shift clustering. Analyze the factors which can affect the clustering methods and try to decrease the use of threshold, and this result can cluster the façade of buildings automatically and offer the after building reconstruction for LOD3.

Les travaux présentés dans ce mémoire de thèse concernent le développement et la mise en œuvre d'un algorithme de filtrage particulaire pour le recalage de la navigation inertielle par mesures altimétriques. Le filtre développé, le MRPF (Mixture Regularized Particle Filter), s'appuie à la fois sur la modélisation de la densité a posteriori sous forme de mélange fini, sur le filtre particulaire régularisé ainsi que sur l'algorithme mean-shiftclustering. Nous proposons également une extension du MRPF au filtre particulaire Rao-Blackwellisé appelée MRBPF (Mixture Rao-Blackwellized ParticleFilter). L'objectif est de proposer un filtre adapté à la gestion des multimodalités dues aux ambiguïtés de terrain. L'utilisation des modèles de mélange fini permet d'introduire un algorithme d'échantillonnage d'importance afin de générer les particules dans les zones d'intérêt. Un second axe de recherche concerne la mise au point d'outils de contrôle d'intégrité de la solution particulaire. En nous appuyant sur la théorie de la détection de changement, nous proposons un algorithme de détection séquentielle de la divergence du filtre. Les performances du MRPF, MRBPF, et du test d'intégrité sont évaluées sur plusieurs scénarios de recalage altimétrique.

This thesis proposes and implements a system for movement analysis of vehicles on crossroads. It detects and tracks the movement of vehicles in the video, gained from the stationary video camera, which has the view of some crossroad. The trajectories are stored and their number and directions are analysed. The detection was made using cascade classifier. A dataset of 10500 positive and 10500 negative samples has been created to train the classifier. Vehicles are tracked using KCF method. For trajectory clustering, needed by analysis, the Mean Shift method is used. Testing showed, that the overall success of vehicle movement analysis is 92.77%.