• ETRI Journal
• /
• v.43 no.1
• /
• pp.74-81
• /
• 2021

The mixture model is a very powerful and flexible tool in clustering analysis. Based on the Dirichlet process and parsimonious Gaussian distribution, we propose a new nonparametric mixture framework for solving challenging clustering problems. Meanwhile, the inference of the model depends on the efficient online variational Bayesian approach, which enhances the information exchange between the whole and the part to a certain extent and applies to scalable datasets. The experiments on the scene database indicate that the novel clustering framework, when combined with a convolutional neural network for feature extraction, has meaningful advantages over other models.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.2
• /
• pp.912-928
• /
• 2019

Crowd behavior analysis research has revealed a central role in helping people to find safety hazards or crime optimistic forecast. Thus, it is significant in the future video surveillance systems. Recently, the growing demand for safety monitoring has changed the awareness of video surveillance studies from analysis of individuals behavior to group behavior. Group detection is the process before crowd behavior analysis, which separates scene of individuals in a crowd into respective groups by understanding their complex relations. Most existing studies on group detection are scene-specific. Crowds with various densities, structures, and occlusion of each other are the challenges for group detection in diverse crowded scenes. Therefore, we propose a group detection approach called Collective Interaction Filtering to discover people motion interaction from trajectories. This approach is able to deduce people interaction with the Expectation-Maximization algorithm. The Collective Interaction Filtering approach accurately identifies groups by clustering trajectories in crowds with various densities, structures and occlusion of each other. It also tackles grouping consistency between frames. Experiments on the CUHK Crowd Dataset demonstrate that approach used in this study achieves better than previous methods which leads to latest results.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.05a
• /
• pp.151-152
• /
• 2012

This paper proposes a Scene Change Detection method using the local decision tree and clustering. The local decision tree detects cluster boundaries wherein local scenes occur, in such a way as to compare time similarity distributions among the difference values between detected scenes and their adjacent frames, and group an unbroken sequence of frames with similarities in difference value into a cluster unit.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.6
• /
• pp.1199-1203
• /
• 2012

This paper proposes a Scene Change Detection method using the local decision tree and clustering. The local decision tree detects cluster boundaries wherein local scenes occur, in such a way as to compare time similarity distributions among the difference values between detected scenes and their adjacent frames, and group an unbroken sequence of frames with similarities in difference value into a cluster unit.

• Journal of Korea Multimedia Society
• /
• v.13 no.8
• /
• pp.1128-1141
• /
• 2010

This paper proposes a method of moving object detection in crowded scene using clustered trajectory. Unlike previous appearance based approaches, the proposed method employes motion information only to isolate moving objects. In the proposed method, feature points are extracted from input frames first and then feature tracking is followed to create feature trajectories. Based on an assumption that feature points originated from the same objects shows similar motion as the object moves, the proposed method detects moving objects by clustering trajectories of similar motions. For this purpose an energy function based on spatial proximity, motion coherence, and temporal continuity is defined to measure the similarity between two trajectories and the clustering is achieved by minimizing the energy function in CRFs (conditional random fields). Compared to previous methods, which are unable to separate falsely merged trajectories during the clustering process, the proposed method is able to rearrange the falsely merged trajectories during iteration because the clustering is solved my energy minimization in CRFs. Experiment results with three different crowded scenes show about 94% detection rate with 7% false alarm rate.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.6 no.12
• /
• pp.3315-3337
• /
• 2012

Crowded scenes analysis is a challenging topic in computer vision field. How to detect diverse motion patterns in crowded scenarios from videos is the critical yet hard part of this problem. In this paper, we propose a novel approach to mining motion patterns by utilizing motion information during both long-term period and short interval simultaneously. To capture long-term motions effectively, we introduce Motion History Image (MHI) representation to access to the global perspective about the crowd motion. The combination of MHI and optical flow, which is used to get instant motion information, gives rise to discriminative spatial-temporal motion features. Benefitting from the robustness and efficiency of the novel motion representation, the following motion pattern mining is implemented in a completely unsupervised way. The motion vectors are clustered hierarchically through automatic hierarchical clustering algorithm building on the basis of graphic model. This method overcomes the instability of optical flow in dealing with time continuity in crowded scenes. The results of clustering reveal the situations of motion pattern distribution in current crowded videos. To validate the performance of the proposed approach, we conduct experimental evaluations on some challenging videos including vehicles and pedestrians. The reliable detection results demonstrate the effectiveness of our approach.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.17 no.11
• /
• pp.2711-2722
• /
• 1993

Most of the current 2-D object recognition systems are model-based. In such systems, the representation of each of a known set of objects are precompiled and stored in a database of models. Later, they are used to recognize the image of an object in each instance. In this thesis, the approach method for the 2-D object recognition is treating an object boundary as a string of structral units and utilizing string matching to analyze the scenes. To reduce string matching time, models are rebuilt by means of fuzzy c-means clustering algorithm. In this experiments, the image of objects were taken at initial position of a robot from the CCD camera, and the models are consturcted by the proposed algorithm. After that the image of an unknown object is taken by the camera at a random position, and then the unknown object is identified by a comparison between the unknown object and models. Finally, the amount of translation and rotation of object from the initial position is computed.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.164-169
• /
• 2009

As one of the most interesting scenes, landmarks constitute a large percentage of the vast amount of scene images available on the web. On the other hand, a specific "landmark" usually has some characteristics that distinguish it from surrounding scenes and other landmarks. These two observations make the task of accurately estimating geographic information from a landmark image necessary and feasible. In this paper, we propose a method to identify landmark location by means of landmark recognition in view of significant viewpoint, illumination and temporal variations. We use GPS-based clustering to form groups for different landmarks in the image dataset. The images in each group rather fully express the possible views of the corresponding landmark. We then use a combination of edge and color histogram to match query to database images. Initial experiments with Zubud database and our collected landmark images show that is feasible.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.7
• /
• pp.3239-3267
• /
• 2018

Image and video dehazing is a popular topic in the field of computer vision and digital image processing. A fast, optimized dehazing algorithm was recently proposed that enhances contrast and reduces flickering artifacts in a dehazed video sequence by minimizing a cost function that makes transmission values spatially and temporally coherent. However, its fixed-size block partitioning leads to block effects. The temporal cost function also suffers from the temporal non-coherence of newly appearing objects in a scene. Further, the weak edges in a hazy image are not addressed. Hence, a video dehazing algorithm based on well designed spectral clustering is proposed. To avoid block artifacts, the spectral clustering is customized to segment static scenes to ensure the same target has the same transmission value. Assuming that edge images dehazed with optimized transmission values have richer detail than before restoration, an edge intensity function is added to the spatial consistency cost model. Atmospheric light is estimated using a modified quadtree search. Different temporal transmission models are established for newly appearing objects, static backgrounds, and moving objects. The experimental results demonstrate that the new method provides higher dehazing quality and lower time complexity than the previous technique.

• Proceedings of the KIEE Conference
• /
• 1988.07a
• /
• pp.671-674
• /
• 1988

This paper describes a matching method to solve the problem of occlusion in a two dimensional scene. The technique consist of three steps: generation of hypotheses, clustering of hypotheses by matching probability, updating of hypotheses. Using this algorithm, simulation results have been tested for 20 scenes contained the 80 models, and have obtained 95% of properly correct recognition rate in average.