Chaehoon Park

Visual tracking for non-rigid objects using Rao-Blackwellized particle filter

Particle filters have been used for visual tracking during long periods because they enable effective estimation for non-linear and non-Gaussian distributions. However, particle filter-based tracking approaches suffer from occlusion and deformation of the target objects, which result in the large difference between the current observations and the target model. Thus, we present a Rao-Blackwellized particle filter (RBPF)-based tracking algorithm that effectively estimates the joint distribution for the target state and the target model; in the proposed method, the target object is tracked by using the particle filter while the target model is simultaneously updated on the basis of the on-line approximation of a mixture of Gaussians. To ensure the robustness to occlusion, we represent the target model by 16 orientation histograms that are spatially divided, and individually update each histogram through a video sequence. We demonstrate the robustness of the proposed method under occlusion and deformation of the target objects.

Vision-based navigation with efficient scene recognition

In this paper, we propose an efficient feature matching method for scene recognition and global localization. The proposed method enables mobile robots to autonomously navigate through the dynamic environment where the robot frequently encounters visual occlusion and kidnapping. For this purpose, we present a scale optimization method to enhance the matching performance with the combination of the FAST detector and integral image-based SIFT descriptors that are computationally efficient. The scale optimization method is required because the FAST detector does not provide scale information to compute descriptors for matching. We evaluate the performance of feature matching using various indoor image sequences and demonstrate the robustness of our navigation system under various conditions.

Robust Binary Feature Using the Intensity Order

Binary features have received much attention with regard to memory and computational efficiency with the emerging demands in the mobile and embedded vision systems fields. In this context, we present a robust binary feature using the intensity order. By analyzing feature regions, we devise a simple but effective strategy to detect keypoints. We adopt an ordinal description and encode the intensity order into a binary descriptor with proper binarization. As a result, our method obtains high repeatability and shows better performance with regard to feature matching with much less storage usage than other conventional features. We evaluate the performance of the proposed binary feature with various experiments, demonstrate its efficiency in terms of storage and computation time, and show its robustness under various geometric and photometric transformations.

Accelerated kmeans Clustering using Binary Random Projection

Codebooks have been widely used for image retrieval and image indexing, which are the core elements of mobile visual searching. Building a vocabulary tree is carried out offline, because the clustering of a large amount of training data takes a long time. Recently proposed adaptive vocabulary trees do not require offline training, but suffer from the burden of online computation. The necessity for clustering high dimensional large data has arisen in offline and online training. In this paper, we present a novel clustering method to reduce the burden of computation without losing accuracy. Feature selection is used to reduce the computational complexity with high dimensional data, and an ensemble learning model is used to improve the efficiency with a large number of data. We demonstrate that the proposed method outperforms the-state of the art approaches in terms of computational complexity on various synthetic and real datasets.

Evaluation of vocabulary trees for localization in robot applications

Vocabulary tree based place recognition is widely used in topological localization and its various applications have been proposed during the past decade. However, the bag-of-words representations from the vocabulary tree, which is trained with fixed training data, are difficult to be optimized to dynamic environments. To solve this problem, an adaptive vocabulary tree has been proposed, but there has been no comparison considering the adaptive properties of the conventional vocabulary tree. This paper provides a performance evaluation of the vocabulary tree and the adaptive vocabulary tree in dynamic scenes. This analysis provides guidance for choosing appropriate vocabulary in robot applications.

Intra-class key feature weighting method for vocabulary tree based image retrieval

With the existing feature weighting methods of image retrieval field, it was impossible to use the fact that images have different key features depending on their classes because the same weight is applied to every image class. We propose a method of indexing features of each class in order of importance and giving them relevant weights, which can be applied to image retrieval. We designed a simple weight mapping function in order to enhance the distinctiveness between the image classes and also proposed a method to re-rank sub-class image set to apply different weight vectors to image retrieval framework. We demonstrated the proposed method on the existing image retrieval framework to compare and verify the performance. Proposed method was evaluated with UKBench Dataset and the result showed a noticeable improvement.

Large object detection in cluttered background using boosted Markov Chain Monte Carlo

In this paper, we present a new object detection method using codebook and boosted Markov Chain Monte Carlo (MCMC) estimation. It is relatively well detected using adaboost and simple Haar-like features for small objects. However, the detection problem is more difficult when object size becomes larger (over 150 × 150) due to different surface markings and clutter. Codebook-based object representation and boosted MCMC method can detect large objects robustly. Experimental results validate convincing detection for large objects.

Product search framework with categorization and identification

When people want to find some products in the Internet. They use query words. In this paper, we propose a product search framework with images instead of words. This is helpful when user doesnpsilat know about a product or want to find similar products. The framework is composed of three parts: (1) classify a category of product (2) find a corresponding product (3) retrieve similar products by its shape or color. We use local features as a visual information and adopt visual words based method in the categorization and identification. The representation of image is a histogram which is made from a set of local features from an image and visual words from image set. As a local feature, we adopt SURF. We use grid sampling method in the categorization part and original Fast-Hessian detector in the identification part. In the similar product search, we use a binary resized image and color information. We validate our product search framework with KAIST-104 DB and our product DB.