• The Transactions of the Korea Information Processing Society
• /
• v.4 no.12
• /
• pp.3211-3218
• /
• 1997

This paper addresses an eye location algorithm which is essential process of human face tracking system for natural video-conferencing. In current video-conferencing systems, user's facial movements are restricted by fixed camera, therefore it is inconvenient to users. We Propose an eye location algorithm for automatic face tracking. Because, locations of other facial features guessed from locations of eye and scale of face in the image can be calculated using inter-ocular distance. Most previous feature extraction methods for face recognition system are approached under assumption that approximative face region or location of each facial feature is known. The proposed algorithm in this paper uses no prior information on the given image. It is not sensitive to backgrounds and lighting conditions. The proposed algorithm uses the valley representation as major information to locate eyes. The experiments have been performed for 213 frames of 17 people and show very encouraging results.

• The KIPS Transactions:PartB
• /
• v.19B no.1
• /
• pp.19-26
• /
• 2012

The rapid increase of information imposes new demands of content management. The purpose of automatic audio segmentation and classification is to meet the rising need for efficient content management. With this reason, this paper proposes a high-accuracy algorithm that segments audio signals and classifies them into different classes such as speech, music, silence, and environment sounds. The proposed algorithm utilizes support vector machine (SVM) to detect audio-cuts, which are boundaries between different kinds of sounds using the parameter sequence. We then extract feature vectors that are composed of statistical data and they are used as an input of fuzzy c-means (FCM) classifier to partition audio-segments into different classes. To evaluate segmentation and classification performance of the proposed SVM-FCM based algorithm, we consider precision and recall rates for segmentation and classification accuracy for classification. Furthermore, we compare the proposed algorithm with other methods including binary and FCM classifiers in terms of segmentation performance. Experimental results show that the proposed algorithm outperforms other methods in both precision and recall rates.

• Journal of Institute of Control, Robotics and Systems
• /
• v.20 no.8
• /
• pp.868-874
• /
• 2014

This paper proposes a novel mapping algorithm in Omni-directional Vision SLAM based on an obstacle's feature extraction using Lucas-Kanade Optical Flow motion detection and images obtained through fish-eye lenses mounted on robots. Omni-directional image sensors have distortion problems because they use a fish-eye lens or mirror, but it is possible in real time image processing for mobile robots because it measured all information around the robot at one time. In previous Omni-Directional Vision SLAM research, feature points in corrected fisheye images were used but the proposed algorithm corrected only the feature point of the obstacle. We obtained faster processing than previous systems through this process. The core of the proposed algorithm may be summarized as follows: First, we capture instantaneous $360^{\circ}$ panoramic images around a robot through fish-eye lenses which are mounted in the bottom direction. Second, we remove the feature points of the floor surface using a histogram filter, and label the candidates of the obstacle extracted. Third, we estimate the location of obstacles based on motion vectors using LKOF. Finally, it estimates the robot position using an Extended Kalman Filter based on the obstacle position obtained by LKOF and creates a map. We will confirm the reliability of the mapping algorithm using motion estimation based on fisheye images through the comparison between maps obtained using the proposed algorithm and real maps.

• Journal of The Korean Astronomical Society
• /
• v.45 no.6
• /
• pp.167-173
• /
• 2012

Inter-granular Bright Points (igBPs) are small-scale objects in the Solar photosphere which can be seen within dark inter-granular lanes. We present a new algorithm to automatically detect and extract igBPs. Laplacian and Morphological Dilation (LMD) technique is employed by the algorithm. It involves three basic processing steps: (1) obtaining candidate "seed" regions by Laplacian; (2) determining the boundary and size of igBPs by morphological dilation; (3) discarding brighter granules by a probability criterion. For validating our algorithm, we used the observed samples of the Dutch Open Telescope (DOT), collected on April 12, 2007. They contain 180 high-resolution images, and each has a $85{\times}68\;arcsec^2$ field of view (FOV). Two important results are obtained: first, the identified rate of igBPs reaches 95% and is higher than previous results; second, the diameter distribution is $220{\pm}25km$, which is fully consistent with previously published data. We conclude that the presented algorithm can detect and extract igBPs automatically and effectively.

• Journal of Biomedical Engineering Research
• /
• v.44 no.5
• /
• pp.346-353
• /
• 2023

In this paper, we propose an innovative approach that leverages deep learning to find optimal reference points for achieving precise tooth segmentation in three-dimensional tooth point cloud data. A dataset consisting of 350 aligned maxillary and mandibular cloud data was used as input, and both end coordinates of individual teeth were used as correct answers. A two-dimensional image was created by projecting the rendered point cloud data along the Z-axis, where an image of individual teeth was created using an object detection algorithm. The proposed algorithm is designed by adding various modules to the Unet model that allow effective learning of a narrow range, and detects both end points of the tooth using the generated tooth image. In the evaluation using DSC, Euclid distance, and MAE as indicators, we achieved superior performance compared to other Unet-based models. In future research, we will develop an algorithm to find the reference point of the point cloud by back-projecting the reference point detected in the image in three dimensions, and based on this, we will develop an algorithm to divide the teeth individually in the point cloud through image processing techniques.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.26 no.2
• /
• pp.149-156
• /
• 2008

In this paper, we propose a method aiming at updating the building information of the digital maps using single high resolution satellite image and digital map. Firstly we produced a digital orthoimage through the automatic co-registration of QuickBird image and 1:1,000 digital map. Secondly we extracted building height information through the template matching of digital map's building vector data and the image's edges obtained by Canny operator. Finally we refined the shape of some buildings by using the result from template matching as the seed polygon of the greedy snake algorithm. In order to evaluate the proposed method's effectiveness, we estimated accuracy of the extracted building information using LiDAR DSM and 1:1,000 digital map. The evaluation results showed the proposed method has a good potential for extraction and revision of building information.

• Proceedings of the KSRS Conference
• /
• 2007.10a
• /
• pp.238-242
• /
• 2007

Today's commercial high resolution satellite imagery such as that provided by IKONOS and QuickBird, offers the potential to extract useful spatial information for geographical database construction and GIS applications. Digital maps supply the most generally used GIS data probiding topography, road, and building information. Currently, the building information provided by digital maps is incompletely constructed for GIS applications due to planar position error and warped shape. We focus on extracting of the accurate building information including position, shape, and height to update the building information of the digital maps and GIS database. In this paper, we propose a new method of 3D building information extraction with a single high resolution satellite image and digital map. Co-registration between the QuickBird image and the 1:1,000 digital maps was carried out automatically using the RPC adjustment model and the building layer of the digital map was projected onto the image. The building roof boundaries were detected using the building layer from the digital map based on the satellite azimuth. The building shape could be modified using a snake algorithm. Then we measured the building height and traced the building bottom automatically using triangular vector structure (TVS) hypothesis. In order to evaluate the proposed method, we estimated accuracy of the extracted building information using LiDAR DSM.

• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.11
• /
• pp.521-526
• /
• 2016

As on-line informal text data have massive in its volume and have unstructured characteristics in nature, there are limitations in applying traditional relational data model technologies for data storage and data analysis jobs. Moreover, using dynamically generating massive social data, social user's real-time reaction analysis tasks is hard to accomplish. In the paper, to capture easily the semantics of massive and informal on-line documents with unsupervised learning mechanism, we design and implement automatic topic extraction systems according to the mass of the words that consists a document. The input data set to the proposed system are generated first, using N-gram algorithm to build multiple words to capture the meaning of the sentences precisely, and Hadoop and Spark (In-memory distributed computing framework) are adopted to run topic model. In the experiment phases, TB level input data are processed for data preprocessing and proposed topic extraction steps are applied. We conclude that the proposed system shows good performance in extracting meaningful topics in time as the intermediate results come from main memories directly instead of an HDD reading.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.12 no.6
• /
• pp.1095-1100
• /
• 2008

In a process of insect foot recognition, footprint segments as basic areas for recognition need to be extracted from scanned insect footprints and appropriate features should be found from the footprint segments in order to discriminate kinds of insects, because the characteristics of the features are important to classify insects. In this paper, we propose methods for automatic footprint segmentation and feature extraction. We use a Trace transform method in order to find out appropriate features from the extracted segments by the above methods. The Trace transform method builds a new type of data structure from the segmented images by functions using parallel trace lines and the new type of data structure has characteristics invariant to translation, rotation and reflection of images. This data structure is converted to Triple features by Diametric and Circus functions, and the Triple features are used for discriminating patterns of insect footprints. In this paper, we show that the Triple features found by the proposed methods are enough distinguishable and appropriate for classifying kinds of insects.

• Journal of KIISE:Software and Applications
• /
• v.29 no.1_2
• /
• pp.42-52
• /
• 2002

Information extraction(IE) is a process of recognizing and fetching particular information fragments from a document. In previous work, most IE systems generate the extraction rules called the wrappers manually, and although this manual wrapper generation may achieve more correct extraction, it reveals some problems in flexibility, extensibility, and efficiency. Some other researches that employ automatic ways of generating wrappers are also experiencing difficulties in acquiring and representing useful domain knowledge and in coping with the structural heterogeneity among different information sources, and as a result, the real-world information sources with complex document structures could not be correctly analyzed. In order to resolve these problems, this paper presents an agent-based information extraction system named XTROS that exploits the domain knowledge to learn from documents in a semi-structured information source. This system generates a wrapper for each information source automatically and performs information extraction and information integration by applying this wrapper to the corresponding source. In XTROS, both the domain knowledge and the wrapper are represented as XML-type documents. The wrapper generation algorithm first recognizes the meaning of each logical line of a sample document by using the domain knowledge, and then finds the most frequent pattern from the sequence of semantic representations of the logical lines. Eventually, the location and the structure of this pattern represented by an XML document becomes the wrapper. By testing XTROS on several real-estate information sites, we claim that it creates the correct wrappers for most Web sources and consequently facilitates effective information extraction and integration for heterogeneous and complex information sources.