• KIPS Transactions on Software and Data Engineering
• /
• v.12 no.8
• /
• pp.371-380
• /
• 2023

Relation extraction is to extract relationships between named entities from text. Traditionally, relation extraction methods only extract relations between predetermined subject and object entities. However, in end-to-end relation extraction, all possible relations must be extracted by considering the positions of the subject and object for each pair of entities, and so this method uses time and resources inefficiently. To alleviate this problem, this paper proposes a method that sets directions based on the positions of the subject and object, and extracts relations according to the directions. The proposed method utilizes existing relation extraction data to generate direction labels indicating the direction in which the subject points to the object in the sentence, adds entity position tokens and entity type to sentences to predict the directions using a pre-trained language model (KLUE-RoBERTa-base, RoBERTa-base), and generates representations of subject and object entities through probabilistic crossover operation. Then, we make use of these representations to extract relations. Experimental results show that the proposed model performs about 3 ~ 4%p better than a method for predicting integrated labels. In addition, when learning Korean and English data using the proposed model, the performance was 1.7%p higher in English than in Korean due to the number of data and language disorder and the values of the parameters that produce the best performance were different. By excluding the number of directional cases, the proposed model can reduce the waste of resources in end-to-end relation extraction.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.29 no.1C
• /
• pp.92-101
• /
• 2004

This paper proposes a algorithm to extract the variety region (object) from video for the real-time encoding of MPEG-4 based. The previous object segmentation methods cannot used the videophone or videoconference required by real-time processing. It is difficult to transfer a video to real-time because it increased complexity for the operation of each pixel on the spatial segmentation and temporal segmentation method proposed by MPEG-4 Working Group. But algorithm proposed for this thesis not operates a pixel unit but operates a macro block unit. Thus this enables real-time transfer. But this algorithm cannot extract several object for a image using proposed algorithm as previous algorithm. On system constructed by encoder and decoder. A proposed algorithm inserted for encoder as pre-process.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.8
• /
• pp.1121-1129
• /
• 2022

Object recognition is carried out using RGB images in various object recognition studies. However, RGB images in dark illumination environments or environments where target objects are occluded other objects cause poor object recognition performance. On the other hand, IR images provide strong object recognition performance in these environments because it detects infrared waves rather than visible illumination. In this paper, we propose an RGB-IR fusion model, high-frequency interchange network (HINet), which improves object recognition performance by combining only the strengths of RGB-IR image pairs. HINet connected two object detection models using a mutual high-frequency transfer (MHT) to interchange advantages between RGB-IR images. MHT converts each pair of RGB-IR images into a discrete cosine transform (DCT) spectrum domain to extract high-frequency information. The extracted high-frequency information is transmitted to each other's networks and utilized to improve object recognition performance. Experimental results show the superiority of the proposed network and present performance improvement of the multispectral object recognition task.

• Proceedings of the Korea Contents Association Conference
• /
• 2009.05a
• /
• pp.839-842
• /
• 2009

It is used to differential image for moving object detection in general. But it is difficult to detect the accurate detection which uses differential image between frames. In this paper, the proposed method overcome the noise that is generated by camera, grabber card, or weather condition. It extract to moving big object such as human or vehicle. The proposed method process morphological filtering and binary for the image with noise, reduce error. We are expect to apply to a real-time moving object detection system at fog condition, pass the limit of the object detection method using the differential image.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.14 no.1
• /
• pp.77-85
• /
• 2018

This paper proposes a robust digital video stabilization algorithm to extract and track an object, which uses a phase correlation-based motion correction. The proposed video stabilization algorithm consists of background stabilization based on motion estimation and extraction of a moving object. The motion vectors can be estimated by calculating the phase correlation of a series of frames in the eight sub-images, which are located in the corner of the video. The global motion vector can be estimated and the image can be compensated by using the multiple local motions of sub-images. Through the calculations of the phase correlation, the motion of the background can be subtracted from the former frame and the compensated frame, which share the same background. The moving objects in the video can also be extracted. In this paper, calculating the phase correlation to track the robust motion vectors results in the compensation of vibrations, such as movement, rotation, expansion and the downsize of videos from all directions of the sub-images. Experimental results show that the proposed digital image stabilization algorithm can provide continuously stabilized videos and tracking object movements.

• The KIPS Transactions:PartB
• /
• v.12B no.7 s.103
• /
• pp.767-774
• /
• 2005

In this paper, we present a snake-based scheme for tracking object contour using disparity information taken from a stereo image sequence with cluttered background. The proposed method is composed of two steps. First, 3-D motion of object is estimated and candidate snake points are selected in disparity space. Second, object contour is extracted by using a modified snake algorithm with disparity information. The proposed algorithm can successfully extract the concave contour of objects and track the object contour in complex image. Performance of the proposed algorithm has been verified by simulation.

• Journal of the Institute of Convergence Signal Processing
• /
• v.4 no.3
• /
• pp.21-26
• /
• 2003

A new method to realize 3-dimensional object pattern recognition system using Fourier-based feature extractor has been proposed. The procedure to obtain the invariant feature vector is as follows ; A closed surface is generated by tracing the surface of object using the 3-dimensional polar coordinate. The centroidal distances between object's geometrical center and each closed surface points are calculated. The distance vector is translation invariant. The distance vector is normalized, so the result is scale invariant. The Fourier spectrum of each normalized distance vector is calculated, and the spectrum is rotation invariant. The Fourier-based feature generating from above procedure completely eliminates the effect of variations in translation, scale, and rotation of 3-dimensional object with closed-surface. The experimental results show that the proposed method has a high accuracy.

• The KIPS Transactions:PartB
• /
• v.13B no.3 s.106
• /
• pp.255-274
• /
• 2006

Moving object boundary is very important for the accurate segmentation of moving object. We extract the moving object boundary from the moving object edge. But the object boundary shows broken boundaries so we develop a novel boundary linking algorithm to link the broken boundaries. The boundary linking algorithm forms a quadrant around the terminating pixel in the broken boundaries and searches for other terminating pixels to link in concentric circles clockwise within a search radius in the forward direction. The boundary linking algorithm guarantees the shortest distance linking. We register the background from the image sequence using the stationary background filtering. We construct two object masks, one object mask from the boundary linking and the other object mask from the initial moving object, and use these two complementary object masks to segment the moving objects. The main contribution of the proposed algorithms is the development of the novel object boundary linking algorithm for the accurate segmentation. We achieve the accurate segmentation of moving object, the segmentation of multiple moving objects, the segmentation of the object which has a hole within the object, the segmentation of thin objects, and the segmentation of moving objects in the complex background using the novel object boundary linking and the background automatically. We experiment the algorithms using standard MPEG-4 test video sequences and real video sequences of indoor and outdoor environments. The proposed algorithms are efficient and can process 70.20 QCIF frames per second and 19.7 CIF frames per second on the average on a Pentium-IV 3.4GHz personal computer for real-time object-based processing.

• Journal of Broadcast Engineering
• /
• v.7 no.3
• /
• pp.248-261
• /
• 2002

In this paper, we proposed a new 3-dimensional(3-D) object-tracking algorithm that can control a stereo camera using a variable window mask supported by which uses ,B-D information and an optical BPEJTC. Hence, three-dimensional information characteristics of a stereo vision system, distance information from the stereo camera to the tracking object. can be easily acquired through the elements of a stereo vision system. and with this information, we can extract an area of the tracking object by varying window masks. This extractive area of the tracking object is used as the next updated reference image. furthermore, by carrying out an optical BPEJTC between a reference image and a stereo input image the coordinates of the tracking objects location can be acquired, and with this value a 3-D object tracking can be accomplished through manipulation of the convergence angie and a pan/tilt of a stereo camera. From the experimental results, the proposed algorithm was found to be able to the execute 3-D object tracking by extracting the area of the target object from an input image that is independent of the background noise in the stereo input image. Moreover a possible implementation of a 3-D tele-working or an adaptive 3-D object tracker, using the proposed algorithm is suggested.

• Journal of Internet Computing and Services
• /
• v.24 no.5
• /
• pp.29-35
• /
• 2023

CCTV is used for various purposes such as crime prevention, public safety reinforcement, and traffic management. However, as the range and resolution of the camera improve, there is a risk of exposing personal information in the video. Therefore, there is a need for new technologies that can identify individuals while protecting personal information in images. In this paper, we propose histogram-based singular value decomposition for object identification and tracking. The proposed method distinguishes different objects present in the image using color information of the object. For object recognition, YOLO and DeepSORT are used to detect and extract people present in the image. Color values are extracted with a black-and-white histogram using location information of the detected person. Singular value decomposition is used to extract and use only meaningful information among the extracted color values. When using singular value decomposition, the accuracy of object color extraction is increased by using the average of the upper singular value in the result. Color information extracted using singular value decomposition is compared with colors present in other images, and the same person present in different images is detected. Euclidean distance is used for color information comparison, and Top-N is used for accuracy evaluation. As a result of the evaluation, when detecting the same person using a black-and-white histogram and singular value decomposition, it recorded a maximum of 100% to a minimum of 74%.