• Journal of Internet Computing and Services
• /
• v.10 no.4
• /
• pp.1-11
• /
• 2009

Recently there has arisen concern in both the database community and the graphics society about data retrieval from large motion databases because the high dimensionality of motion data implies high costs. In this circumstance, finding an effective distance measure and an efficient query processing method for such data is a challenging problem. This paper presents an elaborate motion query processing system, SMoFinder (Similar Motion Finder), which incorporates a novel kinematic distance measure and an efficient indexing strategy via adaptive frame segmentation. To this end, we regard human motions as multi-linkage kinematics and propose the weighted Minkowski distance metric. For efficient indexing, we devise a new adaptive segmentation method that chooses representative frames among similar frames and stores chosen frames instead of all frames. For efficient search, we propose a new search method that processes k-nearest neighbors queries over only representative frames. Our experimental results show that the size of motion databases is reduced greatly (${\times}1/25$) but the search capability of SMoFinder is equal to or superior to that of other systems.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.30 no.2
• /
• pp.163-171
• /
• 2012

High resolution satellite images have been used for building and road system analysis, landscape analysis, and ecological assessment for several years. However, in high resolution satellite images, shadows are necessarily cast by manmade objects such as buildings and over-pass bridges. This paper develops the shadow extraction procedures in urban area including various land-use classes, and the extracted shadow areas are evaluated by a manually digitized shadow map. For the shadow extraction, the Canny edge operator and the dilation filter are applied to make building edge buffer area. Also, the object-based segmentation was performed using Gram-Schmitt fusion image, and spectral and spatial parameters are calculated from the segmentation results. Finally, we proposed appropriate parameters and extraction rules for the shadow extraction. The accuracy of the shadow extraction results from the various assessment indices is 80% to 90%.

• Journal of KIISE:Computing Practices and Letters
• /
• v.8 no.6
• /
• pp.778-786
• /
• 2002

Scene boundary detection is important in the understanding of semantic structure from video data. However, it is more difficult than shot change detection because scene boundary detection needs to understand semantics in video data well. In this paper, we propose a new approach to scene segmentation using contextual information in video data. The contextual information is divided into two categories: local and global contextual information. The local contextual information refers to the foreground regions' information, background and shot activity. The global contextual information refers to the video shot's environment or its relationship with other video shots. Coherence, interaction and the tempo of video shots are computed as global contextual information. Using the proposed contextual information, we detect scene boundaries. Our proposed approach consists of three consecutive steps: linking, verification, and adjusting. We experimented the proposed approach using TV dramas and movies. The detection accuracy of correct scene boundaries is over than 80%.

• Journal of Broadcast Engineering
• /
• v.20 no.4
• /
• pp.557-567
• /
• 2015

Over the past decades, a variety of spatial saliency methods have been introduced. Recently, motion saliency has gained much interests, where motion data estimated from an image sequence are utilized. In general, motion saliency requires reliable motion data as well as image segmentation for producing satisfactory saliency map which poses difficulty in most natural images. To overcome this, we propose a motion-based saliency generation that enhances the spatial saliency based on the combination of spatial and motion saliencies as well as motion complexity without the consideration of complex motion classification and image segmentation. Further, an affinity model is integrated for the purpose of connecting close-by pixels with different colors and obtaining a similar saliency. In experiment, we performed the proposed method on eleven test sets. From the objective performance evaluation, we validated that the proposed method produces better result than spatial saliency based on objective evaluation as well as ROC test.

• Korean Journal of Remote Sensing
• /
• v.20 no.2
• /
• pp.139-152
• /
• 2004

This paper proposes to improve the results of image classification with spatial region growing segmentation by using an MRF-based classifier. The proposed approach is to re-classify the pixels in the boundary area, which have high probability of having classification error. The MRF-based classifier performs iteratively classification using the class parameters estimated from the region growing segmentation scheme. The proposed method has been evaluated using simulated data, and the experiment shows that it improve the classification results. But, conventional MRF-based techniques may yield incorrect results of classification for remotely-sensed images acquired over the ground area where has complicated types of land-use. A multistage MRF-based iterative class-modification in boundary is proposed to alleviate difficulty in classifying intricate land-cover. It has applied to remotely-sensed images collected on the Korean peninsula. The results show that the multistage scheme can produce a spatially smooth class-map with a more distinctive configuration of the classes and also preserve detailed features in the map.

• Journal of the Korean Geotechnical Society
• /
• v.34 no.12
• /
• pp.145-154
• /
• 2018

The damage investigation and inspection methods performed in concrete facilities such as bridges, tunnels, retaining walls and so on, are usually visually examined by the inspector using the surveying tool in the field. These methods highly depend on the subjectivity of the inspector, which may reduce the objectivity and reliability of the record. Therefore, the new image processing techniques are necessary in order to automatically detect the cracks and objectively analyze the characteristics of cracks. In this study, deep learning and image processing technique were developed to detect cracks and analyze characteristics in images for concrete facilities. Two-stage image processing pipeline was proposed to obtain crack segmentation and its characteristics. The performance of the method was tested using various crack images with a label and the results showed over 90% of accuracy on crack classification and segmentation. Finally, the crack characteristics (length and thickness) of the crack image pictured from the field were analyzed, and the performance of the developed technique was verified by comparing the actual measured values and errors.

• Korean Journal of Remote Sensing
• /
• v.38 no.5_2
• /
• pp.685-694
• /
• 2022

Building information is essential for various urban spatial analyses. For this reason, continuous building monitoring is required, but it is a subject with many practical difficulties. To this end, research is being conducted to extract buildings from satellite images that can be continuously observed over a wide area. Recently, deep learning-based semantic segmentation techniques have been used. In this study, a part of the structure of the context-based ResU-Net was modified, and training was conducted to automatically extract a building from a 30 cm Worldview-3 RGB image using SpaceNet's building v2 free open data. As a result of the classification accuracy evaluation, the f1-score, which was higher than the classification accuracy of the 2nd SpaceNet competition winners. Therefore, if Worldview-3 satellite imagery can be continuously provided, it will be possible to use the building extraction results of this study to generate an automatic model of building around the world.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2023.05a
• /
• pp.137-137
• /
• 2023

해양쓰레기는 해안의 심미적 가치 저하뿐만 아니라 생태계 파괴, 유령 어업에 따른 수산업 피해 등의 사회적·환경적 문제를 발생시키며, 그중 70% 이상은 육상 기인으로 플라스틱 및 기타 쓰레기가 주를 이루는 해외와 달리 국내의 경우 다량의 초목류를 포함하고 있다. 다양한 부유쓰레기에 대한 기존의 해양쓰레기량 추정의 한계와 하천·하구 쓰레기 수거의 효율화를 위해 해양으로 유입되는 부유쓰레기 방지를 위한 실효성 있는 대책 수립이 필요한 실정이다. 본 연구는 해양 유입 전 하천의 차단시설에 차집된 부유쓰레기의 수거 효율화 및 지속가능한 해양쓰레기 데이터 구축을 위해 AI기반의 기술을 통해 부유쓰레기 성상 분석 기법(Object Detection)과 차집량 분석 기법(Semantic Segmentation)을 활용하였다. 실제와 유사한 데이터 수집을 위해 다양한 하천 환경(정수조, 소하천, 급경사수로)에 대해 탁도(녹조, 유사), 광량, 쓰레기형상, 초목류 함량, 날씨(소하천), 유속(급경사수로) 등의 실험조건에 대하여 해양쓰레기 분류 기준 및 통계를 바탕으로 부유쓰레기 종류 선정하여 학습을 위한 데이터를 수집하였다. 학습 목적에 따라 구분하여 라벨링(Bounding box, Polygon)을 수행하고, 각 분석 기법별 전이학습을 통해 Phase 1(정수조), Phase 2(소하천), Phase 3(급경사수로) 순서로 모델을 고도화하였다. 성상 분석을 위해 YOLO v4를 활용하여 Train, Test DataSet(9:1)을 구성하고 학습 및 평가는 Iteration마다의 mAP, loss 값을 통해 비교하였으며, 학습 Phase에 따라 모델 고도화로 Test Set의 mAP 값이 성상별로 높아짐을 확인하였으며, 차집량 분석을 위해 Unet을 활용하여 Train, Test, Validation DataSet(8.5:1:0.5)을 구성하고 epoch별 IoU(intersection over Union), F1-score, loss 값을 비교하여 정성적, 정량적 평가 모두 Phase 3에서 가장 높은 성능을 확인하였다. 향후 하천 환경에서의 다양한 영양인자별 분석을 통해 주요 영향인자 도출 및 Hyper Parameter 최적화를 통한 모델 고도화로 인해 활용성이 높아질 것으로 판단된다.

• Journal of Biomedical Engineering Research
• /
• v.45 no.2
• /
• pp.90-94
• /
• 2024

Diagnosing wounds presents a significant challenge in clinical settings due to its complexity and the subjective assessments by clinicians. Wound deep learning algorithms quantitatively assess wounds, overcoming these challenges. However, a limitation in existing research is reliance on specific datasets. To address this limitation, we created a comprehensive dataset by combining open dataset with self-produced dataset to enhance clinical applicability. In the annotation process, machine learning based on Gradient Vector Flow (GVF) was utilized to improve objectivity and efficiency over time. Furthermore, the deep learning model was equipped U-net with residual blocks. Significant improvements were observed using the input dataset with images cropped to contain only the wound region of interest (ROI), as opposed to original sized dataset. As a result, the Dice score remarkably increased from 0.80 using the original dataset to 0.89 using the wound ROI crop dataset. This study highlights the need for diverse research using comprehensive datasets. In future study, we aim to further enhance and diversify our dataset to encompass different environments and ethnicities.

• Journal of the Korean Society of Radiology
• /
• v.18 no.1
• /
• pp.21-26
• /
• 2024

The one of the famous deep learning models for object detection task is you only look once version 5 (YOLOv5) framework based on the one stage architecture. In addition, YOLOv5 model indicated high performance for accurate lesion detection using the bottleneck CSP layer and skip connection function. The purpose of this study was to evaluate the performance of YOLOv5 framework according to various hyperparameters in position emission tomogrpahy (PET) phantom images. The dataset was obtained from QIN PET segmentation challenge in 500 slices. We set the bounding box to generate ground truth dataset using labelImg software. The hyperparameters for network train were applied by changing optimization function (SDG, Adam, and AdamW), activation function (SiLU, LeakyRelu, Mish, and Hardwish), and YOLOv5 model size (nano, small, large, and xlarge). The intersection over union (IOU) method was used for performance evaluation. As a results, the condition of outstanding performance is to apply AdamW, Hardwish, and nano size for optimization function, activation function and model version, respectively. In conclusion, we confirmed the usefulness of YOLOv5 network for object detection performance in nuclear medicine images.