• Journal of the Society of Disaster Information
• /
• v.20 no.2
• /
• pp.302-314
• /
• 2024

Purpose: This study proposes a fall detection model based on a top-down deep learning pose estimation model to automatically determine falls of multiple workers in an underground utility tunnel, and evaluates the performance of the proposed model. Method: A model is presented that combines fall discrimination rules with the results inferred from YOLOv8-pose, one of the top-down pose estimation models, and metrics of the model are evaluated for images of standing and falling two or fewer workers in the tunnel. The same process is also conducted for a bottom-up type of pose estimation model (OpenPose). In addition, due to dependency of the falling interference of the models on worker detection by YOLOv8-pose and OpenPose, metrics of the models for fall was not only investigated, but also for person. Result: For worker detection, both YOLOv8-pose and OpenPose models have F1-score of 0.88 and 0.71, respectively. However, for fall detection, the metrics were deteriorated to 0.71 and 0.23. The results of the OpenPose based model were due to partially detected worker body, and detected workers but fail to part them correctly. Conclusion: Use of top-down type of pose estimation models would be more effective way to detect fall of workers in the underground utility tunnel, with respect to joint recognition and partition between workers.

• Journal of Satellite, Information and Communications
• /
• v.8 no.3
• /
• pp.52-57
• /
• 2013

This paper describes the color comparison analysis of flame in each standard color model in order to propose the optimal color model for image processing based flame detection algorithm. Histogram intersection values were used to analyze the separation characteristics between color of flame and color of non-flame in each standard color model which are RGB, YCbCr, CIE Lab, HSV. Histogram intersection value in each color model and components is evaluated for objective comparison. The analyzed result shows that YCbCr color model is the most suitable for flame detection by average HI value of 0.0575. Among the 12 components of standard color models, each Cb, R, Cr component has respectively HI value of 0.0433, 0.0526, 0.0567 and they have shown the best flame separation characteristics.

• Smart Media Journal
• /
• v.11 no.11
• /
• pp.92-98
• /
• 2022

Identifying defects in textiles is a key procedure for quality control. This study attempted to create a model that detects defects by analyzing the images of the fabrics. The models used in the study were deep learning-based VGGNet and ResNet, and the defect detection performance of the two models was compared and evaluated. The accuracy of the VGGNet and the ResNet model was 0.859 and 0.893, respectively, which showed the higher accuracy of the ResNet. In addition, the region of attention of the model was derived by using the Grad-CAM algorithm, an eXplainable Artificial Intelligence (XAI) technique, to find out the location of the region that the deep learning model recognized as a defect in the fabric image. As a result, it was confirmed that the region recognized by the deep learning model as a defect in the fabric was actually defective even with the naked eyes. The results of this study are expected to reduce the time and cost incurred in the fabric production process by utilizing deep learning-based artificial intelligence in the defect detection of the textile industry.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.45 no.6
• /
• pp.38-43
• /
• 2008

Skin color information is an important feature for face region detection in color images. This can detect face region using statistical skin color model who is created from skin color information. However, due to the including of different race of people's skin color points, this general statistical model is not accurate enough to detect each specific image as we expected. This paper proposes method to detect correctly face region in various color image that other complexion part is included. In this method set face candidate region applying complexion Gausian distribution based on YCbCr skin color model and applied mathematical morphology to remove noise part and part except face region in color image. And achieved correct face region detection because using Haar-like feature. This approach is capable to distinguish face region from extremely similar skin colors, such as neck skin color or am skin color. Experimental results show that our method can effectively improve face detection results.

• Transactions on Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.17-20
• /
• 2024

This paper presents a novel lightweight object detection model tailored for low-powered edge devices, addressing the limitations of traditional resource-intensive computer vision models. Our proposed detector, inspired by the Single Shot Detector (SSD), employs a compact yet robust network design. Crucially, it integrates an 'enhancer block' that significantly boosts its efficiency in detecting smaller objects. The model comprises two primary components: the Light_Block for efficient feature extraction using Depth-wise and Pointwise Convolution layers, and the Enhancer_Block for enhanced detection of tiny objects. Trained from scratch on the Udacity Annotated Dataset with image dimensions of 300x480, our model eschews the need for pre-trained classification weights. Weighing only 5.5MB with approximately 0.43M parameters, our detector achieved a mean average precision (mAP) of 27.7% and processed at 140 FPS, outperforming conventional models in both precision and efficiency. This research underscores the potential of lightweight designs in advancing object detection for edge devices without compromising accuracy.

• Proceedings of the Korean Information Science Society Conference
• /
• 2001.10b
• /
• pp.379-381
• /
• 2001

본 논문은 카메라의 움직임이 있는 영상에서 움직이는 사람의 얼굴을 검출하는 방법을 제안한다. 제안된 방법에서, 얼굴 영역을 찾기 위해 피부 색깔 정보와 움직임 정보를 이용한다. 카메라의 움직임을 어파인 모션 모델(Affine Motion Model)을 이용해 제거한 후, 적응적 임계치(adaptive thresholding)를 통해 얻어진 움직임 영역 내에서만 피부 색깔 모델(skin color model)을 이용해 얼굴 영역을 검출한다. 제안된 방법은 시간에 따라 조명이 변하거나 잡음이 포함된 영상에서도 좋은 결과를 얻을 수 있다.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2014.11a
• /
• pp.160-163
• /
• 2014

본 논문에서는 프레임 내 전체 배경 모델을 도입하여 기존 배경 분리 알고리즘에서의 오검출을 줄여 정확도를 개선하고자 한다. 기존의 알고리즘은 프레임 간의 정보만을 이용하여 배경 확률 모델을 만들고 배경을 제외한 전경만을 검출한다. 제안하는 알고리즘에서는 먼저 기존의 알고리즘을 통해 프레임 간의 정보를 이용하여 간단하게 배경과 전경을 분리한다. 그 후 프레임 내 정보를 통해 전체 배경 모델을 만들고, 앞의 결과에서 한번 더 배경을 제외함으로써 검출 정확도를 개선하고자 한다. 실험결과에서 Change Detection Workshop dataset에 대해 실험을 한 후 결과 영상 비교 및 F-measure 를 통해 개선된 결과를 확인할 수 있다.

• Annual Conference of KIPS
• /
• 2010.11a
• /
• pp.679-682
• /
• 2010

피부 영역 검출 기술은 생체 인식 기술의 하나로서 얼굴 자동 인식 혹은 손 모양 자동 인식 등을 위해 사용되고 있다. 일반적으로 색상을 이용하여 피부 영역을 검출하기 위해서는 다양한 피부색 샘플을 이용해 구해진 피부색 모델을 이용한다. 하지만 피부색은 사람마다 다르고, 조명과 같은 주변 환경의 영향도 받기 때문에 다양한 영상에 하나의 고정된 피부색 모델을 적용하여 피부 영역을 검출하기에는 한계가 있다. 이러한 문제점을 해결하기 위해 본 논문에서는 영상 피드백 방법을 이용하여 영상에 적응적인 피부색 모델을 구한 뒤 이를 적용하여 피부 영역을 추출하는 방법을 제안한다.

• The Journal of Engineering Geology
• /
• v.25 no.4
• /
• pp.587-598
• /
• 2015

We examined various existing threshold methods for the determination of the first arrival time of acoustic emission (AE), and developed a new variable threshold method that could determine the first arrival time of AE more accurately and more quickly than existing methods. The new method, a modification of an existing threshold method, does not fix the threshold, but applies variable thresholds for the AE signals according to noise analysis. Two- and three-dimensional models were established to test the effectiveness of the new method. It could determine source locations of AE in a two-dimensional model 38.3% more accurately than the pre-existing threshold methods. Its accuracy improvement over the existing methods in a three-dimensional model was about 15.2%. A practical test involved measuring the source locations of AE during three-point bending tests of granite cores. The new method placed the sources closer to the fracture plane than did the pre-existing methods, indicating its superior (and quicker) ability to determine the source locations of AE.

• Proceedings of the Korea Contents Association Conference
• /
• 2004.11a
• /
• pp.377-383
• /
• 2004

In this paper, we propose an application model of RFID that needs not be worried about battery degradation problems and offers convenience of hand-carry. We design the RFID application model which can track medium distance target RFID Tag and detect the signal difference between multi-detectors in a room area using 900MHz frequency band. In the case of multi-ID Tag in one detector, we approach TDMA scheme.