• Journal of Broadcast Engineering
• /
• v.26 no.4
• /
• pp.419-428
• /
• 2021

Shadow removal plays a key role for the pre-processing of image processing techniques such as object tracking and detection. With the advances of image recognition based on deep convolution neural networks, researches for shadow removal have been actively conducted. In this paper, we propose a novel method for shadow removal, which utilizes self attention distillation to extract semantic features. The proposed method gradually refines results of shadow detection, which are extracted from each layer of the proposed network, via top-down distillation. Specifically, the training procedure can be efficiently performed by learning the contextual information for shadow removal without shadow masks. Experimental results on various datasets show the effectiveness of the proposed method for shadow removal under real world environments.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.6
• /
• pp.615-621
• /
• 2020

In this paper, we propose a multi-site based earthquake event classification method using graph convolution networks. In the traditional earthquake event classification methods using deep learning, they used single-site observation to estimate seismic event class. However, to achieve robust and accurate earthquake event classification on the seismic observation network, the method using the information from the multi-site observations is needed, instead of using only single-site data. Firstly, our proposed model employs convolution neural networks to extract informative embedding features from the single-site observation. Secondly, graph convolution networks are used to integrate the features from several stations. To evaluate our model, we explore the model structure and the number of stations for ablation study. Finally, our multi-site based model outperforms up to 10 % accuracy and event recall rate compared to single-site based model.

• Korean Journal of Remote Sensing
• /
• v.35 no.6_1
• /
• pp.999-1009
• /
• 2019

The importance of high-resolution sea ice maps of the Arctic Ocean is increasing due to the possibility of pioneering North Pole Routes and the necessity of precise climate prediction models. In this study,sea ice classification algorithms for two deep learning models were examined using Sentinel-1 A/B SAR data to generate high-resolution sea ice classification maps. Based on current ice charts, three classes (Open Water, First Year Ice, Multi Year Ice) of training data sets were generated by Arctic sea ice and remote sensing experts. Ten sea ice classification algorithms were generated by combing two deep learning models (i.e. Simple CNN and Resnet50) and five cases of input bands including incident angles and thermal noise corrected HV bands. For the ten algorithms, analyses were performed by comparing classification results with ground truth points. A confusion matrix and Cohen's kappa coefficient were produced for the case that showed best result. Furthermore, the classification result with the Maximum Likelihood Classifier that has been traditionally employed to classify sea ice. In conclusion, the Convolutional Neural Network case, which has two convolution layers and two max pooling layers, with HV and incident angle input bands shows classification accuracy of 96.66%, and Cohen's kappa coefficient of 0.9499. All deep learning cases shows better classification accuracy than the classification result of the Maximum Likelihood Classifier.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.25 no.4
• /
• pp.476-481
• /
• 2019

Early detection of fire is an important measure for minimizing the loss of life and property damage. However, fire and smoke need to be simultaneously detected. In this context, numerous studies have been conducted on image-based fire detection. Conventional fire detection methods are compute-intensive and comprise several algorithms for extracting the flame and smoke characteristics. Hence, deep learning algorithms and convolution neural networks can be alternatively employed for fire detection. In this study, recorded image data of fire in a ship engine room were analyzed. The flame and smoke characteristics were extracted from the outer box, and the YOLO (You Only Look Once) convolutional neural network algorithm was subsequently employed for learning and testing. Experimental results were evaluated with respect to three attributes, namely detection rate, error rate, and accuracy. The respective values of detection rate, error rate, and accuracy are found to be 0.994, 0.011, and 0.998 for the flame, 0.978, 0.021, and 0.978 for the smoke, and the calculation time is found to be 0.009 s.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.07a
• /
• pp.47-50
• /
• 2023

본 논문에서는 딥러닝 CNN 알고리즘을 이용하여 사람의 얼굴 이미지를 학습한 다음 나이와 성별을 예측하는 시스템을 제안한다. 이 시스템은 개개인 마다 각기 다른 외형적 특성을 고려하여 이를 분석한 다음 이에 맞는 헤어 스타일, 옷차림을 추천할 수 있다. 해당 기술을 활용하여 메타버스 아바타 생성에 사용자의 얼굴과 같은 신체적 특성을 고려할 수 있다. 향후에는 신체 전체를 이미지화하여 보다 더 다양한 정보를 인식할 수 있도록 연구를 진행할 것이다.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.7
• /
• pp.1100-1104
• /
• 2017

A detector-based approach shows the limited performances for the defect inspections such as shallow fine cracks and indistinguishable defects from background. Deep learning technique is widely used for object recognition and it's applications to detect defects have been gradually attempted. Deep learning requires huge scale of learning data, but acquisition of data can be limited in some industrial application. The possibility of applying CNN which is one of the deep learning approaches for surface defect inspection is investigated for industrial parts whose detection difficulty is challenging and learning data is not sufficient. VOV is adopted for pre-processing and to obtain a resonable number of ROIs for a data augmentation. Then CNN method is applied for the classification. Three CNN networks, AlexNet, VGGNet, and mofified VGGNet are compared for experiments of defects detection.

• Proceedings of the Korean Society of Computer Information Conference
• /
• 2023.01a
• /
• pp.369-372
• /
• 2023

본 논문에서는 삼각형 구조로 구성된 3차원 메쉬(Mesh)에서 합성곱 신경망(Convolution Neural Network, CNN)을 응용하여 정확도가 높은 새로운 학습 표현 기법을 제시한다. 우리는 메쉬를 구성하고 있는 폴리곤의 edge와 face의 로컬 특징을 기반으로 학습을 진행한다. 일반적으로 딥러닝은 인공신경망을 수많은 계층 형태로 연결한 기법을 말하며, 주요 처리 대상은 1, 2차원 데이터 형태인 오디오 파일과 이미지였다. 인공지능에 대한 연구가 지속되면서 3차원 딥러닝이 도입되었지만, 기존의 학습과는 달리 3차원 딥러닝은 데이터의 확보가 쉽지 않다. 혼합현실과 메타버스 시장의 확대로 인해 3차원 모델링 시장이 증가하고, 기술의 발전으로 데이터를 획득할 수 있는 방법이 생겼지만, 3차원 데이터를 직접적으로 학습에 이용하는 방식으로 적용하는 것은 쉽지 않다. 그렇게 때문에 본 논문에서는 산업 현장에서 이용되는 데이터인 메쉬 구조를 폴리곤의 최소 단위인 삼각형 형태로 구성하여 학습 데이터를 구성해 기존의 방법보다 정확도가 높은 학습 기법을 제안한다.

• Journal of Institute of Control, Robotics and Systems
• /
• v.21 no.11
• /
• pp.1008-1016
• /
• 2015

This paper proposes a method for detecting the front side of vehicles. The method can find the car side with a license plate even with complicated and cluttered backgrounds. A convolutional neural network (CNN) is used to solve the detection problem as a unified framework combining feature detection, classification, searching, and localization estimation and improve the reliability of the system with simplicity of usage. The proposed CNN structure avoids sliding window search to find the locations of vehicles and reduces the computing time to achieve real-time processing. Multiple responses of the network for vehicle position are further processed by a weighted clustering and probabilistic threshold decision method. Experiments using real images in parking lots show the reliability of the method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.18 no.2
• /
• pp.456-477
• /
• 2024

With information technology's rapid development, the Internet faces serious security problems. Studies have shown that malware has become a primary means of attacking the Internet. Therefore, adversarial samples have become a vital breakthrough point for studying malware. By studying adversarial samples, we can gain insights into the behavior and characteristics of malware, evaluate the performance of existing detectors in the face of deceptive samples, and help to discover vulnerabilities and improve detection methods for better performance. However, existing adversarial sample generation methods still need help regarding escape effectiveness and mobility. For instance, researchers have attempted to incorporate perturbation methods like Fast Gradient Sign Method (FGSM), Projected Gradient Descent (PGD), and others into adversarial samples to obfuscate detectors. However, these methods are only effective in specific environments and yield limited evasion effectiveness. To solve the above problems, this paper proposes a malware adversarial sample generation method (PixGAN) based on the pixel attention mechanism, which aims to improve adversarial samples' escape effect and mobility. The method transforms malware into grey-scale images and introduces the pixel attention mechanism in the Deep Convolution Generative Adversarial Networks (DCGAN) model to weigh the critical pixels in the grey-scale map, which improves the modeling ability of the generator and discriminator, thus enhancing the escape effect and mobility of the adversarial samples. The escape rate (ASR) is used as an evaluation index of the quality of the adversarial samples. The experimental results show that the adversarial samples generated by PixGAN achieve escape rates of 97%, 94%, 35%, 39%, and 43% on the Random Forest (RF), Support Vector Machine (SVM), Convolutional Neural Network (CNN), Convolutional Neural Network and Recurrent Neural Network (CNN_RNN), and Convolutional Neural Network and Long Short Term Memory (CNN_LSTM) algorithmic detectors, respectively.

• Journal of Advanced Navigation Technology
• /
• v.28 no.5
• /
• pp.601-609
• /
• 2024

In this paper, we present the design and implementation results of a system that classifies drones from other objects using an FMCW (frequency-modulated continuous wave) radar sensor. The proposed system detects various objects through a four-stage signal processing procedure, consisting of FFT, CFAR, clustering, and tracking, using signals received from the radar sensor. Subsequently, a deep learning process is conducted to classify the detected objects as either drones or other objects. To mitigate the high computational demands and extensive memory requirements of deep learning, a BNN (binary neural network) structure was applied, binarizing the CNN (convolutional neural network) operations. The performance evaluation and verification results demonstrated a drone classification accuracy of 89.33%, with a total execution time of 4 ms, confirming the feasibility of real-time operation.