Earticle

To address the high complexity, poor real-time performance, and the prevalence of false positives and false negatives in current algorithms for detecting small-target pollutants on UAV-based building facades, this study proposes SDS-YOLOv8. The spatial pyramid pooling structure in the backbone is enhanced to improve feature representation. DySample is incorporated into the neck to adaptively adjust sampling points based on the image feature distribution. Additionally, the SCAM module is introduced to improve the memory of important information, and the loss function is further optimized. Experimental results demonstrate that the accuracy of the proposed algorithm is significantly improved, exhibiting strong generalization capability. ⓒ2025 The Korean Institute of Communications and Information Sciences. Publishing Services by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

This study proposes a smart camera system for detecting fires in Building-Integrated Photovoltaic (BIPV) systems. Traditional fire detection methods, such as smoke detectors, temperature sensors, thermal cameras, and videobased surveillance systems, each have specific advantages and disadvantages depending on the environment. However, due to the unique characteristics of BIPV systems, a more sophisticated and integrated detection system is required. In this study, we developed a smart camera system using an NVIDIA Jetson Orin Nano board, a YOLOv5 model, and a Lepton thermal camera to monitor and provide early warnings of fires in BIPV systems in real time. Experimental results across various environments demonstrated that the proposed system offers higher accuracy and reliability compared to existing methods.

The purpose of this study is to provide the damage detection method on shear building structures by the damage index directly induced from dynamic equation. The provided damage index could be estimated from measured mode shape of undamaged structure and frequency difference between undamaged and damaged structure. The damage detection method is applied to numerical analysis model such as MATLAB and MIDAS GENw for the verification. The damage index at damaged story represents (-) sign and 15 times than other undamaged stories.

The purpose of this study is to present the damage detection method on shear building structures by mode shape. The damage location index using 1st mode shape is observed theoretically to find out damage location. The damage detection method is applied to numerical analysis model such as MATLAB and MIDAS GENw for the verification. Finally the shacking table test on 3 story shear building is performed for the examination of the damage detection method.

본 논문에서는 LSTM(Long Short-Term Memory)을 기반으로 하는 Deep Learning 모델을 구축하여 인간의 습관적 특성을 고려한 악성 도메인 탐지 방법을 제시한다. DGA(Domain Generation Algorithm) 악성 도메인은 인간의 습관적 인 실수를 악용하여 심각한 보안 위협을 초래한다. 타이포스쿼팅을 통한 악성 도메인의 변화와 은폐 기술에 신속히 대 응하고, 정확하게 탐지하여 보안 위협을 최소화하는 것이 목표이다. LSTM 기반 Deep Learning 모델은 악성코드별 특 징을 분석하고 학습하여, 생성된 도메인을 악성 또는 양성으로 자동 분류한다. ROC 곡선과 AUC 정확도를 기준으로 모 델의 성능 평가 결과, 99.21% 이상 뛰어난 탐지 정확도를 나타냈다. 이 모델을 활용하여 악성 도메인을 실시간 탐지할 수 있을 뿐만 아니라 다양한 사이버 보안 분야에 응용할 수 있다. 본 논문은 사용자 보호와 사이버 공격으로부터 안전 한 사이버 환경 조성을 위한 새로운 접근 방식을 제안하고 탐구한다.

This paper proposes a method for detecting malicious domains considering human habitual characteristics by buil ding a Deep Learning model based on LSTM (Long Short-Term Memory). DGA (Domain Generation Algorithm) m alicious domains exploit human habitual errors, resulting in severe security threats. The objective is to swiftly and accurately respond to changes in malicious domains and their evasion techniques through typosquatting to minimize security threats. The LSTM-based Deep Learning model automatically analyzes and categorizes generated domains as malicious or benign based on malware-specific features. As a result of evaluating the model's performance based on ROC curve and AUC accuracy, it demonstrated 99.21% superior detection accuracy. Not only can this model dete ct malicious domains in real-time, but it also holds potential applications across various cyber security domains. Thi s paper proposes and explores a novel approach aimed at safeguarding users and fostering a secure cyber environm ent against cyber attacks.

기존의 범죄 예측 모델은 정형 데이터에 편중되어 대중 담론 내부에 잠재된 전조 신 호를 파악하는 데 제약이 따른다. 이에 본 연구는 비정형 빅데이터에 텍스트마이닝 기 법을 도입하여 폭력 범죄 관련 온라인 담론의 사회 변동의 초기 징후를 탐지하는 사회 적 위험의 선제적 인지 시스템을 제안하였다. 2021년부터 2025년까지 네이버 뉴스 기사 10,825건 및 댓글 371,292건을 분리 분석한 결과 기사 담론은 치안 정책·수사 절차 개편 맥락의 어휘가 약신호로 부상한 반면 댓글 담론은 사법 불신과 응보적 열망이 강신호 로 형성되는 담론 구조의 괴리가 확인되었다. 정치면 기사의 기사당 댓글 수가 사회면 의 3.6배에 달하는 플랫폼 구조가 댓글 기반 신호의 정치화를 추동하고 있음도 실증하 였다. 본 연구는 Pre-CAS 고도화와 거버넌스의 대응 전략 제고를 위한 실무적 함의를 제시하며 향후 KICS 데이터와의 인과관계 검증을 통한 후속 연구를 제언한다.

Existing crime prediction models are biased toward structured data, limiting their ability to detect subtle precursor signals embedded in public discourse. This study proposes a Proactive Social Risk Recognition System that applies text mining techniques to unstructured big data to detect early indicators of social change in violent crime-related online discourse. Analyzing 10,825 news articles and 371,292 comments collected from Naver News between 2021 and 2025, the results reveal a clear discrepancy in discourse structure: article discourse showed policing policy and investigative procedure reform as emerging weak signals, whereas comment discourse formed strong signals centered on judicial distrust and retributive demands. The study also demonstrates that the platform structure—where political section articles receive 3.6 times more comments per article than social section articles—structurally drives the politicization of comment-based signals. The study presents practical implications for advancing Pre-CAS and improving governance response strategies, and recommends follow-up research through causal verification with KICS data.

본 연구는 다중이용시설 이용자들의 쾌적함과 안전 그리고 시설내부 에너지 사용량의 최적 절감을 위하여 이 용자수를 분석예측한 정보에 따른 공기질품질제어시스템 운영을 통해 국민 중심의 안전하고 쾌적한 서비스를 제 공할 필요로 수행되었다. 이를 위하여 실내유동인구수를 카운팅하는 로컬시스템을 제작하고 수집된 유동인구 카 운팅 정보를 시계열 모델링을 기반으로 분석예측하는 연구를 진행하였다. 개발된 시스템 성능평가 결과 유동인구 카운팅시스템은 95% 이상 정확도를 보여주었고, 예측시스템은 83~95% 정확도를 확보하였다. 본 연구결과 개발 된 시스템은 다중이용시설에 즉시 적용가능하며 향후 남녀노소 인식을 진행하고 이를 예측한 정보에 의한 보다 다양한 서비스 개발을 추진할 계획이다.

This paper presents two methods for buildings extraction in Very High Resolution (VHR) remotely sensed multispectral (4 band) images based on supervised and unsupervised segmentation using different image properties. The proposed approach for unsupervised or automatic building detection involves four stages, primarily, filtering to smoothen and enhance the objects present and sharpen the details. Secondly, a binary mask creation over which edge detection is applied. Edge linking is done to preserve information about the object. Lastly region properties like area, perimeter, etc are applied on the prepared mask and buildings are detected. The semi-automated or supervised method uses advanced color based segmentation algorithm to extract the buildings tops. This technique creates a number of masks based on segmentation and uses region properties based on color. Experiments are made on VHR images captured from satellites of commercial companies like Digital Globe and Geo Eye. Results of both methods are compared with respect to various accuracy measures at the end. The results illustrate that supervised algorithm using color property produces more accurate object delineation.

This paper presents a new Discriminative Random Fields (DRFs) framework. Based on the DRFs framework proposed by Kumar and Hebert, the following improvements have been conducted. Firstly, the interaction potential and the associated potential model are simplified. Secondly, we reduce the dimension of the multi-scale features, re-definedimension of the single-scale feature, and increase the color feature of Building. Thirdly,the quasi-Newton method with linear search and gradient descent method are adopted to solve parameters, whichget a simple model and achieve good performance. Finally, the partition function of the DRF is eliminatedby using Pseudo-likelihood method for parameter learning. The simulation results show thatthe proposed method’s false positive rate is lower than the method from Kumar and Hebert, while the correct rate and detection ratearehigher than their experimental effects after these improvements.

Most health problems of building structures are accumulative damages which are difficult to detect, and it is more difficult to monitor the structure health due to the complexity of the practical structure and the environment noise, and the existing methods need lots of data for model training but it is very complicated to mark the data in practice. In order to solve above problems, the wireless sensor network is configured and the sparse encoding method is adopted to monitor the bridge structure health, and meanwhile the sparse encoding algorithm is adopted for training on the basis of the characteristic extraction of many unlabeled instances, thus to compress data dimensionality and preprocess unlabeled data. Then, the deep learning algorithm is adopted to predict the bridge structure health monitoring type, and meanwhile Hessian optimization is improved on the basis of the linear conjugate gradient in order to replace uncertain Hessian matrix by positive semidefinite Gaussian - Newton curvature matrix for secondary objective combination, thus to improve the efficiency of the deep learning algorithm. The experiment result shows that the security detection of the bridge structure based on deep learning algorithm can monitor the high-accuracy structure health conditions under the sparse encoding of the environment noise.

Since the very beginning of the industrial revolution the work-force (employees) is forced to perform repeated laborious tasks round the clock. To make things worse, most of the tasks are non-intellectual. With the passage of time while the employee is gaining experience and becoming more productive, he is losing human characteristics. And gradually transforming into a circus animal. Being inherently irregular and unpredictable employee (humans) require a system that can not only curb their irregularities but also, reminds them of their expected location via vocal announcements in real time. To meet these requirements, we present an efficient system that tracks the where about of employees in indoor environment. And keeps a compact tempo-spatial track records of each employee. It helps them to avoid irregularities in real time which otherwise may lead to accumulation of a considerable debt at the end of the month. The proposed system is both employee and organization friendly. The employee benefits because he/she is not deprived of human comfort level. And the organization benefits because the efficiency of employee is increase in real time.

Datasets are a foundational step in the development of any Artificial Intelligence (AI) powered solutions. In cybersecurity, especially in malware detection and mitigation, cybersecurity AI datasets focusing on malware can play a critical role in improving accuracy and efficiency of AI models. In this paper we explore several recent techniques used in construction of malware AI datasets, identify gaps and recommend practical solutions to address them. Specifically, we explore various frameworks and techniques for improving data collection, preprocessing and dataset validation. Furthermore, we explore various recent approaches applied in AI based malware detection. In a special way we examine shallow learning, deep learning, bio-inspired computing, behavior-based detection, heuristic-based approaches, and hybrid approaches. We then draw our observations and recommend specific strategies for improving the process of malware AI dataset construction as well as detection techniques. Through our research we also contribute to the ongoing much needed efforts for combating malware attacks by providing a framework for building quality malware focused cybersecurity AI datasets, there by improving the current state of the art AI-powered malware detection systems.

Recently, due to decrease hardware prices and the development of technology, analog signage has been changing to digital signage for providing content such as advertisements, videos. Furthermore, in order to provide advertisements and contents to users more effectively, technical researches are being conducted in various industries. In addition, including digital signage that uses displays, it can be seen that it provides advertisements and contents using diverse devices such as LED signage, smart pads, and smart phones. However, most digital signage is installed in one place to provide contents and provides interactivity through simple events such as manual content provision or touch. So, in this paper, we suggest a new object detection algorithm based on an adjacent frames based image correction algorithm for interactive building signage.

This paper presents an algorithm for building detection from aerial image using segment measure function and line relation. In the detection algorithm proposed, edge detection, linear approximation and line linking are used and then line measure function is applied to each line segment in order to improve the accuracy of linear approximation. Parallelisms, orthogonalities are applied to the extracted liner segments to extract building. The algorithm was applied to aerial image and the buildings were accurately detected.

Building detection is a fundamental and challenging task in satellite image analysis and has significant importance in the wide range of applications such as creation and update of maps and GIS database, urban monitoring, and planning application. This paper presents an approach for building detection in high resolution remotely sensed images. The proposed approach firstly segmented the pan-sharpened image by Automatic Histogram based Fuzzy C-Means (AHFCM) algorithm. Secondly the vegetation and shadow regions are eliminated by using Normalized Difference Vegetation Index (NDVI) and ratio map in HIS model respectively. Finally the buildings and non building regions are classified by using Principal Component Analysis (PCA) and outcomes are further refined by morphological operations.

This paper presents a new method for building detection and reconstruction from aerial images. In our approach, we extract useful building location information from the generated disparity map to segment the interested objects and consequently reduce unnecessary line segments extracted in the low level feature extraction step. Hypothesis selection is carried out by using an undirected graph, in which close cycles represent complete rooftops hypotheses. We test the proposed method with the synthetic images generated from Avenches dataset of Ascona aerial images. The experiment result shows that the extracted 3D line segments of the reconstructed buildings have an average error of 1.69m and our method can be efficiently used for the task of building detection and reconstruction from aerial images.

In this letter, we propose a novel approach to detecting and tracking apartment buildings for the development of a video-based navigation system that provides augmented reality representation of guidance information on live video sequences. For this, we propose a building detector and tracker. The detector is based on the AdaBoost classifier followed by hierarchical clustering. The classifier uses modified Haar-like features as the primitives. The tracker is a motion-adjusted tracker based on pyramid implementation of the Lukas-Kanade tracker, which periodically confirms and consistently adjusts the tracking region. Experiments show that the proposed approach yields robust and reliable results and is far superior to conventional approaches.

In this paper the advanced method of geodesic active contours was developed for the task of building detection from aerial and satellite images. Automatic extraction of man-made structures including buildings, building blocks or roads from remote sensing data is useful for land use mapping, scene understanding, robotic navigation, image retrieval, surveillance, emergency management procedures, cadastral etc. A level set method based on a region-driven segmentation model was implemented with which building boundaries were detected, through this curve propagation technique. The essence of this approach is to optimize the position and the geometric form of the curve by measuring information along that curve, and within the regions that compose the image partition. To this end, one can consider uniform intensities inside objects and the background. Thus, given an initial position of the curve, one can determine global, region-driven functions and provide a statistical description of the inside and outside object area. The calculus of variations and a gradient descent method was used to optimize the variational functional by an iterative steady state process. Experimental results demonstrate the potential of the proposed processing scheme.

Many building detection methods mainly rely on line segments extracted from aerial or satellite imagery. Building detection methods based on line segments, however, are difficult to succeed in high resolution satellite imagery such as IKONOS imagery, for most buildings in IKONOS imagery have small size of roofs with low contrast between roof and background. In this paper, we propose an efficient method to extract line segments and group them at the same time. First, edge preserving filtering is applied to the imagery to remove the noise. Second, we segment the imagery by watershed method, which collects the pixels with similar intensities to obtain homogeneous region. The boundaries of homogeneous region are not completely coincident with roof boundaries due to low contrast in the vicinity of the roof boundaries. Finally, to resolve this problem, we set up snake model with segmented region boundaries as initial snake's positions. We used a greedy algorithm to fit a snake to roof boundary. Experimental results show our method can obtain more .correct roof boundary with small size and low contrast from IKONOS imagery. Snake algorithm, building roof detection, watershed segmentation, edge-preserving filtering