• 전기전자학회논문지
• /
• 제24권4호
• /
• pp.1148-1155
• /
• 2020

Fire must be extinguished as quickly as possible because they cause a lot of economic loss and take away precious human lives. Especially, the detection of smoke, which tends to be found first in fire, is of great importance. Smoke detection based on image has many difficulties in algorithm research due to the irregular shape of smoke. In this study, we introduce a new real-time smoke detection algorithm that reduces the detection of false positives generated by irregular smoke shape based on faster r-cnn of factory-installed surveillance cameras. First, we compute the global frame similarity and mean squared error (MSE) to detect the movement of smoke from the input surveillance camera. Second, we use deep learning algorithm (Faster r-cnn) to extract deferred candidate regions. Third, the extracted candidate areas for acting are finally determined using space and temporal features as smoke area. In this study, we proposed a new algorithm using the space and temporal features of global and local frames, which are well-proposed object information, to reduce false positives based on deep learning techniques. The experimental results confirmed that the proposed algorithm has excellent performance by reducing false positives of about 99.0% while maintaining smoke detection performance.

• 한국HCI학회:학술대회논문집
• /
• 한국HCI학회 2008년도 학술대회 1부
• /
• pp.215-217
• /
• 2008

동영상 내용검색을 위해서 가장 많이 사용되고 있는 기술은 컷 추출에 의한 내용비교 방법이다. 그러나 컷 추출을 위해 사용되는 CHD(Color Histogram Difference)나 ECR(Edge Change Ratio)등은 영상물의 Cropping, Resizing Low bit rate등의 변화에 대해 대단히 취약하다. 본 방법은 이러한 변형에 강인하도록 상대적으로 변형이 적은 오디오정보를 이용하여 Indexing과 Searching을 수행하였다. 특히 변형에 강인한 Searching을 위해 오디오의 장면(Scene)을 검출하였고 장면을 중심으로 Time-frequency domain에서 각각의 Frequency bin. 에 대한 스펙트럴 파워를 파워임계값을 중심으로 이진화(Binary)하였다. 제안된 방법으로 Cropping, clipping, Lowbit rate, Additive Frame 등의 변형본에 대한 검색을 시도한 결과 False posit ive Error 와 True Negative Error 에 대해 각각 1%미만의 오탐지 결과를 얻었다.

• 한국HCI학회:학술대회논문집
• /
• 한국HCI학회 2009년도 학술대회
• /
• pp.193-196
• /
• 2009

단일 영상에서 피부색 영역을 추출하기 위해서 기존의 많은 방법들이 하나의 고정된 피부색 모델을 사용한다. 그러나 영상에 특성에 따라 영상에 포함된 피부색의 분포가 다양하기 때문에 이러한 방법을 이용하여 피부색을 검출할 경우 낮은 검출율이나 높은 긍정 오류율이 발생할 수 있다. 따라서 영상의 특징에 따라 적응적으로 피부색 영역을 추출할 수 있는 방법이 필요하다. 이에 본 논문에서는 영상의 특징에 따라 2단계의 과정을 거쳐 피부색 모델을 수정하는 방법으로, 다양한 조명과 환경 조건에서 높은 검출율과 낮은 긍정 오류율을 동시에 가지는 알고리즘을 제안한다.

• Journal of the Optical Society of Korea
• /
• 제14권4호
• /
• pp.451-457
• /
• 2010

Increasing demands on the safety of public train services have led to the development of various types of security monitoring systems. Most of the surveillance systems are focused on the estimation of crowd level in the platform, thereby yielding too many false alarms. In this paper, we present a novel security monitoring system to detect critically dangerous situations such as when a passenger falls from the station platform, or when a passenger walks on the rail tracks. The method is composed of two stages of detecting dangerous situations. Objects falling over to the dangerous zone are detected by motion tracking. 3D depth information retrieved by the stereo vision is used to confirm fallen events. Experimental results show that virtually no error of either false positive or false negative is found while providing highly reliable detection performance. Since stereo matching is performed on a local image only when potentially dangerous situations are found; real-time operation is feasible without using dedicated hardware.

• 대한건축학회논문집:구조계
• /
• 제36권3호
• /
• pp.113-120
• /
• 2020

The purpose of this study is to compare the models of Deep Learning-based Convolution Neural Network(CNN) for concrete crack detection. The comparison models are AlexNet, GoogLeNet, VGG16, VGG19, ResNet-18, ResNet-50, ResNet-101, and SqueezeNet which won ImageNet Large Scale Visual Recognition Challenge(ILSVRC). To train, validate and test these models, we constructed 3000 training data and 12000 validation data with 256×256 pixel resolution consisting of cracked and non-cracked images, and constructed 5 test data with 4160×3120 pixel resolution consisting of concrete images with crack. In order to increase the efficiency of the training, transfer learning was performed by taking the weight from the pre-trained network supported by MATLAB. From the trained network, the validation data is classified into crack image and non-crack image, yielding True Positive (TP), True Negative (TN), False Positive (FP), False Negative (FN), and 6 performance indicators, False Negative Rate (FNR), False Positive Rate (FPR), Error Rate, Recall, Precision, Accuracy were calculated. The test image was scanned twice with a sliding window of 256×256 pixel resolution to classify the cracks, resulting in a crack map. From the comparison of the performance indicators and the crack map, it was concluded that VGG16 and VGG19 were the most suitable for detecting concrete cracks.

• 대한세포병리학회지
• /
• 제19권2호
• /
• pp.111-118
• /
• 2008

Background : Cervicovaginal cytology is a screening test of uterine cervical cancer. The sensitivity of cervicovaginal cytology is less than 50%, but studies of cytologic/histologic correlation are limited. We analyzed the diagnostic accuracy of cervicovaginal cytology in the detection of the squamous epithelial lesions of the uterine cervix and investigate the cause of diagnostic discordance. Materials and Methods : We collected a total of 481 sets of cervicovaginal cytology and biopsies over 5 years. The cytologic diagnoses were categorized based on The Bethesda System and the histologic diagnoses were classified as negative, flat condyloma, cervical intraepithelial neoplasia (CIN) I, CIN II, CIN III, or squamous cell carcinoma. Cytohistologic discrepancies were reviewed. Results: The concordance rate between the cytological and the histological diagnosis was 79.0%. The sensitivity and specificity of cervicovaginal cytology were 80.6% and 92.6%, respectively. Its positive predictive value and negative predictive value were 93.7% and 77.7%, respectively. The false negative rate was 19.4%. Among 54 false negative cytology cases, they were confirmed by histology as 50 flat condylomas, 2 CIN I, 1 CIN III, and 1 squamous cell carcinoma. The causes of false negative cytology were sampling errors in 75.6% and interpretation errors in 24.4%. The false positive rate was 7.4%. Among 15 false positive cytology cases, they were confirmed by histology as 12 atypical squamous cells of undetermined significance (ASCUS) and 3 low grade squamous intraepithelial lesions (LSIL). The cause of error was interpretation error in all cases. The overall diagnostic accuracy of cervicovaginal cytology was 85.7%. Conclusions : Cervicovaginal cytology shows high overall diagnostic accuracy and is a useful primary screen of uterine cervical cancer.

• 산업경영시스템학회지
• /
• 제41권2호
• /
• pp.9-15
• /
• 2018

This study is intended to investigate that it is possible to analyze the public awareness and satisfaction of the weather forecast service provided by the Korea Meteorological Administration (KMA) through social media data as a way to overcome limitations of the questionnaire-based survey in the previous research. Sentiment analysis and association rule mining were used for Twitter data containing opinions about the weather forecast service. As a result of sentiment analysis, the frequency of negative opinions was very high, about 75%, relative to positive opinions because of the nature of public services. The detailed analysis shows that a large portion of users are dissatisfied with precipitation forecast and that it is needed to analyze the two kinds of error types of the precipitation forecast, namely, 'False alarm' and 'Miss' in more detail. Therefore, association rule mining was performed on negative tweets for each of these error types. As a result, it was found that a considerable number of complaints occurred when preventive actions were useless because the forecast predicting rain had a 'False alarm' error. In addition, this study found that people's dissatisfaction increased when they experienced inconveniences due to either unpredictable high winds and heavy rains in summer or severe cold in winter, which were missed by weather forecast. This study suggests that the analysis of social media data can provide detailed information about forecast users' opinion in almost real time, which is impossible through survey or interview.

• 한국산업보건학회지
• /
• 제21권4호
• /
• pp.222-226
• /
• 2011

This study was conducted to identify the characteristics of analytical errors shown in the Korean quality control program on bulk asbestos analyses using polarized light microscopy (PLM). 179 participating laboratories were required to analyze 4 samples respectively and asked to classify each test sample as asbestos-containing (positive) or non-asbestos-containing (negative). For positive samples, participants were also asked to identify the type and semiquantitate the contents of asbestos present. The test results showed 21 (4%) false negative errors among 562 samples, 9 (6%) false positive errors among 154 samples and 53 (9%) asbestos identification errors among 562 samples. Most of false negative and positive errors were observed in a few types of samples. Higher frequencies of asbestos identification errors were shown in samples containing two or more types of asbestos and samples containing anthophyllite, tremolite or actinolite asbestos. For semiquantitative analyses, the ratios of mean to nominal weight contents were 2.1 for chrysotile and 2.9 for amphiboles. A tendency of over-estimation was observed in semiquantitative analyses using the visual estimation technique and higher in case of analyzing samples containing amphiboles than chrysotile. Coefficients of variation (CVs) of semiquantitative analytical results were 0.44~0.83 and 0.5~1.14 for samples containing chrysotile and amphibole asbestos, respectively.

• 대한세포병리학회지
• /
• 제8권1호
• /
• pp.11-19
• /
• 1997

Clinical lymphadenopathies are subjected to fine needle aspiration cytology(FNAC) for diagnosing not only benign lesions but also malignant ones, as the first diagnostic procedure. While the diagnostic reliability in metastatic carcinoma is high, it is difficult to differentiate malignant lymphoma from reactive conditions. We evaluated the diagnostic reliability of FNAC in lymphadenopathy, and discuss the diagnostic limitation and its place in clinical practice in this study, Over 8 years from January 1988, FNAC of 1,216 lymphadenopathies were analyzed and among them 170 cases were compared with histopathology. The results are as follows. 1. Of ail the cases, 890 cases(73.2%) were diagnosed cytologically as benign, 312 cases(25.7%) as malignant, and 14 cases(1.1%) as unsatisfactory material. Reactive hyperplasia was diagnosed in 585 cases(65.7%) of the benign lesions, and among the malignant diseases, metastatic carcinoma was diagnosed in 248 cases(79.5%), and malignant lymphoma in 62 cases(19.9%). 2. The overall diagnostic accuracy was 89.2%, and no false positive case and 9 false negative results were observed among 170 cases which were proven by histopathology. Six cases of sampling error of false negative diagnoses included 3 of metastatic carcinomas and 3 of malignant lymphomas. The causes were difference between aspiration and biopsy site, poor fixation, or scanty cellularity with bloody smear. All 3 cases of misinterpretation error were malignant lympliomas, one of mixed type on biopsy which was diagnosed as reactive hyperplasia cytologically. In summary, FNAC technique is thought to be useful in the initial diagnosis of lymphadenopathies as well as in the follow-up of patients with known malignancy. Although the results of malignant lymphoma was less accurate than other malignant lesions, the application of strict cytologic criteria or lymphoid marker studies of aspiration material will reduce the false negative rate.

• 전자공학회논문지
• /
• 제53권7호
• /
• pp.49-55
• /
• 2016

디지털 포렌식 영상은 여러 가지 영상타입으로 위 변조되어 유통되는 심각한 문제가 대두되어 있다. 이러한 문제를 해결하기 위하여, 본 논문에서는 포렌식 영상의 분류 알고리즘을 제안한다. 제안된 알고리즘은 여러 가지 영상타입의 그레이 레벨 co-occurrence 행렬의 특성 중에서 콘트라스트와 에너지 그리고 영상의 엔트로피로 21-dim.의 특징벡터를 추출하고, 결정나무 플랜에서 분류학습을 위하여 PPCA를 이용하여 2-dim.으로 차원을 축소한다. 포렌식 영상의 분류 테스트는 영상 타입들의 전수조합에서 수행되었다. 실험을 통하여, TP (True Positive)와 FN (False Negative)을 검출하고, 제안된 알고리즘의 성능평가에서 민감도 (Sensitivity)와 1-특이도 (1-Specificity)의 AUROC (Area Under Receiver Operating Characteristic) 커브 면적은 0.9980으로 'Excellent(A)' 등급임을 확인하였다. 산출된 최소평균 판정에러 0.0179에서 분류할 포렌식 영상타입이 모두 포함되어 분류 효율성이 높다.