• KIPS Transactions on Software and Data Engineering
• /
• v.5 no.11
• /
• pp.511-520
• /
• 2016

Automatic License Plate Detection (ALPD) is a key technology for a efficient traffic control. It is used to improve work efficiency in many applications such as toll payment systems and parking and traffic management. Until recently, the hand-crafted features made for image processing are used to detect license plates in most studies. It has the advantage in speed. but can degrade the detection rate with respect to various environmental changes. In this paper, we propose a way to utilize a Faster Region based Convolutional Neural Networks (Faster R-CNN) and a Conventional Convolutional Neural Networks (CNN), which improves the computational speed and is robust against changed environments. The module based on Faster R-CNN is used to detect license plate candidate regions from images and is followed by the module based on CNN to remove False Positives from the candidates. As a result, we achieved a detection rate of 99.94% from images captured under various environments. In addition, the average operating speed is 80ms/image. We implemented a fast and robust Real-Time License Plate Detection System.

• Convergence Security Journal
• /
• v.18 no.5_1
• /
• pp.45-51
• /
• 2018

As data utilization and importance becomes important, data-related accidents and damages are gradually increasing. Especially, insider threats are the most harmful threats. And these insider threats are difficult to detect by traditional security systems, so rule-based abnormal behavior detection method has been widely used. However, it has a lack of adapting flexibly to changes in new attacks and new environments. Therefore, in this paper, we propose an adaptive anomaly movement detection framework based on a statistical Markov model to detect insider threats in advance. This is designed to minimize false positive rate and false negative rate by adopting environment factors that directly influence the behavior, and learning data based on statistical Markov model. In the experimentation, the framework shows good performance with a high F2-score of 0.92 and suspicious behavior detection, which seen as a normal behavior usually. It is also extendable to detect various types of suspicious activities by applying multiple modeling algorithms based on statistical learning and environment factors.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.12
• /
• pp.4008-4023
• /
• 2022

Since machine learning was introduced into cross-site scripting (XSS) attack detection, many researchers have conducted related studies and achieved significant results, such as saving time and labor costs by not maintaining a rule database, which is required by traditional XSS attack detection methods. However, this topic came across some problems, such as poor generalization ability, significant false negative rate (FNR) and false positive rate (FPR). Moreover, the automatic clustering property of graph convolutional networks (GCN) has attracted the attention of researchers. In the field of natural language process (NLP), the results of graph embedding based on GCN are automatically clustered in space without any training, which means that text data can be classified just by the embedding process based on GCN. Previously, other methods required training with the help of labeled data after embedding to complete data classification. With the help of the GCN auto-clustering feature and labeled data, this research proposes an approach to detect XSS attacks (called GCNXSS) to mine the dependencies between the units that constitute an XSS payload. First, GCNXSS transforms a URL into a word homogeneous graph based on word co-occurrence relationships. Then, GCNXSS inputs the graph into the GCN model for graph embedding and gets the classification results. Experimental results show that GCNXSS achieved successful results with accuracy, precision, recall, F1-score, FNR, FPR, and predicted time scores of 99.97%, 99.75%, 99.97%, 99.86%, 0.03%, 0.03%, and 0.0461ms. Compared with existing methods, GCNXSS has a lower FNR and FPR with stronger generalization ability.

• Journal of dental hygiene science
• /
• v.20 no.4
• /
• pp.206-212
• /
• 2020

Background: A Teachable Machine is a kind of machine learning web-based tool for general persons. In this paper, the feasibility of Google's Teachable Machine (ver. 2.0) was studied in the diagnosis of the tooth-marked tongue. Methods: For machine learning of tooth-marked tongue diagnosis, a total of 1,250 tongue images were used on Kaggle's web site. Ninety percent of the images were used for the training data set, and the remaining 10% were used for the test data set. Using Google's Teachable Machine (ver. 2.0), machine learning was performed using separated images. To optimize the machine learning parameters, I measured the diagnosis accuracies according to the value of epoch, batch size, and learning rate. After hyper-parameter tuning, the ROC (receiver operating characteristic) analysis method determined the sensitivity (true positive rate, TPR) and specificity (false positive rate, FPR) of the machine learning model to diagnose the tooth-marked tongue. Results: To evaluate the usefulness of the Teachable Machine in clinical application, I used 634 tooth-marked tongue images and 491 no-marked tongue images for machine learning. When the epoch, batch size, and learning rate as hyper-parameters were 75, 0.0001, and 128, respectively, the accuracy of the tooth-marked tongue's diagnosis was best. The accuracies for the tooth-marked tongue and the no-marked tongue were 92.1% and 72.6%, respectively. And, the sensitivity (TPR) and specificity (FPR) were 0.92 and 0.28, respectively. Conclusion: These results are more accurate than Li's experimental results calculated with convolution neural network. Google's Teachable Machines show good performance by hyper-parameters tuning in the diagnosis of the tooth-marked tongue. We confirmed that the tool is useful for several clinical applications.

• Journal of Korean Academy of Oral and Maxillofacial Radiology
• /
• v.28 no.1
• /
• pp.37-46
• /
• 1998

The purpose of this study was to evaluate the diagnostic performance of the senior dental students for the proximal dental caries on intraoral radiographs and to compare it with the dental hospital residents, the reference group. It was also investigated the diagnostic performance according to the carious lesion depth. Thirty-five intraoral periapical and bitewing radiographs with 213 proximal surfaces included in this study were selected from the dental patients at Chonnam National University Hospital. The observers were 181 senior dental students from 5 dental schools and 40 dentists who were second year resident from 5 dental hospitals. They were asked to evaluate the presence or the absence of the proximal dental caries. The results were as follows: 1. The mean of the hitting rate for the overall observers was 184.51 surfaces and the diagnostic accuracy was 86.62％. 2. The diagnostic performance of the sound proximal tooth surfaces was very high, i.e., 91.5％ true negative rate and 8.5％ false positive rate. 3. The diagnostic performance of the dentist group was higher than the student group(P<0.05). 4. The proximal dental caries perceptibility increased as the lesion depth increased significantly(P<0.001) except no difference between the carious lesion depth III and IV (P>0.001).

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.7
• /
• pp.1117-1122
• /
• 2017

Pedestrian detection is a very important and valuable part of artificial intelligence and computer vision. It can be used in various areas for example automatic drive, video analysis and others. Many works have been done for the pedestrian detection. The accuracy of pedestrian detection on multiple pedestrian image has reached high level. It is not easily get more progress now. This paper proposes a new structure based on the idea of HOG and convolutional filters to do the pedestrian detection in single pedestrian image. It can be a method to increase the accuracy depend on the high accuracy in single pedestrian detection. In this paper, we use Multiblock HOG and magnitude of the pixel as the feature and use convolutional filter to do the to extract the feature. And then use NEWFM to be the classifier for training and testing. We use single pedestrian image of the INRIA data set as the data set. The result shows that the Convolutional Multiblock HOG we proposed get better performance which is 0.015 miss rate at 10-4 false positive than the other detection methods for example HOGLBP which is 0.03 miss rate and ChnFtrs which is 0.075 miss rate.

• Convergence Security Journal
• /
• v.21 no.4
• /
• pp.109-115
• /
• 2021

Currently, with the development of 5G and IoT technology, it is being used in connection with the things used in real life through a network. However, attempts to use networked computers for malicious purposes are increasing, and attacks using malicious codes that infringe the confidentiality and integrity of user information are becoming more intelligent. As a countermeasure to this, research is being conducted on a method of detecting malicious packets using a security control system and AI technology, supervised learning. The cyber security control system is being operated inefficiently in terms of manpower and cost. In addition, in the era of the COVID-19 pandemic, remote work has increased, making it difficult to respond immediately. In addition, malicious code detection using the existing AI technology, supervised learning, does not detect variant malicious code, and has an inaccurate malicious code detection rate depending on the quantity and quality of data. Therefore, in this study, by converging malicious packet detection technologies through various machine learning and deep learning models, the accuracy of malicious packet detection is increased, the false positive rate and the false positive rate are reduced, and a new type of malicious packet can be efficiently detected when intrusion. We propose a malicious packet detection technology.

• The Korean Journal of Nuclear Medicine
• /
• v.37 no.6
• /
• pp.374-381
• /
• 2003

Purpose: Determining an appropriate thresholding is crucial for PDG PET analysis since strong control of Type I error could fail to find pathological differences between eariy Alzheimer' disease (AD) patients and healthy normal controls. We compared the SPM results on FDG PET imaging of early AD using uncorrected p-value, random-field based corrected p-value and false discovery rate (FDR) control. Materials and Methods: Twenty-eight patients ($66{\pm}7$ years old) with early AD and 18 age-matched normal controls ($68{\pm}6$ years old) underwent FDG brain PET. To identify brain regions with hypo-metabolism in group or individual patient compared to normal controls, group images or each patient's image was compared with normal controls usingthe same fixed p-value of 0.001 on uncorrected thresholding, random-field based corrected thresholding and FDR control. Results: The number of hypo-metabolic voxels was smallest in corrected p-value method, largest in uncorrected p-value method and intermediate in FDG thresholding in group analysis. Three types of result pattern were found. The first was that corrected p-value did not yield any voxel positive but FDR gave a few significantly hypometabolic voxels (8/28, 29%). The second was that both corrected p-value and FDR did not yield any positive region but numerous positive voxels were found with the threshold of uncorrected p-values (6/28, 21%). The last was that FDR was detected as many positive voxels as uncorrected p-value method (14/28, 50%). Conclusions FDR control could identify hypo-metaboiic areas in group or individual patients with early AD. We recommend FDR control instead of uncorrected or random-field corrected thresholding method to find the areas showing hypometabolism especially in small group or individual analysis of FDG PET.

• Korean Journal of Bronchoesophagology
• /
• v.15 no.2
• /
• pp.30-35
• /
• 2009

Lung cancer could be developed through a series of morphological changes from dysplasia to carcinoma in situ and then invasive cancer. However, precancerous lesions are generally a few cell layers thick and are detected only by chance. Autofluorescence bronchoscopy(AFB) is one of the newly developed diagnostic tools to detect the pre-cancerous lesions m the bronchial tissue. Several studies have shown that AFB improved the rate of detection of cancer and dysplastic lesions of the airway, especially those in intraepithelial stage. However, there were high rates of false positive with AFB, and it is also important to develop non-biopsy methods because of lack of accurate information of variable course of preneoplastic lesions regarding progression. So, many other technologies were developed, such as narrow band imaging(NBI), endobronchoscopic ultrasound, optical coherence tomography, and confocal fluorescence microendoscopy. Among the new machines, NBI is a new optical technology that can clearly visualize the microvascular structure m the mucosal layer. NBI seems to increase specificity without compromising sensitivity. In the future such techniques would make it possible to precisely study in detail the natural history of the premalignant epithelium.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.108-113
• /
• 2009

Significant attention has recently been drawn to digital home photo albums that use face detection technology. The tendency can be found in home photo albums that people prefer to allocate concerned objects in the center of the image rather than the boundary when they take a picture. To improve detection performance and speed that are important factors of face detection task, this paper proposes a face detection method that takes spatial context information into consideration. Experiments were performed to verify the usefulness of the proposed method and results indicate that the proposed face detection method can efficiently reduce the false positive rate as well as the runtime of face detection.