• Proceedings of the Korean Society of Propulsion Engineers Conference
• /
• 2010.11a
• /
• pp.499-505
• /
• 2010

The UAV(Unmanned Aerial Vehicle) which is remotely operating with long endurance in high altitude must have a very reliable propulsion system. The precise fault diagnostic system of the turboprop engine as a propulsion system of this type UAV can promote reliability and availability. This work proposes a diagnostic method which can identify the faulted components from engine measuring parameter changes using Fuzzy Logic and quantify its faults from the identified fault pattern using Neural Network Algorithms. It is found by evaluation examples that the proposed diagnostic method can detect well not only single type faults but also multiple type faults.

• Convergence Security Journal
• /
• v.15 no.5
• /
• pp.37-45
• /
• 2015

In IT system, logs are an indicator of the previous key events. Therefore, when a security problem occurs in the system, logs are used to find evidence and solution to the problem. So, it is important to ensure the integrity of the stored logs. Existing schemes have been proposed to detect tampering of the stored logs after the key has been exp osed. Existing schemes are designed separately in terms of log transmission and storage. We propose a new log sys tem for integrating log transmission with storage. In addition, we prove the security requirements of the proposed sc heme and computational efficiency with existing schemes.

• Proceedings of the KIEE Conference
• /
• summer
• /
• pp.1291-1293
• /
• 2004

The control method of programmable dynamometer for overall test of machine is to load the reference torque which is computed from torque transducer into motor under test. But the torque information detected from torque transducer have a lot of noise when the load torque of meter is a small quantity or changing. Thus, torque transducer must have a low pass filter to detect a definite torque information. But The torque delay generated by filter with torque transducer occur a torque trouble for moter torque of programmable dynamometer. Therefore, this kind of system could not perform dynamic and nonlinear load. In this paper, the control method using the load torque observer without a measure for torque transducer is proposed. The proposed system improved the problem of the torque measuring delay with torque transducer, and the load torque is estimated by the minimal order state observer based on the torque component of the vector control induction meter. Therefore, the torque controller is not affected by a load torque disturbance.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.17 no.2
• /
• pp.703-711
• /
• 2016

IDS (Intrusion Detection System) is used to detect network attacks through network data analysis. The system requires a high accuracy and detection rate, and low false alarm rate. In addition, the system uses a range of techniques, such as expert system, data mining, and state transition analysis to analyze the network data. The purpose of this study was to compare the performance of two data mining methods for detecting network attacks. They are Support Vector Machine (SVM) and a neural network called Forward Additive Neural Network (FANN). The well-known KDD Cup 99 training and test data set were used to compare the performance of the two algorithms. The accuracy, detection rate, and false alarm rate were calculated. The FANN showed a slightly higher false alarm rate than the SVM, but showed a much higher accuracy and detection rate than the SVM. Considering that treating a real attack as a normal message is much riskier than treating a normal message as an attack, it is concluded that the FANN is more effective in intrusion detection than the SVM.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.38 no.3
• /
• pp.185-197
• /
• 2001

Recently, the design method by VHDL coding and synthesis has been used widely. As the integration ratio increases, the amount design by VHDL at a time also increases so many coding errors occur in a design. Thus, lots of time and effort is dissipated to detect those coding errors. This paper proposed a method to verify the coding errors in VHDL behavioral-level designs. As the methodology, we chose the method to detect the coding error by applying the generated set of verifying patterns and comparing the responses from the error-free case(gold unit) and the real design. Thus, we proposed an algorithm to generate the verifying pattern set for the coding errors. Verifying pattern generation is peformed for each code and the coding errors are classified as two kind: condition errors and assignment errors. To generate the patterns, VHDL design is first converted into the corresponding CDFG(Control & Data Flow Graph) and the necessary information is extracted by searching the paths in CDFG. Path searching method consists of forward searching and backward searching from the site where it is assumed that coding error occurred. The proposed algorithm was implemented with C-language. We have applied the proposed algorithm to several example VHDL behavioral-level designs. From the results, all the patterns for all the considered coding errors in each design could be generated and all the coding errors were detectable. For the time to generate the verifying patterns, all the considered designed took less than 1 ［sec］ of CPU time in Pentium-II 400MHz environments. Consequently, the verification method proposed in this paper is expected to reduce the time and effort to verify the VHDL behavioral-level designs very much.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.654-655
• /
• 2003

Korea has experienced 10 a Cochlodinium polykrikoides red tide outbreaks during the last 10 years (1993-2002). The monitoring activities at National Fisheries Research and Development Institute (NFRDI) in Korea have been extended to all the coastal waters after the worst of fish killing by C. polykrikoides blooms in 1995. NFRDI is looking forward to finding out the feasibility of red tide detection around Korean waters using satellite remote sensing of NOAA/AVHRR, Orbview-2/SeaWiFS, IRS-P4/OCM and Terra/MODIS on real time base. In this study, we used several alternative methods including climatological analysis, spectral and optical methods which may offer a potential detection of the major species of red tide in Korean waters. The relationship between the distribution of SST and C. polykrikoides bloom areas was studied. In climatological analysis, NOAA, SeaWiFS, OCM satellite data in 20th and 26th August 2001 were chosen using the known C. polykrikoides red tide bloom area mapped by helicopter reconnaissance and ground observation. The 26th August, 2001 SeaWiFS chlorophyll a anomaly imageries against the imageries of non-occurring red tide for August 20, 2001 showed the areas C. polykrikoides occurred. The anomalies of chlorophyll a concentration from satellite data between before and after red tide outbreaks showed the similar distribution of C. polykrikoides red tide in 26th August, 2001. The distribution of the difference in SST between daytime and nighttime also showed the possibility of red tide detection. We used corrected vegetation index (CVI) to detect floating vegetation and submerged vegetation containing algal blooms. The simple result of optical absorption from C. polykrikoides showed that if we use the optical characteristics of each red tide we will be able to get the feasibility of the red tide detection.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.7 no.4
• /
• pp.163-172
• /
• 2012

A driver assistance system is critical to improve a convenience and stability of vehicle driving. Several systems have been already commercialized such as adaptive cruise control system and forward collision warning system. Efficient vehicle detection is very important to improve such driver assistance systems. Most existing vehicle detection systems are based on a radar system, which measures distance between a host and leading (or oncoming) vehicles under various weather conditions. However, it requires high deployment cost and complexity overload when there are many vehicles. A camera based vehicle detection technique is also good alternative method because of low cost and simple implementation. In general, night time vehicle detection is more complicated than day time vehicle detection, because it is much more difficult to distinguish the vehicle's features such as outline and color under the dim environment. This paper proposes a method to detect vehicles at night time using analysis of a captured color space with reduction of reflection and other light sources in images. Four colors spaces, namely RGB, YCbCr, normalized RGB and Ruta-RGB, are compared each other and evaluated. A suboptimal threshold value is determined by Otsu algorithm and applied to extract candidates of taillights of leading vehicles. Statistical features such as mean, variance, skewness, kurtosis, and entropy are extracted from the candidate regions and used as feature vector for SVM(Support Vector Machine) classifier. According to our simulation results, the proposed statistical feature based SVM provides relatively high performances of leading vehicle detection with various distances in variable nighttime environments.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.8
• /
• pp.1693-1700
• /
• 2012

Recently, the warning system to aid drivers for safe driving is being developed. The system estimates the distance between the driver's car and the car before it and informs him of safety distance. In this paper, we designed and implemented the collision warning system which detects the car in front on the actual road situation and measures the distance between the cars in order to detect the risk situation for collision and inform the driver of the risk of collision. First of all, using the forward-looking camera, it extracts the interest area corresponding to the road and the cars from the image photographed from the road. From the interest area, it extracts the object of the car in front through the analysis on the critical value of the shadow of the car in front and then alerts the driver about the risk of collision by calculating the distance from the car in front. Based on the results of detecting driving cars and measuring the distance between cars, the collision warning system was designed and realized. According to the result of applying it in the actual road situation and testing it, it showed very high accuracy; thus, it has been verified that it can cope with safe driving.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.5
• /
• pp.1433-1449
• /
• 2023

In recent years, the rapid development of social networks has led to a rapid increase in the amount of information available on the Internet, which contains a large amount of sensitive information related to pornography, politics, and terrorism. In the aspect of sensitive image detection, the existing machine learning algorithms are confronted with problems such as large model size, long training time, and slow detection speed when auditing and supervising. In order to detect sensitive images more accurately and quickly, this paper proposes a multiclassification sensitive image detection method based on lightweight Convolutional Neural Network. On the basis of the EfficientNet model, this method combines the Ghost Module idea of the GhostNet model and adds the SE channel attention mechanism in the Ghost Module for feature extraction training. The experimental results on the sensitive image data set constructed in this paper show that the accuracy of the proposed method in sensitive information detection is 94.46% higher than that of the similar methods. Then, the model is pruned through an ablation experiment, and the activation function is replaced by Hard-Swish, which reduces the parameters of the original model by 54.67%. Under the condition of ensuring accuracy, the detection time of a single image is reduced from 8.88ms to 6.37ms. The results of the experiment demonstrate that the method put forward has successfully enhanced the precision of identifying multi-class sensitive images, significantly decreased the number of parameters in the model, and achieved higher accuracy than comparable algorithms while using a more lightweight model design.

• Composites Research
• /
• v.13 no.6
• /
• pp.47-54
• /
• 2000

One promising method for impact detection of composite structures is based on the use of piezopolymer thin fim (PVDf) sensor. In this paper, the relationship between the contact force and the signals of the attached strain gage and PVDF sensor to the composite plate subjected to low-velocity impact were derived. The relation for the open circuit and short circuit voltage of PVDF sensor was derived based on the equivalent circuit model of the piezoelectric sensor. The work was then extended to include experimental investigation into the use of short circuit voltage of PVDF sensor without using charge amplifier to detect low-velocity impact. The natural frequencies and damping ratio of the composite plate obtained from the vibration test were used to modify the analytical model and therefore the differences between measured and simulated signal of the modified analytical model in both forward and backward problem were considerably reduced. The reconstructed contact force and simulated sensor signals agreed well with the measured contact force, strain gage signal, and PVDF sensor singanl.