Earticle

인공지능 캐릭터에 대한 연구는 게임뿐만 아니라 애니메이션 그래픽스 분야에서 가장 많이 연구되고 논의되어 왔 던 이론 중의 하나이다. 게임 환경의 인공지능 캐릭터인 NPC는 감각 시스템을 이용하여 환경을 인지하고 기억 시스 템에 저장한다. 기억 체계는 감각 기억과 인지 기억으로 구성되고 인지 기억은 공간 데이터베이스 이론에 기반을 둔 공간에 존재하는 정보들 간의 위치 관계를 이용하여 정보를 저장한다. 본 논문에서는 NPC의 효율적인 인지 기억 기 법으로 공간 데이터베이스 이론과 퍼지이론을 적용하여 인지한 객체 간의 관계그래프, 즉 인지 맵을 만들어 기억한다. 즉 기억된 정보간의 인지 맵을 표현하기 위해 퍼지함수 이론을 적용한다. NPC는 이 인지 맵을 통해 객체 간의 개략 적인 거리와 방향을 자신의 인지 메모리에 저장 가능하다. 실험 결과 인지 맵을 이용한 결과 NPC의 목표 객체 탐색 시간이 기존 NPC와 비교하여 1.5배 향상되었다.

Most autonomous virtual characters use the programmed information and always repeatedly do actions with predefined patterns. In this thesis, we present a cognitive memory model for autonomous virtual characters, which manages the recognized information by visual attention. The cognitive memory of autonomous virtual character is comprised of visual memory and spatial relationship memory. Virtual characters take adaptive actions for the situation by utilizing the stored information in the memory system. We have implemented our cognitive memory system for the experiment to confirm its feasibility and availability. As a result of the experiment, we positively verify that autonomous virtual character perceives the static and dynamic object of 3D virtual world by visual attention mechanism with about 1.6 times improvement of time complexity, and also that the virtual character utilizes the information of memory and searches the target object and sets up a path plan.

Responding to changes in artificial intelligence models and the data environment is crucial for increasing data-learning accuracy and inference stability of industrial applications. A learning model that is overfitted to specific training data leads to poor learning performance and a deterioration in flexibility. Therefore, an early stopping technique is used to stop learning at an appropriate time. However, this technique does not consider the homogeneity and independence of the data collected by heterogeneous nodes in a differential network environment, thus resulting in low learning accuracy and degradation of system performance. In this study, the generalization performance of neural networks is maximized, whereas the effect of the homogeneity of datasets is minimized by achieving an accuracy of 99.7%. This corresponds to a decrease in delay time by a factor of 2.33 and improvement in performance by a factor of 2.5 compared with the conventional method.

This study proposed a novel framework for plant leaf disease identification. The proposed model consists of four steps including pre-processing, segmentation, feature extraction, and classification. At first, the unwanted noise and overfitting are removed, and also image contrast level is enhanced. Secondly, the Fuzzy C-Means (FCM) based Chameleon Swarm Algorithm (CSA) named as (FCM-CSA) is used for plant leaf diseased part segmentation. In the third stage, the feature extraction is performed using a fast GLCM feature extraction model. Finally, the Progressive Neural Architecture Search (PNAS) is used for plant leaf disease identification. The experimental investigations are carried out using MATLAB software with the Mendeley database. From this dataset, we have used Apple Cedar Apple Rust (ACAR), Cherry Powdery Mildew (CPM), Corn Common Rust (CCCR), Apple Healthy (AH), Grape Black Rot (GBR), Pepper Bell Bacterial Spot (PBBS), Potato Late Blight (PLB) and Tomato Leaf Mold (TLM) disease images. Different measures such as precision, recall, sensitivity, specificity, and accuracy results are used to validate the performance of the proposed model.

Windshield wipers play a key role in assuring the driver's safety during precipitation. The traditional wiper systems, however, requires driver's constant attention in adjusting the wiper speed and the intermittent wiper interval because the amount of precipitation on the windshield constantly varies according to time and vehicle's speed. Because the manual adjustment of the wiper distracts driver's attention, which may be a direct cause of traffic accidents, many companies have developed automatic wiper systems using some optical sensors with various levels of success. This paper presents the development of vision-based smart windshield wiper system that can automatically adjust its speed and intermittent interval according to the amount of water drops on the windshield. The system employs various image processing algorithms to detect water drops and fuzzy logic to determine the speed and the interval of the wiper.

Non-local means (NLM) algorithm is an effective and successful denoising method, but it is computationally heavy. To deal with this obstacle, we propose a novel NLM algorithm with fuzzy metric (FM-NLM) for image denoising in this paper. A new feature metric of visual features with fuzzy metric is utilized to measure the similarity between image pixels in the presence of Gaussian noise. Similarity measures of luminance and structure information are calculated using a fuzzy metric. A smooth kernel is constructed with the proposed fuzzy metric instead of the Gaussian weighted L2 norm kernel. The fuzzy metric and smooth kernel computationally simplify the NLM algorithm and avoid the filter parameters. Meanwhile, the proposed FM-NLM using visual structure preferably preserves the original undistorted image structures. The performance of the improved method is visually and quantitatively comparable with or better than that of the current state-of-the-art NLM-based denoising algorithms.

In the conventional clustering algorithms, an object could be assigned to only one group. However, this is sometimes not the case in reality, there are cases where the data do not belong to one group. As against, the fuzzy clustering takes into consideration the degree of fuzzy membership of each pixel relative to different classes. In order to overcome some shortcoming with traditional clustering methods, such as slow convergence and their sensitivity to initialization values, we have used the Harmony Search algorithm. It is based on the population metaheuristic algorithm, imitating the musical improvisation process. The major thrust of this algorithm lies in its ability to integrate the key components of population-based methods and local search-based methods in a simple optimization model. We propose in this paper a new unsupervised clustering method called the Fuzzy Harmony Search-Fourier Transform (FHS-FT). It is based on hybridization fuzzy clustering and the harmony search algorithm to increase its exploitation process and to further improve the generated solution, while the Fourier transform to increase the size of the image's data. The results show that the proposed method is able to provide viable solutions as compared to previous work.

The chest X-ray image cannot be focused in the same manner that optical lenses are and the resultant image generally tends to be slightly blurred. Therefore, the methods to improve the quality of chest X-ray image have been studied. In this paper, the inherent noises of the input images are suppressed by adding the Laplacian image to the original. First, the chest X-ray image using an Gaussian high pass filter and an optimal high frequency emphasis filter has shown improvements in the edges and contrast of flat areas. Second, using fuzzy logic_histogram equalization, each pixel of the chest X-ray image shows the normal distribution of intensities that are not overexposed. As a result, the proposed method has shown the enhanced edge and contrast of the images with the noise canceling effect.

Chest X-ray image cannot be focused in the same manner as optical lenses and the resultant image generally tends to be slightly blurred. Therefore, appropriate methods to improve the quality of chest X-ray image have been studied in this paper. As the frequency domain filters work well for slight blurring and moderate levels of additive noises, we propose an algorithm that is particularly suitable for enhancing chest image. First, the chest image using Gaussian high pass filter and the optimal high frequency emphasis filter shows improvements in the edges and contrast of the flat areas. Second, as compared to using histogram equalization where each pixel of chest image is characterized by a loss of detail and much noises, in using fuzzy logic, each pixel of chest image shows the detail preservation and little noise.

This paper proposes a fuzzy correction algorithm that can correct the distorted medical image caused by the scanning nonlinear velocity of the Digital X-ray Scanner System (DX-Scanner) using the Multichannel Ionization Chamber (MIC). In the DX-Scanner, the scanned medical image is distorted for reasons of unsuitable integration time at the nonlinear acceleration period of the AC servo motor during the inspection of patients. The proposed algorithm finds the nonlinear motor velocity modeling through fuzzy system by clustering and reconstructs the normal medical image lines by calculating the suitable moving distance with the velocity of the motor using the modeling, acceleration time and integration time. In addition, several image processing is included in the algorithm. This algorithm analyzes exact pixel lines by comparing the distance of the acceleration period with the distance of the uniform velocity period in every integration time and is able to compensate for the velocity of the acceleration period. By applying the proposed algorithm to the test pattern for checking the image resolution, the effectiveness of this algorithm is verified. The corrected image obtained from distorted image is similar to the normal and better image for a doctor's diagnosis.

The resistance Spot Welding is widely used in the field of assembling the plates. However we don't still have any satisfactory solution, which is non-destructive quality evaluation in real-time or on-line, against it. Moreover, even though the rate of welding under the condition of expulsion has been high until now, quality control of welding against expulsion hasn't still been established. In this paper, it was proposed on the quality assurance technique of resistance spot welding pieces using Neuro-Fuzzy algorithm. Four parameters from electrode separation signal in the case of non-expulsion, and dynamic resistance patterns in the case of expulsion are selected as fuzzy input parameters. The parameters consist of Fuzzy Inference System are determined through Neuro-Learning algorithm. And then, fuzzy Inference System is constructed. It was confirmed that the fuzzy inference values of strength have within ${\pm}$4% error specimen in comparison with real strength for the total strength range, and the specimen percent having within ${\pm}$1% error was 88.8%. According to KS(Korean Industrial Standard), tensile-shear strength limit for electric coated of zinc is 400kgf/mm2. Judging to the quality of welding is good or bad, according to this criterion and the results of inference, the probability of misjudgement that good quality is valuated into poor one was 0.43%, on contrary it was 2.59%. Finally, the proposed Neuro-Fuzzy Inference System can infer the tensile-shear strength of resistance spot welding pieces with high performance for all cases-non-expulsion and expulsion. And On-Line Welding Quality Inspection System will be realized sooner or later.

In this paper, the development of a new group controller for high-speed elevator is carried out utilizing the approach of an adaptive dual fuzzy logic. Some goals of control are the minimization of waiting time, mean-waiting time and long-waiting time in a high building, when a new hall call is generated, adaptive dual fuzzy controller evaluate traffic pattern and change appropriately the membership function of a fuzzy rule base. Controls for co-operation among elevators in a group control algorithm arte essential, and the most critical control function in the group controller is an effective and proper hall call assignment of elevators. The group elevator system utilizing adaptive dual fuzzy control reveals a great deal of improvement on its performance.

Crane operation for transporting heavy loads causes swinging motion at the loads due to crane's acceleration and deceleration. This sway causes the suspension ropes to leave their grooves and leads to possibility of serious damages. So, this swing of the objects is a serious problem and the goal of crane system is transporting to a goal position as soon as possible without the oscillation of the rope. Generally crane is operated by expert's knowledge. Therefore, a satisfactory control method to supress object sway during transport is indispensible. The dynamic behavior of the crane shows nonlinear characteristics. when the length of the rope is changed the crane is time varying system and the design of anti-sway controller is very difficult. In this paper, the nonlinear dynamic model for the industrial overhead crane whose girder, trolley and hoister move simultaneously is derived. and the Fuzzy logic controller based on the expert experiments during acceleration, constant velocity, deceleration and stop position period is proposed to supress the swing motion and control the position of the crane. The performance of the fuzzy controller for the nonlinear crane model is simulated on the personal computer.

Robot manipulators, which are nonlinear structures and have uncertain system parameters, have complex in dynamics when are operated in unknown environment. To compensate for estimate errors of the uncertain system parameters and to accomplish the desired trajectory tracking, nonlinear robust controllers are appropriate. However, when estimation errors or tracking errors are large, they require large input torques, which may not be satisfied due to torque limits of actuators. As a result, their stability can not be guaranteed. In this paper, a new robust control scheme is presented to solve stability problem and to achieve fast trajectory tracking in the presence of torque limits. By using fuzzy logic, new desired trajectories which can be reduced are generated based on the initial desired trajectory, and torques of the robust controller are regulated to not exceed torque limits. Numerical examples are shown to validate the proposed controller using an uncertain two degree-of-freedom underwater robot manipulator.

This thesis develops a fuzzy-PID control algorithm for control the filament winding tension. It is developed by applying classical PID control technique to a fuzzy logic controller. It is composed of a fuzzy-PI controller and a fuzzy-D controller. The fuzzy-PI controller uses error and integrated error as inputs, and the fuzzy-D controller uses derivative of error as input. The fuzzy-PI controller uses Takagi-Sugeno fuzzy inference system, and the fuzzy-D controller uses Mamdani fuzzy inference system. The fuzzy rule base for the fuzzy-PI controller is designed using 19 rules, and the fuzzy rule base for the fuzzy-D controller is designed using 5 rules. A test-bed is set-up for verifying the effectiveness of the developing control algorithm in control the filament winding tension. It is composed of a mandrel, a carriage, a force sensor, a driving roller, nip rollers, a creel, and a real-time control system. Nip rollers apply a vertical force to a filament, and the driving roller drives it. The real-time control system is developed by using MATLAB/xPC Target. First, experiments for showing the inherent problems of an open-loop control scheme in a filament winding are performed. Then, experiments for showing the robustness of the developing fuzzy-PID control algorithm are performed under various working conditions occurring in a filament winding such as mandrel rotating speed change, carriage traversing, spool radius change, and reference input change.

본 논문은 가중 퍼지소속함수 기반 신경망 (Neural Network with Weighted Fuzzy Membership functions, NEWFM)과 심박수 변이도(Heart Rate Variability, HRV)를 이용하여 우울증 진단알고리즘을 제안하고 있다. 본 알고리즘에서 사용할 NEWFM의 입력특징을 추출하기 위해서 주파수도메인 특징추출, 시간도메인 특징추출, 웨이블릿변환 특징추출, 포인케어변환 특징추출 방법을 이용하여 22개의 초기 HRV 특징들을 추출하였다. 또한 NEWFM에서 제공하는 비중복면적 분산측정법 (Non-overlap Area Distribution Measurement, NADM)에 의해 입력특징의 중요도를 평가하여 22개의 초기특징으로부터 중요도가 가장 높은 6개 최적입력특징을 선택하였다. 이 6개 특징을 이용하여 우울증을 진단한 결과는 95.8% 의 정확도를 나타내었다.

This paper presents an algorithm for depression diagnosis using the Neural Network with Weighted Fuzzy Membership functions (NEWFM) and heart rate variability (HRV). In the algorithm, 22 different features were initially extracted from the HRV signal by frequency domain, time domain, wavelet transformed, and Poincar$\acute{e}$ transformed feature extraction methods; of these 6 optimal features were selected by significance evaluation using Non-overlap Area Distribution Measurement (NADM) based on NEWFM. The proposed algorithm uses these 6 optimal features to diagnose depression with an accuracy of 95.83%.

Quality control is essential to the final product in BGA-type PCB fabrication. So, many automatic vision systems have been developed to achieve speedy, low cost and high quality inspection. A multiple threshold selection algorithm is a very important technique for machine vision based inspection. In this paper, an inspected image is modeled by using fuzzy sets and then the parameters of specified membership functions are estimated to be in maximum fuzzy entropy with the probability of the fuzzy sets, using the exhausted search method. Fuzzy c-partitions with the estimated parameters are automatically generated, and then multiple thresholds are selected as the crossover points of the fuzzy sets that form the estimated fuzzy partitions. Several experiments related to flip chip BGA images show that the proposed algorithm outperforms previous ones using both entropy and variance, and also can be successfully applied to AVI systems.

In this paper, a motion control algorithm is developed using a fuzzy control and the optimization of performance function, which makes a robot arm avoid an unexpected obstacle when the end-effector of the robot arm is moving to the goal position. During talc motion, if there exists no obstacle, the end-effector of the robot arm moves along the predefined path. But if these exists an obstacle and close to talc robot arm, the fuzzy motion controller is activated to adjust the path of the end-effector of the robot arm. Then, the robot arm takes the optimal posture far collision avoidance with the obstacle. To show the feasibility of the developed algorithm, numerical simulations are carried out with changing both the positions and sites of obstacles. It was concluded that the proposed algorithm gives a good performance for obstacle avoidance.

최근 전기기기로 인한 화재발생이 급증함에 따라 기기에 대한 안전진단의 필요성이 높아지고 있다. 본 연 구는 지능형의 Fuzzy기술을 이용한 전기기기의 안전진단에 관한 것으로 기기의 사용전류특성, 누적사용시간, 열화특성 및 Arc특성 등의 복합적인 전기안전 요인을 검출하여 진단한다. 이들 안전요인을 실시간으로 추출하기 위하여 각종 Sensor회로, DSP(Digital Signal Processor) 신호처리회로, 무선통신회로 등으로 구성된 Board를 설계하였고, 추출된 4가지 진단정보를 이용하여, 기기의 안전정도를 퍼지수치 값으로 표시하기 위하여 각 정보마다 Gaussian function을 사용한 퍼지 알고리즘을 설계하고 DSP에 실장 하였다. 지능적인 퍼지알고리즘은 4가지의 진단정보를 입력받아 퍼지엔진으로 추론하고 해당기기의 종합적인 안전 상태를 사람의 감성에 익숙한 100단계의 아날로그 퍼지 값으로 출력한다. 본 연구에서 구현된 DSP 하드웨어와 퍼지 알고리즘을 융합한 보드의 실험을 통하여, 전기기기의 운전 중 실시간 안전 상태를 복합적으로 검출하고, 사람에게 친화적인 감성적 퍼지 값으로 진단결과를 출력하는 기능을 소형의 DSP Board에서 구현할 수 있음을 입증하였다. 향후 인공지능 전용 Micom이 출시된다면 지능을 바탕으로 보다 진보된 진단 시스템을 연구할 수 있을 것으로 기대한다.

Recently, the necessity of safety diagnosis of electrical devices has been increasing as the fire caused by electric devices has increased rapidly. This study is concerned with the safety diagnosis of electric equipment using intelligent Fuzzy technology. It is used as a diagnostic input for the multiple electrical safety factors such as the use current, cumulative use time, deterioration and arc characteristics inherent to the equipment. In order to extract these information in real time, a device composed of various sensor circuits, DSP signal processing, and communication circuit is implemented. The fuzzy logic algorithm using the Gaussian function for each information is designed and compiled to be implemented on a small DSP board. The fuzzy logic receives the four diagnostic information, deduces it by the fuzzy engine, and outputs the overall safety status of the device as a 100-step analog fuzzy value familiar to human sensibility. By experiments of a device that combines hardware and fuzzy algorithm implemented in this study, it is verified that it can be implemented in a small DSP board with human-friendly fuzzy value, diagnosing real-time safety conditions during operation of electric equipment. In the future, we expect to be able to study more intelligent diagnostic systems based on artificial intelligent with AI dedicated Micom.

분포형 모형이 개발되어 지면서 이러한 유역의 공간적인 특성을 고려한 정확한 강우 자료와 조밀한 계측망의 요구는 더욱 커지고 있다. 그러나 현실적으로 조밀한 계측망에 의해 측정된 정확한 강우 자료를 얻기는 쉽지 않다. 일반적으로 강우관측소가 적정 밀도를 가지고 유역을 대표 하도록 설치되어 있으나 부족한 실정이고, 설치되어 있더라도 강우의 시․공간적 변동성을 반영하기가 쉽지 않다. 또한 여러 가지 이유로 결측이 되는 경우도 있다. 강우는 측정된 점 관측 자료를 이용해 유역의 평균 강우분포를 추정하게 된다. 따라서 결측 강우자료는 시간의 연속성 측면에서 그 보정이 반드시 필요하며 보정 후 강우자료의 공간적 분포를 산정할 수 있을 것이다. 본 연구에서는 결측 강우량의 보정을 위하여 퍼지-유전자 알고리즘을 이용하였는데 이 방법을 기존의 방법 즉, 산술평균법, 역거리법, 년정상강우량법, 거리-고도비율법과 비교하였다. 보정결과 기존의 방법은 실측의 70∼80%의 정확도를 보였으나 퍼지-유전자 알고리즘은 90%정도의 정확도를 보였다. 특히, 민감도 분석 결과를 바탕으로 수평거리와 고도차에 대한 적정 차수를 제안하였다.