• 한국빅데이터학회지
• /
• 제9권1호
• /
• pp.61-73
• /
• 2024

클래스 불균형은 딥러닝 작업에서 중요한 문제 중 하나이며, 이는 특히 데이터가 제한적인 분야에서 두드러진다. 본 연구에서는 토마토 잎의 병해를 효과적으로 분류하기 위해 최소한의 라벨 데이터만을 활용하는 새로운 접근법을 제안한다. 이를 위해 YOLOv8 모델을 사용한 재귀적 학습 방식을 도입하였다. 학습 데이터에 대한 이미지 탐지 예측 결과를 다시 학습 데이터로 활용함으로써 라벨 데이터의 개수를 점진적으로 증가시켰다. 이 방식은 기존의 데이터 증강 및 업-다운 샘플링 기법과는 달리 실제 데이터의 활용도를 극대화하여 클래스 불균형 문제를 보다 근본적으로 해결하려 한다. 이를 통해 확보된 라벨 데이터를 바탕으로, 토마토잎을 추출하고 EfficientNet 모델을 이용해 병해를 분류했다. 이 과정을 통해 98.92%라는 높은 정확도를 달성하였다. 특히, 가장 적은 데이터를 가진 클래스인 잎마름역병 병해에서 기존 대비 12.9% 향상된 결과를 확인할 수 있었다. 이 연구는 데이터 불균형 문제를 해결하는 동시에, 높은 정확도로 병해를 분류할 수 있는 방법론을 제시함으로써 다른 작물에서도 적용될 수 있을 것으로 기대된다.

• 대한전기학회논문지
• /
• 제39권7호
• /
• pp.720-728
• /
• 1990

A learning control method is proposed in this paper, using a knowledge base which contains control rules, data, and patterns of the past experience of a plant. The knowledge for plant control is retrieved from measurement data during operation and continually modified after control performance evaluation. A control method is proposed using tinually modified after control performance evaluation. A control method is proposed using fuzzy model of the plant and a recursive statistic decision method of fuzzy subset for control rule generation. Also, the resulting knowledge-based control algorithm has been applied to aprocess and its performance improvement and proper generation of appropriate control rules have been verified.

• Journal of applied mathematics & informatics
• /
• 제7권2호
• /
• pp.629-640
• /
• 2000

this paper present an efficient line-drawing algorithm that reduces the amount of space required, Because of its efficiency , this line-drawing algorithm is faster than the Bresenham algorithm or the recursive bisection method. this efficiency was achieved through a new data structure; namely , the modified segment tree (MST). Using the modified segment tree and the distribution rule suggested in this paper, we dra lines without generating the recursive calls used in [3] and without creating the binary operation used in [4]. we also show that line accuracy improves in proportion to the display resolution . In practice, we can significantly improve the algorithm's performance with respect to time and space, This improvement offer an increase in speed, specially with lines at or near horizontal, diagonal. or vertical ; that is, this algorithm requires the time complexity of (n) and the space complexity O(2k+1), where n is the number of pixels and k is a level of the modified segment tree.

• Structural Engineering and Mechanics
• /
• 제11권2호
• /
• pp.133-149
• /
• 2001

A high-fidelity model of a tracked vehicle traversing a flexible ground terrain with a varying profile is presented here. In this work, we employed a recursive formulation to model the track subsystem. This method yields a minimal set of coordinates and hence, computationally more efficient than conventional approaches. Also, in the vehicle subsystem, the undercarriage frame is assumed to be connected to the chassis by a revolute joint and a spring-damper unit. This increase in system mobility makes the model more realistic. To capture the vehicle-ground interaction, a Winkler-type foundation with springs-dampers is used. Simulation runs of the integrated tracked vehicle system for vibrations for four varying ground profiles are provided.

• 한국자동차공학회논문집
• /
• 제19권1호
• /
• pp.51-57
• /
• 2011

Information of the lateral velocity and the sideslip angle in a vehicle is very useful in many active vehicle safety applications such as yaw stability control and rollover prevention. Because cost-effective sensors to measure the lateral velocity and the sideslip angle are not available, reliable algorithms to estimation them are necessary. In this paper, a sliding mode observer is designed to estimate the lateral velocity. The side slip angle is estimated using the recursive least square with the disturbance observer and the pseudo integral. The estimated parameters from the combined estimation method are updated recursively to minimize the discrepancy between the model and the physical plant, and any possible effects caused by disturbances. The performance of the proposed monitoring system is evaluated through simulations and experiments.

• 한국기계가공학회지
• /
• 제1권1호
• /
• pp.123-132
• /
• 2002

In this study, the static and dynamic characteristics of endmilling process was modelled and the analytic realization of chatter mechanism was discussed. In this regard, We have discussed on the comparative assessment of recursive time series modeling algorithms that can represent the machining process and detect the abnormal machining behaviors in precision endmilling operation. In this study, simulation and experimental work were performed to show the malfunctional behaviors. For this purpose, new recursive least square method (RLSM) were adopted for the on-line system identification and monitoring of a machining process, we can apply these new algorithms in real process for detection of abnormal chatter. Also, The stability lobe of chatter was analysed by varying parameter of cutting dynamices in regenerative chatter mechanics.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2011년도 제42회 하계학술대회
• /
• pp.1738-1739
• /
• 2011

In this paper, the range difference based the moving target tracking problem using multiple UAVs is solved within the new framework of linear robust state estimation. To do this, the relative kinematics is modeled as an uncertain linear system containing stochastic parametric uncertainties in its measurement matrix. Applying the non-conservative robust Kalman filter for the uncertain system, a quasi-optimal linear target tracking filter is designed. For its recursive linear filter structure, the proposed method can ensure the fast convergence and reliable target tracking performance. Moreover, it is suitable for real-time applications using multiple UAVs.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 2011년도 제42회 하계학술대회
• /
• pp.1788-1789
• /
• 2011

In order to develop a one-axis gimbaled acoustic source tracker for mobile robots, a pseudo-linear direction of arrival(DOA) estimator is proposed using a linear ultrasonic sensor array. Under the assumption that the sensor measurement errors are negligible, a linear measurement model is derived using the linear prediction relation of the received sinusoidal acoustic signals. Applying the Kalman filtering technique for this model, the linear recursive DOA estimator is designed. For its linear recursive filter structure, it is preferable for real-time implementation on a commercial DSP. Through the experiments, the effectiveness of the suggested method is demonstrated.

• 전기학회논문지
• /
• 제67권11호
• /
• pp.1536-1541
• /
• 2018

For effective electricity consumption in urban railway station such as peak load shaving, it is important to know each electrical load pattern by various usage. The total electricity consumption in the urban railway substation is already measured in Korea, but the electricity consumption for each usage is not measured. The author proposed the deep learning method to estimate the electrical load pattern for each usage in the urban railway substation with public data such as weather data. GRU (gated recurrent unit), a variation on the LSTM (long short-term memory), was used, which aims to solve the vanishing gradient problem of standard a RNN (recursive neural networks). The optimal model was found and the estimation results with that were assessed.

• Journal of the Korean Society for Industrial and Applied Mathematics
• /
• 제8권2호
• /
• pp.75-86
• /
• 2004

In this study, a recursive algorithm for generating the equations of motion of a chain of rigid rods is presented. The methods rests upon the idea of replacing the rigid body by a dynamically equivalent constrained system of particles. The concepts of linear and angular momentums are used to generate the rigid body equations of motion without either introducing any rotational coordinates or the corresponding transformation matrices. For open-chain, the equations of motion are generated recursively along the serial chains. For closed-chain, the system is transformed to open-chain by cutting suitable kinematic joints with the addition of cut-joints kinematic constraints. An example of a closed-chain of rigid rods is chosen to demonstrate the generality and simplicity of the proposed method.