• Journal of the Korean Institute of Intelligent Systems
• /
• v.15 no.2
• /
• pp.163-172
• /
• 2005

When tracking a human face in the moving pictures with complex background under irregular lighting conditions, the detected face can be larger including background or smaller including only a part of the face. Even background can be detected as a face area. To solve these problems, this paper proposes a new face tracking method using a block difference image and a Kalman estimator. The block difference image allows us to detect even a small motion of a human and the face area is selected using the skin color inside the detected motion area. If the pixels with skin color inside the detected motion area, the boundary of the area is represented by a code sequence using the 8-neighbor window and the head area is detected analysing this code. The pixels in the head area is segmented by colors and the region most similar with the skin color is considered as a face area. The detected face area is represented by a rectangle including the area and its four vertices are used as the states of the Kalman estimator to trace the motion of the face area. It is proved by the experiments that the proposed method increases the accuracy of face detection and reduces the fare detection time significantly.

• Journal of the Institute of Electronics Engineers of Korea SD
• /
• v.41 no.10
• /
• pp.95-104
• /
• 2004

This paper propose a motion estimator architecture to reduce the power consumption of the most-power-consuming motion estimation method when designing multimedia SoC with deep submicron technologies below 0.13${\mu}{\textrm}{m}$. The proposed architecture considers both dynamic and static power consumption so that it is suitable for large leakage process technologies, while conventional architectures consider only dynamic power consumption. Consequently, it is suitable for mobile information terminals such as mobile videophone where efficient power management is essential. It exploits full search method for simple hardware implementation. It also exploits early break-off method to reduce dynamic power consumption. To reduce static power consumption, megablock shutdown method considering power line noise is also employed. To evaluate the proposed architecture when applied multimedia SoC, system-level control flow and low-power control algorithm are developed and the power consumption was calculated based on thor From the simulation results, power consumption was reduced to about 60%. Considering the line width reduction and increased leakage current due to heat dissipation in chip core, the proposed architecture shows steady power reduction while it goes worse in conventional architectures.

• 제어로봇시스템학회:학술대회논문집
• /
• 1998.10a
• /
• pp.301-307
• /
• 1998

The goal of this paper is to describe an advanced method of the fault diagnois using Control Theory with reference to a crack detection, a new way to localize the crack position under infulence of the plant disturbance and white measurement noise on a rotating shaft. As a first step, the shaft is physically modelled with a finite element method as usual and the dynamic mathematical model is derived from it using the Hamilton - principle and in this way the system is modelled by various subsystems. The equations of motion with crack is established by adaption of the local stiffness change through breathing and gaping from the crack to the equation of motion with un-damaged shaft. This is supposed to be regarded as reference for the given system. Based on the fictitious model of the time behaviour induced from vibration phenomena measured at the bearings, a nonlinear State Observer is designed in order to detect the crack on the shaft. This is elementary NL- observer(EOB). Using the elementary observer, an Estimator(Observer) Bank is established and arranged at the certain position on the shaft. In case a crack is found and its position is known, the procedure for the estimation of the depth is going to begin.

• 제어로봇시스템학회:학술대회논문집
• /
• 1997.10a
• /
• pp.524-527
• /
• 1997

Conventional target tracking algorithms based on the linear estimation techniques perform quite efficiently when the target motion does not involve maneuvers. Target maneuvers involving short term accelerations, however, cause a bias in the measurement sequence. Accurate compensation for the bias requires processing more samples of which adds to the computational complexity. The primary motivation for employing a neural network for this task comes from the efficiency with which more features can be as inputs for bias compensation. A system architecture that efficiently integrates the fusion capabilities of a trained multilayer neural net with the tracking performance of a Kalman filter is described. The parallel processing capability of a properly trained neural network can permit fast processing of features to yield correct acceleration estimates and hence can take the burden off the primary Kalman filter which still provides the target position and velocity estimates.

• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.10
• /
• pp.747-754
• /
• 2003

An enhancement of the probabilistic data association filter is presented for tracking a single maneuvering target in clutter environment. The use of the variable dimensional structure leads the probabilistic data association filter to adjust to real motion of a target. The detection of the maneuver for the model switching is performed by the acceleration estimates taken from a bias estimator of the two stage Kalman filter. The proposed algorithm needs low computational power since it is implemented with a single filtering procedure. A simple Monte Carlo simulation was performed to compare the performance of the proposed algorithm and the IMMPDA filter.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.2033-2036
• /
• 2004

A novel approach to estimate the real time moving trajectory of an object is proposed in this paper. The object position is obtained from the image data of a CCD camera, while a state estimator predicts the linear and angular velocities of the moving object. To overcome the uncertainties and noises residing in the input data, a Kalman filter and neural networks are utilized. Since the Kalman filter needs to approximate a non-linear system into a linear model to estimate the states, there always exist errors as well as uncertainties again. To resolve this problem, the neural networks are adopted in this approach, which have high adaptability with the memory of the input-output relationship. Kohonen Network(Self-Organized Map) is selected to learn the motion trajectory since it is spatially oriented. The superiority of the proposed algorithm is demonstrated through the real experiments.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.1375-1379
• /
• 2004

Unmanned Helicopter has several abilities such as vertical Take off, hovering, low speed flight at low altitude. Such vehicles are becoming popular in actual applications such as search and rescue, aerial reconnaissance and surveillance. These vehicles also used under risky environments without threatening the life of a pilot. Since a small aerial vehicle is very sensitive to environmental conditions, it is generally known that the flight control is very difficult problems. In this paper, a flight control system was designed for an unmanned helicopter. This paper was concentrated on describing the mechanical design, electronic equipments and their interconnections for acquiring autonomous flight. The design methodologies and performance of the helicopter were illustrated and verified with a linearized equation of motion. The LQG based estimator and controller was designed and tested for this unmanned helicopter.

• Proceedings of the IEEK Conference
• /
• 2004.08c
• /
• pp.616-621
• /
• 2004

This paper investigated the speed sensorless indirect vector control of a two-phase induction motor to implement adjustable-speed drive for low-power applications. The sliding mode observer estimates rotor speed. The convergence of the nonlinear time-varying observer along with the asymptotic stability of the controller was analyzed. To define the control action which maintains the motion on the sliding manifold, an 'equivalent control' concept was used. It was simulated and implemented on a sensorless indirect vector drive for 150W two-phase induction motor. The simulation and experimental results demonstrated effectiveness of the estimation method.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.12 no.2
• /
• pp.693-709
• /
• 2018

In this paper, we propose a compact Convolutional Neural Network (CNN)-based tracker in conjunction with a particle filter architecture, in which the CNN model operates as an accurate candidates estimator, while the particle filter predicts the target motion dynamics, lowering the overall number of calculations and refines the resulting target bounding box. Experiments were conducted on the Online Object Tracking Benchmark (OTB) [34] dataset and comparison analysis in respect to other state-of-art has been performed based on accuracy and precision, indicating that the proposed algorithm outperforms all state-of-the-art trackers included in the OTB dataset, specifically, TLD [16], MIL [1], SCM [36] and ASLA [15]. Also, a comprehensive speed performance analysis showed average frames per second (FPS) among the top-10 trackers from the OTB dataset [34].

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2010.07a
• /
• pp.104-107
• /
• 2010

본 논문은 H.264/AVC 디코더의 하드웨어 구현 시 가장 많은 시간을 소비하는 부분이 움직임 추정기를 하드웨어로 구현하였다. 움직임 추정을 함에 있어서 외부메모리 Access 량을 줄이고, SAD연산을 수행할 때 Clock의 손실 없이 계산을 하는 움직임 예측기를 제안한다. 제안한 구조는 재탐색 구간에서 이전 탐색 범위와 공통부분을 이루는 부분을 레지스터에 따로 저장해 두었다가, 재탐색시에 이전 Data를 사용하는 방법을 이용하였다. 움직임 추정을 수행할 때의 SAD (Sum of absolute differences)연산 부분과 Adder-tree를 묶은 PU Array와 SAD 누적기, 선택기를 Pipelining을 통하여 Clock의 손실 없이 연속적으로 계산하는 움직임 예측기를 설계하였다. 구현한 하드웨어는 최대 446.43MHz의 주파수에서 동작할 수 있었고, 탐색영역 64${\times}$64, 참조 프레임 3, 그리고 영상크기 1920${\times}$1080 기준으로 구현한 결과 50 프레임을 처리할 수 있는 성능을 보였다.