• Journal of the Korean Institute of Intelligent Systems
• /
• v.17 no.4
• /
• pp.511-518
• /
• 2007

This paper presents an efficient face recognition method by using both centroid shift and mutual information estimation of images. The centroid shift is to move an image to center coordinate calculated by first moment, which is applied to improve the recognition performance by excluding the needless backgrounds in face image. The mutual information which is a measurements of correlations, is applied to efficiently measure the similarity between images. Adaptive partition mutual information(AP-MI) estimation is especially applied to find an accurate dependence information by equally partitioning the samples of input image for calculating the probability density function(PDF). The proposed method has been applied to the problem for recognizing the 48 face images(12 persons * 4 scenes) of 64*64 pixels. The experimental results show that the proposed method has a superior recognition performances(speed, rate) than a conventional method without centroid shift. The proposed method has also robust performance to changes of facial expression, position, and angle, etc. respectively.

• The Journal of The Korea Institute of Intelligent Transport Systems
• /
• v.21 no.3
• /
• pp.83-95
• /
• 2022

This paper proposes a system that performs object detection and distance estimation for application to autonomous vehicles. Object detection is performed by a network that adjusts the split grid to the input image ratio using the characteristics of the recently actively used deep learning model YOLOv4, and is trained to a custom dataset. The distance to the detected object is estimated using a bounding box and homography. As a result of the experiment, the proposed method improved in overall detection performance and processing speed close to real-time. Compared to the existing YOLOv4, the total mAP of the proposed method increased by 4.03%. The accuracy of object recognition such as pedestrians, vehicles, construction sites, and PE drums, which frequently occur when driving to the city center, has been improved. The processing speed is approximately 55 FPS. The average of the distance estimation error was 5.25m in the X coordinate and 0.97m in the Y coordinate.

• ETRI Journal
• /
• v.40 no.1
• /
• pp.101-110
• /
• 2018

In wireless positioning systems using range measurements non-line-of-sight (NLOS) links cause estimation errors. Several studies have attempted to improve the positioning performance by mitigating these NLOS errors. These studies, however, have focused on the performance of a dataset consisting of three or more links. Therefore, measurement errors induced by links are averaged, and a reliable link is not fully utilized in the dataset. This paper proposes a Link Reliability based on Range Measurement (LRRM) scheme, which specifies the relative reliability of each link using residuals. The link reliability becomes the input to a Link Residual Weighting (LRW) scheme, which is also proposed as a weighted positioning scheme. Moreover, LRRM and LRW constitute new turbo positioning, where the estimation errors are reduced considerably by iterative updates.

• Journal of IKEEE
• /
• v.24 no.2
• /
• pp.497-501
• /
• 2020

Dehaze is essential for autonomous driving and intelligent CCTV to operate normally even in foggy weather. The method of airlight estimation is particularly important in dehaze technology. In this paper, we propose a patch-based airlight estimation algorithm and hardware structure that can reduce the amount of unnecessary computation and effectively estimate the airlight in various input images. Proposed algorithm is compared with the popular quad-tree method, and the hardware design is implemented by using XILINX's xc7z045-ffg900 target board as a structure that can satisfy to international standard 4K video in real time.

• IEIE Transactions on Smart Processing and Computing
• /
• v.5 no.6
• /
• pp.383-389
• /
• 2016

This paper presents a novel automatic white balance (AWB) algorithm for consumer imaging devices. While existing AWB methods require reference white patches to correct color, the proposed method performs the AWB function using only an input image in two steps: i) white point detection, and ii) color constancy gain computation. Based on the dark channel prior assumption, a white point or region can be accurately extracted, because the intensity of a sufficiently bright achromatic region is higher than that of other regions in all color channels. In order to finally correct the color, the proposed method computes color constancy gain values based on the Y component in the XYZ color space. Experimental results show that the proposed method gives better color-corrected images than recent existing methods. Moreover, the proposed method is suitable for real-time implementation, since it does not need a frame memory for iterative optimization. As a result, it can be applied to various consumer imaging devices, including mobile phone cameras, compact digital cameras, and computational cameras with coded color.

• 제어로봇시스템학회:학술대회논문집
• /
• 2000.10a
• /
• pp.506-506
• /
• 2000

To improve control performance of a non-linear system, many other researches have used the sliding mode control algorithm. The sliding mode controller is known to be robust against nonlinear and unmodeled dynamic terms. However. this algorithm raises the inherent chattering caused by excessive switching inputs around the sliding surface. Therefore, in order to solve the chattering problem and improve control performance, this study has developed the sliding mode controller with a perturbation estimator using the observer-based fuzzy adaptive network generates the control input for compensating unmodeled dynamics terms and disturbance. And, the weighting parameters of the fuzzy adaptive network are updated on-line by adaptive law in order to force the estimation errors to converge to zero. Therefore, the combination of sliding mode control and fuzzy adaptive network gives rise to the robust and intelligent routine. For evaluating control performance of the proposed approach. tracking control simulation is carried out for the hydraulic motion simulator which is a 6-degree of freedom parallel manipulator.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.21 no.1
• /
• pp.80-85
• /
• 2011

This paper presents the design of the power-assisted controller for the in-wheel type smart wheelchair by using torque estimation that is predicted by relationship between input voltage and output wheel angular velocity. Nowadays, interest of the moving assistant aids is increased according to the increase in population of the elderly and the handicapped person. However some of the moving assistant aids have problems. For example, manual wheelchair has difficulty moving at the slope, because users lack the muscular strength of their arm. In electric wheelchair case, users should be weak by being decreased muscles of upper body. To overcome these problems, power-assisted electric wheelchair are proposed. Most of the power-assisted electric wheelchair have the special rims that can measure the user's power. In here, the rims have to be designed to install the sensors to measure user's power. In this paper, we don't design the rim to measure the man power. To predict the man power, we propose a control algorithm of the in-wheeled electric wheelchair by using torque estimation from the wheel. First, we measure the wheel velocity and voltage at the in-wheel electric wheelchair. And then we extract driving will forces by using proposed mathematical model. Also they are applied at the controller as the control input, we verify to be able to control in-wheel type smart wheelchair by using simulation.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2001.12a
• /
• pp.1-4
• /
• 2001

The classification performance of Artificial Neural Network (ANN) and RBF-NN was compared for Landsat TM image. The RBF-NN was validated for three unique landuse types (e.g. Mixed landuse area, Cultivated area, Urban area), different input band combinations and classification class. The bootstrap resampling technique was employed to estimate the confidence intervals and distribution for unit load, The pollutant generation was varied significantly according to the classification accuracy and percentile unit load applied. Especially in urban area, where mixed landuse is dominant, the difference of estimated pollutant load is largely varied.

• Transactions on Control, Automation and Systems Engineering
• /
• v.3 no.2
• /
• pp.117-123
• /
• 2001

To improve control performance of a non-linear system, many other reserches have used the sliding model control algorithm. The sliding mode controller is known to be robust against nonlinear and unmodeled dynamic terms. However, this algorithm raises the inherent chattering caused by excessive switching inputs around the sliding surface. Therefore, in order to solve the chattering problem and improve control performance, this study has developed the sliding mode controller with a perturbation estimator using the observer-based fuzzy adaptive network. The perturbation estimator based on the fuzzy adaptive network generates the control input of compensating unmodeled dynamics terms and disturbance. And the weighting parameters of the fuzzy adaptive network are updated on-line by adaptive law in order to force the estimation errors converge to zero. Therefore, the combination of sliding mode control and fuzzy adaptive network gives rise to the robust and intelligent routine. For evaluation control performance of the proposed approach, tracking control simulation is carried is carried out for the hydraulic motion simulator which is a 6-degree of freedom parallel manipulator.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.16 no.1
• /
• pp.27-35
• /
• 2016

Black-box classifiers, such as artificial neural network and support vector machine, are a popular classifier because of its remarkable performance. They are applied in various fields such as inductive inferences, classifications, or regressions. However, by its characteristics, they cannot provide appropriate explanations how the classification results are derived. Therefore, there are plenty of actively discussed researches about interpreting trained black-box classifiers. In this paper, we propose a method to make a fuzzy logic-based classifier using extracted rules from the artificial neural network and support vector machine in order to interpret internal structures. As an object of classification, an anomalous propagation echo is selected which occurs frequently in radar data and becomes the problem in a precipitation estimation process. After applying a clustering method, learning dataset is generated from clusters. Using the learning dataset, artificial neural network and support vector machine are implemented. After that, decision trees for each classifier are generated. And they are used to implement simplified fuzzy logic-based classifiers by rule extraction and input selection. Finally, we can verify and compare performances. With actual occurrence cased of the anomalous propagation echo, we can determine the inner structures of the black-box classifiers.