• Journal of Korea Multimedia Society
• /
• v.9 no.8
• /
• pp.1045-1053
• /
• 2006

We introduce a new method of painterly rendering. Instead of using the gradient direction of the source image to generate a brush stroke, we extract regions that can be drawn in one stroke using MHN filtering followed by identification of connected components, and make a brush stroke from each, based on an approximation to the medial axis. This method results in realistic-looking brush strokes of varying width that have an irregular directions where necessary.

• Korean Journal of Remote Sensing
• /
• v.27 no.2
• /
• pp.151-161
• /
• 2011

PGA(Phase Gradient Autofocus) is a representative autofocus technique to improve the SAR(Synthetic Aperture Radar) image quality. PGA can estimate high order phase errors and have good robustness in noisy environments. However, PGA is not suitable to apply to the stripmap mode data directly because it is based on the spotlight mode operation. In this paper, the PGA implementation technique for stripmap mode data and the method of ROI(Region of Interest) selection that affects severely on PGA performance have been proposed. The proposed technique was verified by the point target simulation first, and was applied to the real SAR signal data acquired by the flight test. Finally, the significant improvements in focusing quality were shown in the processed SAR images using the proposed method.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2000.11a
• /
• pp.687-690
• /
• 2000

Lately, many studies have been progressed for the protection human's lives and property as holding in check accidents happened by human's carelessness or mistakes. One part of these is the development of an autonomouse vehicle. General control method of vision-based autonomous vehicle system is to determine the navigation direction by analyzing lane images from a camera, and to navigate using proper control algorithm. In this paper, characteristic points are abstracted from lane images using lane recognition algorithm with sobel operator. And then the vehicle is controlled using two proposed auto-steering algorithms. Two steering control algorithms are introduced in this paper. First method is to use the geometric relation of a camera. After transforming from an image coordinate to a vehicle coordinate, a steering angle is calculated using Ackermann angle. Second one is using a neural network algorithm. It doesn't need to use the geometric relation of a camera and is easy to apply a steering algorithm. In addition, It is a nearest algorithm for the driving style of human driver. Proposed controller is a multilayer neural network using Levenberg-Marquardt backpropagation learning algorithm which was estimated much better than other methods, i.e. Conjugate Gradient or Gradient Decent ones.

• Journal of the Korean Institute of Telematics and Electronics
• /
• v.26 no.1
• /
• pp.78-87
• /
• 1989

In this paper, a spatio-temporal gradient (STG) method with generalized least square estimation (GLSE) is proposed for the detection of an object motion in an image sequence corrupted by white Gaussian noise. The proposed method is applied to an automatic target tracker using a high speed 16-bit microprocessor in order to track one moving target in real time. Experimental results show that the proposed method has much better performance over the conventional one with least square estimation (LSE).

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.64 no.7
• /
• pp.1064-1073
• /
• 2015

In this study, we introduce the pedestrian detection system by using the feature of HOG-PCA and RBFNNs pattern classifier. HOG(Histogram of Oriented Gradient) feature is extracted from input image to identify and recognize a object. And a dimension is reduced for improving performance as well as processing speed by using PCA which is a typical dimensional reduction algorithm. So, the feature of HOG-PCA through the dimensional reduction by using PCA leads to the improvement of the detection rate. FCM clustering algorithm is used instead of gaussian function to apply the characteristic of input data as well and connection weight is used by polynomial expression such as constant, linear, quadratic and modified quadratic. Finally, INRIA person database known as one of the benchmark dataset used for pedestrian detection is applied for the performance evaluation of the proposed classifier. The experimental result of the proposed classifier are compared with those studied by Dalal.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.32 no.10C
• /
• pp.1000-1009
• /
• 2007

This paper presents a content adaptive interpolation (CAI) for intra deinterlacing. The CAI consists of three steps: pre-processing, content classification, and adaptive interpolation. There are also three main interpolation methods in our proposed CAI, i.e. modified edge-based line averaging (M-ELA), gradient directed interpolation (GDI), and window matching method (WMM). Each proposed method shows different performances according to spatial local features. Therefore, we analyze the local region feature using the gradient detection and classify each missing pixel into four categories. And then, based on the classification result, a different do-interlacing algorithm is activated in order to obtain the best performance. Experimental results demonstrate that the CAI method performs better than previous techniques.

• International Journal of Automotive Technology
• /
• v.3 no.2
• /
• pp.71-78
• /
• 2002

A new control method far vision-based autonomous vehicle is proposed to determine navigation direction by analyzing lane information from a camera and to navigate a vehicle. In this paper, characteristic featured data points are extracted from lane images using a lane recognition algorithm. Then the vehicle is controlled using new Levenberg-Marquardt neural network algorithm. To verify the usefulness of the algorithm, another algorithm, which utilizes the geometric relation of a camera and vehicle, is introduced. The second one involves transformation from an image coordinate to a vehicle coordinate, then steering is determined from Ackermann angle. The steering scheme using Ackermann angle is heavily depends on the correct geometric data of a vehicle and a camera. Meanwhile, the proposed neural network algorithm does not need geometric relations and it depends on the driving style of human driver. The proposed method is superior than other referenced neural network algorithms such as conjugate gradient method or gradient decent one in autonomous lateral control .

• Korean Journal of Remote Sensing
• /
• v.19 no.5
• /
• pp.393-402
• /
• 2003

In times of the civil uses of commercialized high-resolution satellite imagery, applications of remote sensing have been widely extended to the new fields or the problem solving beyond traditional application domains. Transportation application of this sensor data, related to the automatic or semiautomatic road extraction, is regarded as one of the important issues in uses of remote sensing imagery. Related to these trends, this study focuses on automatic road extraction using Gradient Direction Profile Algorithm (GDPA) scheme, with IKONOS panchromatic imagery having 1 meter resolution. For this, the GDPA scheme and its main modules were reviewed with processing steps and implemented as a prototype software. Using the extracted bi-level image and ground truth coming from actual GIS layer, overall accuracy evaluation and ranking error-assessment were performed. As the processed results, road information can be automatically extracted; by the way, it is pointed out that some user-defined variables should be carefully determined in using high-resolution satellite imagery in the dense or low contrast areas. While, the GDPA method needs additional processing, because direct results using this method do not produce high overall accuracy or ranking value. The main advantage of the GDPA scheme on road features extraction can be noted as its performance and further applicability. This experiment study can be extended into practical application fields related to remote sensing.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.29 no.1
• /
• pp.117-125
• /
• 2019

Deep Learning, which is being used in various fields recently, is being threatened with Adversarial Attack. In this paper, we experimentally verify that the classification accuracy is lowered by adversarial samples generated by malicious attackers in image classification models. We used MNIST dataset and measured the detection accuracy by injecting adversarial samples into the Autoencoder classification model and the CNN (Convolution neural network) classification model, which are created using the Tensorflow library and the Pytorch library. Adversarial samples were generated by transforming MNIST test dataset with JSMA(Jacobian-based Saliency Map Attack) and FGSM(Fast Gradient Sign Method). When injected into the classification model, detection accuracy decreased by at least 21.82% up to 39.08%.

• Nuclear Engineering and Technology
• /
• v.53 no.2
• /
• pp.522-531
• /
• 2021

We describe a new deep learning model - Convolutional eXtreme Gradient Boosting (ConvXGB) for classification problems based on convolutional neural nets and Chen et al.'s XGBoost. As well as image data, ConvXGB also supports the general classification problems, with a data preprocessing module. ConvXGB consists of several stacked convolutional layers to learn the features of the input and is able to learn features automatically, followed by XGBoost in the last layer for predicting the class labels. The ConvXGB model is simplified by reducing the number of parameters under appropriate conditions, since it is not necessary re-adjust the weight values in a back propagation cycle. Experiments on several data sets from UCL Repository, including images and general data sets, showed that our model handled the classification problems, for all the tested data sets, slightly better than CNN and XGBoost alone and was sometimes significantly better.