• Proceedings of the KSRS Conference
• /
• 2005.10a
• /
• pp.575-578
• /
• 2005

Topographical images, in case of aerial or satellite images, are usually similar in colors and textures, and complex in shapes. Thus we have to use shape features of images for efficiently retrieving a query image from topographical image databases. In this paper, we propose a shape feature extraction method which is suitable for topographical images. This method, which improves the existing projection in the Cartesian coordinates, performs the projection operation in the polar coordinates. This method extracts three attributes, namely the number of region pixels, the boundary pixel length of the region from the centroid, the number of alternations between region and background, along each angular direction of the polar coordinates. It extracts the features of complex shape objects which may have holes and disconnected regions. An advantage of our method is that it is invariant to rotation/scale/translation of images. Finally we show the advantages of our method through experiments by comparing it with CSS which is one of the most successful methods in the area of shape feature extraction

• 제어로봇시스템학회:학술대회논문집
• /
• 2001.10a
• /
• pp.23.4-23
• /
• 2001

This paper proposes a new approach to unsupervised classification of remotely sensed images. Fusion of optic images (Landsat TM) and radar data (SAR) has beer used to increase the accuracy of classification. Number of clusters is estimated using generalized Dunns measure. Performance of the proposed method is best observed comparing the classified images with classified aerial images.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2006.11a
• /
• pp.203-207
• /
• 2006

Nowadays, Cars are continuing to grow at an alarming rate but they also cause many problems such as traffic accident, pollutions and so on. One of the most effective methods that prevent traffic accidents is the use of traffic monitoring systems, which are already widely used in many countries. The monitoring system is beginning to be used in domestic recently. An intelligent monitoring system generates photo images of cars as well as identifies cars by recognizing their plates. That is, the system automatically recognizes characters of vehicle plates. An automatic vehicle plate recognition consists of two main module: a vehicle plate locating module and a vehicle plate number identification module. We study for a vehicle plate number identification module in this paper. We use image preprocessing, feature extraction, multi-layer neural networks for recognizing characters of vehicle plates and we present a feature-comparison method for improving the performance of vehicle plate number identification module. In the experiment on identifying vehicle plate number, 300 images taken from various scenes were used. Of which, 8 images have been failed to identify vehicle plate number and the overall rate of success for our vehicle plate recognition algorithm is 98%.

• Journal of Korea Multimedia Society
• /
• v.22 no.12
• /
• pp.1447-1456
• /
• 2019

As artificial intelligence technology is actively used in image processing, it is possible to generate high-quality fake images based on deep learning. Fake images generated using GAN(Generative Adversarial Network), one of unsupervised learning algorithms, have reached levels that are hard to discriminate from the naked eye. Detecting these fake images is required as they can be abused for crimes such as illegal content production, identity fraud and defamation. In this paper, we develop a deep-learning model based on CNN(Convolutional Neural Network) for the detection of StyleGAN fake images. StyleGAN is one of GAN algorithms and has an excellent performance in generating face images. We experiment with 48 number of experimental scenarios developed by combining parameters of the proposed model. We train and test each scenario with 300,000 number of real and fake face images in order to present a model parameter that improves performance in the detection of fake faces.

• Journal of Biosystems Engineering
• /
• v.37 no.4
• /
• pp.271-278
• /
• 2012

Worldwide trends in animal welfare have resulted in an increased interest in individual management of sows housed in groups within hog barns. Estrus detection has been shown to be one of the greatest determinants of sow productivity. Purpose: We conducted this study to develop a method that can automatically detect the estrus state of a sow by selecting optimal texture parameters from images of a sow's pudendum and by optimizing the number of neurons in the hidden layer of an artificial neural network. Methods: Texture parameters were analyzed according to changes in a sow's pudendum in estrus such as mucus secretion and expansion. Of the texture parameters, eight gray level co-occurrence matrix (GLCM) parameters were used for image analysis. The image states were classified into ten grades for each GLCM parameter, and an artificial neural network was formed using the values for each grade as inputs to discriminate the estrus state of sows. The number of hidden layer neurons in the artificial neural network is an important parameter in neural network design. Therefore, we determined the optimal number of hidden layer units using a trial and error method while increasing the number of neurons. Results: Fifteen hidden layers were determined to be optimal for use in the artificial neural network designed in this study. Thirty images of 10 sows were used for learning, and then 30 different images of 10 sows were used for verification. Conclusions: For learning, the back propagation neural network (BPN) algorithm was used to successful estimate six texture parameters (homogeneity, angular second moment, energy, maximum probability, entropy, and GLCM correlation). Based on the verification results, homogeneity was determined to be the most important texture parameter, and resulted in an estrus detection rate of 70%.

• International journal of advanced smart convergence
• /
• v.10 no.2
• /
• pp.108-122
• /
• 2021

With the development of 3D graphics software and the speed of computer hardware, it is an era that can be realistically expressed not only in movie visual effects but also in console games. In the production of such realistic 3D models, 3D scans are increasingly used because they can obtain hyper-realistic results with relatively little effort. Among the various 3D scanning methods, photogrammetry can be used only with a camera. Therefore, no additional hardware is required, so its demand is rapidly increasing. Most 3D artists shoot as many images as possible with a video camera, etc., and then calculate using all of those images. Therefore, the photogrammetry method is recognized as a task that requires a lot of memory and long hardware operation. However, research on how to obtain precise results with 3D photogrammetry scans is insufficient, and a large number of photos is being utilized, which leads to increased production time and data capacity and decreased productivity. In this study, point cloud data generated according to changes in the number and resolution of photographic images were produced, and an experiment was conducted to compare them with original data. Then, the precision was measured using the average distance value and standard deviation of each vertex of the point cloud. By comparing and analyzing the difference in the precision of the 3D photogrammetry scans according to the number and resolution of images, this paper presents a direction for obtaining the most precise and effective results to 3D artists.

• Journal of Korea Society of Digital Industry and Information Management
• /
• v.18 no.4
• /
• pp.1-11
• /
• 2022

Recently, research on license plate recognition, which is a core technology of an intelligent transportation system(ITS), is being actively conducted. In this paper, we propose a method to extract numbers and characters from low-quality license plate images by applying the YOLOv4 algorithm. YOLOv4 is a one-stage object detection method using convolution neural network including BACKBONE, NECK, and HEAD parts. It is a method of detecting objects in real time rather than the previous two-stage object detection method such as the faster R-CNN. In this paper, we studied a method to directly extract number and character regions from low-quality license plate images without additional edge detection and image segmentation processes. In order to evaluate the performance of the proposed method we experimented with 500 license plate images. In this experiment, 350 images were used for training and the remaining 150 images were used for the testing process. Computer simulations show that the mean average precision of detecting number and character regions on vehicle license plates was about 93.8%.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.65 no.9
• /
• pp.1531-1540
• /
• 2016

In this paper, we propose vehicle license plate recognition system based on Radial Basis Function Neural Networks (RBFNNs) with the use of local binarization functions and canny edge algorithm. In order to detect the area of license plate and also recognize license plate numbers, binary images are generated by using local binarization methods, which consider local brightness, and canny edge detection. The generated binary images provide information related to the size and the position of license plate. Additionally, image warping is used to compensate the distortion of images obtained from the side. After extracting license plate numbers, the dimensionality of number images is reduced through Principal Component Analysis (PCA) and is used as input variables to RBFNNs. Particle Swarm Optimization (PSO) algorithm is used to optimize a number of essential parameters needed to improve the accuracy of RBFNNs. Those optimized parameters include the number of clusters and the fuzzification coefficient used in the FCM algorithm, and the orders of polynomial of networks. Image data sets are obtained by changing the distance between stationary vehicle and camera and then used to evaluate the performance of the proposed system.

• Journal of the Korean Society of Radiology
• /
• v.17 no.7
• /
• pp.1017-1023
• /
• 2023

The purpose of this study was to improve the accuracy of effective atomic number (EAN) and relative electron density (RED) using a polynomial-based calibration method using dual-energy CT images. A phantom composed of 11 tissue-equivalent materials was acquired with dual-energy CT to obtain low- and high-energy images. Using the acquired dual-energy images, the ratio of attenuation of low- and high-energy images for EAN was calibrated based on Stoichiometric, Quadratic, Cubic, Quartic polynomials. EAN and RED were extracted using each calibration method. As a result of the experiment, the average error of EAN using Cubic polynomial-based calibration was minimum. Even in the RED image extracted using EAN, the error of the Cubic polynomial-based RED was minimum. Cubic polynomial-based calibration contributes to improving the accuracy of EAN and RED, and would like to contribute to accurate diagnosis of lesions in CT examinations or quantification of various materials in the human body.

• Proceedings of the KSR Conference
• /
• 2011.10a
• /
• pp.3306-3311
• /
• 2011

Last Subway video surveillance system consists of a network device that is used. Through the network to transmit video data to digital conversion of analog video via a process server or a PC video to a split-screen in various forms is expressed. In recent years, multi-monitor video cameras from the same pop-up or more, such as history, structure expressed on a variety of video is required by express. The problem with these systems, video compression and transmission of many cameras, and this image data received from the server or PC to take out all the images you want to watch to occur when in order to express all of the images because of the need to decode most of the program per limit of number of channels is positioned. This limited number of channels to have a video that nothing forced, but it is likely to do so in the future performance of the hardware evolves gradually channeled images available number of channels will increase proportionately. However, as the development of hardware required for a single screen video channel will be more gradual capital. The hardware rather than relying solely on the performance of the decoded video data on the screen in order to express a more efficient utilization of shared memory for video surveillance software will provide the operating plan.