• Journal of Institute of Control, Robotics and Systems
• /
• v.9 no.12
• /
• pp.1033-1041
• /
• 2003

A new visual tracking scheme is proposed for a mobile robot that tracks a moving object in 3D space in real time. Visual tracking is to control a mobile robot to keep a moving target at the center of input image at all time. We made it possible by simplifying the relationship between the 2D image frame captured by a single camera and the 3D workspace frame. To precisely calculate the input vector (orientation and distance) of the mobile robot, the speed vector of the target is determined by eliminating the speed component caused by the camera motion from the speed vector appeared in the input image. The problem of temporary disappearance of the target form the input image is solved by selecting the searching area based on the linear prediction of target motion. The experimental results have shown that the proposed scheme can make a mobile robot successfully follow a moving target in real time.

• International Journal of Advanced Culture Technology
• /
• v.6 no.3
• /
• pp.142-150
• /
• 2018

This paper describes the algorithm that lowers the dimension, maintains the object recognition and significantly reduces the eigenspace configuration time by combining the edge orientation histogram and principle component analysis. By using the detected object region as a recognition input image, in this paper the object recognition method combined with principle component analysis and the multi-layer network which is one of the intelligent classification was suggested and its performance was evaluated. As a pre-processing algorithm of input object image, this method computes the eigenspace through principle component analysis and expresses the training images with it as a fundamental vector. Each image takes the set of weights for the fundamental vector as a feature vector and it reduces the dimension of image at the same time, and then the object recognition is performed by inputting the multi-layer neural network.

• Journal of Institute of Control, Robotics and Systems
• /
• v.6 no.5
• /
• pp.390-395
• /
• 2000

We present a fingerprint identification algorithm using the wavelet transform and correlation. The wavelet transform is used because of its simple operation to extract fingerprint minutiaes features for fingerprint classification. We perform the rowwise 1-D wavelet transform for a $256\times256$ fingerprint image to get a $1\times256$ column vector using the Haar wavelet and repeat 1-D wavelet transform for a 1$\times$256 column vector to get a $1\times4$ feature vector. Using PNN(Probabilistic Neural Network), we select the possible candidates from the stored feature vectors for fingerprint images. For those candidates, we compute the correlation between the input binary image and the target binary image to find the most similar fingerprint image. The proposed algorithm may be the key to a low cost fingerprint identification system that can be operated on a small computer because it does not need a large memory size and much computation.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 2009.01a
• /
• pp.435-440
• /
• 2009

The scalable vector graphics (SVG) standard has allowed the complex bitmap images to be represented by vector based graphics and provided some advantages over the raster based graphics in applications, for example, where scalability is required. This paper presents an algorithmto convert bitmap images into SVG format. The algorithm is an integration of pixel-level triangulation, data dependent triangulation, a new image mesh simplification algorithm, and a polygonization process. Both triangulation techniques enable the image quality (especially the edge features) to be preserved well in the reconstructed image and the simplification and polygonization procedures reduce the size of the SVG file. Experiments confirm the effectiveness of the proposed algorithm.

• KIEE International Transaction on Systems and Control
• /
• v.12D no.1
• /
• pp.7-11
• /
• 2002

In this study, fur the shape-based image retrieval, a method using local differential invariants is proposed. This method calculates the differential invariant feature vector at every feature point extracted by Harris comer point detector. Then through vector quantization using LBG algorithm, all feature vectors are represented by a codebook index. All images are indexed by the histogram of codebook index, and by comparing the histograms the similarity between images is obtained. The proposed method is compared with the existing method by performing experiments for image database including various 1100 trademarks.

• Korean Journal of Remote Sensing
• /
• v.25 no.3
• /
• pp.233-242
• /
• 2009

This paper presents two different types of supervised classifiers such as support vector machine (SVM) and spectral angle mapper (SAM). The Compact Airborne Spectrographic Imager (CASI) high resolution aerial image was classified with the above two classifier. The image was classified into eight land use /land cover classes. Accuracy assessment and Kappa statistics were estimated for SVM and SAM separately. The overall classification accuracy and Kappa statistics value of the SAM were 69.0% and 0.62 respectively, which were higher than those of SVM (62.5%, 0.54).

• International Journal of Computer Science & Network Security
• /
• v.21 no.6
• /
• pp.1-6
• /
• 2021

Recently, pre-trained convolutional neural network CNNs have been widely used and applied for medical image classification. These models can utilised in three different ways, for feature extraction, to use the architecture of the pre-trained model and to train some layers while freezing others. In this study, the ResNet18 pre-trained CNNs model is used for feature extraction, followed by the support vector machine for multiple classes to classify medical images from multi-classes, which is used as the main classifier. Our proposed classification method was implemented on Kvasir and PH2 medical image datasets. The overall accuracy was 93.38% and 91.67% for Kvasir and PH2 datasets, respectively. The classification results and performance of our proposed method outperformed some of the related similar methods in this area of study.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.25 no.1B
• /
• pp.120-127
• /
• 2000

In this paper, we propose multispectral image compression using classified interband prediction and vector quantization in wavelet domain. This method classifies each region considering reflection characteristics of each band in image data. In wavelet domain, we perform the classified intraband VQ to remove intraband redundancy for a reference band image that has the lowest spatial variance and the best correlation with other band. And in wavelet domain, we perform the classifled interband prediction to remove interband redundancy for the remaining bands. Then error wavelet coefficients between original image and predicted image are intraband vector quantized to reduce prediction error. Experiments on remotely sensed satellite image show that coding efficiency of theproposed method is better than that of the conventional method.

• Journal of the Institute of Convergence Signal Processing
• /
• v.3 no.1
• /
• pp.41-47
• /
• 2002

Fractal image compression can reduce the size of image data by the contractive mapping that is affine transformation to find the block(called as range block) which is the most similar to the original image. Even though fractal image compression is regarded as an efficient way to reduce the data size, it has high distortion rate and requires long encoding time. In this paper, we presented a hybrid fractal image compression system with the modified SOFM Vector Quantizer which uses improved competitive learning method. The simulation results showed that the VQ hybrid fractal using improved competitive loaming SOFM has better distortion rate than the VQ hybrid fractal using normal SOFM.

• Journal of the Institute of Convergence Signal Processing
• /
• v.8 no.4
• /
• pp.293-296
• /
• 2007

In this paper, we propose the systolic array architecture for the vector median filter. In the color image processing, the vector signal (i.e. the color) consists of three elements, red, green and blue. The vector median filter is very effective to utilize the correlation among red, green and blue elements. The computational complexity of the proposed architecture for computing the vector median of N vector signals is (N+2) clock periods compared to the (3N+1) clock periods in the previous method. In addition to, the input vector signals can be loaded in serial in the proposed architecture. In the previous method, N input vector signals should be loaded to the vector median filter in parallel at the first clock. The proposed architecture is implemented with FPGA.