• Journal of the Korean Society of Industry Convergence
• /
• v.23 no.3
• /
• pp.515-521
• /
• 2020

Deep learning methods have been effectively used to provide great improvement in various research fields such as machine learning, image processing and computer vision. One of the most frequently used deep learning methods in image processing is the convolutional neural networks. Compared to the traditional artificial neural networks, convolutional neural networks do not use the predefined kernels, but instead they learn data specific kernels. This property makes them to be used as feature extractors as well. In this study, we compared the quality of CNN features for traditional texture feature extraction methods. Experimental results demonstrate the superiority of the CNN features. Additionally, the recognition process and result of a pioneering CNN on MNIST database are presented.

• Transactions of the Korean Society of Mechanical Engineers
• /
• v.17 no.11
• /
• pp.2711-2722
• /
• 1993

Most of the current 2-D object recognition systems are model-based. In such systems, the representation of each of a known set of objects are precompiled and stored in a database of models. Later, they are used to recognize the image of an object in each instance. In this thesis, the approach method for the 2-D object recognition is treating an object boundary as a string of structral units and utilizing string matching to analyze the scenes. To reduce string matching time, models are rebuilt by means of fuzzy c-means clustering algorithm. In this experiments, the image of objects were taken at initial position of a robot from the CCD camera, and the models are consturcted by the proposed algorithm. After that the image of an unknown object is taken by the camera at a random position, and then the unknown object is identified by a comparison between the unknown object and models. Finally, the amount of translation and rotation of object from the initial position is computed.

• Proceedings of the Korean Institute of Navigation and Port Research Conference
• /
• 2023.05a
• /
• pp.252-253
• /
• 2023

Container terminal automation might offer many potential benefits, such as increased productivity, reduced cost, and improved safety. Autonomous trucks can lead to more efficient container transport. A robust lane detection method is proposed using score-based generative modeling through stochastic differential equations for image-to-image translation. Image processing techniques are combined with Density-Based Spatial Clustering of Applications with Noise (DBSCAN) and Genetic Algorithm (GA) to ensure lane positioning robustness. The proposed method is validated by a dataset collected from the port terminals under different environmental conditions and tested the robustness of the lane detection method with stochastic noise.

• 제어로봇시스템학회:학술대회논문집
• /
• 2004.08a
• /
• pp.875-878
• /
• 2004

Researching in text detection and segmentation has been done for a long period in the OCR area. However, there is some other area that the text detection and segmentation from images can be very useful. In this report, we first propose the design of a mobile translator system which helps non-native speakers to understand the foreign language using ubiquitous mobile network and camera mobile phones. The main focus of the paper will be the algorithm in detecting and segmenting texts embedded in the natural scenes from taken images. The image, which is captured by a camera mobile phone, is transmitted to a translator server. It is initially passed through some preprocessing processes to smooth the image as well as suppress noises. A threshold is applied to binarize the image. Afterward, an edge detection algorithm and connected component analysis are performed on the filtered image to find edges and segment the components in the image. Finally, the pre-defined layout relation constraints are utilized in order to decide which components likely to be texts in the image. A preliminary experiment was done and the system yielded a recognition rate of 94.44% on a set of 36 various natural scene images that contain texts.

• Journal of the Korea Society of Computer and Information
• /
• v.27 no.7
• /
• pp.27-33
• /
• 2022

In this paper, we propose a novel generator architecture for multi-domain style transfer method not an image to image translation, as a method of generating a styled image by transfering a style to the content image. A latent vector and Gaussian noises are added to the generator of GAN so that a high quality image is generated while considering the characteristics of various data distributions for each domain and preserving the features of the content data. With the generator architecture of the proposed GAN, networks are configured and presented so that the content image can learn the styles for each domain well, and it is applied to the domain composed of images of the four seasons to show the high resolution style transfer results.

• Proceedings of the KIEE Conference
• /
• 1993.07a
• /
• pp.418-420
• /
• 1993

In this paper, we proposed a new FMS Landmark recognition algorithm using color images. Firstly, a NTSC image fame is captured, and then it is converted to a field image in order to reduce the image blurring from the AGV motion. Secondly, the landmark is detected via the comparison of the color vectors of image pixels with the landmark color. Finally, the identification of FMS landmark is executed using a newly designed landmark pattern with a set of reference points. The landmark pattern is normalized against its translation, rotation, and scaling. And then, its vertical projection data are fisted for the pattern classification using the standard data set. Experimental results show that our scheme performs well.

• Proceedings of the KIEE Conference
• /
• 1994.11a
• /
• pp.353-355
• /
• 1994

This work describes a hardware implementation of a precision image tracking system which employs a CCD camera mounted on pan/tilt device. Unknown translation between two successive images of a moving object is estimated by using a generalized least-squares method. Estimated position error obtained by the tracking algorithm is used to drive DC motors built in the pan/tilt device for the camera to follow the image. An experimental result shows a sub-resolution tracking error for a image moving with a uniform velocity.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2022.05a
• /
• pp.497-500
• /
• 2022

High-quality images and videos are being generated as technologies for deep learning-based image style translation and conversion of static images into dynamic images have developed. However, it takes a lot of time and resources to manually transform images, as well as professional knowledge due to the difficulty of natural image transformation. Therefore, in this paper, we study natural style mixing through a style conversion network using GAN and natural dynamic image generation using the First Order Motion Model network (FOMM).

• Proceedings of the IEEK Conference
• /
• 2000.11d
• /
• pp.251-254
• /
• 2000

This paper proposes an enhanced algorithm for person identification system utilizing hand vein pattern. The conventional algorithm does not cope with distortion caused by image rotation caused by misplaced hands on the imaging device. A straightforward approach to consider the rotaional compensation required too much computational load, thus, we devised an approach to expect the rotation direction along with image translation, reducing the compuational requirement dramatically In this paper, we present the details of the algorithm with experimental results with the new algorithm.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.33B no.5
• /
• pp.73-82
• /
• 1996

In this paper, an efficient algorithm for recognizing and positioning occuluded objects in a two-dimensional plane is presented. Model objects and unknown input image are approximated by polygonal boundaries, which are compactly represented by shape functions of the polygons. The input image is partitioned into measningful segments whose end points are at the locations of possible occlusion - i.e. at concave vertices. Each segment is matched against known model objects by calculating a matching measure, which is defined as the minimum euclidean distance between the shape functions. An O(mm(n+m) algorithm for computing the measure is presentd, where n and m are the number of veritces for a model and an unknown object, respectively. Match results from aprtial segments are combined based on mutual compatibility, then are verified using distance transformation and translation vector to produce the final recognition. The proposed algorithm is invariant under translation and rotation of objects, which has been shown by experimental results.