• Journal of Information Processing Systems
• /
• v.17 no.2
• /
• pp.227-241
• /
• 2021

Object detection techniques based on deep learning such as YOLO have high detection performance and precision in a single channel video stream. In order to expand to multiple channel object detection in real-time, however, high-performance hardware is required. In this paper, we propose a novel back-end server framework, a real-time AI vision platform (RAVIP), which can extend the object detection function from single channel to simultaneous multi-channels, which can work well even in low-end server hardware. RAVIP assembles appropriate component modules from the RODEM (real-time object detection module) Base to create per-channel instances for each channel, enabling efficient parallelization of object detection instances on limited hardware resources through continuous monitoring with respect to resource utilization. Through practical experiments, RAVIP shows that it is possible to optimize CPU, GPU, and memory utilization while performing object detection service in a multi-channel situation. In addition, it has been proven that RAVIP can provide object detection services with 25 FPS for all 16 channels at the same time.

• The Journal of Korea Robotics Society
• /
• v.12 no.1
• /
• pp.33-41
• /
• 2017

We present a region-based approach for accurate pose estimation of small mechanical components. Our algorithm consists of two key phases: Multi-view object co-segmentation and pose estimation. In the first phase, we explain an automatic method to extract binary masks of a target object captured from multiple viewpoints. For initialization, we assume the target object is bounded by the convex volume of interest defined by a few user inputs. The co-segmented target object shares the same geometric representation in space, and has distinctive color models from those of the backgrounds. In the second phase, we retrieve a 3D model instance with correct upright orientation, and estimate a relative pose of the object observed from images. Our energy function, combining region and boundary terms for the proposed measures, maximizes the overlapping regions and boundaries between the multi-view co-segmentations and projected masks of the reference model. Based on high-quality co-segmentations consistent across all different viewpoints, our final results are accurate model indices and pose parameters of the extracted object. We demonstrate the effectiveness of the proposed method using various examples.

• The Transactions of the Korea Information Processing Society
• /
• v.3 no.6
• /
• pp.1493-1505
• /
• 1996

As Internet and its users grow dramatically multi-media data are getting common, many types of electronic mail applications are showing up. Internet s/w developers and users would like to electronic mail system for commercial purposes. They also would like to protect their electronic mail somehow. For these purpose, the security feature using cryptography is one of the essential functions. In this paper, we describe the library function development for handing MOSS(MIME Object Security Service), the security version of MINE(Multi -Purpose Internet Mail Extension). For these purpose, we explain the security features and structures of MOSS and defined the necessary function and we also discussed design issues for the MOSS implementation.

• Journal of the Korean Society for Precision Engineering
• /
• v.16 no.10
• /
• pp.194-202
• /
• 1999

A method for pattern recognition based on wavelet transform is proposed in this paper. The boundary of the object to be recognized includes shape information for object of machine parts. The contour is first represented using a one-dimensional signal and normalized about translation, rotation and scale, then is used to build the wavelet transform representation of the object. Wavelets allow us to decompose a function into multi-resolution hierarchy of localized frequency bands. The recognition of 2-dimensional object based on the wavelet is described to analyze the shape of analysis technique; the discrete wavelet transform(DWT). The feature vectors obtained using wavelet analysis is classified using a multi-layer neural network. The results show that, compared with the use of fourier descriptors, recognition using wavelet is more stable and efficient representation. And particularly the performance for objects corrupted with noise is better than that of other method.

• Journal of Korea Multimedia Society
• /
• v.11 no.5
• /
• pp.623-632
• /
• 2008

In this paper, we used the color transfer function and the opacity transfer function for the internal 3D object visualization of an image volume data. In transfer function, creating values of between boundaries generally is ambiguous. We concentrated to extract boundary features for segmenting the visual volume rendering objects. Consequently we extracted an image gradient feature in spatial domain and created a multi-dimensional transfer function according to the GPU efficient improvement. Finally using these functions we obtained a good research result as an implementing object boundary visualization of the graphic hardware based 3D texture mapping.

• Journal of the Korean Society for Precision Engineering
• /
• v.14 no.12
• /
• pp.24-29
• /
• 1997

Recently, Genetic Algorithm(GA), which is a stochastic direct search strategy that mimics the process of genetic evolution, is widely adapted into a search procedure for structural optimization. Contrast to traditional optimal design techniques which use design sensitivity analysis results, GA is very simple in their algorithms and there is no need of continuity of functions(or functionals) any more in GA. So, they can be easily applicable to wide area of design optimization problems. Also, owing to multi-point search procedure, they have higher porbability of convergence to global optimum compared to traditional techniques which take one-point search method. The methods consist of three genetics opera- tions named selection, crossover and mutation. In this study, a method of finding the omtimum size and topology of drilling machine is proposed by using the GA, For rapid converge to optimum, elitist survival model,roulette wheel selection with limited candidates, and multi-point shuffle cross-over method are adapted. And pseudo object function, which is the combined form of object function and penalty function, is used to include constraints into fitness function. GA shows good results of weight reducing effect and convergency in optimal design of drilling machine.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.48 no.2
• /
• pp.155-164
• /
• 2012

The CSR rule was defined by IACS as the unified rule for a commercial ship like a bulk carrier and a tanker. It have been required more strict conditions for various parts like loading conditions, the local and girder strength, fatigue strength, FEM for the ship rule. It was changed in many parts of the ship rules. In this paper, the mid-parts of 17.5K DWT bulk carrier were optimized by the CSR rule. On the other hand, the modified artificial life algorithms with multi-object functions were developed for optimizing the scantling. It is possible to find multi-global optimum solutions in the multi-object functions. And it is faster and efficient than the artificial life algorithm. First, to be optimizing the scantling and the weight by CSR rule, that is calculated by the CSR rule. The next, the result is re-calculated by the modified artificial life algorithm with multi-object functions. The optimized results which are satisfied with the CSR rule like the minimum size and the thickness of stiffener and the minimum cost have been searched by the optimizing algorithm. And the results have been compared with the non-optimizing results.

• Smart Media Journal
• /
• v.4 no.1
• /
• pp.16-24
• /
• 2015

In this paper, we propose a method for multi-touch tracking using MRF-based particle filters and gesture likelihood estimation Each touch (of one finger) is considered to be one object. One of frequently occurring issues is the hijacking problem which means that an object tracker can be hijacked by neighboring object. If a predicted particle is close to an adjacent object then the particle's weight should be lowered by analysing the influence of neighboring objects for avoiding hijacking problem. We define a penalty function to lower the weights of those particles. MRF is a graph representation where a node is the location of a target object and an edge describes the adjacent relation of target object. It is easy to utilize MRF as data structure of adjacent objects. Moreover, since MRF graph representation is helpful to analyze multi-touch gestures, we describe how to define gesture likelihoods based on MRF. The experimental results show that the proposed method can avoid the occurrence of hijacking problems and is able to estimate gesture likelihoods with high accuracy.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.21 no.6
• /
• pp.141-152
• /
• 2011

Growing efforts and interests of security techniques in a diverse surveillance environment, the intelligent surveillance system, which is capable of automatically detecting and tracking target objects in multi-cameras environment, is actively developing in a security community. In this paper, we propose an effective visual surveillance system that is avaliable to track objects continuously in multiple non-overlapped cameras. The proposed object tracking scheme consists of object tracking module and tracking management module, which are based on hand-off scheme and protocol. The object tracking module, runs on IP camera, provides object tracking information generation, object tracking information distribution and similarity comparison function. On the other hand, the tracking management module, runs on video control server, provides realtime object tracking reception, object tracking information retrieval and IP camera control functions. The proposed object tracking scheme allows comprehensive framework that can be used in a diverse range of application, because it doesn't rely on the particular surveillance system or object tracking techniques.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.2 no.3
• /
• pp.215-220
• /
• 2002

This paper proposes a method to apply the Backpropagation(BP) algorithm of MVL(Multi-Valued Logic) Neural Network to pattern recognition. It extracts the property of an object density about an original pattern necessary for pattern processing and makes the property of the object density mapped to MVL. In addition, because it team the pattern by using multiple valued logic, it can reduce time f3r pattern and space fer memory to a minimum. There is, however, a demerit that existed MVL cannot adapt the change of circumstance. Through changing input into MVL function, not direct input of an existed Multiple pattern, and making it each variable loam by neural network after calculating each variable into liter function. Error has been reduced and convergence speed has become fast.