• Proceedings of the Korea Information Processing Society Conference
• /
• 2016.10a
• /
• pp.506-509
• /
• 2016

We present real-time pedestrian detection that exploit accuracy of Faster R-CNN network. Faster R-CNN has shown to success at PASCAL VOC multi-object detection tasks, and their ability to operate on raw pixel input without the need to design special features is very engaging. Therefore, in this work we apply and adjust Faster R-CNN to single object detection, which is pedestrian detection. The drawback of Faster R-CNN is its failure when object size is small. Previously, small sized object problem was solved by Scale-aware Network. We incorporate Scale-aware Network to Faster R-CNN. This made our method Scale-aware Faster R-CNN (DF R-CNN) that is both fast and very accurate. We separated Faster R-CNN networks into two sub-network, that is one for large-size objects and another one for small-size objects. The resulting approach achieves a 28.3% average miss rate on the Caltech Pedestrian detection benchmark, which is competitive with the other best reported results.

• The Journal of Korea Robotics Society
• /
• v.4 no.3
• /
• pp.185-191
• /
• 2009

This paper presents a new shape-based algorithm using affine category shape model for object category recognition and model learning. Affine category shape model is a graph of interconnected nodes whose geometric interactions are modeled using pairwise potentials. In its learning phase, it can efficiently handle large pose variations of objects in training images by estimating 2-D homography transformation between the model and the training images. Since the pairwise potentials are defined on only relative geometric relationship betweenfeatures, the proposed matching algorithm is translation and in-plane rotation invariant and robust to affine transformation. We apply spectral matching algorithm to find feature correspondences, which are then used as initial correspondences for RANSAC algorithm. The 2-D homography transformation and the inlier correspondences which are consistent with this estimate can be efficiently estimated through RANSAC, and new correspondences also can be detected by using the estimated 2-D homography transformation. Experimental results on object category database show that the proposed algorithm is robust to pose variation of objects and provides good recognition performance.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2006.05a
• /
• pp.193-194
• /
• 2006

A solid freeform fabrication (SFF) system using selective laser sintering (SLS) is currently recognized as a leading process and SLS extends the applications to machinery and automobiles due to the variousmaterials employed. In order to develop a more elaborate and rapid system for fabricating large objects compared to existing SLS, this study employs a new selective dual-laser sintering (SDLS) process. Also, this paper will address development of an SFF system which employs the dual laser system and the unique scanning device. Experiments were performed to evaluate the effect of a scanning path and fabrication parameters on sintering process and to fabricate the various 3D objects using polymer powder.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.11 no.5
• /
• pp.2523-2538
• /
• 2017

In most of the existing literature, the definition of the class label has the following characteristics. First, the class label of the samples from the same object has an absolutely fixed value. Second, the difference between class labels of the samples from different objects should be maximized. However, the appearance of a face varies greatly due to the variations of the illumination, pose, and expression. Therefore, the previous definition of class label is not quite reasonable. Inspired by discriminative least squares regression algorithm (DLSR), a noisy label based discriminative least squares regression algorithm (NLDLSR) is presented in this paper. In our algorithm, the maximization difference between the class labels of the samples from different objects should be satisfied. Meanwhile, the class label of the different samples from the same object is allowed to have small difference, which is consistent with the fact that the different samples from the same object have some differences. In addition, the proposed NLDLSR is expanded to the kernel space, and we further propose a novel kernel noisy label based discriminative least squares regression algorithm (KNLDLSR). A large number of experiments show that our proposed algorithms can achieve very good performance.

• KIEAE Journal
• /
• v.14 no.5
• /
• pp.75-80
• /
• 2014

An attempt has been made to prove the so-called 'feature accumulation theory'. It is the theory describing that people tend to feel the same space with more identifiable objects much larger than that with fewer objects. Applying this theory to our cognition of spatial size, this paper made an experiment. Students were asked that if the lecture room they are sitting becomes a module (module 1), then how large are the questioned spaces in the campus. The result was striking. Through the mental image processing, they answered that the library and the architecture building looks much smaller than they actually are, and more surprisingly the basketball field much more smaller than it really is. This experiment shows that there is a strong tendency by which people regard the space much larger when there are more occupiable or behavior-causing elements in the space. In the case of basketball field, since there is nothing that can be occupied, this open space is seen as a small space for the subjects. This line of cognitive perception can be applied to the practice of urban planning and architectural planning. With the same size of given space, we can make it feel more rich and larger.

• Proceedings of the KSME Conference
• /
• 2001.06b
• /
• pp.216-221
• /
• 2001

The equivalent source method(ESM) is used for the calculation of the internal pressure field for an enclosure which can have arbitrary boundary conditions and may include internal objects which scatter the sound field. The advantage of using ESM is that it requires relatively low computing cost and is easy to model the internal diffracting objects. In the ESM modelling, some of the equivalent positions are chosen to be the same as the first order images of the source inside the enclosure, some are positioned on a spherical surface some distance outside the enclosure. The normal velocity on the surfaces of the enclosure walls is evaluated at a larger number of positions than there are equivalent sources. The sum of the squared difference between this velocity and the expected is minimized by adjusting the strength of the equivalent sources. This study is on the optimal equivalent source positions, the far field sources. Typically, the far field sources are evenly distributed on a surface of a virtual sphere which is centered at the enclosure with a sufficiently large radius. In this study, optimal far field source locations are searched using simulated annealing method and simulation results showed that optimally located sources gave better accuracy even with a smaller number of far field sources.

• Publications of The Korean Astronomical Society
• /
• v.30 no.2
• /
• pp.461-462
• /
• 2015

Quasars are among the farthest and brightest objects known in the universe. Because quasars are mostly observed in the redshift range between 1 and 3, they can be used to study large scale structure in the universe, and its evolution over the past billion years. An important issue is the evolution of the quasar luminosity function, which has been investigated for relative small samples of the 2QZ catalog. Here we extend the study to 3 quasar samples, the most recent data of the Milliquas, Master and 2QZ quasar catalogs to determine the luminosity function of quasars and its evolution, using the Standard cosmological ${\Lambda}CDM$ model with ${\Omega}_{\Lambda}=0.73$, ${\Omega}_M=0.27$, and $H_0=70kms^{-1}Mpc^{-1}$. For the purpose of this analysis we initially used 0.25-mag bins and approximately 0.180-redshift bins, then calculated the comoving distance and comoving volume for each bin of redshift and calculated the number of objects in each bin per unit volume, in order to find the number density per absolute magnitude bin. Our analysis on the basis of these new and much more complete datasets is largely in agreement with earlier studies of the luminosity evolution of quasars.

• IE interfaces
• /
• v.16 no.1
• /
• pp.70-84
• /
• 2003

Recently, object-oriented techniques have been used more and more for developments of an information system. But, established object-oriented methodologies are hard to express a business process of various abstract degrees in the analysis level and independent components of the system. They have difficulties in developing a large-scale information system of manufacturing industry such as PDM and CIM. This paper proposes an information system development methodology that imports the object-oriented IDEF0 (OOIDEF0) function model in analysis level. This methodology is made up of requirements gathering, system analysis, system design, and implementation. In requirements gathering level, organization diagram and interview technique are used for input data of OOIDEF0 function model. The OOIDEF0 function model, the interface model and function descriptions are made out in analysis level. Information objects and implementation objects are designed on the basis of the OOIDEF0 function model in design level. The database is built and programming is accomplished in implementation level. In order to prove the consistency and efficiency of the proposed methodology, the PDM system for ship production is modeled and prototyped.

• Proceedings of the KAIS Fall Conference
• /
• 2003.11a
• /
• pp.118-126
• /
• 2003

In data mining, having access to large amount of data sets for the purpose of predictive data does not guarantee good method, even where the size of Real data is Mobile commerce unlimited. In addition to searching expected Goods objects for Users, it becomes necessary to develop a recommendation service based on XML. In this paper, we design the optimized XML Recommender product data. Efficient XML data preprocessing is required, include of formatting, structural, and attribute representation with dependent on User Profile Information. Our goal is to find a relationship among user interested products from E-Commerce and M-Commerce to XDB. Firstly, analyzing user profiles information. In the result creating clusters with analyzed user profile such as with set of sex, age, job. Secondly, it is clustering XML data which are associative products classify from user profile in shopping mall. Thirdly, after composing categories and goods data in which associative objects exist from the first clustering, it represent categories and goods in shopping mall and optimized clustering XML data which are personalized products. The proposed personalized user profile clustering method has been designed and simulated to demonstrate it's efficient.

• Journal of Information Processing Systems
• /
• v.13 no.5
• /
• pp.1299-1319
• /
• 2017

This paper presents a scalable multiple camera collaboration strategy for active tracking applications in large areas. The proposed approach is based on distributed mechanism but emulates the master-slave mechanism. The master and slave cameras are not designated but adaptively determined depending on the object dynamic and density distribution. Moreover, the number of cameras emulating the master is not fixed. The collaboration among the cameras utilizes global and local sectors in which the visual correspondences among different cameras are determined. The proposed method combines the local information to construct the global information for emulating the master-slave operations. Based on the global information, the load balancing of active tracking operations is performed to maximize active tracking coverage of the highly dynamic objects. The dynamics of all objects visible in the local camera views are estimated for effective coverage scheduling of the cameras. The active tracking synchronization timing information is chosen to maximize the overall monitoring time for general surveillance operations while minimizing the active tracking miss. The real-time simulation result demonstrates the effectiveness of the proposed method.