• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.8 no.2
• /
• pp.483-503
• /
• 2014

In this paper, human action recognition using pyramid histograms of oriented gradients and collaborative multi-task learning is proposed. First, we accumulate global activities and construct motion history image (MHI) for both RGB and depth channels respectively to encode the dynamics of one action in different modalities, and then different action descriptors are extracted from depth and RGB MHI to represent global textual and structural characteristics of these actions. Specially, average value in hierarchical block, GIST and pyramid histograms of oriented gradients descriptors are employed to represent human motion. To demonstrate the superiority of the proposed method, we evaluate them by KNN, SVM with linear and RBF kernels, SRC and CRC models on DHA dataset, the well-known dataset for human action recognition. Large scale experimental results show our descriptors are robust, stable and efficient, and outperform the state-of-the-art methods. In addition, we investigate the performance of our descriptors further by combining these descriptors on DHA dataset, and observe that the performances of combined descriptors are much better than just using only sole descriptor. With multimodal features, we also propose a collaborative multi-task learning method for model learning and inference based on transfer learning theory. The main contributions lie in four aspects: 1) the proposed encoding the scheme can filter the stationary part of human body and reduce noise interference; 2) different kind of features and models are assessed, and the neighbor gradients information and pyramid layers are very helpful for representing these actions; 3) The proposed model can fuse the features from different modalities regardless of the sensor types, the ranges of the value, and the dimensions of different features; 4) The latent common knowledge among different modalities can be discovered by transfer learning to boost the performance.

• ETRI Journal
• /
• v.38 no.6
• /
• pp.1114-1123
• /
• 2016

Multi-view video plus depth (MVD) has been widely used owing to its effectiveness in three-dimensional data representation. Using MVD, color videos with only a limited number of real viewpoints are compressed and transmitted along with captured or estimated depth videos. Because the synthesized views are generated from decoded real views, their original reference views do not exist at either the transmitter or receiver. Therefore, it is challenging to define an efficient metric to evaluate the quality of synthesized images. We propose a novel metric-the reduced-reference quality metric. First, the effects of depth distortion on the quality of synthesized images are analyzed. We then employ the high correlation between the local depth distortions and local color characteristics of the decoded depth and color images, respectively, to achieve an efficient depth quality metric for each real view. Finally, the objective quality metric of the synthesized views is obtained by combining all the depth quality metrics obtained from the decoded real views. The experimental results show that the proposed quality metric correlates very well with full reference image and video quality metrics.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.2 no.4
• /
• pp.33-39
• /
• 2003

The stiffness of multi-articulated industrial robots is very weak, because their structure is an articulated type with some links and joints. Thus it is known that cutting processes for metals are not accepted in machine shop well, but they have a lot of merits for cutting processes, for example, drilling, tapping, and engraving etc, because of the characteristics of their high degree of freedom. The temptation of cutting aluminium was carried out to investigate the feasibility and the limitations or constraints for cutting metals on them. First the mode shapes of 6 degree of freedom welding robot were taken and analysed, and next the cutting processes were practically carried out on it. The results of study were found out to show the feasibility of cutting processes at drilling under 6mm of tool diameter, as well as to have some limitations and constraints, for examples, feed rate, depth of cut and cutting force etc.

• Journal of Broadcast Engineering
• /
• v.12 no.4
• /
• pp.333-341
• /
• 2007

Multi-view video has recently gained much attraction from academic and commercial fields because it can deliver the immersive viewing of natural scenes. This paper presents multi-view video processing being composed of intermediate view reconstruction (IVR), graphics composition, and multi-view video viewer. First we generate virtual views between multi-view cameras using depth and texture images of the input videos. Then we mix graphic objects to the generated view images. The multi-view video viewer is developed to examine the reconstructed images and composite images. As well, it can provide users with some special effects of multi-view video. We present experimental results that validate our proposed method and show that graphic objects could become the inalienable part of the multi-view video.

• Journal of Korean Society for Atmospheric Environment
• /
• v.34 no.5
• /
• pp.735-747
• /
• 2018

Rainfall depth is an important meteorological information. Generally, high spatial resolution rainfall data such as road-level rainfall data are more beneficial. However, it is expensive to set up sufficient Automatic Weather Systems to get the road-level rainfall data. In this paper, we propose to use deep learning to recognize rainfall depth from road surveillance videos. To achieve this goal, we collect a new video dataset and propose a procedure to calculate refined rainfall depth from the original meteorological data. We also propose to utilize the differential frame as well as the optical flow image for better recognition of rainfall depth. Under the Temporal Segment Networks framework, the experimental results show that the combination of the video frame and the differential frame is a superior solution for the rainfall depth recognition. The final model is able to achieve high performance in the single-location low sensitivity classification task and reasonable accuracy in the higher sensitivity classification task for both the single-location and the multi-location case.

• Proceedings of the Korean Geotechical Society Conference
• /
• 2005.10a
• /
• pp.405-416
• /
• 2005

The rubblization technique is breaking the aged concrete pavement slab into rubblized concrete aggregate, and use it as an base material at its original position, then builds overlay above the rubblized base. This method has been successively used in USA due to the advantage of good contructibility, cost-effectiveness as well as the capability of preventing of reflection cracks. However, constructibility and economic performance of rubblization on typical Korean concrete pavements needed to be investigate since to typical Korean concrete pavements have thick slab, as well include lean concrete subbase course. This stud explored optimum breaking depth and suggested minimum 10cm based on reflection crack simulation test. Also proper head shape and impact energy were investigated based on small breaking field tests. It was found that $127kg/cm^2$ of stress with 52.3% of head contact area are breaking requirement. Also, Multi-head type breaker suitable for Korean condition was designed and developed. This multi-head type breaker was designed to rubblize old concrete to the suggested optimum rubblized-depth and rubblized-concrete-aggregate size to prevent reflection crack and maintain high bearing capacity. This machine was used for the test of rubblization of old concrete pavement on a non-use old concrete and a in-serviced road. In these two tests, engineering properties of rubblized base and constructability and cost were investigated. In both tests, the old concrete rubblized to targeted size and depth, and high-level bearing capacity was achieved. Also, superior constructability and lower cost compared with traditional reconstruction was examined.

• Geophysics and Geophysical Exploration
• /
• v.8 no.1
• /
• pp.26-33
• /
• 2005

Inversion of multi-mode surface-wave phase velocity for shallow engineering site investigation has received much attention in recent years. A sensitivity analysis and inversion of both synthetic and field data demonstrates the greater effectiveness of this method over employing the fundamental mode alone. Perturbation of thickness and shear-wave velocity parameters in multi-modal Rayleigh wave phase velocities revealed that the sensitivities of higher modes: (a) concentrate in different frequency bands, and (b) are greater than the fundamental mode for deeper parameters. These observations suggest that multi-mode phase velocity inversion can provide better parameter discrimination and imaging of deep structure, especially with a velocity reversal, than can inversion of fundamental mode data alone. An inversion of the theoretical phase velocities in a model with a low velocity layer at 20 m depth can only image the soft layer when the first higher mode is incorporated. This is especially important when the lowest measurable frequency is only 6 Hz. Field tests were conducted at sites surveyed by borehole and PS logging. At the first site, an array microtremor survey, often used for deep geological surveying in Japan, was used to survey the soil down to 35 m depth. At the second site, linear multichannel spreads with a sledgehammer source were recorded, for an investigation down to 12 m depth. The f-k power spectrum method was applied for dispersion analysis, and velocities up to the second higher mode were observed in each test. The multi-mode inversion results agree well with PS logs, but models estimated from the fundamental mode alone show f large underestimation of the depth to shallow soft layers below artificial fill.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.26 no.1D
• /
• pp.81-87
• /
• 2006

The rubblization technique is breaking the aged concrete pavement slab into rubblized concrete aggregate, and use it as an base material at its original position, then builds overlay above the rubblized base. This method has been successively used in USA due to the advantage of good constructibility, cost-effectiveness as well as the capability of preventing of reflection cracks. However, constructibility and economic performance of rubblization on typical Korean concrete pavements needed to be investigate since to typical Korean concrete pavements have thick slab, as well include lean concrete subbase course. Multi-head type breaker suitable for Korean condition was designed and developed. This multi-head type breaker was designed to rubblize old concrete to the suggested optimum rubblized-depth and rubblized-concrete-aggregate size to prevent reflection crack and maintain high bearing capacity. This machine was used for the test of rubblization of old concrete pavement on a non-use old concrete and a in-serviced road. In these two tests, engineering properties of rubblized base and constructability and cost were investigated. In both tests, the old concrete rubblized to targeted size and depth, and high-level bearing capacity was achieved. Also, superior constructability and lower cost compared with traditional reconstruction was examined.

• Korean Institute of Interior Design Journal
• /
• v.21 no.4
• /
• pp.12-20
• /
• 2012

The aim of this study is to review phenomenological characteristics in Steven Holl's architecture through his design of St. Ignatius Chapel at Seattle University. To obtain this purpose, an analytical frame based on Holl's theory of phenomenological architecture is suggested to have a systematic study for St. Ignatius chapel. This chapel can be a good example of phenomenological architecture in that it is based on the concept of 'A Gathering of Different Lights' related to phenomenology and considered perception including multi sensory (as well as vision) as primary factors from site and program interpretation to spatial configuration. Unprecedented exterior of St. Ignatius chapel reflected on characteristics and function of rooms to magnify user's spatial experiences through inducing natural light and spatial effect. Holl used various openings and screen for natural light with colors to invoke religious inspiration. He also try to give spatial depth and multi foci for experiencing space through various ceiling forms. These phenomenological features originated in client's strong will as well as appropriateness of the function of facility's religious experiences through building to the purpose of phenomenological architecture.

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.49 no.2
• /
• pp.83-93
• /
• 2012

This paper proposes a method to efficiently encode multi-view color videos and depth maps. The proposed coding method for multi-view color videos and depth maps can improve the coding efficiency by additional inter-view prediction, as well as inter-frame prediction. By means of the proposed method, we achieved the coding gain of approximately 55% for 2-view color videos and approximately 12% for 2-view depth maps. For 3-view case, we found that the proposed system yields 54% of coding gain from outer view color videos and 56% of coding gain from center view color videos, respectively. Moreover, for 3-view depth map case, approximately 11% of coding gain from outer view and 13% of coding gain from center view are obtained with the proposed coder, respectively.