Heegon Kim

CPU-GPU hybrid computing for feature extraction from video stream

In this paper, we propose a way to distribute the video analytics workload into both the CPU and GPU, with a performance prediction model including characteristics of feature extraction from the video stream data. That is, we estimate the total execution time of a CPU-GPU hybrid computing system with the performance prediction model, and determine the optimal workload ratio and how to use the CPU cores for the given workload. Based on experimental results, we confirm that our proposed method can improve the speedups of three typical workload distributions: CPU-only, GPU-only, or CPU-GPU hybrid computing with a 50:50 workload ratio.

Energy efficient image/video data transmission on commercial multi-core processors.

In transmitting image/video data over Video Sensor Networks (VSNs), energy consumption must be minimized while maintaining high image/video quality. Although image/video compression is well known for its efficiency and usefulness in VSNs, the excessive costs associated with encoding computation and complexity still hinder its adoption for practical use. However, it is anticipated that high-performance handheld multi-core devices will be used as VSN processing nodes in the near future. In this paper, we propose a way to improve the energy efficiency of image and video compression with multi-core processors while maintaining the image/video quality. We improve the compression efficiency at the algorithmic level or derive the optimal parameters for the combination of a machine and compression based on the tradeoff between the energy consumption and the image/video quality. Based on experimental results, we confirm that the proposed approach can improve the energy efficiency of the straightforward approach by a factor of 2∼5 without compromising image/video quality.

Depth-Based Detection of Standing-Pigs in Moving Noise Environments

In a surveillance camera environment, the detection of standing-pigs in real-time is an important issue towards the final goal of 24-h tracking of individual pigs. In this study, we focus on depth-based detection of standing-pigs with “moving noises”, which appear every night in a commercial pig farm, but have not been reported yet. We first apply a spatiotemporal interpolation technique to remove the moving noises occurring in the depth images. Then, we detect the standing-pigs by utilizing the undefined depth values around them. Our experimental results show that this method is effective for detecting standing-pigs at night, in terms of both cost-effectiveness (using a low-cost Kinect depth sensor) and accuracy (i.e., 94.47%), even with severe moving noises occluding up to half of an input depth image. Furthermore, without any time-consuming technique, the proposed method can be executed in real-time.

Detection of Low-Weight Pigs by Using a Top-View Camera

Caring weaning pigs is important in the management of a group-housed pig farm. In this study, we propose an automatic method for detecting low-weight pigs in a pigsty. We install a top-view camera in a room of weaning pigs to detect the motion area of each pig from the video obtained. Then, we automatically detect a low-weight pig by comparing the size of each pig. Based on the experimental results, we confirm that the proposed method can automatically detect relatively low-weight pigs without any manual inspection or measurement of actual weight by a farm administrator.

Image Segmentation of Adjoining Pigs Using Spatio-Temporal Information

Recently, automatic video monitoring of individual pigs is emerging as an important issue in the management of group-housed pigs. Although a rich variety of studies have been reported on video monitoring techniques in intensive pig farming, it still requires further elaboration. In particular, when there exist adjoining pigs in a crowd pig room, it is necessary to have a way of separating adjoining pigs from the perspective of an image processing technique. In this paper, we propose an efficient image segmentation solution using both spatio-temporal information and region growing method for the identification of individual pigs in video surveillance systems. The experimental results with the videos obtained from a pig farm located in Sejong illustrated the efficiency of the proposed method.

Parallel Processing of Multimedia Data in a Heterogeneous Computing Environment

Recently, many multimedia applications can be parallelized by using multicore platforms such as CPU and GPU. In this paper, we propose a parallel processing approach for a multimedia application by using both CPU and GPU. Instead of distributing the parallelizable workload to either CPU or GPU(i.e., homogeneous computing), we distribute the workload simultaneously into both CPU and GPU(i.e., heterogeneous computing) by using OpenCL. Based on the experimental results with a photomosaic application, we confirm that the proposed parallel processing approach can provide better performance than the typical parallel processing approach by utilizing the given resource maximally.

Investigation of the mechanism of chromium removal in (3-aminopropyl)trimethoxysilane functionalized mesoporous silica

We are proposed that a possible mechanism for Cr(VI) removal by functionalized mesoporous silica. Mesoporous silica was functionalized with (3-aminopropyl)trimethoxysilane (APTMS) using the post-synthesis grafting method. The synthesized materials were characterized using transmission electron microscopy (TEM), X-ray diffraction (XRD), N2 adsorption-desorption analysis, Fourier-transform infrared (FT-IR), thermogravimetric analyses (TGA), and X-ray photoelectron spectroscopy (XPS) to confirm the pore structure and functionalization of amine groups, and were subsequently used as adsorbents for the removal of Cr(VI) from aqueous solution. As the concentration of APTMS increases from 0.01 M to 0.25 M, the surface area of mesoporous silica decreases from 857.9 m2/g to 402.6 m2/g. In contrast, Cr(VI) uptake increases from 36.95 mg/g to 83.50 mg/g. This indicates that the enhanced Cr(VI) removal was primarily due to the activity of functional groups. It is thought that the optimum concentration of APTMS for functionalization is approximately 0.05 M. According to XPS data, NH3+ and protonated NH2 from APTMS adsorbed anionic Cr(VI) by electrostatic interaction and changed the solution pH. Equilibrium data are well fitted by Temkin and Sips isotherms. This research shows promising results for the application of amino functionalized mesoporous silica as an adsorbent to removal Cr(VI) from aqueous solution.

Segmentation of Touching Objects by using Motion Information

It is important to segment and track objects automatically in many monitoring applications. When the objects as monitored are close each other, however, it is challenging to segment each object from the touching group. Especially, if the number of touching objects is large, the segmentation accuracy degrades significantly. In this paper, we propose a method of reducing the number of touching objects which should be separated. We first detect objects by using the color information and then extract the motion information of a touching group by using GMM. By excluding the non-moving objects from the touching group, we can reduce the number of touching objects which should be separated. The experimental results show that the proposed method can exclude the non-moving objects from a touching group and thus make the segmentation problem of touching objects liable to be solved.

Automatic Identification of a Coughing Animal using Audio and Video Data

Automatic detection of individual animal’s health condition based on a 24-hour monitoring system is important for efficient management of a livestock farm. In particular, the wasting disease is highly contagious in the group-based environment. The early detection of wasting disease is very important in order to minimize possible damage to the farm. In this paper, we propose a method to detect the wasting disease automatically by using both audio and video data. When a cough sound is detected with audio analysis, the Motion History Image-based video analysis will identify the coughing animal. By computing the change of MHI motion and the moving distance of each moving animal, the coughing animal can be identified. Based on the experimental results, we confirm that the Motion History Image-based method can discriminate the shaking motion of a coughing from other movement motions with careful motion pattern analysis.

Efficient Workload Distribution of Photomosaic Using OpenCL into a Heterogeneous Computing Environment

ABSTRACT Recently, parallel processing methods with accelerator have been introduced into a high performance computing and a mobile computing. The photomosaic application can be parallelized by using inherent data parallelism and accelerator. In this paper, we propose a way to distribute the workload of the photomosaic application into a CPU and GPU heterogeneous computing environment. That is, the photomosaic application is parallelized using both CPU and GPU resource with the asynchronous mode of OpenCL, and then the optimal workload distribution rate is estimated by measuring the execution time with CPU-only and GPU-only distribution rates. The proposed approach is simple but very effective, and can be applied to parallelize other applications on a CPU and GPU heterogeneous computing environment. Based on the experimental results, we confirm that the performance is improved by 141% into a heterogeneous computing environment with the optimal workload distribution compared with using GPU-only method.Keywords:Heterogeneous Computing, OpenCL, Photomosaic