• Korean Journal of Remote Sensing
• /
• v.33 no.4
• /
• pp.423-436
• /
• 2017

This paper presents a deep learning-based road segmentation framework from very high-resolution orthophotos. The proposed method uses Deep Convolutional Autoencoders for end-to-end mapping of orthophotos to road segmentations. In addition, a set of post-processing steps were applied to make the model outputs GIS-ready data that could be useful for various applications. The optimization of the model's parameters is explained which was conducted via grid search method. The model was trained and implemented in Keras, a high-level deep learning framework run on top of Tensorflow. The results show that the proposed model with the best-obtained hyperparameters could segment road objects from orthophotos at an average accuracy of 88.5%. The results of optimization revealed that the best optimization algorithm and activation function for the studied task are Stochastic Gradient Descent (SGD) and Exponential Linear Unit (ELU), respectively. In addition, the best numbers of convolutional filters were found to be 8 for the first and second layers and 128 for the third and fourth layers of the proposed network architecture. Moreover, the analysis on the time complexity of the model showed that the model could be trained in 4 hours and 50 minutes on 1024 high-resolution images of size $106{\times}106pixels$, and segment road objects from similar size and resolution images in around 14 minutes. The results show that the deep learning models such as Convolutional Autoencoders could be a best alternative to traditional machine learning models for road segmentation from aerial photographs.

• Transactions of the Korean Society of Automotive Engineers
• /
• v.11 no.1
• /
• pp.128-136
• /
• 2003

In today's design trend of vehicle structure, crush zone is fiequently reinforced by adding a box-shaped sub-frame in order to avoid an excessive deformation against a high-speed offset barrier such as EU Directive 96/97 EC, IIHS offset test. That kind of vehicle structure design results in a relatively monotonic crash pulse for airbag ECU(Electronic Control Unit) located at non-crush zone. As for an angular crash event, the measured crash signal using a single-axis accelerometer in a longitudinal direction is usually weaker than that of frontal barrier crash. Therefore, it is not so easy task to achieve a satisfactory crash discrimination performance for offset and angular crash events. In this paper, we introduce a new crash discrimination algorithm using an electronic X-Y 2-axis accelerometer in order to improve crash discrimination performance especially for those crash events. The proposed method uses a crash signal in lateral direction(Y-axis) as well as in longitudinal direction(X-axis). A crash severity measure obtained from Y-axis acceleration is used to improve the discrimination between fire and no-fire events. The result obtained by the proposed measure is logically ORed with an existing algorithm block using X-axis crash signal. Simulation and pulse injection test have been conducted to verify the performance of proposed algorithm by using real crash data of a 2,000cc passenger vehicle.

• KIISE Transactions on Computing Practices
• /
• v.23 no.8
• /
• pp.498-503
• /
• 2017

SIMD (Single Instruction Multiple Data) is a representative parallelization architecture that processes multiple data loaded in a SIMD register with a single instruction. Quicksort is a sorting algorithm that picks an element as a pivot from the array and reorders the array such that all elements having the values less than the pivot value are located in the left side on the pivot as well as all elements having the value greater than the pivot value are located in the right side on the pivot and then the algorithm performs the same task on both sublist recursively. In this paper, we propose an efficient Quicksort algorithm applying the SIMD instructions which minimally invokes conditional branches to avoid the performance degradation incurred by branch misprediction in a pipeline architecture. In addition, we improve the performance of the Quicksort algorithm by fetching data into a SIMD register as a byte unit to apply VBP (Vertical Bit Parallel) and the early pruning technique.

• Journal of the Korean Chemical Society
• /
• v.61 no.3
• /
• pp.104-111
• /
• 2017

In this study, we analyzed verbal interactions in small group science learning using smart devices by the level of prior achievement. Four heterogeneous groups at a coed high school in Seoul participated. Verbal interactions during small group science learning were audio- and video-taped, transcribed, and analyzed. Verbal interactions were analyzed at the levels of a turn and an interaction unit. The results revealed that the frequencies of verbal interactions were high in task category, especially at information explanation, information question, and reflection on standards subcategories. Furthermore, the frequencies of high-level students at direction explanation, reflection on standards and progress subcategories were higher than those of low-level students, and the frequencies of low-level students at direction question and information explanation subcategories were higher than their counterpart. In the analyses of the interaction units, the frequencies in symmetric elaborated interaction were high, especially at cumulative and evaluative subcategories.

• Korean Journal of Occupational Health Nursing
• /
• v.17 no.2
• /
• pp.216-223
• /
• 2008

Purpose: This study was performed to evaluate the worker's job stress status in the workplace of a local area. Method: Data were collected from October to December, 2007. The subjects were 208 workers at 2 work sites in Busan Metropolitan area, who were examined using Job-Strain-Model Questionnaire. Data were analyzed by SPSS 12.0 Win Program to get the percentage, mean, standard deviation, t-test, ${\chi}^2$-test, ANOVA. Results: The results of this study were as follows: In the mean sub-factors job stress level, the mean of job demand was $28.7{\pm}4.4$(median 29.0), the mean of job discretion was $54.7{\pm}8.2$(median 54.0), the mean of social support was $21.8{\pm}2.9$(range:8-32). This study's subjects were appeared as active group with relatively higher score of job demand and job discretion than the average value of those. There was no statistically significant difference of general characteristics among the different job strain groups. There was statistically significant difference with of social supports among the different job strain groups. Conclusion: In conclusion, the subjects of this study's were active group. Thus, it is suggested that it is be necessary to repeated the education of the job task work for active group with high score of job demand and job discretion.

• Journal of the Korean Home Economics Association
• /
• v.38 no.8
• /
• pp.13-27
• /
• 2000

The background of single animal figures was originated from those northern mounted nomadic groups, which was Scythe style Culture. The art of the nomads working in the Scythian idiom was small in size and essentially decorative in intention, yet practically every object which can be associated with any unit in this group of people possesses many of the attributes essential to a real work of art. Clarity of conception, purity of form, co-ordination of rhythm and balance, and not least, an understanding and respect for the material employed were triumphantly blended by the Eurasian nomads to produce a distinctive style. In Scythian art the multitude of animal representations well illustrates the reoccupation of this nomadic people with animals in their environment. Usually only wild animals are represented. Commonly depicted are: stags and deer, lions or other large cats, eagles, birds heads (perhaps of ravens), griffins, snakes, hares, fish, goats, rams, boars, moose (elk), yak, sheep and bears. The occasional exception to the wild animal rule is domesticated horses-important because the Scythians were horse bleeders and their whole culture revolved around their dependence on the horse. The nomads had little reason to create object in honour of gods or men, but they had an instinct for beauty and the wish to surround themselves with the animal forms in which they had come to delight The Scytians tried to combine in a single rendering all the salient points of the animal they were delineating. They archived considerable success in the difficult task of showing in a single image the various and often incompatible poses assumed by a single animal in the course of its life. Zoomorphic motifs were used not simple for decorative effect, but to trim the object into amulets, with magical power to assist in hunting, and to protect the owner from harm.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.16 no.3
• /
• pp.605-614
• /
• 2012

Nowadays u-healthcare which is very sensitive to the character of user's information among other ubiquitous computing field is popular in medical field. u-healthcare deals extremely personal information including personal health/medical information so it is exposed to various weaknees and threats in the part of security and privacy. In this paper, RFID based patient's information protecting protocol that prevents to damage the information using his or her mobile unit illegally by others is proposed. The protocol separates the authority of hospital(doctor, nurse, pharmacy) to access to patient's information by level of access authority of hospital which is registered to management server and makes the hospital do the minimum task. Specially, the management server which plays the role of gateway makes access permission key periodically not to be accessed by others about unauthorized information except authorized information and improves patient's certification and management.

• Smart Media Journal
• /
• v.4 no.4
• /
• pp.56-63
• /
• 2015

Modern GPU can execute mass parallel computation by exploiting many GPU core. GPGPU architecture, which is one of approaches exploiting outstanding computational resources on GPU, executes general-purpose applications as well as graphics applications, effectively. In this paper, we investigate the impact of memory-efficiency and performance according to number of CTAs(Cooperative Thread Array) on a SM(Streaming Multiprocessors), since the analysis of relation between number of CTA on a SM and them provides inspiration for researchers who study the GPU to improve the performance. Our simulation results show that almost benchmarks increasing the number of CTAs on a SM improve the performance. On the other hand, some benchmarks cannot provide performance improvement. This is because the number of CTAs generated from same kernel is a little or the number of CTAs executed simultaneously is not enough. To precisely classify the analysis of performance according to number of CTA on a SM, we also analyze the relations between performance and memory stall, dram stall due to the interconnect congestion, pipeline stall at the memory stage. We expect that our analysis results help the study to improve the parallelism and memory-efficiency on GPGPU architecture.

• Korean System Dynamics Review
• /
• v.16 no.3
• /
• pp.5-29
• /
• 2015

The importance of the concept of safety culture has increased in the security of high-risk facility after Chernobyl accident in 1986. This paper elaborated the concept of safety culture and its main factors by Causal Loop Diagram. Due to the decline of safety culture, the occurrence of incidents and accidents require more and more corrective actions to the members of high-risk facilities and thereby increasing their workloads. Employees who must complete the task within the given time have to have time pressures and don't comply with the rules and procedures. Also, a schedule pressure is a big stress for employees, causing mistakes in precision work. In order to improve these problems, CLD of the safety culture in this paper suggests hiring more workers, re-allocation of given workloads and strengthen the learning, communication capabilities and safety leadership. In addition, the two real accident cases were analyzed to test the feasibility of the System Dynamic simulation model through the process of structuring the fault trees on the stationary black out accident in Kori unit 1 in South Korea and Kleen Energy power station explosion in US. The simulation results show that the various safety factors cause the serious accident combined with mechanical failure and safety culture will reduce the possibility of the accidents in these high-risk organizations. This simulation model can contribute to analyzing the impact of the organizational and human factors of safety culture and can provide the alternatives in high-risk facilities.

• Smart Structures and Systems
• /
• v.6 no.8
• /
• pp.939-951
• /
• 2010

A low-cost wireless sensor network (WSN) solution with highly expandable super and simple nodes was developed. The super node was designed as a sensing unit as well as a receiving terminal with low energy consumption. The simple node was designed to serve as a cheaper alternative for large-scale deployment. A 12-bit ADC inputs and DAC outputs were reserved for sensor boards to ease the sensing integration. Vibration and thermal field tests of the Chi-Lu Bridge were conducted to evaluate the WSN's performance. Integral acceleration, temperature and tilt sensing modules were constructed to simplify the task of long-term environmental monitoring on this bridge, while a star topology was used to avoid collisions and reduce power consumption. We showed that, given sufficient power and additional power amplifier, the WSN can successfully be active for more than 7 days and satisfy the half bridge 120-meter transmission requirement. The time and frequency responses of cables shocked by external force and temperature variations around cables in one day were recorded and analyzed. Finally, guidelines on power characterization of the WSN platform and selection of acceleration sensors for structural health monitoring applications were given.