• Proceedings of the KIEE Conference
• /
• 2002.06a
• /
• pp.41-45
• /
• 2002

In industrial computer vision systems, the image quality is dependent on the parameters such as light source, illumination method, optics, and surface properties. Most of them are related with the lighting system, which is designed in heuristic, based on the designer's experimental knowledge, In this paper, a design code by which the optimal lighting method and light source for computer vision systems can be found are suggested based on experimental results, To prove the usefulness of the design code, it is applied to the lighting system design of the transistor marking inspection system and the results are presented.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.9 no.5
• /
• pp.163-168
• /
• 2009

Most image segmentation methods are to represent observed feature vectors at each pixel, which are assumed as appropriated probability models. These models can be used by statistical estimating or likelihood clustering algorithms of feature vectors. EM algorithms have some calculation problems of maximum likelihood for unknown parameters from incomplete data and maximum value in post probability distribution. First, the performance is dependent upon starting positions and likelihood functions are converged on local maximum values. To solve these problems, we mixed the Gausian function and histogram at all the level values at the image, which are proposed most suitable image segmentation methods. This proposed algoritms are confirmed to classify most edges clearly and variously, which are implemented to MFC programs.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.8 no.3
• /
• pp.95-103
• /
• 1998

This paper deals with a problem occuring in recognition of tactile images due to the effects of imposed force at a me urement moment. Tactile image of a contact surface, used for recognition of the surface type, varies depending on the forces imposed so that a false recognition may result in. This paper fuzzifies two parameters of the contour of a tactile image with the membership function formed by considering the imposed force. Two fuzzifed paramenters are fused by the average Minkowski's dist; lnce. The proposed algorithm was implemented on the multisensor system cnmposed of an optical tact le sensor and a 6 axes forceltorque sensor. By the experiments, the proposed algorithm has shown average recognition ratio greater than 869% over all imposed force ranges and object models which is about 14% enhancement comparing to the case where only the contour information is used. The pro- ~oseda lgorithm can be used for end-effectors manipulating a deformable or fragile objects or for recognition of 3D objects by implementing on multi-fingered robot hand.

• Journal of Broadcast Engineering
• /
• v.24 no.6
• /
• pp.919-927
• /
• 2019

In this paper, we propose an accurate camera calibration method for acquiring multiview stereoscopic images. Generally, camera calibration is performed by using checkerboard structured patterns. The checkerboard pattern simplifies feature point extraction process and utilizes previously recognized lattice structure, which results in the accurate estimation of relations between the point on 2-dimensional image and the point on 3-dimensional space. Since estimation accuracy of camera parameters is dependent on feature matching, accurate detection of checkerboard corner is crucial. Therefore, in this paper, we propose the method that performs accurate camera calibration method through accurate detection of checkerboard corners. Proposed method detects checkerboard corner candidates by utilizing 1-dimensional gaussian filters with succeeding corner refinement process to remove outliers from corner candidates and accurately detect checkerboard corners in sub-pixel unit. In order to verify the proposed method, we check reprojection errors and camera location estimation results to confirm camera intrinsic parameters and extrinsic parameters estimation accuracy.

• Management & Information Systems Review
• /
• v.12
• /
• pp.137-160
• /
• 2003

The association between accounting earnings and the stock price of an entity is the subject that has been most heavily researched during the past 25 years in accounting literature. Researcher's common finding is that there are positive relationships between accounting earnings and stock prices. However, the explanatory power of accounting earnings which was measured by $R^2$ of regression functions used was rather low. To be connected with these low results, The prior studies propose that there will be additional information, errors in variables. This study investigates empirically determinants of earnings response coefficients(ERCs), which measure the correlation between earnings and stock prices, using earnings level / change, as the dependent variable in the return/earnings regression. Specifically, the thesis tests whether the factors such as earnings persistence, growth, systematic risk, image, information asymmetry and firm size. specially, the determinable variables of ERC are explained in detail. The image / information asymmetry variables are selected to be connected with additional information stand point, The debt / growth variables are selected to be connected with errors in variables. In this study, The sample of firms, listed in Korean Stock Exchange was drawn from the KIS-DATA and was required to meet the following criteria: (1) Annual accounting earnings were available over the 1986-1999 period on the KIS-FAS to allow computation of variables parameter; (2) sufficient return data for estimation of market model parameters were available on the KIS-SMAT month returns: (3) each firm had a fiscal year ending in December throughout the study period. Implementation of these criteria yielded a sample of 1,141 firm-year observation over the 10-year(1990-1999) period. A conventional regression specification would use stock returns(abnormal returns) as a dependent variable and accounting earnings(unexpected earnings) changes interacted with other factors as independent variables. In this study, I examined the relation between other factors and the RRC by using reverse regression. For an empirical test, eight hypotheses(including six lower-hypotheses) were tested. The results of the performed empirical analysis can be summarized as follows; The first, The relationship between persistence of earnings and ERC have significance of each by itself, this result accord with one of the prior studies. The second, The relationship between growth and ERC have not significance. The third, The relationship between image and ERC have significance of each by itself, but a forecast code doesn't present. This fact shows that image cost does not effect on market management share, is used to prevent market occupancy decrease. The fourth, The relationship between information asymmetry variable and ERC have significance of each by. The fifth, The relationship between systematic risk$(\beta)$ and ERC have not significance. The sixth, The relationship between debt ratio and ERC have significance of each by itself, but a forecast code doesn't present. This fact is judged that it is due to the effect of financial leverage effect and a tendency of interest.

• Journal of Korean Society of Forest Science
• /
• v.96 no.5
• /
• pp.591-601
• /
• 2007

Spatial information on forest biomass is an important factor to evaluate the capability of forest as a carbon sequestrator and is a core independent variable required to drive models which describe ecological processes such as carbon budget, hydrological budget, and energy flow. The objective of this study is to understand the relationship between satellite image and field data, and to quantitatively estimate and map the spatial distribution of forest biomass. Landsat Enhanced Thematic Mapper (ETM+) derived vegetation indices and field survey data were applied to estimate the biomass distribution of mountainous forest located in Gwangneung Experimental Forest (230 ha). Field survey data collected from the ground plots were used as the dependent variable, forest biomass, while satellite image reflectance data (Band 1~5 and Band 7), Normalized Difference Vegetation Index (NDVI), Soil-Adjusted Vegetation Index (SAVI), and RVI (Ratio Vegetation Index) were used as the independent variables. The mean and total biomass of Gwangneung catchment area were estimated to be about 229.5 ton/ha and $52.8{\times}10^3$ tons respectively. Regression analysis revealed significant relationships between the measured biomass and Landsat derived variables in both of deciduous forest ($R^2=0.76$, P < 0.05) and coniferous forest ($R^2=0.75$, P < 0.05). However, there still exist many uncertainties in the estimation of forest ecosystem parameters based on vegetation remote sensing. Developing remote sensing techniques with adequate filed survey data over a long period are expected to increase the estimation accuracy of spatial information of the forest ecosystem.

• Applied Microscopy
• /
• v.28 no.3
• /
• pp.415-424
• /
• 1998

The dependency of long-range order parameter on the ordered structure of $Pb(Mg_{1/3}Nb_{2/3})O_3$ solid solutions has been investigated by means of computer simulations of high-resolution lattice images. The computer image simulations were performed by the multislice method in a wide range of sample thicknesses, defocusing values, and long-range order parameters. It was found that the lattice images of the ordered structures were predominantly dependent on the long-range order parameter, The lattice images in a complete ordered structure showed a pseudo-hexagonal pattern. As the order parameter decreases, the simulated images changed slowly from a pseudo-hexagonal pattern to a rectangular pattern. The lattice images in a complete disordered structure showed a rectangular pattern. Also, the simulated images of the $Pb(Mg_xNb_{1-x})O_3$ with different Mg:Nb ratios obtained at a given long-range order parameter were basically the same patterns, irrespective of Mg and Nb compositions.

• Asian Pacific Journal of Cancer Prevention
• /
• v.15 no.4
• /
• pp.1831-1835
• /
• 2014

Background: To investigate the relationship between extracellular matrix parameters and texture of prostatic lesions evaluated by transrectal real-time tissue elastography (TRTE). Methods: 120 patients suspicious for prostate cancer underwent TRTE. Targeted biopsies were carried out after 12-core systematic biopsy. Epithelia were stained with hematoxylin-eosin, and Victoria blue and Ponceau S were used to stain elastic-collagen fibers, and picric acid-sirius red for visualization of collagen type I (Col1) and III (Col3). Smooth muscles were visualized by immunohistochemistry. All image analyses were performed in a blind manner using Image Pro Plus 6.0, and the area ratios of epithelium, elastic fibers, collagen fibers and Col1/Col3 were determined. Results: 42 patients with typical elastograms were included in the final data analysis. Significant differences were detected between the benign and malignant groups in the area ratios of epithelium (P = 0.01), smooth muscles and Col1/Col3 (P = 0.04, P = 0.02, respectively). There were no significant differences in the area ratios of epithelium, smooth muscle and elastic fibers between the stiff and soft lesion groups. The area ratio of Col1 was ($0.05{\pm}0.03$) in the stiff group, and ($0.02{\pm}0.01$) in the soft group (P= 0.00). However, the area ratio of Col3 was ($0.03{\pm}0.02$) in the stiff group, and ($0.05{\pm}0.04$) in the soft group (P = 0.16). Col1/Col3 in the stiff group ($1.99{\pm}1.59$) was greater than in the soft group ($0.71{\pm}0.64$) (P = 0.01). Conclusions: Tissue hardness of prostatic tumors was mainly dependent on the Col1 content, Col1/Col3 being higher in malignant than in benign lesions, so the prostate tissue texture can be used as a target for distinguishing between the two with TRTE.

• Proceedings of the KSRS Conference
• /
• 2009.03a
• /
• pp.97-100
• /
• 2009

The surface parameters associated with the land are usually dependent on the climate, and many physical processes that are displayed in the image sensed from the land then exhibit temporal variation with seasonal periodicity. An adaptive feedback system proposed in this study reconstructs a sequence of images remotely sensed from the land surface having the physical processes with seasonal periodicity. The harmonic model is used to track seasonal variation through time, and a Gibbs random field (GRF) is used to represent the spatial dependency of digital image processes. In this study, the Normalized Difference Vegetation Index (NDVI) was computed for one week composites of the Advanced Very High Resolution Radiometer (AVHRR) imagery over the Korean peninsula for 1996 and 2000 using a dynamic technique, and the adaptive reconstruction of harmonic model was then applied to the NDVI time series for tracking changes on the ground surface. The results show that the adaptive approach is potentially very effective for continuously monitoring changes on near-real time.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.3
• /
• pp.89-100
• /
• 2023

Recently, on-device artificial intelligence (AI) solutions using mobile devices and embedded edge devices have emerged in various fields, such as computer vision, to address network traffic burdens, low-energy operations, and security problems. Although vision transformer deep learning models have outperformed conventional convolutional neural network (CNN) models in computer vision, they require more computations and parameters than CNN models. Thus, they are not directly applicable to embedded edge devices with limited hardware resources. Many researchers have proposed various model compression methods or lightweight architectures for vision transformers; however, there are only a few studies evaluating the effects of model compression techniques of vision transformers on performance. Regarding this problem, this paper presents a performance evaluation of vision transformers on embedded platforms. We investigated the behaviors of three vision transformers: DeiT, LeViT, and MobileViT. Each model performance was evaluated by accuracy and inference time on edge devices using the ImageNet dataset. We assessed the effects of the quantization method applied to the models on latency enhancement and accuracy degradation by profiling the proportion of response time occupied by major operations. In addition, we evaluated the performance of each model on GPU and EdgeTPU-based edge devices. In our experimental results, LeViT showed the best performance in CPU-based edge devices, and DeiT-small showed the highest performance improvement in GPU-based edge devices. In addition, only MobileViT models showed performance improvement on EdgeTPU. Summarizing the analysis results through profiling, the degree of performance improvement of each vision transformer model was highly dependent on the proportion of parts that could be optimized in the target edge device. In summary, to apply vision transformers to on-device AI solutions, either proper operation composition and optimizations specific to target edge devices must be considered.