• 전자공학회논문지
• /
• 제51권11호
• /
• pp.146-154
• /
• 2014

텍스쳐 분류에 사용되는 방식 중 하나인 지역적 이진화 패턴은 일반적으로 영상 내의 평탄한 부분, 에지, 코너의 분포를 사용한다. 그러나 영상이 가지는 방향성을 고려하지 않고, 단순히 크고 작음만을 비교하는 지역적 이진화 패턴의 특성때문에 화소간 차이를 반영하지 못하는 문제점이 있다. 또한 영상의 분포를 사용하기 때문에 작은 크기의 영상에 대해서는 분류 성능이 저하된다. 이런 문제를 해결하기 위해 본 논문에서는 영상의 방향성 분포와 고유치 행렬을 이용한 세부 분류 기법을 제안한다. 지역적 이진화 패턴으로 초기 분류에서 누락된 텍스쳐 영상에 대하여 두 가지 특징을 이용하여 세부적으로 분류한다. 첫째, 영상이 가질 수 있는 방향을 여덟 가지로 양자화하고 그 방향들의 분포를 계산한다. 둘째, 구조 행렬을 이용하여 나온 고유치 중 큰 값의 분포를 구한다. 모의 실험을 통해 지역적 이진화 패턴만을 사용하였을 때 대비 제안 방법이 약 8% 정도 분류 정확도가 향상됨을 보였다.

• 한국정보과학회논문지:소프트웨어및응용
• /
• 제29권7호
• /
• pp.487-497
• /
• 2002

컴포넌트 기반의 재사용 기술이 소개되면서 소프트웨어 컴포넌트의 유통이 인터넷을 통한 온라인 기반의 유통 형태로 변하게 되었다. 이를 위해서는 유통 모델이 필요하며, 유통 시스템의 구축이 이루어져야 한다. 더욱이 유통 시스템이 효율적으로 운영되기 위해서는 컴포넌트들을 효율적으로 관리, 검색하기 위한 분류 체계가 마련되어야 한다. 본 논문은 이러한 유통 시스템 구축에 필요한 컴포넌트 분류 체계를 6가지 관점을 기반으로 한 컴포넌트 분류체계를 제시하고 BNF 표기법을 이용하여 명세한다. 제시된 분류체계의 효율성을 검증하고 기존의 분류체계들과 비교하기 위해 개발된 컴포넌트들을 적용하여 적중율과 정확도를 측정하여 실험 및 평가한다. 본 논문에서 제시한 기법이 기존의 분류기법에 비해서 여러 각도에서 분류하기 때문에 컴포넌트의 검색이나 등록이 효율적으로 이루어질 수 있도록 한다.

• Communications for Statistical Applications and Methods
• /
• 제19권4호
• /
• pp.629-638
• /
• 2012

We propose some methods to obtain optimal thresholds and classification functions by using various cutoff criterion based on the bivariate ROC curve that represents bivariate cumulative distribution functions. The false positive rate and false negative rate are calculated with these classification functions for bivariate normal distributions.

• 대한전기학회논문지:전기물성ㆍ응용부문C
• /
• 제49권1호
• /
• pp.19-23
• /
• 2000

Partial discharges(PD)in air insulated electric power apparatus often lead to deterioration of solid insulation by electron bombardments and electrochemical reaction. The PD caused to reduce the life time of power apparatus and to increase power losses. Thus understanding and classification of PD patterns in air are very important to discern sources of PD. In this paper, PD in air by using statistical methods was investigated. We classified air discharges, corona, surface discharges and cavity discharges by Kohonen network. For classification of PD patterns, we used statistical operators and parameters such as skewness$(S^+,\; S^-),\; kurtosis(K^+, K^-),\; mean phase(AP^+, AP^-)$, cross-correlation factor(CC) and asymmetry derived from the mean pulse-height phase distribution$(H_{avg}(\phi))$, the max pulse-height phase distribution $(H_{qmax}(\phi))$, the pulse count phase distribution $(H_n(\phi))$ and the pulse height vs. Repetition rate $(H_q(n))$.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제17권12호
• /
• pp.3383-3397
• /
• 2023

Scene graphs serve as semantic abstractions of images and play a crucial role in enhancing visual comprehension and reasoning. However, the performance of Scene Graph Generation is often compromised when working with biased data in real-world situations. While many existing systems focus on a single stage of learning for both feature extraction and classification, some employ Class-Balancing strategies, such as Re-weighting, Data Resampling, and Transfer Learning from head to tail. In this paper, we propose a novel approach that decouples the feature extraction and classification phases of the scene graph generation process. For feature extraction, we leverage a transformer-based architecture and design an adaptive calibration function specifically for predicate classification. This function enables us to dynamically adjust the classification scores for each predicate category. Additionally, we introduce a Distribution Alignment technique that effectively balances the class distribution after the feature extraction phase reaches a stable state, thereby facilitating the retraining of the classification head. Importantly, our Distribution Alignment strategy is model-independent and does not require additional supervision, making it applicable to a wide range of SGG models. Using the scene graph diagnostic toolkit on Visual Genome and several popular models, we achieved significant improvements over the previous state-of-the-art methods with our model. Compared to the TDE model, our model improved mR@100 by 70.5% for PredCls, by 84.0% for SGCls, and by 97.6% for SGDet tasks.

• Journal of Biosystems Engineering
• /
• 제37권4호
• /
• pp.215-224
• /
• 2012

Purpose: This study was conducted to classify the energy efficiency of 131 tractor models tested during from 2006 to 2010 in Korea. Methods: Four sub-indexes were developed using the fuel consumptions at 60% and 90% of rated speed with partial loads and at pull speeds of 3.0 km/h and 7.5 km/h with maximum drawbar pull. Weighting factors of the sub-indexes were also considered to reflect the characteristics of tractor's actual working hours in Korea. Four sub-indexes were integrated into a classification index. Using the developed classification index, a five-classification system was made on the basis of normal distribution of tractors over the classification range. Percentage of $1^{st}$ grade interval was expected to be close to 15%, $2^{nd}$ grade 20%, $3^{rd}$ grade 30%, $4^{th}$ grade 20%, $5^{th}$ grade 15%. Results: Number of $1^{st}$ grade was 21, $2^{nd}$ grade 23, $3^{rd}$ grade 39, $4^{th}$ grade 33, $5^{th}$ grade 15 among 131 models. Conclusions: Classification index was developed by integrating four sub-indexes. By the classification method using developed index, distribution of classified tractors was acceptable for practical application.

• 환경영향평가
• /
• 제11권4호
• /
• pp.247-258
• /
• 2002

This study aimed at verifying the grain-sized distribution of surface deposits in a tidal flat using multi-spectral Landsat TM. In this study, we employed the grain-sized analysis, PCA and unsupervised classification techniques for analyzing the distribution of deposits. As a result in this study, the unsupervised classification method using PCA image was found to be most useful in classifying tidal flat deposits using satellite data. This method is considerably effective in analyzing not only the aspects of distribution in terms of accumulated deposits and erosion, but also the changes in seaside topography and shoreline. The grain-sized distribution analysis indicates that the mud flat inside the Cheonsu-bay tidal flat is distributed, the mixed flat located in the middle, and the sand flat distributed near the sea. The sand flat is dominant around the southern part of Seomot isle and its beach. On the other hand, the mud and mixed flat is dominant on the western part. Likewise, the western coast of Seomot isle and its beach is significantly affected by waves facing the offshore. However, the eastern side of the bay could be a site for the evolution of tidal flat made of fine materials where it is less affected by ocean waves. These results show that multi-spectral satellite data are effective for the classification of distribution materials and environmental impact assessment and continuous monitoring. In particular, the research on environmental deposits can provide important decision-supporting information for decision-making on seaside development, by analyzing the progress of deposits and environmental changes.

• 한국약용작물학회지
• /
• 제21권4호
• /
• pp.268-275
• /
• 2013

This text was analyzed and investigated the distribution of medicinal plants in Cheongoksan Bonghwagun Korea, in order to search the medicinal resources that are used in modern medicine. Medicinal plants of the Korean Pharmacopoeia (10th edition) distributed in Cheongoksan Bonghwagun were consisted of 93 taxa ; 82 species, 10 varieties, 1 forma of 79 genus, 50 families. In medicinal plants of the Korean Pharmacopoeia, rate of native species and exotic species was 89.2% (83 taxa) and 10.8% (10 taxa) respectively. Family classification was the most of compositae of 8 taxa, and life form classification was most of herb of hemicryptophyte species. The classification by using parts were 34 taxa of root use and the classification of efficacy utilization was 24 taxa of Cheongyeolyak (heat-clearing drug) use.

• 대한전기학회논문지:시스템및제어부문D
• /
• 제54권6호
• /
• pp.383-389
• /
• 2005

This paper proposes a pattern classification algorithm for the defect distribution of Hard Disk Drive (HDD). In the HDD production, the defect pattern of defective HDD set is important information to diagnosis of defective HDD set. In this paper, 5 characteristics are determined for the classification to six standard defect pattern classes. A fuzzy inference system is proposed, the inputs of which are 5 characteristic values and the outputs are the possibilities that the input pattern is classified to standard patterns. Therefore, classification result is the pattern with maximum possibility. The proposed algorithm is implemented with the PC system for defective HDD sets and shows its effectiveness.

• 한국전기전자재료학회:학술대회논문집
• /
• 한국전기전자재료학회 1999년도 춘계학술대회 논문집
• /
• pp.704-707
• /
• 1999

Partial discharge(PD) in air insulated electric power systems is responsible for considerable power lossesfrom high voltage transmission lines. PD in air often leads to deterioration of insulation by the combined action of the discharge ions bombarding the surface and the action of chemical compounds that are formed by the discharge and may give rise to interference in ommunication systems. PD can indicate incipient failure. Thus understanding and classification of PD in air is very important to discern source of PD. In this paper, we investigated PD in air by using statical method. We classified air discharge with corona, surface discharge and cavity discharge by source of discharge. we used the mean pulse-height phase distribution $H_{qmean}(\psi)$, the max pulse-height phase distribution $H_{qmax}(\psi)$ , the pulse count phase distribution $H_n(\psi)$ and the max pulse height vs. repetition rate $H_{q}(n)$ for analysis PD pattern. We used statistical operators, such as skewness(S+. S-1, kurtosis(K+, K-), mean phase(AP+. AP-), cross-correlation factor(CC) and asymmetry from the distribution.