• Journal of Korea Multimedia Society
• /
• v.11 no.11
• /
• pp.1615-1623
• /
• 2008

This paper proposes methods to classify scanned insect footprints. We propose improved SOM and ART2 algorithms for extracting segments, basic areas for feature extraction, and utilize Trace transform and fuzzy weighted mean methods for extracting feature values for classification of the footprints. In the proposed method, regions are extracted by a morphological method in the beginning, and then improved SOM and ART2 algorithms are utilized to extract segments regardless of kinds of insects. Next, A Trace transform method is used to find feature values suitable for various kinds of deformation of insect footprints. In the Trace transform method, Triple features from reconstructed combination of diverse functions, are used to classify the footprints. In general, it is very difficult to decide automatically whether the extracted footprint segment is meaningful for classification or not. So we use a fuzzy weighted mean method for not excluding uncertain footprint segments because the uncertain footprint segments may be possible candidates for classification. We present experimental results of footprint segment extraction and segment classification by the proposed methods.

• Journal of Broadcast Engineering
• /
• v.25 no.4
• /
• pp.569-575
• /
• 2020

The serial arc is one of factors causing electrical fires. Over past decades, various researches have been carried out to detect arc occurrences. Even though frequency analysis, wavelet, and statistical features have been used, additional steps such as transformation and feature extraction are required. On the contrary, deep learning models directly use the raw data without any feature extraction processes. Therefore, the usage of time-domain data is preferred, but the performance is not satisfactory. To solve this problem, subsequent 1-D signals are transformed into 2-D data that can feed into a convolutional neural network (CNN). Experiments validated that CNN model outperforms deep neural network (DNN) by the classification accuracy of 8.6%. In addition, data augmentation is utilized, resulting in the accuracy improvement by 14%.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.24 no.6
• /
• pp.622-627
• /
• 2014

This paper proposes a new method which improves recognition rate on the gait recognition system using wavelet transform, linear discriminant analysis and genetic algorithm. We use wavelet transform to obtain the four sub-bands from the gait energy image. In order to extract feature data from sub-bands, we use linear discriminant analysis. Distance values between training data and four sub-band data are calculated and four weights which are calculated by genetic algorithm is assigned at each sub-band distance. Based on a new fusion distance value, we conducted recognition experiments using k-nearest neighbors algorithm. Experimental results show that the proposed weight fusion method has higher recognition rate than conventional method.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2006.05a
• /
• pp.477-480
• /
• 2006

To represent the accurate feature of signal and system, many researches have been done in many fields of basic and engineering science which led a great development of modem society. Even until currently, in order to acquire useful information from signals at high speed, many methods and transforms have been processed. In these methods, the Fourier transform which represents signal as the combination of the frequency component has been applied to the most fields. But as transform not to consider time information, the Fourier transform does not provide time information of the time and presents only overall features of signals. The wavelet transform, which is proposed to overcome this problem and recently expands the range of the application, presents time-frequency localization and many kinds of the wavelet can be applied according to the environment of application. In this paper, we detect the features of signals using the function which is considered as the wavelet and do research for filtering locally in the wavelet plane.

• Proceedings of the Korean Institute of Communication Sciences Conference
• /
• 1994.06a
• /
• pp.367-370
• /
• 1994

• Proceedings of the Korean Information Science Society Conference
• /
• 2004.10b
• /
• pp.742-744
• /
• 2004

본 논문은 복잡한 배경에서의 얼굴 검출에 있어서 다중 특징 추출 데이터로 학습한 계단식 분류기에 의한 방법을 제안한다 얼굴 검출에서 얼굴의 패턴은 상당히 다양한 영상 표현으로 나타나기 때문에 하나의 특징 추출 방법은 사람의 얼굴을 모델링 하기에는 부족하다. 따라서 여기서는 얼굴의 전체적인 지역적인 특징을 나타내는 Subregion과, 얼굴의 주파수 특성에 따라 좀 더 세밀하고 다양한 속성들을 나타내는 Haar 웨이블릿 변환을 이용하여 다중으로 특징을 추출하여 효과적인 모델링을 시도하였다. 특징을 추출한 얼굴과 비얼굴의 패턴(pattern)을 구분하기 위해서 패턴들의 통계적인 특성을 이용하여 각 추출방법에 맞게 학습된 Bayesian 분류기를 직렬로 연결하여 사용하였으며 비얼굴은 얼굴과 유사한 비얼굴(face-like nonface) 패턴들을 사용하여 모델링 하였다. 제안한 얼굴 검출 방식의 성능은 MIT-CMU 시험 영상들을 이용하여 평가하였다. 그 결과 한 가지 특징 추출을 사용하는 것 보다 두 가지 특징 추출을 병행한 계단식 구성이 더 정확한 검출 결과를 나타내었다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2003.04c
• /
• pp.576-578
• /
• 2003

본 논문에서는 이미지 검색을 위한 새로운 검색 기법을 제시한다. 기존의 특징기반 검색 기법이나 주석기반 검색 기법은 특징이나 주석에 대하여 색인 형태나 질의 형태가 동일하였다. 그러나, 제안하는 검색 기법은 위의 두 전형적인 검색기법을 혼합한 것으로, 텍스트로 질의하면 질의 텍스트를 질의처리를 통해 텍스트에 포함된 특징용어를 추출하고 특징용어를 이미지가 본질적으로 가지는 특징(color, shape, texture)으로 변환한 다음 그 특징을 질의로 이용하여 특징기반 검색을 하는 기법이다. 이러한 기법은 현재 사용자에게 친숙한 텍스트 질의를 유지할 수 있게 해 주며 앞으로 음성인식을 통한 음성 질의인터페이스가 적용될 경우 더욱 효과적으로 사용될 수 있을 것이다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2004.11a
• /
• pp.887-890
• /
• 2004

임의 영상에서 얼굴 영역을 검출하고 얼굴 특징점 정보를 획득하는 기술은 얼굴 인식 및 표정 인식 시스템에서 중요한 역할을 한다. 본 논문은 색도 정보와 Top-hat 연산을 이용함으로써 얼굴의 유효 특징점을 효과적으로 검출할 수 있는 방법을 제안한다. 제안한 방법은 얼굴 영역 검출, 눈/눈썹 특징추출, 입술 특징추출의 세 과정으로 나눈다. 얼굴 영역은 $YC_{b}C_{r}$을 이용하여 피부색 영역을 추출한 후 모폴로지 연산과 분할을 통해 획득하고, 눈/눈썹 특징점은 BWCD(Black & White Color Distribution) 변환과 Top-hat 연산을 이용하며. 입술 특징점은 눈/눈썹과의 지정학적 상관관계와 입술 색상분포를 이용하는 방법을 사용한다. 실험을 수행한 결과. 제안한 방법이 다양한 영상에 대해서도 효과적으로 얼굴의 유효 특징점을 검출할 수 있음을 확인하였다.

• The KIPS Transactions:PartB
• /
• v.11B no.6
• /
• pp.667-674
• /
• 2004

Extraction of texts from images is a fundamental and important problem to understand the images. This paper suggests a text region verification method by statistical means of stroke features of the characters. The method extracts 36 dimensional features from $16\times16$sized text and non-text images using wavelet transform - these 36 dimensional features express stroke and direction of characters - and select 12 sub-features out of 36 dimensional features which yield adequate separation between classes. After selecting the features, SVM trains the selected features. For the verification of the text region, each $16\times16$image block is scanned and classified as text or non-text. Then, the text region is finally decided as text region or non-text region. The proposed method is able to verify text regions which can hardly be distin guished.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.28 no.3
• /
• pp.353-359
• /
• 2010

In this paper, we propose the automatic image-to-image registration of high resolution satellite images using local properties of tie points to improve the registration accuracy. A spatial distance between interest points of reference and sensed images extracted by Scale Invariant Feature Transform(SIFT) is additionally used to extract tie points. Coefficients of affine transform between images are extracted by invariant descriptor based matching, and interest points of sensed image are transformed to the reference coordinate system using these coefficients. The spatial distance between interest points of sensed image which have been transformed to the reference coordinates and interest points of reference image is calculated for secondary matching. The piecewise linear function is applied to the matched tie points for automatic registration of high resolution images. The proposed method can extract spatially well-distributed tie points compared with SIFT based method.