• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.40 no.2
• /
• pp.91-108
• /
• 2022

In most cases, optical images have been used as training data of DL (Deep Learning) models for object detection, recognition, identification, classification, semantic segmentation, and instance segmentation. However, properties of 3D objects in the real-world could not be fully explored with 2D images. One of the major sources of the 3D geospatial information is DSM (Digital Surface Model). In this matter, characteristic information derived from DSM would be effective to analyze 3D terrain features. Especially, man-made objects such as buildings having geometrically unique shape could be described by geometric elements that are obtained from 3D geospatial data. The background and motivation of this paper were drawn from concept of the intrinsic image that is involved in high-level visual information processing. This paper aims to extract buildings after classifying terrain features by training DL model with DSM-derived information including slope, aspect, and SRI (Shaded Relief Image). The experiments were carried out using DSM and label dataset provided by ISPRS (International Society for Photogrammetry and Remote Sensing) for CNN-based SegNet model. In particular, experiments focus on combining multi-source information to improve training performance and synergistic effect of the DL model. The results demonstrate that buildings were effectively classified and extracted by the proposed approach.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.23 no.3
• /
• pp.145-151
• /
• 2023

3D-CNN is one of the deep learning techniques for learning time series data. Such three-dimensional learning can generate many parameters, so that high-performance machine learning is required or can have a large impact on the learning rate. When learning dynamic hand-gestures in spatiotemporal domain, it is necessary for the improvement of the efficiency of dynamic hand-gesture learning with 3D-CNN to find the optimal conditions of input video data by analyzing the learning accuracy according to the spatiotemporal change of input video data without structural change of the 3D-CNN model. First, the time ratio between dynamic hand-gesture actions is adjusted by setting the learning interval of image frames in the dynamic hand-gesture video data. Second, through 2D cross-correlation analysis between classes, similarity between image frames of input video data is measured and normalized to obtain an average value between frames and analyze learning accuracy. Based on this analysis, this work proposed two methods to effectively select input video data for 3D-CNN deep learning of dynamic hand-gestures. Experimental results showed that the learning interval of image data frames and the similarity of image frames between classes can affect the accuracy of the learning model.

• Journal of IKEEE
• /
• v.25 no.3
• /
• pp.493-500
• /
• 2021

In this paper, we propose the development of deep learning structure to improve quality of polygonal containers. The deep learning structure consists of a convolution layer, a bottleneck layer, a fully connect layer, and a softmax layer. The convolution layer is a layer that obtains a feature image by performing a convolution 3x3 operation on the input image or the feature image of the previous layer with several feature filters. The bottleneck layer selects only the optimal features among the features on the feature image extracted through the convolution layer, reduces the channel to a convolution 1x1 ReLU, and performs a convolution 3x3 ReLU. The global average pooling operation performed after going through the bottleneck layer reduces the size of the feature image by selecting only the optimal features among the features of the feature image extracted through the convolution layer. The fully connect layer outputs the output data through 6 fully connect layers. The softmax layer multiplies and multiplies the value between the value of the input layer node and the target node to be calculated, and converts it into a value between 0 and 1 through an activation function. After the learning is completed, the recognition process classifies non-circular glass bottles by performing image acquisition using a camera, measuring position detection, and non-circular glass bottle classification using deep learning as in the learning process. In order to evaluate the performance of the deep learning structure to improve quality of polygonal containers, as a result of an experiment at an authorized testing institute, it was calculated to be at the same level as the world's highest level with 99% good/defective discrimination accuracy. Inspection time averaged 1.7 seconds, which was calculated within the operating time standards of production processes using non-circular machine vision systems. Therefore, the effectiveness of the performance of the deep learning structure to improve quality of polygonal containers proposed in this paper was proven.

• Journal of Intelligence and Information Systems
• /
• v.25 no.3
• /
• pp.161-177
• /
• 2019

In this paper, we study the performance improvement of the answer extraction in Question-Answering system by using sentence dependency parsing result. The Question-Answering (QA) system consists of query analysis, which is a method of analyzing the user's query, and answer extraction, which is a method to extract appropriate answers in the document. And various studies have been conducted on two methods. In order to improve the performance of answer extraction, it is necessary to accurately reflect the grammatical information of sentences. In Korean, because word order structure is free and omission of sentence components is frequent, dependency parsing is a good way to analyze Korean syntax. Therefore, in this study, we improved the performance of the answer extraction by adding the features generated by dependency parsing analysis to the inputs of the answer extraction model (Bidirectional LSTM-CRF). The process of generating the dependency graph embedding consists of the steps of generating the dependency graph from the dependency parsing result and learning the embedding of the graph. In this study, we compared the performance of the answer extraction model when inputting basic word features generated without the dependency parsing and the performance of the model when inputting the addition of the Eojeol tag feature and dependency graph embedding feature. Since dependency parsing is performed on a basic unit of an Eojeol, which is a component of sentences separated by a space, the tag information of the Eojeol can be obtained as a result of the dependency parsing. The Eojeol tag feature means the tag information of the Eojeol. The process of generating the dependency graph embedding consists of the steps of generating the dependency graph from the dependency parsing result and learning the embedding of the graph. From the dependency parsing result, a graph is generated from the Eojeol to the node, the dependency between the Eojeol to the edge, and the Eojeol tag to the node label. In this process, an undirected graph is generated or a directed graph is generated according to whether or not the dependency relation direction is considered. To obtain the embedding of the graph, we used Graph2Vec, which is a method of finding the embedding of the graph by the subgraphs constituting a graph. We can specify the maximum path length between nodes in the process of finding subgraphs of a graph. If the maximum path length between nodes is 1, graph embedding is generated only by direct dependency between Eojeol, and graph embedding is generated including indirect dependencies as the maximum path length between nodes becomes larger. In the experiment, the maximum path length between nodes is adjusted differently from 1 to 3 depending on whether direction of dependency is considered or not, and the performance of answer extraction is measured. Experimental results show that both Eojeol tag feature and dependency graph embedding feature improve the performance of answer extraction. In particular, considering the direction of the dependency relation and extracting the dependency graph generated with the maximum path length of 1 in the subgraph extraction process in Graph2Vec as the input of the model, the highest answer extraction performance was shown. As a result of these experiments, we concluded that it is better to take into account the direction of dependence and to consider only the direct connection rather than the indirect dependence between the words. The significance of this study is as follows. First, we improved the performance of answer extraction by adding features using dependency parsing results, taking into account the characteristics of Korean, which is free of word order structure and omission of sentence components. Second, we generated feature of dependency parsing result by learning - based graph embedding method without defining the pattern of dependency between Eojeol. Future research directions are as follows. In this study, the features generated as a result of the dependency parsing are applied only to the answer extraction model in order to grasp the meaning. However, in the future, if the performance is confirmed by applying the features to various natural language processing models such as sentiment analysis or name entity recognition, the validity of the features can be verified more accurately.

• Journal of KIISE:Software and Applications
• /
• v.30 no.10
• /
• pp.904-911
• /
• 2003

In this paper, we propose an effective method for the distinction between machine-printed and handwritten Korean address images. It is important to know whether an input image is handwritten or machine-printed, because methods for handwritten image are quite different from those of machine-printed image in such applications as address reading, form processing, FAX routing, and so on. Our method consists of three blocks: valid connected components grouping, feature extraction, and classification. Features related to width and position of groups of valid connected components are used for the classification based on a neural network. The experiment done with live Korean address images has demonstrated the superiority of the proposed method. The correct classification rate for 3,147 testing images was about 98.85%.

• Journal of Physiology & Pathology in Korean Medicine
• /
• v.23 no.1
• /
• pp.97-103
• /
• 2009

The status and the property of a tongue are the important indicators to diagnose one's health like physiological and clinicopathological changes of inner organs. However, the tongue diagnosis is affected by examination circumstances like a light source, patient's posture, and doctor's condition. To develop an automatic tongue diagnosis system for an objective and standardized diagnosis, classifying tongue coating is inevitable but difficult since the features like color and texture of the tongue coatings and substance have little difference, especially in the neighborhood on the tongue surface. The proposed method has two procedures; the first is to acquire the color table to classify tongue coatings and substance by automatically separating coating regions marked by oriental medical doctors, decomposing the color components of the region into hue, saturation and brightness and obtaining the 2nd order discriminant with statistical data of hue and saturation corresponding to each kind of tongue coatings, and the other is to apply the tongue region in an input image to the color table, resulting in separating the regions of tongue coatings and classifying them automatically. As a result, kinds of tongue coatings and substance were segmented from a face image corresponding to regions marked by oriental medical doctors and the color table for classification took hue and saturation values as inputs and produced the classification of the values into white coating, yellow coating and substance in a digital tongue diagnosis system. The coating regions classified by the proposed method were almost the same to the marked regions. The exactness of classification was 83%, which is the degree of correspondence between what Oriental medical doctors diagnosed and what the proposed method classified. Since the classified regions provide effective information, the proposed method can be used to make an objective and standardized diagnosis and applied to an ubiquitous healthcare system. Therefore, the method will be able to be widely used in Oriental medicine.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.19 no.10
• /
• pp.1910-1924
• /
• 1994

Curved 3D objects represented by range data contain large amounts of information compared with planar objects, but do not have distinct features for matching to those of object models. This makes it difficult to represent and identify a general 3D curved object. This paper introduces a new view-point independent approach to recognizing general 3D curved objects using range data. Our approach makes use of the relative geometric differences between particular points on the object surface and some model points. The model points are prespecified arbitrarily and keeping the task in mind so that the following task can be easily described using the model points. Our approach has several advantages. Since model points are specified arbitrarily and task dependently, further processing can be reduced in application by locating the model points at places which are useful for further operations in the task. The knowledge base is simple with less storage requirement. And, it is easy to compensate the uncertainties of positions estimation caused by noise and quantization error.

• Journal of KIISE:Software and Applications
• /
• v.36 no.8
• /
• pp.633-641
• /
• 2009

In this paper, we proposed a new pattern classifier which can be incrementally learned, be added new class in learning time, and handle with analog data. Proposed pattern classifier has hierarchical structure and the classification rate is improved by using different metric for each levels. Proposed model is based on the Gaussian ARTMAP which is an artificial neural network model for the pattern classification. We hierarchically constructed the Gaussian ARTMAP and proposed the Principal Component Emphasis(P.C.E) method to be learned different features in each levels. And we defined new metric based on the P.C.E. P.C.E is a method that discards dimensions whose variation are small, that represents common attributes in the class. And remains dimensions whose variation are large. In the learning process, if input pattern is misclassified, P.C.E are performed and the modified pattern is learned in sub network. Experimental results indicate that Hierarchical Gaussian ARTMAP yield better classification result than the other pattern recognition algorithms on variable data set including real applicable problem.

• The Journal of Korean Medicine Ophthalmology and Otolaryngology and Dermatology
• /
• v.22 no.3
• /
• pp.153-166
• /
• 2009

Objectives : This study was planed to evaluate clinical status of the alopecia patients who had visited Korean medicine clinic. And the result from this study would provide a standard in Korean medical diagnostic and classification method of alopecia. Methods : Clinical records of 183 patients with alopecia seen from January 2004 to April 2005 at Korean medical clinic was examined. They were classified into 4 different types according to chief complains besides alopecia by 2 Korean medical doctors. Results and conclusions : We made clinical analysis of patients of alopecia from January 2004 to April 2005. Among the alopecia patients who visit Korean medical clinic, people age between 20 and 30 had high ratio. The duration from the recognition of initial hair loss to the time of the first visit to the Korean medical clinic was less than 12 months in 20.8%(38/138), and less than 60 months in 72.2% (132/183). The condition of alopecia was more worse than other alopecia patients who visit the west medical clinic. Also the ratio with increased temperature of face or scalp is chief complaint except alopecia in alopecia patients was high in men and the ratio with dysfunction of digestive system or chronic weakness was high in women. Among the incidence of alopecia, the androgenic alopecia was most in number; 43.7%(80/183) and the sex distribution showed 83 men and 100 women.

• Fashion & Textile Research Journal
• /
• v.6 no.1
• /
• pp.71-77
• /
• 2004

Various types of industry, from manufacturing industry to service industry, can be a subject of exhibition. So, subdivided studies about exhibition for each type of industry or field are necessary. However, there are still insufficient studies about them. Also, fashion industry related exhibitions should be classified differently from other industries' exhibitions because of their special features. Therefore, this study examines that how an exhibition is utilized as a promotion means of a company and the present condition of a fashion exhibition. It also compares a fashion exhibition with other industrial exhibitions to suggest the effective operations and the progressive promotion of the fashion exhibition. This study uses questionnaire from 140 companies for 5 exhibitions, which participated fashion related exhibitions. Data of this study is statistically analyzed using SPSS for window ver. 10.0 program. It also uses frequency, cross tabs, paired t-test, Independent Samples t-test. As a result of this study, fashion related companies' recognition of exhibitions has been increased. However; in terms of exhibition participation, attitudes such as pre-promotion or pre-education for the staff, who are in charge of the exhibition booths, and the outcome of the exhibition aren't maximized because the right understanding concerning exhibitions is lacking and long-term preparation is insufficient. Also, advertisement for the exhibitions isn't enough because of insufficient preparation and absence of pre-promotion, so consultation with promising buyers can't be activated. Even though there are consultations, the consultations don't have much effect.