• Journal of Korea Multimedia Society
• /
• v.21 no.2
• /
• pp.98-107
• /
• 2018

In this paper, we propose a method to classify the skeletal maturity with a small amount of hand wrist X-ray image using deep learning-based meta-learning. General deep-learning techniques require large amounts of data, but in many cases, these data sets are not available for practical application. Lack of learning data is usually solved through transfer learning using pre-trained models with large data sets. However, transfer learning performance may be degraded due to over fitting for unknown new task with small data, which results in poor generalization capability. In addition, medical images require high cost resources such as a professional manpower and mcuh time to obtain labeled data. Therefore, in this paper, we use meta-learning that can classify using only a small amount of new data by pre-trained models trained with various learning tasks. First, we train the meta-model by using a separate data set composed of various learning tasks. The network learns to classify the bone maturity using the bone maturity data composed of the radiographs of the wrist. Then, we compare the results of the classification using the conventional learning algorithm with the results of the meta learning by the same number of learning data sets.

• Nuclear Engineering and Technology
• /
• v.53 no.2
• /
• pp.522-531
• /
• 2021

We describe a new deep learning model - Convolutional eXtreme Gradient Boosting (ConvXGB) for classification problems based on convolutional neural nets and Chen et al.'s XGBoost. As well as image data, ConvXGB also supports the general classification problems, with a data preprocessing module. ConvXGB consists of several stacked convolutional layers to learn the features of the input and is able to learn features automatically, followed by XGBoost in the last layer for predicting the class labels. The ConvXGB model is simplified by reducing the number of parameters under appropriate conditions, since it is not necessary re-adjust the weight values in a back propagation cycle. Experiments on several data sets from UCL Repository, including images and general data sets, showed that our model handled the classification problems, for all the tested data sets, slightly better than CNN and XGBoost alone and was sometimes significantly better.

• Journal of Information Processing Systems
• /
• v.12 no.1
• /
• pp.149-168
• /
• 2016

The Active Appearance Model (AAM) is a class of deformable models, which, in the segmentation process, integrates the priori knowledge on the shape and the texture and deformation of the structures studied. This model in its sequential form is computationally intensive and operates on large data sets. This paper presents another framework to implement the standard version of the AAM model. We suggest a distributed and parallel approach justified by the characteristics of the model and their potentialities. We introduce a schema for the representation of the overall model and we study of operations that can be parallelized. This approach is intended to exploit the benefits build in the area of advanced image processing.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.395-398
• /
• 2003

Two component analysis methods based on the fuzzy relational calculus are proposed in the setting of the ordered structure. First component analysis is based on a decomposition of fuzzy relation into fuzzy bases, using gradient method. Second one is a component analysis based on the eigen fuzzy sets of fuzzy relation. Through experiments using the test images extracted from SIDBA and View Sphere Database, the effectiveness of the proposed component analysis methods is confirmed. Furthermore, improvements of the image compression/reconstruction and image retrieval based on ordered structure are also indicated.

• Wind and Structures
• /
• v.21 no.6
• /
• pp.709-726
• /
• 2015

In this study, a novel vision-based bolt-loosening monitoring technique is proposed for bolted joints connecting tubular steel segments of the wind turbine tower (WTT) structure. Firstly, a bolt-loosening detection algorithm based on image processing techniques is developed. The algorithm consists of five steps: image acquisition, segmentation of each nut, line detection of each nut, nut angle estimation, and bolt-loosening detection. Secondly, experimental tests are conducted on a lab-scale bolted joint model under various bolt-loosening scenarios. The bolted joint model, which is consisted of a ring flange and 32 sets of bolt and nut, is used for simulating the real bolted joint connecting steel tower segments in the WTT. Finally, the feasibility of the proposed vision-based technique is evaluated by bolt-loosening monitoring in the lab-scale bolted joint model.

• Journal of the Korean Society of Manufacturing Process Engineers
• /
• v.11 no.5
• /
• pp.142-147
• /
• 2012

Machining tool conditions directly affect to quality of product and productivity of manufacturing. Many researches performed for tool condition monitoring in machining process to improve quality and productivity. Conventional methods use characteristics of signal for cutting force, motor current consumption, vibration of machine tools and machining sound. Recently, diameter of machining tool is become smaller for minimizing of mechanical parts. Tool condition monitoring using conventional methods are relatively difficult because micro machining using small diameter tool has low machining load and high cutting speed. These days, the direct monitoring for tool conditions using vision system is performed actively. But, vision system is affected by external conditions such as back ground of image and illumination. In this study, minimizing technology of external conditions using distribution analysis of image data are developed in micro machining using small diameter drill and tap. The image data is gathered from vision system. Several sets of experiment results are performed to verify the characteristics of the proposed machining technology.

• Journal of the Korea Society of Computer and Information
• /
• v.22 no.12
• /
• pp.55-61
• /
• 2017

In this paper, we propose a novel tracking method using target separation and detection that are based on discriminative correlation filter (DCF), which is studied a lot recently. 'Retainability' is one of the most important factor of tracking. There are some factors making retainability of tracking worse. Especially, fast movement and occlusion of a target frequently occur in image data, and when it happens, it would make target lost. As a result, the tracking cannot be retained. For maintaining a robust tracking, in this paper, separation of a target is used so that normal tracking is maintained even though some part of a target is occluded. The detection algorithm is executed and find new location of the target when the target gets out of tracking range due to occlusion of whole part of a target or fast movement speed of a target. A variety of experiments with various image data sets are conducted. The algorithm proposed in this paper showed better performance than other conventional algorithms when fast movement and occlusion of a target occur.

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.4 no.2
• /
• pp.142-148
• /
• 2004

In this paper, we propose the color fuzzy decision algorithm to face segmentation in a color image. Our algorithm can segment without the user's interaction by fuzzy decision marking. And it removes small parts such as a noise using wavelet morphology in the image obtained by applying the fuzzy decision algorithm. Also, it merges and chooses the face region in each quantization image through rough sets. This video object division algorithm is shown to be superior to a conventional algorithm.

• Proceedings of the KIEE Conference
• /
• 1996.11a
• /
• pp.294-296
• /
• 1996

The process of detecting edge in an image is an important component of many Pattern Recognition and Computer Vision applications. In many practical cases, there exist blurred images due to defocussing, movement of an object and so on. In addition, local perturbation noise can be added to the images. We propose the edge detection technique in blurred and noisy image. For this, we use Fuzzy pyramid linking mothod to remove noise and enhance the edge in images. We develop contrast intensifier using the concept of Fuzzy sets as a postprocessing.

• Proceedings of the Korean Society of Broadcast Engineers Conference
• /
• 1999.06a
• /
• pp.97.1-102
• /
• 1999

This paper porposes and efficient coding scheme for remote medical communication systems, or“telemedicine systems”. These systems require a technique which is able to transfer large volume of data such as X-ray images effectively. We have already developed a hierarchical image coding and transmission scheme (HITS), which achieves an efficient transmission of medical images simply[1]. In this paper, a new coding scheme for HITS is proposed, which used hierarchical context modeling for the purpose of improving the coding efficiency. The hierarchical context modeling divides wavelet coefficients into several sets by the value of a correspondent coefficient in their higher class, or“a parent”, optimizes a Golomb-Rice (GR) code parameter in each set, and then encodes the coefficients with the parameter. Computer simulation shows that the proposed scheme is effective with simple implementation. This is due to fact that a wavelet coefficient has dependence on its parent. As a result, high speed data transmission is achieved even if the telemedicine system consists of simple personal computers.