• Journal of the Semiconductor & Display Technology
• /
• v.17 no.3
• /
• pp.113-116
• /
• 2018

This paper proposes a mobile image search system which uses a sensor information of smart phone, and enables running in a variety of environments, which is implemented on Android platform. The implemented system deals with a new image descriptor using combination of the visual feature (CEDD) with EXIF attributes in the target of JPEG image, and image matching scheme, which is optimized to the mobile platform. Experimental result shows that the proposed method exhibited a significant improved searching results of around 80% in precision in the large image database. Considering the performance such as processing time and precision, we think that the proposed method can be used in other application field.

• Journal of Korea Multimedia Society
• /
• v.25 no.1
• /
• pp.52-60
• /
• 2022

Few-shot learning is a sub-area of machine learning problems, which aims to classify target images that only contain a few labeled samples for training. As a representative few-shot learning method, the Prototypical network has been received much attention due to its simplicity and promising results. However, the Prototypical network uses the sample mean of samples from the same class as the prototypes of that class, which easily results in learning uncharacteristic features in the low-data scenery. In this study, we propose to use local descriptors (i.e., patches along the channel within feature maps) from the same class to explicitly obtain more representative prototypes for Prototypical Network so that significant intra-class feature information can be maintained and thus improving the classification performance on few-shot learning tasks. Experimental results on various benchmark datasets including mini-ImageNet, CUB-200-2011, and tiered-ImageNet show that the proposed method can learn more discriminative intra-class features by the local descriptors and obtain more generic prototype representations under the few-shot setting.

• Proceedings of the KSRS Conference
• /
• 2002.10a
• /
• pp.34-39
• /
• 2002

Road information is very important for topographic mapping, transportation application, urban planning and other related application fields. Therefore, automatic detection of road networks from spatial imagery, such as aerial photos and satellite imagery can play a central role in road information acquisition. In this paper, we use least squares correlation matching alone for road center tracking and show that it works. We assumed that (bright) road centerlines would be visible in the image. We further assumed that within a same road segment, there would be only small differences in brightness values. This algorithm works by defining a template around a user-given input point, which shall lie on a road centerline, and then by matching the template against the image along the orientation of the road under consideration. Once matching succeeds, new match proceeds by shifting a matched target window further along road orientation at the target window. By repeating the process above, we obtain a series of points, which lie on a road centerline successively. A 1m resolution IKONOS images over Seoul and Daejeon were used for tests. The results showed that this algorithm could extract road centerlines in any orientation and help in fast and exact he ad-up digitization/vectorization of cartographic images.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.17 no.3
• /
• pp.701-720
• /
• 2023

Aerial objects are more challenging to segment than normal objects, which are usually smaller and have less textural detail. In the process of segmentation, target objects are easily omitted and misdetected, which is problematic. To alleviate these issues, we propose local aggregation feature pyramid networks (LAFPNs) and pyramid integrated context modules (PICMs) for aerial object segmentation. First, using an LAFPN, while strengthening the deep features, the extent to which low-level features interfere with high-level features is reduced, and numerous dense and small aerial targets are prevented from being mistakenly detected as a whole. Second, the PICM uses global information to guide local features, which enhances the network's comprehensive understanding of an entire image and reduces the missed detection of small aerial objects due to insufficient texture information. We evaluate our network with the MS COCO dataset using three categories: airplanes, birds, and kites. Compared with Mask R-CNN, our network achieves performance improvements of 1.7%, 4.9%, and 7.7% in terms of the AP metrics for the three categories. Without pretraining or any postprocessing, the segmentation performance of our network for aerial objects is superior to that of several recent methods based on classic algorithms.

• KIPS Transactions on Computer and Communication Systems
• /
• v.9 no.10
• /
• pp.231-238
• /
• 2020

With a large number of documents appearing on the web, document-level novelty detection has become important since it can reduce the efforts of finding novel documents by discarding documents sharing redundant information already seen. A recent work proposed a convolutional neural network (CNN)-based novelty detection model with significant performance improvements. We observed that it has a restriction of using document-level information in determining novelty but assumed that the document-level information is more important. As a solution, this paper proposed two methods of effectively incorporating document-level information using a CNN-based novelty detection model. Our methods focus on constructing a feature vector of a target document to be classified by extracting relative information between the target document and source documents given as evidence. A series of experiments showed the superiority of our methods on a standard benchmark collection, TAP-DLND 1.0.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.11 no.4
• /
• pp.1272-1277
• /
• 2010

Target detection and tracking wireless sensor network must have various signal processing ability. Wireless sensor nodes need to light weight signal processing algorithm because of energy constraints and communication bandwidth constraints. General signal processing algorithm of wireless sensor node consists of de-noising, received signal strength computation, feature extraction and signal compression. Wireless sensor network life-time and performance of target detection and classification depend on sensor node signal processing. In this paper, we propose energy efficient signal processing algorithm using wavelet transform. The proposed method estimates exact target position.

• Journal of Internet Computing and Services
• /
• v.5 no.4
• /
• pp.125-134
• /
• 2004

Recently a lot of researches on the component-based software product lines and on applying generative programming into software product lines are being performed actively. This paper proposes an automatic component reconfiguration tool that could be applied in constructing the component-based software product lines. Our tool accepts the reuser's requirement via a feature model which is the main result of the domain engineering, and makes the feature configuration from this requirement. Then it generates the source code of the reconfigured component according to this feature configuration. To accomplish this process, the component family in our tool should have the architecture of GenVoca that is one of the most influential generative programming approaches. In addition, XSLT scripts provide the code templates for implementation elements which are the ingredients of the target component. Taking the ‘Bank Account' component family as our example, we showed that our component reconfiguration tool produced automatically the component source code that the reuser wants to create. The result of this paper would be applied extensively for creasing the productivity of building the software product lines.

• Proceedings of the IEEK Conference
• /
• 2000.09a
• /
• pp.157-160
• /
• 2000

In the generally a voice conversion has used VQ(Vector Quantization) for partitioning the spectral feature and has performed by adding an appropriate offset vector to the source speaker's spectral vector. But there is not represented the target speaker's various characteristics because of discrete characteristics of transformed parameter. In this paper, these problems are solved by using the LMR(Linear Multivariate Regression) instead of the mapping codebook which is determined to the relationship of source and target speaker vocal tract characteristics. Also we propose the method for solved the discontinuity which is caused by applying to time aligned parameters using Dynamic Time Warping the time or pitch-scale modified speech. In our proposed algorithm for overcoming the transitional discontinuities, first of all, we don't change time or pitch scale and by using the LMR change a speaker's vocal tract characteristics in speech with non-modified time or pitch. Compared to existed methods based on VQ and LMR, we have much better voice quality in the result of the proposed algorithm.

• Journal of Broadcast Engineering
• /
• v.26 no.7
• /
• pp.868-876
• /
• 2021

Deploying deep neural network model training performs remarkable performance in the fields of Object detection and Instance segmentation. To train these models, features are first extracted from the input image using a backbone network. The extracted features can be reused by various tasks. Research has been actively conducted to serve various tasks by using these learned features. In this process, standardization discussions about encoding, decoding, and transmission methods are proceeding actively. In this scenario, it is necessary to analyze the response characteristics of features against various distortions that may occur in the data transmission or data compression process. In this paper, experiment was conducted to inject various distortions into the feature in the object recognition task. And analyze the mAP (mean Average Precision) metric between the predicted value output from the neural network and the target value as the intensity of various distortions was increased. Experiments have shown that features are more robust to distortion than images. And this points out that using the feature as transmission means can prevent the loss of information against the various distortions during data transmission and compression process.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.26 no.11A
• /
• pp.1946-1956
• /
• 2001

In this paper, we propose a novel 2-D partial matching algorithm based on model-based stochastic analysis of feature correspondences in a relation vector space, which is quite robust to shape variations as well as invariant to geometric transformations. We represent an object using the ARG (Attributed Relational Graph) model with features of a set of relation vectors. In addition, we statistically model the partial occlusion or noise as the distortion of the relation vector distribution in the relation vector space. Our partial matching algorithm consists of two-phases. First, a finite number of candidate sets areselected by using logical constraint embedding local and structural consistency Second, the feature loss detection is done iteratively by error detection and voting scheme thorough the error analysis of relation vector space. Experimental results on real images demonstrate that the proposed algorithm is quite robust to noise and localize target objects correctly even inseverely noisy and occluded scenes.