• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.22 no.3
• /
• pp.471-479
• /
• 2018

Recently, the spread of smart TV based Android iOS Set Top box has become common. This paper propose a new approach to control the TV using gestures away from the era of controlling the TV using remote control. In this paper, the AdaBoost algorithm is applied to gesture recognition by using a mono camera. First, we use Camshift-based Body tracking and estimation algorithm based on Gaussian background removal for body coordinate extraction. Using global and local feature vectors, we recognized gestures with speed change. By tracking the time interval trajectories of hand and wrist, the AdaBoost algorithm with CART algorithm is used to train and classify gestures. The principal component feature vector with high classification success rate is searched using CART algorithm. As a result, 24 optimal feature vectors were found, which showed lower error rate (3.73%) and higher accuracy rate (95.17%) than the existing algorithm.

• The Journal of Bigdata
• /
• v.5 no.2
• /
• pp.53-67
• /
• 2020

This study employees a supervised learning prediction model to detect nonconformity in advance of processed food manufacturing and processing businesses. The study was conducted according to the standard procedure of machine learning, such as definition of objective function, data preprocessing and feature engineering and model selection and evaluation. The dependent variable was set as the number of supervised inspection detections over the past five years from 2014 to 2018, and the objective function was to maximize the probability of detecting the nonconforming companies. The data was preprocessed by reflecting not only basic attributes such as revenues, operating duration, number of employees, but also the inspections track records and extraneous climate data. After applying the feature variable extraction method, the machine learning algorithm was applied to the data by deriving the company's risk, item risk, environmental risk, and past violation history as feature variables that affect the determination of nonconformity. The f1-score of the decision tree, one of ensemble models, was much higher than those of other models. Based on the results of this study, it is expected that the official food control for food safety management will be enhanced and geared into the data-evidence based management as well as scientific administrative system.

• Journal of Broadcast Engineering
• /
• v.26 no.1
• /
• pp.26-38
• /
• 2021

In this paper, DLKF (Deep Learning Keypoint Filtering), the deep learning-based keypoint filtering method for the rapidization of the image registration method for remote sensing images is proposed. The complexity of the conventional feature-based image registration method arises during the feature matching step. To reduce this complexity, this paper proposes to filter only the keypoints detected in the artificial structure among the keypoints detected in the keypoint detector by ensuring that the feature matching is matched with the keypoints detected in the artificial structure of the image. For reducing the number of keypoints points as preserving essential keypoints, we preserve keypoints adjacent to the boundaries of the artificial structure, and use reduced images, and crop image patches overlapping to eliminate noise from the patch boundary as a result of the image segmentation method. the proposed method improves the speed and accuracy of registration. To verify the performance of DLKF, the speed and accuracy of the conventional keypoints extraction method were compared using the remote sensing image of KOMPSAT-3 satellite. Based on the SIFT-based registration method, which is commonly used in households, the SURF-based registration method, which improved the speed of the SIFT method, improved the speed by 2.6 times while reducing the number of keypoints by about 18%, but the accuracy decreased from 3.42 to 5.43. Became. However, when the proposed method, DLKF, was used, the number of keypoints was reduced by about 82%, improving the speed by about 20.5 times, while reducing the accuracy to 4.51.

• Journal of Internet Computing and Services
• /
• v.23 no.2
• /
• pp.61-70
• /
• 2022

General users who use smartphone apps often use the Vault app to protect personal information such as photos and videos owned by individuals. However, there are increasing cases of criminals using the Vault app function for anti-forensic purposes to hide illegal videos. These apps are one of the apps registered on Google Play. This paper proposes a methodology for extracting feature points through XML-based keyword frequency analysis to explore Vault apps used by criminals, and text mining techniques are applied to extract feature points. In this paper, XML syntax was compared and analyzed using strings.xml files included in the app for 15 hidden Vault anti-forensics apps and non-hidden Vault apps, respectively. In hidden Vault anti-forensics apps, more hidden-related words are found at a higher frequency in the first and second rounds of terminology processing. Unlike most conventional methods of static analysis of APK files from an engineering point of view, this paper is meaningful in that it approached from a humanities and sociological point of view to find a feature of classifying anti-forensics apps. In conclusion, applying text mining techniques through XML parsing can be used as basic data for exploring hidden Vault anti-forensics apps.

• Journal of the Korea Convergence Society
• /
• v.12 no.11
• /
• pp.45-51
• /
• 2021

With the development of deep learning, semantic segmentation methods are being studied in various fields. There is a problem that segmenation accuracy drops in fields that require accuracy such as medical image analysis. In this paper, we improved PSPNet, which is a deep learning based segmentation method to minimized the loss of features during semantic segmentation. Conventional deep learning based segmentation methods result in lower resolution and loss of object features during feature extraction and compression. Due to these losses, the edge and the internal information of the object are lost, and there is a problem that the accuracy at the time of object segmentation is lowered. To solve these problems, we improved PSPNet, which is a semantic segmentation model. The multi-scale attention proposed to the conventional PSPNet was added to prevent feature loss of objects. The feature purification process was performed by applying the attention method to the conventional PPM module. By suppressing unnecessary feature information, eadg and texture information was improved. The proposed method trained on the Cityscapes dataset and use the segmentation index MIoU for quantitative evaluation. As a result of the experiment, the segmentation accuracy was improved by about 1.5% compared to the conventional PSPNet.

• Asia-pacific Journal of Multimedia Services Convergent with Art, Humanities, and Sociology
• /
• v.7 no.4
• /
• pp.551-559
• /
• 2017

In this paper, a real-time training method to improve the performance of bearing fault diagnosis. The traditional bearing fault diagnosis cannot classify a condition which is not trained by the classifier. The proposed 4-step method trains and recognizes new condition in real-time, thereby it can classify the condition accurately. In the first step, we calculate the maximum distance value for each class by calculating a Euclidean distance between a feature vector of each class and a centroid of the corresponding class in the training information. In the second step, we calculate a Euclidean distance between a feature vector of new acquired data and a centroid of each class, and then compare with the allowed maximum distance of each class. In the third step, if the distance between a feature vector of new acquired data and a centroid of each class is larger than the allowed maximum distance of each class, we define that it is data of new condition and increase count of new condition. In the last step, if the count of new condition is over 10, newly acquired 10 data are assigned as a new class and then conduct re-training the classifier. To verify the performance of the proposed method, bearing fault data from a rotating machine was utilized.

• Journal of Korea Multimedia Society
• /
• v.13 no.7
• /
• pp.1062-1075
• /
• 2010

In the medical image application the difference of intensity is widely used for the image segmentation and feature extraction, and a well known method is the threshold technique that determines a threshold value and generates a binary image based on the threshold. A frequently-used threshold technique is the Otsu algorithm that provides efficient processing and effective selection criterion for choosing the threshold value. However, we cannot get good segmentation results by applying the Otsu algorithm to chest X-ray images. It is because there are various organic structures around lung regions such as ribs and blood vessels, causing unclear distribution of intensity levels. To overcome the ambiguity, we propose in this paper an effective algorithm to extract pulmonary regions that utilizes the Otsu algorithm after removing the background of an X-ray image, constructs intensity-level maps, and uses them for segmenting the X-ray image. To verify the effectiveness of our method, we compared it with the existing 1-dimensional and 2-dimensional Otsu algorithms, and also the results by expert's naked eyes. The experimental result showed that our method achieved the more accurate extraction of pulmonary regions compared to the Otsu methods and showed the similar result as the naked eye's one.

• Journal of the Korea Convergence Society
• /
• v.10 no.6
• /
• pp.7-14
• /
• 2019

In this paper, we propose a Cascade Classifier face detection method using the Haar-like feature, which is complemented by the Flood Fill algorithm for lossy areas due to illumination and shadow in YCbCr color space extraction. The Cascade Classifier using Haar-like features can generate noise and loss regions due to lighting, shadow, etc. because skin color extraction using existing YCbCr color space in image only uses threshold value. In order to solve this problem, noise is removed by erosion and expansion calculation, and the loss region is estimated by using the Flood Fill algorithm to estimate the loss region. A threshold value of the YCbCr color space was further allowed for the estimated area. For the remaining loss area, the color was filled in as the average value of the additional allowed areas among the areas estimated above. We extracted faces using Haar-like Cascade Classifier. The accuracy of the proposed method is improved by about 4% and the detection rate of the proposed method is improved by about 2% than that of the Haar-like Cascade Classifier by using only the YCbCr color space.

• Korean Journal of Remote Sensing
• /
• v.38 no.6_1
• /
• pp.1505-1514
• /
• 2022

Image matching is a crucial preprocessing step for effective utilization of multi-temporal and multi-sensor very high resolution (VHR) satellite images. Deep learning (DL) method which is attracting widespread interest has proven to be an efficient approach to measure the similarity between image pairs in quick and accurate manner by extracting complex and detailed features from satellite images. However, Image matching of VHR satellite images remains challenging due to limitations of DL models in which the results are depending on the quantity and quality of training dataset, as well as the difficulty of creating training dataset with VHR satellite images. Therefore, this study examines the feasibility of DL-based method in matching pair extraction which is the most time-consuming process during image registration. This paper also aims to analyze factors that affect the accuracy based on the configuration of training dataset, when developing training dataset from existing multi-sensor VHR image database with bias for DL-based image matching. For this purpose, the generated training dataset were composed of correct matching pairs and incorrect matching pairs by assigning true and false labels to image pairs extracted using a grid-based Scale Invariant Feature Transform (SIFT) algorithm for a total of 12 multi-temporal and multi-sensor VHR images. The Siamese convolutional neural network (SCNN), proposed for matching pair extraction on constructed training dataset, proceeds with model learning and measures similarities by passing two images in parallel to the two identical convolutional neural network structures. The results from this study confirm that data acquired from VHR satellite image database can be used as DL training dataset and indicate the potential to improve efficiency of the matching process by appropriate configuration of multi-sensor images. DL-based image matching techniques using multi-sensor VHR satellite images are expected to replace existing manual-based feature extraction methods based on its stable performance, thus further develop into an integrated DL-based image registration framework.

• Journal of the Korea Institute of Building Construction
• /
• v.17 no.6
• /
• pp.545-557
• /
• 2017

The study is about the efficient alternative to concrete surface in the field of visual inspection technology for deteriorated infrastructure. By combining industrial drones and deep learning based image analysis techniques with traditional visual inspection and research, we tried to reduce manpowers, time requirements and costs, and to overcome the height and dome structures. On board device mounted on drones is consisting of a high resolution camera for detecting cracks of more than 0.3 mm, a lidar sensor and a embeded image processor module. It was mounted on an industrial drones, took sample images of damage from the site specimen through automatic flight navigation. In addition, the damege parts of the site specimen was used to measure not only the width and length of cracks but white rust also, and tried up compare them with the final image analysis detected results. Using the image analysis techniques, the damages of 54ea sample images were analyzed by the segmentation - feature extraction - decision making process, and extracted the analysis parameters using supervised mode of the deep learning platform. The image analysis of newly added non-supervised 60ea image samples was performed based on the extracted parameters. The result presented in 90.5 % of the damage detection rate.