• 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.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.9
• /
• pp.2882-2903
• /
• 2022

Image super-resolution (SR) processing is of great value in the fields of digital image processing, intelligent security, film and television production and so on. This paper proposed a densely connected deep learning network based on cascade architecture, which can be used to solve the problem of super-resolution in the field of image quality enhancement. We proposed a more efficient residual scaling dense block (RSDB) and the multi-channel cascade architecture to realize more efficient feature reuse. Also we proposed a hybrid loss function based on L₁ error and L_∞ error to achieve better L_∞ error performance. The experimental results show that the overall performance of the network is effectively improved on cascade architecture and residual scaling. Compared with the residual dense net (RDN), the PSNR / SSIM of the new method is improved by 2.24% / 1.44% respectively, and the L_∞ performance is improved by 3.64%. It shows that the cascade connection and residual scaling method can effectively realize feature reuse, improving the residual convergence speed and learning efficiency of our network. The L_∞ performance is improved by 11.09% with only a minimal loses of 1.14% / 0.60% on PSNR / SSIM performance after adopting the new loss function. That is to say, the L_∞ performance can be improved greatly on the new loss function with a minor loss of PSNR / SSIM performance, which is of great value in L_∞ error sensitive tasks.

• International Journal of Internet, Broadcasting and Communication
• /
• v.12 no.4
• /
• pp.83-92
• /
• 2020

A recent study to determine the fall is focused on analyzing fall motions using a recurrent neural network (RNN), and uses a deep learning approach to get good results for detecting human poses in 2D from a mono color image. In this paper, we investigated the improved detection method to estimate the position of the head and shoulder key points and the acceleration of position change using the skeletal key points information extracted using PoseNet from the image obtained from the 2D RGB low-cost camera, and to increase the accuracy of the fall judgment. In particular, we propose a fall detection method based on the characteristics of post-fall posture in the fall motion analysis method and on the velocity of human body skeleton key points change as well as the ratio change of body bounding box's width and height. The public data set was used to extract human skeletal features and to train deep learning, GRU, and as a result of an experiment to find a feature extraction method that can achieve high classification accuracy, the proposed method showed a 99.8% success rate in detecting falls more effectively than the conventional primitive skeletal data use method.

• Journal of Korea Multimedia Society
• /
• v.22 no.11
• /
• pp.1242-1250
• /
• 2019

Super resolution technique aims to convert a low-resolution image with coarse details to a corresponding high-resolution image with refined details. In the past decades, the performance is greatly improved due to progress of deep learning models. However, universal solution for various objects is a still challenging issue. We observe that learning super resolution with a general dataset has poor performance on faces. In this paper, we propose a super resolution fusion scheme that works well for both general- and face datasets to achieve more universal solution. In addition, object-specific feature extractor is employed for better reconstruction performance. In our experiments, we compare our fusion image and super-resolved images from one- of the state-of-the-art deep learning models trained with DIV2K and FFHQ datasets. Quantitative and qualitative evaluates show that our fusion scheme successfully works well for both datasets. We expect our fusion scheme to be effective on other objects with poor performance and this will lead to universal solutions.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.15 no.3
• /
• pp.521-526
• /
• 2020

This paper presents an improved HOLP neural network that adds 25 average values to a typical HOLP neural network using 25 feature vector values as input values by applying high-order local autocorrelation function, which is excellent for extracting immutable feature values of iris images. Compared with deep learning structures with different types, we compared the recognition rate of iris recognition using Back-Propagation neural network, which shows excellent performance in voice and image field, and synthetic product neural network that integrates feature extractor and classifier.

• Journal of Korea Multimedia Society
• /
• v.24 no.7
• /
• pp.890-902
• /
• 2021

Although the object detection accuracy with still images has been significantly improved with the advance of deep learning techniques, the object detection problem with video data remains as a challenging problem due to the real-time requirement and accuracy drop with occlusion. In this research, we propose a method in pig detection for video monitoring environment. First, we determine a motion, from a video data obtained from a tilted-down-view camera, based on the average size of each pig at each location with the training data, and extract key frames based on the motion information. For each key frame, we then apply YOLO, which is known to have a superior trade-off between accuracy and execution speed among many deep learning-based object detectors, in order to get pig's bounding boxes. Finally, we merge the bounding boxes between consecutive key frames in order to reduce false positive and negative cases. Based on the experiment results with a video data set obtained from a pig farm, we confirmed that the pigs could be detected with an accuracy of 97% at a processing speed of 37fps.

• Journal of Biomedical Engineering Research
• /
• v.43 no.6
• /
• pp.441-447
• /
• 2022

In this paper, we propose a pill identification model using engraved text feature and image feature such as shape and color, and compare it with an identification model that does not use engraved text feature to verify the possibility of improving identification performance by improving recognition rate of the engraved text. The data consisted of 100 classes and used 10 images per class. The engraved text feature was acquired through Keras OCR based on deep learning and 1D CNN, and the image feature was acquired through 2D CNN. According to the identification results, the accuracy of the text recognition model was 90%. The accuracy of the comparative model and the proposed model was 91.9% and 97.6%. The accuracy, precision, recall, and F1-score of the proposed model were better than those of the comparative model in terms of statistical significance. As a result, we confirmed that the expansion of the range of feature improved the performance of the identification model.

• East Asian mathematical journal
• /
• v.39 no.2
• /
• pp.93-117
• /
• 2023

The aim of this study is to explore the pre-service math teachers' characteristics of education to develop their AI literacy and SW competency, and to derive some implications. We conducted a 14-hours AI and SW education program for pre-service teachers with theory and practice, and an analysis on class observation data, video frames of classes and interview, Python programming assignments and papers. The results of this case study for 3 pre-service teachers are as follows. First, two students understood artificial neural network and deep learning system accurately, furthermore, all students conducted a couple of explorations related with performance improvement of deep learning system with interest. Second, coding and exploration activities using Python improved students' computational thinking as well as SW competency, which help them give convergence education in the future. Third, they responded positively to the necessity of AI literacy and SW competency development, and to applying coding to math class. Lastly, it's necessary to endeavor to give a coding education to the student's eye level according to his or her prerequisite and to ease the burden of student's studying AI technology.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.18 no.2
• /
• pp.51-58
• /
• 2023

This paper proposes a deep learning method for estimating the heart rate from facial videos. Our proposed method estimates remote photoplethysmography (rPPG) signals to predict the heart rate. Although there have been proposed several methods for estimating rPPG signals, most previous methods can not be utilized in low-power single board computers due to their computational complexity. To address this problem, we construct a lightweight student model and employ a knowledge distillation technique to reduce the performance degradation of a deeper network model. The teacher model consists of 795k parameters, whereas the student model only contains 24k parameters, and therefore, the inference time was reduced with the factor of 10. By distilling the knowledge of the intermediate feature maps of the teacher model, we improved the accuracy of the student model for estimating the heart rate. Experiments were conducted on the UBFC-rPPG dataset to demonstrate the effectiveness of the proposed method. Moreover, we collected our own dataset to verify the accuracy and processing time of the proposed method on a real-world dataset. Experimental results on a NVIDIA Jetson Nano board demonstrate that our proposed method can infer the heart rate in real time with the mean absolute error of 2.5183 bpm.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.38 no.5
• /
• pp.443-453
• /
• 2020

The automatic land category extraction method was proposed, and the accuracy was evaluated by learning the aerial photo characteristics by land category in the border area with various restrictions on the acquisition of geospatial data. As experimental data, this study used four years' worth of published aerial photos as well as serial cadastral maps from the same time period. In evaluating the results of land category extraction by learning features from different temporal and spatial ranges of aerial photos, it was found that land category extraction accuracy improved as the temporal and spatial ranges increased. Moreover, the greater the diversity and quantity of provided learning images, the less the results were affected by the quality of images at a specific time to be extracted, thus generally demonstrating accurate and practical land category feature extraction.