• The Journal of the Korea Contents Association
• /
• v.21 no.1
• /
• pp.552-560
• /
• 2021

Currently, the domestic aquaculture industry is pursuing smartization, but it is still proceeding with human subjective judgment in many processes in the aquaculture stage. The prerequisite for the smart aquaculture industry is to effectively grasp the condition of fish in the farm. If real-time monitoring is possible by identifying the number of fish populations, size, pathways, and speed of movement, various forms of automation such as automatic feed supply and disease determination can be carried out. In this study, we proposed an algorithm to identify the state of fish in real time using underwater video data. The fish detection performance was compared and evaluated by applying the latest deep learning-based object detection models, and an algorithm was proposed to measure fish object identification, path tracking, and moving speed in continuous image frames in the video using the fish detection results. The proposed algorithm showed 92% object detection performance (based on F1-score), and it was confirmed that it effectively tracks a large number of fish objects in real time on the actual test video. It is expected that the algorithm proposed in this paper can be effectively used in various smart farming technologies such as automatic feed feeding and fish disease prediction in the future.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.39 no.3
• /
• pp.141-148
• /
• 2021

This paper presents a method to detect zebra-crossing using deep learning which combines SegNet and ResNet. For the blind, a safe crossing system is important to know exactly where the zebra-crossings are. Zebra-crossing detection by deep learning can be a good solution to this problem and robotic vision-based assistive technologies sprung up over the past few years, which focused on specific scene objects using monocular detectors. These traditional methods have achieved significant results with relatively long processing times, and enhanced the zebra-crossing perception to a large extent. However, running all detectors jointly incurs a long latency and becomes computationally prohibitive on wearable embedded systems. In this paper, we propose a model for fast and stable segmentation of zebra-crossing from captured images. The model is improved based on a combination of SegNet and ResNet and consists of three steps. First, the input image is subsampled to extract image features and the convolutional neural network of ResNet is modified to make it the new encoder. Second, through the SegNet original up-sampling network, the abstract features are restored to the original image size. Finally, the method classifies all pixels and calculates the accuracy of each pixel. The experimental results prove the efficiency of the modified semantic segmentation algorithm with a relatively high computing speed.

• Journal of the Korean Society for Aeronautical & Space Sciences
• /
• v.48 no.12
• /
• pp.977-985
• /
• 2020

KOMPSAT-3A launched in 2015 provides Middle InfraRed(MIR) images with 3.3~5.2㎛. Though the satellite provide high resolution images for estimating bright temperature of ground objects, it is different from existing satellites developed for natural science purposes. An atmospheric compensation process is essential in order to estimate the surface brightness temperature from a single channel MIR image of KOMPSAT-3A. However, even after the atmospheric compensation process, there is a brightness temperature error due to various factors. In this paper, we analyzed the cause of the brightness temperature estimation error by tracking signal flow from camera physical characteristics to image processing. Also, we study on possibility of improvement of MIR brightness temperature bias error of KOMPSAT-3A using GEOKOMPSAT-2A. After bias compensation of a real nighttime image with a large bias error, it was confirmed that the surface brightness temperature of KOMPSAT-3A and GEOKOMPSAT-2A have correlation. We expect that the GEOKOMPSAT-2A images will be helpful to improve MIR brightness temperature bias error of KOMPSAT-3A.

• Journal of Cadastre & Land InformatiX
• /
• v.51 no.2
• /
• pp.73-82
• /
• 2021

According to the government's announcement of the safety management enhancement policy for small and medium-sized private construction sites, the subject of mandatory CCTV installation has been expanded from large construction sites to small and medium-sized construction sites. However, since the existing CCTV at construction sites has been used for simple control for safety management, so research is needed for monitoring of construction sites. Therefore, in this study, three vanishing points were calculated based on a single image taken with a monocular camera, and then a camera matrix containing interior orientation parameters information was determined. And the accuracy was verified by calculating the height of the target object from the height of the reference object. Through height determination experiments using vanishing points based on a monocular camera, it was possible to determine the height of target objects only with a single image without separately surveying of ground control points. As a result of the accuracy evaluation, the root mean square error was ±0.161m. Therefore, it is determined that the progress of construction work at the construction sites can be monitored through the single image taken using the single camera.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.22 no.2
• /
• pp.63-70
• /
• 2022

Although 2D Object detection has been largely improved in the past years with the advance of deep learning methods and the use of large labeled image datasets, 3D object detection from 2D imagery is a challenging problem in a variety of applications such as robotics, due to the lack of data and diversity of appearances and shapes of objects within a category. Google has just announced the launch of Objectron that has a novel data pipeline using mobile augmented reality session data. However, it also is corresponding to 2D-driven 3D object detection technique. This study explores more mature 2D object detection method, and applies its 2D projection to Objectron 3D lifting system. Most object detection methods use bounding boxes to encode and represent the object shape and location. In this work, we explore a stochastic representation of object regions using Gaussian distributions. We also present a similarity measure for the Gaussian distributions based on the Hellinger Distance, which can be viewed as a stochastic Intersection-over-Union. Our experimental results show that the proposed Gaussian representations are closer to annotated segmentation masks in available datasets. Thus, less accuracy problem that is one of several limitations of Objectron can be relaxed.

• Tunnel and Underground Space
• /
• v.32 no.2
• /
• pp.131-143
• /
• 2022

The lunar exploration autonomous vehicle operates based on the lunar topography information obtained from real-time image characterization. For highly accurate topography characterization, a large number of training images with various background conditions are required. Since the real lunar topography images are difficult to obtain, it should be helpful to be able to generate mimic lunar image data artificially on the basis of the planetary analogs site images and real lunar images available. In this study, we aim to artificially create lunar topography images by using the location information-based style transfer algorithm known as Wavelet Correct Transform (WCT2). We conducted comparative experiments using lunar analog site images and real lunar topography images taken during China's and America's lunar-exploring projects (i.e., Chang'e and Apollo) to assess the efficacy of our suggested approach. The results show that the proposed techniques can create realistic images, which preserve the topography information of the analog site image while still showing the same condition as an image taken on lunar surface. The proposed algorithm also outperforms a conventional algorithm, Deep Photo Style Transfer (DPST) in terms of temporal and visual aspects. For future work, we intend to use the generated styled image data in combination with real image data for training lunar topography objects to be applied for topographic detection and segmentation. It is expected that this approach can significantly improve the performance of detection and segmentation models on real lunar topography images.

• KIPS Transactions on Computer and Communication Systems
• /
• v.11 no.11
• /
• pp.403-410
• /
• 2022

Non-Fungible Tokens (NFTs) created on the basis of blockchain have their own unique value, so they cannot be forged or exchanged with other tokens or coins. Using these characteristics, NFTs can be issued to digital assets such as images, videos, artworks, game characters, and items to claim ownership of digital assets among many users and objects in cyberspace, as well as proving the original. However, interest in NFTs exploded from the beginning of 2020, causing a lot of load on the blockchain network, and as a result, users are experiencing problems such as delays in computational processing or very large fees in the mining process. Additionally, all actions of users are stored in the blockchain, and digital assets are stored in a blockchain-based distributed file storage system, which may unnecessarily expose the personal information of users who do not want to identify themselves on the Internet. In this paper, we propose an NFT platform using cloud computing, access gate, conversion table, and cloud ID to improve usability and privacy problems that occur in existing system. For performance comparison between local and cloud systems, we measured the gas used for smart contract deployment and NFT-issued transaction. As a result, even though the cloud system used the same experimental environment and parameters, it saved about 3.75% of gas for smart contract deployment and about 4.6% for NFT-generated transaction, confirming that the cloud system can handle computations more efficiently than the local system.

• Journal of Space Technology and Applications
• /
• v.1 no.1
• /
• pp.121-130
• /
• 2021

Since the 1960s, the United States has developed and implemented policies to encourage commercial space launches. Specifically, national policies have been implementing to expand the role of commercial space actors, which required establishing a process for private space launches. In the early days of the space age, private launches accounted for a small portion of the total launch rate, but, since the 1990s, the proportion has exploded, with private space companies presenting large projects one after another, accounting for more than 50% of the total launch rate. This diversification of space actors and the increase in orbital space objects have led to changes in the perspectives of existing space environmental management processes. During and after the Cold War, when the space age began, civilian actors' actions were limited, and policies limited their actions, too. So they had little impact on government space activities. However, space technology's entry barrier has lowered since, and policies to facilitate commercial space launches have been implemented for a long, and the accumulated amount of space waste over the past 60 years is also threatening the safety, stability, and sustainability of space use. This paper examined how the United States, the most active country in commercial space launches, has managed commercial space launches. The United States has a Space Traffic Management (STM), distributed to departments such as the Department of Defense, Department of Commerce, Department of Transport, NASA, etc. A review of changes in U.S STM management policy could also provide implications for us to manage commercial space launches in Korea.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.1-16
• /
• 2022

Despite recent breakthroughs in deep learning and computer vision fields, the pixel-wise identification of tiny objects in high-resolution images with complex disturbances remains challenging. This study proposes a modified U-net for tiny crack segmentation in real-world steel-box-girder bridges. The modified U-net adopts the common U-net framework and a novel Self-Attention-Self-Adaption (SASA) neuron as the fundamental computing element. The Self-Attention module applies softmax and gate operations to obtain the attention vector. It enables the neuron to focus on the most significant receptive fields when processing large-scale feature maps. The Self-Adaption module consists of a multiplayer perceptron subnet and achieves deeper feature extraction inside a single neuron. For data augmentation, a grid-based crack random elastic deformation (CRED) algorithm is designed to enrich the diversities and irregular shapes of distributed cracks. Grid-based uniform control nodes are first set on both input images and binary labels, random offsets are then employed on these control nodes, and bilinear interpolation is performed for the rest pixels. The proposed SASA neuron and CRED algorithm are simultaneously deployed to train the modified U-net. 200 raw images with a high resolution of 4928 × 3264 are collected, 160 for training and the rest 40 for the test. 512 × 512 patches are generated from the original images by a sliding window with an overlap of 256 as inputs. Results show that the average IoU between the recognized and ground-truth cracks reaches 0.409, which is 29.8% higher than the regular U-net. A five-fold cross-validation study is performed to verify that the proposed method is robust to different training and test images. Ablation experiments further demonstrate the effectiveness of the proposed SASA neuron and CRED algorithm. Promotions of the average IoU individually utilizing the SASA and CRED module add up to the final promotion of the full model, indicating that the SASA and CRED modules contribute to the different stages of model and data in the training process.

• Science of Emotion and Sensibility
• /
• v.25 no.2
• /
• pp.23-30
• /
• 2022

The weight of an object is an important research topic in terms of sense, and objects have size-weight, color-weight, and material-weight illusions due to the influence of size, color, and material and the weight of the object. Although temperature is a very important environmental factor in our daily life, there is a dearth of studies on how the temperature of an object affects its perceived weight. This study analyzed the effect of an object's temperature on the perceived weight. A stainless steel cup, in which weight and temperature are easily adjustable, was selected as a perceived weight measurement object; 5 temperature levels (0, 9, 20, 40, 70 degrees) and 2 weight levels (250, 400 g) were set. A total of 40 healthy men and women in their 20s participated in the experiment. The weight of the given cup compared to the reference cup was evaluated according to the modulus method. The study revealed that both temperature and weight significantly affected the perceived weight. The effect of temperature on the weight was different depending on the weight of the object. When the cup's weight was small (250 g), the temperature of the cup did not affect the weight. However, the perceived weight of a large cup (400 g) increased at a low temperature. This result suggests that the effect of temperature on the weight of an object depends on the size-weight illusion.