• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2022.05a
• /
• pp.396-398
• /
• 2022

The endoscopic pathological findings of gastrointestinal tract (GIT) are important in the early diagnosis of colorectal cancer. Deep learning based on convolutional nueral network (CNN) has been implemented to solve the subjective analysis problem and to increase the performance of early detection of pathological findings. However, the desired performance is yet to be achieved and CNNs are computationally complex. To solve these problems, in this paper, we propose a vision transformer based endoscopic pathological findings classification for the early detection of colorectal cancer. Publicly available endoscopic images with three pathological findings, including esophagitis, polyps, and ulcerative colitis, each with 1000 images were used. Using our approach, we have achieved a test accuracy of 98% in classifying the three pathological findings.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.16 no.5
• /
• pp.1597-1610
• /
• 2022

At present, the main method of high-speed train chassis detection is using computer vision technology to extract keypoints from two related chassis images firstly, then matching these keypoints to find the pixel-level correspondence between these two images, finally, detection and other steps are performed. The quality and accuracy of image matching are very important for subsequent defect detection. Current traditional matching methods are difficult to meet the actual requirements for the generalization of complex scenes such as weather, illumination, and seasonal changes. Therefore, it is of great significance to study the high-speed train image matching method based on deep learning. This paper establishes a high-speed train chassis image matching dataset, including random perspective changes and optical distortion, to simulate the changes in the actual working environment of the high-speed rail system as much as possible. This work designs a convolutional neural network to intensively extract keypoints, so as to alleviate the problems of current methods. With multi-level features, on the one hand, the network restores low-level details, thereby improving the localization accuracy of keypoints, on the other hand, the network can generate robust keypoint descriptors. Detailed experiments show the huge improvement of the proposed network over traditional methods.

• Journal of Yeungnam Medical Science
• /
• v.40 no.2
• /
• pp.212-217
• /
• 2023

We report a case of transient osteoporosis of the hip with a femoral neck fracture found during follow-up. A 53-year-old man presented with left hip pain without trauma. The pain did not improve after 2 weeks and he was brought to our hospital by ambulance. Magnetic resonance imaging (MRI) of the left hip joint showed diffuse edema in the bone marrow, which was identified by low signal intensity on T1-weighted images, high signal intensity on T2-weighted images, and increased signal intensity on short tau inversion recovery. This edema extended from the femoral head and neck to the intertrochanteric area. He was diagnosed with transient osteoporosis of the left hip. Rest gradually improved his pain; however, 3 weeks later, his left hip pain worsened without trauma. X-ray, computed tomography, and MRI results of the hip joint demonstrated a left femoral neck fracture, and osteosynthesis was performed. Differential diagnoses included avascular necrosis of the femoral head, infection, complex regional pain syndrome, rheumatoid arthritis, leukemia, and other cancers. Transient osteoporosis of the hip generally has a good prognosis with spontaneous remission within a few months to 1 year. However, a sufficient length of follow-up from condition onset to full recovery is necessary to avoid all probable complications such as fractures.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.5
• /
• pp.737-745
• /
• 2023

This paper discusses a study on developing an artificial intelligence model to detect incidents of fires in distribution switchboard using thermal images. The objective of the research is to preprocess collected thermal images into suitable data for object detection models and design a model capable of determining the occurrence of fires within distribution panels. The study utilizes thermal image data from AI-HUB's industrial complex for training. Two CNN-based deep learning object detection algorithms, namely Faster R-CNN and RetinaNet, are employed to construct models. The paper compares and analyzes these two models, ultimately proposing the optimal model for the task.

• Journal of Animal Science and Technology
• /
• v.65 no.6
• /
• pp.1254-1269
• /
• 2023

This study aims to predict the change in corn share according to the grazing of 20 gestational sows in a mature corn field by taking images with a camera-equipped unmanned air vehicle (UAV). Deep learning based on convolutional neural networks (CNNs) has been verified for its performance in various areas. It has also demonstrated high recognition accuracy and detection time in agricultural applications such as pest and disease diagnosis and prediction. A large amount of data is required to train CNNs effectively. Still, since UAVs capture only a limited number of images, we propose a data augmentation method that can effectively increase data. And most occupancy prediction predicts occupancy by designing a CNN-based object detector for an image and counting the number of recognized objects or calculating the number of pixels occupied by an object. These methods require complex occupancy rate calculations; the accuracy depends on whether the object features of interest are visible in the image. However, in this study, CNN is not approached as a corn object detection and classification problem but as a function approximation and regression problem so that the occupancy rate of corn objects in an image can be represented as the CNN output. The proposed method effectively estimates occupancy for a limited number of cornfield photos, shows excellent prediction accuracy, and confirms the potential and scalability of deep learning.

• ETRI Journal
• /
• v.46 no.2
• /
• pp.165-174
• /
• 2024

We propose a method to suppress the speckle noise and blur effects of the light field extracted from a hologram using a deep-learning technique. The light field can be extracted by bandpass filtering in the hologram's frequency domain. The extracted light field has reduced spatial resolution owing to the limited passband size of the bandpass filter and the blurring that occurs when the object is far from the hologram plane and also contains speckle noise caused by the random phase distribution of the three-dimensional object surface. These limitations degrade the reconstruction quality of the hologram resynthesized using the extracted light field. In the proposed method, a deep-learning model based on a generative adversarial network is designed to suppress speckle noise and blurring, resulting in improved quality of the light field extracted from the hologram. The model is trained using pairs of original two-dimensional images and their corresponding light-field data extracted from the complex field generated by the images. Validation of the proposed method is performed using light-field data extracted from holograms of objects with single and multiple depths and mesh-based computer-generated holograms.

• Spatial Information Research
• /
• v.21 no.1
• /
• pp.63-73
• /
• 2013

Geographic features are reflected in satellite images, which contain characteristic elements. Information on changes can be obtained through a comparison of images taken at different times. If multi-temporal images can be classified through the use of an unsupervised method, this is likely to improve the accuracy of image classification and contribute to various applications. A rule-based image classification algorithm for automatic processing without human involvement has been developed, but it must be verified that its results are not affected by imperfect elements. In this study, Landsat images of Jeju Island were used to carry out a rule-based image classification. The application results were examined for complex cases, including the presence of clouds in the images, different photographed times, and the type of target area, such as city, mountain, or field. The presence of clouds did not affect calculations, and appropriate classification rules were applied, depending on the different photographed times. The expansion of the urban areas of Jeju and the increase of facilities such as vinyl greenhouses in Seoguipo were identified. Furthermore, space information changes and accurate classifications for Jeju Island were obtained. With the goal of performing high-quality unsupervised classifications, measures to generalize and improve the methods employed were searched for. The findings of this study could be used in time-series analyses of images for various applications, including urban development and environmental change monitoring.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.6 no.1
• /
• pp.124-132
• /
• 2002

In this paper, we present a vision algorithm and method for input image improvement and preprocessing of dented and raised characters on the sidewall of tires. we define optical condition between reflect coefficient and reflectance by the physical vector calculate. On the contrary this work will recognize the engraved characters using the computer vision technique. Tire input images have all most same grey levels between the characters and backgrounds. The reflectance is little from a tire surface. therefore, it's very difficult segment the characters from the background. Moreover, one side of the character string is raised and the other is dented. So, the captured images are varied with the angle of camera and illumination. For optimum Input images, the angle between camera and illumination was found out to be with in 90$^{\circ}$. In addition, We used complex filtering with low-pass and high-pass band filters to improve input images, for clear input images. Finally we define equation reflect coefficient and reflectance. By doing this, we obtained good images of tires for pattern recognition.

• 대한예방의학회:학술대회논문집
• /
• 1994.02b
• /
• pp.169-179
• /
• 1994

B-mode ultrasound is being used to assess carotid atherosclerosis in epidemiological studies and clinical trials. Recently the interpretation of measurements made from ultrasound images has been questioned. This study examines the anatomical correlates of B-mode ultrasound of carotid arteries in vitro and in situ in cadavers. Twenty-seven segments of human carotid artery were collected at autopsy. pressure perfusion fixed in buffered 2.5% gluteraldehyde and 4% paraformaldehyde and imaged using an ATL UM-8 (10 MHz single crystal mechanical probe). Each artery was then frozen, sectioned and stained with van Gieson or elastin van Gieson. The thickness of the intima. media and adventitia were measured 'to an accuracy of 0.01 mm from histological sections using a calibrated eye graticule on a light microscope. Shrinkage artifact induced by histological preparation was determined to be 7.8%. Digitised ultra sound images of the artery wall were analysed off-line. The distance from the leading edge of the first interface ($LE_{1}$) to the leading edge of the second interface ($LE_2$) was measured using a dedicated programme. $LE_{1}$-$LE_{2}$ measurements were correlated against histological measurements corrected for shrinkage. Mean values for the far wall were: ultra sound $LE_{1}$-$LE_{2}$ (0.97 mm, S.D. 0.26), total wall thickness (1.05 mm, S.D. 0.37), adventitia (0.35 mm, S.D. 0.16), media (0.61 mm, S.D. 0.18). intima (0.09 mm, S.D. 0.13). Ultrasound measurements corresponded best with total wall thickness, rather than elastin or the intima-media complex. Excision of part of the intima plus media or removal of the adventitia resulted in a corresponding decrease in the $LE_{1}$-$LE_{2}$ distance of the B-mode image. Furthermore. increased wall thickness due to intimal atherosclerotic thickening correlated well with $LE_{1}$-$LE_{2}$ distance of the B-mode images. B-mode images obtained from the carotid arteries in situ in four cadavers also corresponded best with total wall thickness measured from histological sections and not with the thickness of the intima plus media. In conclusion, the $LE_{1}$-$LE_{2}$ distance measured on B-mode images of the carotid artery best represents total wall thickness of intima plus media plus adventitia and not intima plus media alone.

• Korean Association for Visual Culture
• /
• v.17
• /
• pp.65-92
• /
• 2011

This thesis investigates the complex functions of the magic lantern in illusionism and enlightment which was the most popular visual media and the direct ancestor of cinema. Especially, the thesis focuses on the characteristics of magic lantern's images which had been varied with the scientific and technological development. During the early period of the magic lantern, from the late 18th century to the beginning of the 19th century, it frightened viewers by showing magic images with ghosts and spectres, 'phantasmagoria', and wondered with images of natural catastropes and interesting stories like fables and fairy tales, which fulfilled the entertainment function. Since the mid 19th century the magic lantern began to show not only pictures of the 'scientific themes' on the earth, nature and human, but also them of the ethnological on the far, exotic worlds like Africa, Amazon and Syberia etc. from the European perspective. These contents conducted the educative function and contributed to the process of Enlightment to the peoples in the pre-modern age. The two functions of the magic lantern such as entertainment and education had been neither historically followed, nor clearly divided, but the one was predominant according to the development of lantern techniques as well as the changes of the world view and the culture of the time. The entertainment function of the magic lantern based on the visual fantacy did exist in the late 19th century further, and also in the late industrial society, even in the age of highly developed science and technology, viewers want rather 're-enchantment' by illusionism than facts and truths on the reality. This is an essential characteristic of the moving image media, as it had already been presented in the images of the magic lantern.