• The Journal of Bigdata
• /
• v.7 no.1
• /
• pp.37-47
• /
• 2022

Autonomous delivery operation data is the key to driving a paradigm shift for last-mile delivery in the Corona era. To bridge the technological gap between domestic autonomous delivery robots and overseas technology-leading countries, large-scale data collection and verification that can be used for artificial intelligence training is required as the top priority. Therefore, overseas technology-leading countries are contributing to verification and technological development by opening AI training data in public data that anyone can use. In this paper, 326 objects were collected to trainn autonomous delivery robots, and artificial intelligence models such as Mask r-CNN and Yolo v3 were trained and verified. In addition, the two models were compared based on comparison and the elements required for future autonomous delivery robot research were considered.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1315-1318
• /
• 2003

In this paper, a neural network based approach using a self-organizing feature map is proposed for the segmentation of X ray images. A number of algorithms based on such approaches as histogram analysis, region growing, edge detection and pixel classification have been proposed for segmentation of general images. However, few approaches have been applied to X ray image segmentation because of blur of the X ray image and vagueness of its edge, which are inherent properties of X ray images. To this end, we develop a new model based on the neural network to detect objects in a given X ray image. The new model utilizes Mumford-Shah functional incorporating with a modified adaptive SOFM. Although Mumford-Shah model is an active contour model not based on the gradient of the image for finding edges in image, it has some limitation to accurately represent object images. To avoid this criticism, we utilize an adaptive self organizing feature map developed earlier by the authors.[1] It's learning rule is derived from Mumford-Shah energy function and the boundary of blurred and vague X ray image. The evolution of the neural network is shown to well segment and represent. To demonstrate the performance of the proposed method, segmentation of an industrial part is solved and the experimental results are discussed in detail.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.13 no.3
• /
• pp.1689-1701
• /
• 2019

In recent years, personal videos have been shared online due to the popular uses of portable devices, such as smartphones and action cameras. A recent report predicted that 80% of the Internet traffic will be video content by the year 2021. Several studies have been conducted on the detection of main video events to manage a large scale of videos. These studies show fairly good performance in certain genres. However, the methods used in previous studies have difficulty in detecting events of personal video. This is because the characteristics and genres of personal videos vary widely. In a research, we found that adding a dataset with the right perspective in the study improved performance. It has also been shown that performance improves depending on how you extract keyframes from the video. we selected frame segments that can represent video considering the characteristics of this personal video. In each frame segment, object, location, food and audio features were extracted, and representative vectors were generated through a CNN-based recurrent model and a fusion module. The proposed method showed mAP 78.4% performance through experiments using LSVC data.

• Progress in Medical Physics
• /
• v.31 no.3
• /
• pp.111-123
• /
• 2020

Deep learning (DL) is a subset of machine learning and artificial intelligence that has a deep neural network with a structure similar to the human neural system and has been trained using big data. DL narrows the gap between data acquisition and meaningful interpretation without explicit programming. It has so far outperformed most classification and regression methods and can automatically learn data representations for specific tasks. The application areas of DL in radiation oncology include classification, semantic segmentation, object detection, image translation and generation, and image captioning. This article tries to understand what is the potential role of DL and what can be more achieved by utilizing it in radiation oncology. With the advances in DL, various studies contributing to the development of radiation oncology were investigated comprehensively. In this article, the radiation treatment process was divided into six consecutive stages as follows: patient assessment, simulation, target and organs-at-risk segmentation, treatment planning, quality assurance, and beam delivery in terms of workflow. Studies using DL were classified and organized according to each radiation treatment process. State-of-the-art studies were identified, and the clinical utilities of those researches were examined. The DL model could provide faster and more accurate solutions to problems faced by oncologists. While the effect of a data-driven approach on improving the quality of care for cancer patients is evidently clear, implementing these methods will require cultural changes at both the professional and institutional levels. We believe this paper will serve as a guide for both clinicians and medical physicists on issues that need to be addressed in time.

• IEMEK Journal of Embedded Systems and Applications
• /
• v.17 no.4
• /
• pp.217-228
• /
• 2022

Currently, there are increasing demands for applying deep neural networks (DNNs) in the embedded domain such as classification and object detection. The DNN processing in embedded domain often requires custom hardware such as NPU for acceleration due to the constraints in power, performance, and area. Processing DNN models requires a large amount of data, and its seamless transfer to NPU is crucial for performance. In this paper, we developed a cycle-accurate NPU simulator to evaluate diverse NPU microarchitectures. In addition, we propose a novel technique for reducing the number of memory accesses when processing convolutional layers in convolutional neural networks (CNNs) on the NPU. The main idea is to reuse data with memory interleaving, which recycles the overlapping data between previous and current input windows. Data memory interleaving makes it possible to quickly read consecutive data in unaligned locations. We implemented the proposed technique to the cycle-accurate NPU simulator and measured the performance with LeNet-5, VGGNet-16, and ResNet-50. The experiment shows up to 2.08x speedup in processing one convolutional layer, compared to the baseline.

• Journal of The Korean Society of Agricultural Engineers
• /
• v.59 no.3
• /
• pp.21-28
• /
• 2017

The main goal of this paper is to assess application of UAV (Unmanned Aerial Vehicle) remote sensing and GIS based images in detection and measuring of rice field drought area in South Korea. Drought is recurring feature of the climatic events, which often hit South Korea, bringing significant water shortages, local economic losses and adverse social consequences. This paper describes the assesment of the near-realtime drought damage monitoring and reporting system for the agricultural drought region. The system is being developed using drought-related vegetation characteristics, which are derived from UAV remote sensing data. The study area is $3.07km^2$ of Wonbuk-myeon, Taean-gun, Chungnam in South Korea. UAV images were acquired three times from July 4 to October 29, 2015. Three images of the same test site have been analysed by object-based image classification technique. Drought damaged paddy rices reached $754,362m^2$, which is 47.1 %. The NongHyeop Agricultural Damage Insurance accepted agricultural land of 4.6 % ($34,932m^2$). For paddy rices by UAV investigation, the drought monitoring and crop productivity was effective in improving drought assessment method.

• Journal of Sensor Science and Technology
• /
• v.32 no.2
• /
• pp.88-94
• /
• 2023

Recently, piezoelectric tactile sensors have garnered considerable attention in the field of texture recognition owing to their high sensitivity and high-frequency detection capability. Despite their remarkable potential, improving their mechanical flexibility to attach to complex surfaces remains challenging. In this study, we present a flexible piezoelectric sensor that can be bent to an extremely small radius of up to 2.5 mm and still maintain good electrical performance. The proposed sensor was fabricated by controlling the thickness that induces internal stress under external deformation. The fabricated piezoelectric sensor exhibited a high sensitivity of 9.3 nA/kPa ranging from 0 to 10 kPa and a wide frequency range of up to 1 kHz. To demonstrate real-time texture recognition by rubbing the surface of an object with our sensor, nine sets of fabric plates were prepared to reflect their material properties and surface roughness. To extract features of the objects from the detected sensing data, we converted the analog dataset to short-term Fourier transform images. Subsequently, texture recognition was performed using a convolutional neural network with a classification accuracy of 97%.

• Proceedings of the Korean Institute of Building Construction Conference
• /
• 2021.11a
• /
• pp.217-218
• /
• 2021

The development of artificial intelligence in the field of construction and construction is revitalizing. The performance and development techniques of artificial intelligence are changing rapidly, but if you look at the cases of domestic construction sites, they are using technologies from 5 to 7 years ago. It is right to follow a stable method in consideration of commercialization, but the previous AI development method requires more manpower and time to develop than the current technology. In addition, in order to actively utilize artificial intelligence technology, customized artificial intelligence is required to be applied to ever-changing changes in construction sites. it is the reality As a result, even if good AI technology is secured at the construction site, it is reluctant to introduce it because there is no advantage in terms of time and cost compared to the existing method to apply it only to some processes. Currently, an AI technique with a faster development process and accurate recognition has been developed to cope with a fluid situation, so it will be important to understand and introduce the rapidly changing AI development method.

• Journal of the Korean Society of Surveying, Geodesy, Photogrammetry and Cartography
• /
• v.28 no.4
• /
• pp.421-430
• /
• 2010

Image segmentation technique is becoming increasingly important in the field of remote sensing image analysis in areas such as object oriented image classification to extract object regions of interest within images. This paper presents a new method for image segmentation in High Resolution Remote Sensing Image based on Improved Seeded Region Growing (ISRG) and Region merging. Firstly, multi-spectral edge detection was done using an entropy operator in pan-sharpened QuickBird imagery. Then, the initial seeds were automatically selected from the obtained multi-spectral edge map. After automatic selection of significant seeds, an initial segmentation was achieved by applying ISRG to consider spectral and edge information. Finally the region merging process, integrating region texture and spectral information, was carried out to get the final segmentation result. The accuracy assesment was done using the unsupervised objective evaluation method for evaluating the effectiveness of the proposed method. Experimental results demonstrated that the proposed method has good potential for application in the segmentation of high resolution satellite images.

• Progress in Superconductivity
• /
• v.3 no.1
• /
• pp.56-59
• /
• 2001

We have constructed a multi-channel SQUID magnetometer system for localization and classification of magnetic targets. Ten SQUID magnetometers were arranged to measure 5 independent components of 3 $\times$ 3 magnetic field gradient tensor. To get gradient from the difference of magnetic field measurements, we carefully balanced magnetometers. SQUIDs with slotted washer were used for operation in an unshielded laboratory environment, and noise characteristic in the laboratory was measured. With the multi-channel SQUID magnetometer system, we have successfully traced the motion of a bar magnet moving around it at a distance of about 1 m. In the urban environment, the drift of uniform magnetic field due to the irregular motion of a large magnetic body at distance and earth field causes an error in the position calculation, and this results in the distortion of the calculated trajectory. In this paper, we present the architecture and the performance of the system.