• Journal of Broadcast Engineering
• /
• v.27 no.3
• /
• pp.283-294
• /
• 2022

The feature map used in the network for deep learning generally has larger data than the image and a higher compression rate than the image compression rate is required to transmit the feature map. This paper proposes a method for transmitting a pyramid feature map with high compression rate, which is used in a network with an FPN structure that has robustness to object size in deep learning-based image processing. In order to efficiently compress the pyramid feature map, this paper proposes a structure that predicts a pyramid feature map of a level that is not transmitted with pyramid feature map of some levels that transmitted through the proposed prediction network to efficiently compress the pyramid feature map and restores compression damage through the proposed reconstruction network. Suggested mAP, the performance of object detection for the COCO data set 2017 Train images of the proposed method, showed a performance improvement of 31.25% in BD-rate compared to the result of compressing the feature map through VTM12.0 in the rate-precision graph, and compared to the method of performing compression through PCA and DeepCABAC, the BD-rate improved by 57.79%.

• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.3
• /
• pp.381-388
• /
• 2022

In instance segmentation, Mask-RCNN is mostly used as a base model. Increasing the performance of Mask-RCNN is meaningful because it affects the performance of the derived model. Mask-RCNN has a transform module for unifying size of input images. In this paper, to improve the Mask-RCNN, we apply deep-learning-based ASSR to the resizing part in the transform module and inject calculated scale information into the model using IM(Integration Module). The proposed IM improves instance segmentation performance by 2.5 AP higher than Mask-RCNN in the COCO dataset, and in the periment for optimizing the IM location, the best performance was shown when it was located in the 'Top' before FPN and backbone were combined. Therefore, the proposed method can improve the performance of models using Mask-RCNN as a base model.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.16 no.6
• /
• pp.465-471
• /
• 2023

This paper proposes a comparator structure that improves the noise and output delay of a single-slope ADC(SS-ADC) used in CMOS Image Sensor (CIS). To improve the noise and delay characteristics of the output, a comparator structure using the miller effect is designed by inserting a capacitor between the output node of the first stage and the output node of the second stage of the comparator. The proposed comparator structure improves the noise, delay of the output, and layout area by using a small capacitor. The CDS counter used in the single slop ADC is designed using a T-filp flop and bitwise inversion circuit, which improves power consumption and speed. The single-slope ADC also performs dual CDS, which combines analog correlated double sampling (CDS) and digital CDS. By performing dual CDS, image quality is improved by reducing fixed pattern noise (FPN), reset noise, and ADC error. The single-slope ADC with the proposed comparator structure is designed in a 0.18-㎛ CMOS process.

• Journal of the Korean Geotechnical Society
• /
• v.40 no.3
• /
• pp.33-39
• /
• 2024

Inspecting cracks to determine a structure's condition is crucial for accurate safety diagnosis. However, visual crack inspection methods can be subjective and are dependent on field conditions, thereby resulting in low reliability. To address this issue, this study automates the detection of concrete cracks in image data using ResNet, FPN, and the Mask R-CNN components as the backbone, neck, and head of a convolutional neural network. The performance of the proposed model is analyzed using the intersection over the union (IoU). The experimental dataset contained 1,203 images divided into training (70%), validation (20%), and testing (10%) sets. The model achieved an IoU value of 95.83% for testing, and there were no cases where the crack was not detected. These findings demonstrate that the proposed model realized highly accurate detection of concrete cracks in image data.

• 제어로봇시스템학회:학술대회논문집
• /
• 2003.10a
• /
• pp.1727-1731
• /
• 2003

In this paper we proposed backward fuzzy Petri-net to diagnoses faults in power systems by using the time sequence information of protection system. As the complexity of power systems increases, especially in the case of multiple faults or incorrect operation of protective devices, fault diagnosis requires new and systematic methods to the reasoning process, which improves both its accuracy and its efficiency. The fuzzy Petri-net models of protection system are composed of the operating process of protective devices and the fault diagnosis process. Fault diagnosis model, which makes use of the nature of fuzzy Petri-net, is developed to overcome the drawbacks of methods that depend on operator knowledge. The proposed method can reduce processing time and increase accuracy when compared with the traditional methods. And also this method covers online processing of real-time data from SCADA (Supervisory Control and Data Acquisition)

• JSTS:Journal of Semiconductor Technology and Science
• /
• v.12 no.4
• /
• pp.388-396
• /
• 2012

In this paper, a $320{\times}240$ pixel, 80 frame/s CMOS image sensor with a low power dual correlated double sampling (CDS) scheme is presented. A novel 8-bit hold-and-go counter in each column is proposed to obtain 10-bit resolution. Furthermore, dual CDS and a configurable counter scheme are also discussed to realize efficient power reduction. With these techniques, the digital counter consumes at least 43% and at most 61% less power compared with the column-counters type, and the frame rate is approximately 40% faster than the double memory type due to a partial pipeline structure without additional memories. The prototype sensor was fabricated in a Samsung $0.13{\mu}m$ 1P4M CMOS process and used a 4T APS with a pixel pitch of $2.25{\mu}m$. The measured column fixed pattern noise (FPN) is 0.10 LSB.

• Journal of Information Technology Applications and Management
• /
• v.12 no.4
• /
• pp.1-12
• /
• 2005

In general, other conventional researches propose the fuzzy Petri net-based fuzzy reasoning algorithms based on the exhaustive search algorithms. If it can allow the certainty factors representing in the fuzzy production rules to use as the heuristic information, then it can allow the reasoning of rule-based systems to perform fuzzy reasoning in more effective manner. This paper presents a fuzzy Petri net(FPN) model to represent the fuzzy production rules of a rule-based system. Based on the fuzzy Petri net model, a weighted fuzzy reasoning algorithm is proposed to Perform the fuzzy reasoning automatically, This algorithm is more effective and more intelligent reasoning than other reasoning methods because it can perform fuzzy reasoning using the certainty factors which are provided by domain experts as heuristic information

• Proceedings of the KIEE Conference
• /
• 2004.11c
• /
• pp.337-339
• /
• 2004

In this paper, we discuss optimal design of Fuzzy Polynomial Neural Networks by means of Genetic Algorithms(GAs). Proceeding the layer, this model creates the optimal network architecture through the selection and the elimination of nodes by itself. So, there is characteristic of flexibility. We use a triangle and a Gaussian-like membership function in premise part of rules and design the consequent structure by constant and regression polynomial (linear, quadratic and modified quadratic) function between input and output variables. GAs is applied to improve the performance with optimal input variables and number of input variables and order. To evaluate the performance of the GAs-based FPNNs, the models are experimented with the use of Medical Imaging System(MIS) data.

• The KSFM Journal of Fluid Machinery
• /
• v.9 no.2 s.35
• /
• pp.19-24
• /
• 2006

Nano fountain-pen is a novel device to make the constant patterning in micro process using new designed probe. Fountain-pen nanolithography (FPN) is applied for constant supply of liquid in conjunction of patterns and surface variation in the micro process. In this study, nuo fountain-pen is composed with reservoir, micro channels, tip and scondary chamber. Instead of traditional method only using capillary force, liquid can be definitely and exactly injected with membrane pumping by the repulse force of tip. It is dfficult to perform experiments in the micro range so that we carried out a numerical analysis for internal flow, using a commercial code, FlUENT, The velocity, pressure and flow rate are obtained under laminar, unsteady, three-dimensional incompressible flow with no-slip condition, and results are graphically described.

• Proceedings of the Korea Inteligent Information System Society Conference
• /
• 2007.05a
• /
• pp.321-328
• /
• 2007

데이터마이닝은 대용량의 데이터에 숨겨진 의미있고 유용한 패턴과 상관관계를 추출하여 의사결정에 활용하는 작업이다. 그 중에서도 고객 트랜잭션의 데이터베이스에서 아이템 사이에 존재하는 연관규칙을 찾는 것은 중요한 일이 되었다. Apriori 알고리즘 이후 연관규칙을 찾기 위해 대용량 데이터베이스로부터 압축된 의미있는 정보를 저장하기 위한 데이터 구조와 알고리즘들이 제안되어 왔다. 본 논문에서는 정점으로 아이템을 표현하고, 간선으로 두 아이템집합을 표현하는 빈발 패턴 네트워크(FPN)이라 불리는 새 자료 구조를 제안한다. 빈발 패턴 네트워크에서 아이템 사이의 연관 관계를 발견하기 위해 이 구조를 어떻게 효율적으로 사용 하느냐에 초점을 두고 있다. 구조의 효율적인 사용을 위하여 한 아이템이 클러스터 내의 아이템과는 유사도가 높고, 다른 클러스터의 아이템과는 유사도가 낮도록 네트워크의 정점을 클러스터링하는 방법을 사용한다. 실험은 신뢰도, 상관관계 그리고 간선 가중치 유사도를 이용하여 네트워크에서 아이템 클러스터링의 정확도를 보여준다. 본 논문의 실험 결과를 통해 신뢰도 유사도가 네트워크의 정점을 클러스터링할 때 클러스터의 정확성에 가장 많은 영향을 미친다는 것을 알 수 있었다.