• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권11호
• /
• pp.4290-4309
• /
• 2020

In recent years, advancements in machine learning capabilities have allowed it to see widespread adoption for tasks such as object detection, image classification, and anomaly detection. However, despite their promise, a limitation lies in the fact that a network's performance quality is based on the data which it receives. A well-trained network will still have poor performance if the subsequent data supplied to it contains artifacts, out of focus regions, or other visual distortions. Under normal circumstances, images of the same scene captured from differing points of focus, angles, or modalities must be separately analysed by the network, despite possibly containing overlapping information such as in the case of images of the same scene captured from different angles, or irrelevant information such as images captured from infrared sensors which can capture thermal information well but not topographical details. This factor can potentially add significantly to the computational time and resources required to utilize the network without providing any additional benefit. In this study, we plan to explore using image fusion techniques to assemble multiple images of the same scene into a single image that retains the most salient key features of the individual source images while discarding overlapping or irrelevant data that does not provide any benefit to the network. Utilizing this image fusion step before inputting a dataset into the network, the number of images would be significantly reduced with the potential to improve the classification performance accuracy by enhancing images while discarding irrelevant and overlapping regions.

• 한국군사과학기술학회지
• /
• 제14권4호
• /
• pp.676-683
• /
• 2011

In this paper, we propose a robust method which can detect and track a high-speed small approaching target in a cluttered environment for Korean Active Protection System. The proposed method uses a temporal and spatial filter, tracking filter to detect and track a single target in consecutive order. And it is comprised of a candidate target detection step, a prior target selection step and a target tracking. Field tests on real infrared image sequences show that the proposed method could stably track a high speed and small target in complex background and target occlusion.

• 제어로봇시스템학회논문지
• /
• 제19권10호
• /
• pp.896-900
• /
• 2013

In this paper, we develop a dust detection sensor system capable of identifying types of dust for an improvement of a robot vacuum cleaner. The dust detection sensor system is composed of a set of infra-red sensors: a single transmitter and multiple receivers. Given the fixed amount of light transmitted from the transmitter, the amount of light coming in multiple receiver sensors varies, depending on the type and density of dust that is passing between the transmitter and the receivers. Therefore, the type of dust can be identified by means of observing the change of the amount of light from the receiver sensors. For experiments, we use two types of dust, rice and sesame, and validate the effectiveness of the proposed method.

• International journal of advanced smart convergence
• /
• 제7권4호
• /
• pp.40-49
• /
• 2018

We propose a more stable robust recognition algorithm which detects faces reliably even in cases where there are changes in lighting and angle of view, as well it satisfies efficiency in calculation and detection performance. We propose detects the face area alone after normalization through pre-processing and obtains a feature vector using (PCA). The feature vector is applied to LDA and using Euclidean distance of intra-class variance and inter class variance in the 2nd dimension, the final analysis and matching is performed. Experimental results show that the proposed method has a wider distribution when the input image is rotated $45^{\circ}$ left / right. We can improve the recognition rate by applying this feature value to a single algorithm and complex algorithm, and it is possible to recognize in real time because it does not require much calculation amount due to dimensional reduction.

• Journal of Biosystems Engineering
• /
• 제25권5호
• /
• pp.409-414
• /
• 2000

A machine vision system was built to obtain single stationary image from an egg. This system includes a CCD camera, an image processing board and a lighting system. A computer program was written to acquire, enhance and get histogram from an image. To minimize the evaluation time, the artificial neural network with the histogram of the image was used for eggshell evaluation. Various artificial neural networks with different parameters were trained and tested. The best network(64-50-1 and 128-10-1) showed an accuracy of 87.5% in evaluating eggshell. The comparison test for the elapsed processing time per an egg spent by this method(image processing and artificial neural network) and by the processing time per an egg spent by this method(image processing and artificial neural network) and by the previous method(image processing only) revealed that it was reduced to about a half(5.5s from 10.6s) in case of cracked eggs and was reduced to about one-fifth(5.5s from 21.1s) in case of normal eggs. This indicates that a fast eggshell evaluation system can be developed by using machine vision and artificial neural network.

• 방송공학회논문지
• /
• 제24권3호
• /
• pp.495-505
• /
• 2019

딥 러닝의 대표 알고리즘에는 영상 인식에 주로 사용되는 CNN(Convolutional Neural Networks), 음성인식 및 자연어 처리에 주로 사용되는 RNN(Recurrent Neural Networks) 등이 있다. 이 중 CNN은 데이터로부터 자동으로 특징을 학습하는 알고리즘으로 특징 맵을 생성하는 필터까지 학습할 수 있어 영상 인식 분야에서 우수한 성능을 보이면서 주류를 이루게 되었다. 이후, 객체 탐지 분야에서는 CNN의 성능을 향상하고자 R-CNN 등 다양한 알고리즘이 등장하였으며, 최근에는 검출 속도 향상을 위해 YOLO(You Only Look Once), SSD(Single Shot Multi-box Detector) 등의 알고리즘이 제안되고 있다. 하지만 이러한 딥러닝 기반 탐지 네트워크는 정지 영상에서 탐지의 성공 여부를 결정하기 때문에 동영상에서의 안정적인 객체 추적 및 탐지를 위해서는 별도의 추적 기능이 필요하다. 따라서 본 논문에서는 동영상에서의 객체 추적 및 탐지 성능 향상을 위해 딥 러닝 기반 탐지 네트워크에 칼만 필터를 결합한 방법을 제안한다. 탐지 네트워크는 실시간 처리가 가능한 YOLO v2를 이용하였으며, 실험 결과 제안한 방법은 기존 YOLO v2 네트워크에 비교하여 7.7%의 IoU 성능 향상 결과를 보였고 FHD 영상에서 20 fps의 처리 속도를 보였다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1987년도 전기.전자공학 학술대회 논문집(II)
• /
• pp.1383-1386
• /
• 1987

This paper concerns, the method for velocity of dynamic Image on two dimensional sequence Image which can be obtained from two sample lines on the street. The velocity of a single moving object Is measured by the number of total frame which Is required when an automobile passes over the second sample line through the first sample line. The measured results show that the velocity error Is less than 5% comparing with the value measured by X-band speed gun.

• International Journal of Contents
• /
• 제8권1호
• /
• pp.16-22
• /
• 2012

In this paper we propose a method for verifying video in a video telephony system implemented in DM6446 DaVinci Processor. Each frame is categorized either error free frame or error frame depending on the predefined criteria. Human face is chosen as a basic means for authenticating the video frame. Skin color based algorithm is implemented for detecting the face in the video frame. The video frame is classified as error free frame if there is single face object with clear view of facial features (eyes, nose, mouth etc.) and the background of the image frame is not different then the predefined background, otherwise it will be classified as error frame. We also implemented the image histogram based NCC (Normalized Cross Correlation) comparison for video verification to speed up the system. The experimental result shows that the system is able to classify frames with 90.83% of accuracy.

• 한국건축시공학회:학술대회논문집
• /
• 한국건축시공학회 2022년도 봄 학술논문 발표대회
• /
• pp.103-104
• /
• 2022

Cracks may occur in reinforced concrete (RC) structures due to various physical and chemical factors, and the growth of cracks causes deterioration of the structure's performance. It is important to prevent the expansion of cracks through periodic diagnosis of cracks in structures. In order to enable free crack exploration even in a narrow space, a construction robot using a Mecanum wheel that can move up, down, left and right and rotate in place was designed. High-quality crack images were periodically collected through the camera, and the image fragments stored during the exploration were combined into a single photo after the exploration was completed. The robot detected cracks with a width of 0.2 mm or more on the concrete probe surface with an accuracy of about 90% or more.

• 한국방송∙미디어공학회:학술대회논문집
• /
• 한국방송∙미디어공학회 2019년도 하계학술대회
• /
• pp.9-12
• /
• 2019

지난 수 년 동안 계속해서 일반 실상 카메라를 이용한 영상분석기술에 대한 연구가 활발히 진행되고 있다. 최근에는 딥러닝 기술을 적용한 지능형 영상분석기술로 발전해 왔으며 국방기지방호, CCTV, 사용자 얼굴인식, 머신비전, 자동차, 드론 산업이 활성화되면서 많은 시너지를 효과를 일으키고 있다. 그러나 어두운 밤과 안개, 날씨, 연기 등 다양한 여건에서 따라서 카메라의 영상분석 정확성 감소와 오류가 수반될 수 있으며 일반적으로 딥러닝 기술을 활용하기 위해서는 고사양의 GPU를 필요로 하기 때문에 다른 추가적인 시스템이 요구된다. 이에 본 연구에서는 열적외선 영상의 객체 검출에 적용하기 위해 SSD(Single Shot MultiBox Detector) 기반의 경량적인 MobilNet 네트워크로 재구성하여, 모바일 기기 등 낮은 사양의 낮은 임베디드 시스템에서도 활용 할 수 있는 방법을 제안한다. 모의 실험결과 제안된 방식의 모델은 적외선 열화상 카메라에서 객체검출과 학습시간이 줄어든 것을 확인 할 수 있었다.