<?xml version="1.0" encoding="UTF-8"?>
<?xml-stylesheet type="text/xsl" href="/resources/xsl/jats-html.xsl"?>
<article article-type="research-article" dtd-version="1.1" xml:lang="ko" xmlns:mml="http://www.w3.org/1998/Math/MathML" 
xmlns:xlink="http://www.w3.org/1999/xlink" 
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance">
<front>
	<journal-meta>
		<journal-id journal-id-type="publisher-id">jkits</journal-id>
		<journal-title-group>
		<journal-title>한국지식정보기술학회 논문지</journal-title>
		<journal-title xml:lang="en">Journal of Knowledge Information Technology and Systems</journal-title>
		</journal-title-group>
		<issn pub-type="ppub">1975-7700</issn>
		<publisher>
		<publisher-name>한국지식정보기술학회</publisher-name>
		<publisher-name xml:lang="en">Korea Knowledge Information Technology Society</publisher-name>
		</publisher>
	</journal-meta>
	<article-meta>
		<article-id pub-id-type="publisher-id">jkits_2020_15_03_323</article-id>
		<article-id pub-id-type="doi">10.34163/jkits.2020.15.3.002</article-id>
		<article-categories>
			<subj-group>
				<subject>Research Article</subject>
			</subj-group>
		</article-categories>
		<title-group>
			<article-title>텐서플로우를 이용한 시정거리 판단 시스템의 구현</article-title>
			<trans-title-group xml:lang="en">
				<trans-title>A Study on the Application of TensorFlow to Determine the Correctional Distance</trans-title>
			</trans-title-group>
		</title-group>
		<contrib-group>
			<contrib contrib-type="author" xlink:type="simple">
				<name-alternatives>
					<name name-style="eastern">
						<surname>한</surname>
						<given-names>익섭</given-names>
					</name>
					<name name-style="western" xml:lang="en">
						<surname>Han</surname>
						<given-names>Eig Seub</given-names>
					</name>
				</name-alternatives>
				<xref ref-type="fn" rid="fn001">*</xref>
			</contrib>
		</contrib-group>
		<aff-alternatives>
			<aff><italic>한국폴리텍대학 정보통신시스템과 초빙교수</italic></aff>
			<aff xml:lang="en"><italic>Department of Information&#x26;Communication System Korea Polytechnics</italic></aff>
		</aff-alternatives>
		<author-notes>
			<fn id="fn001"><label>*</label><p>Corresponding author is with the Department of Information&#x26;Communication System, Korea Polytechnics, 352-21 Uamro Dong-gu Daejeon, 34503, KOREA.</p><p><italic>E-mail address</italic>: <email>eshan1627@kopo.ac.kr</email></p></fn>
		</author-notes>
		<pub-date pub-type="ppub">
			<month>06</month>
			<year>2020</year>
		</pub-date>
		<volume>15</volume>
		<issue>3</issue>
		<fpage>323</fpage>
		<lpage>329</lpage>
		<history>
			<date date-type="received">
				<day>27</day>
				<month>04</month>
				<year>2020</year>
			</date>
			<date date-type="rev-recd">
				<day>04</day>
				<month>06</month>
				<year>2020</year>
			</date>
			<date date-type="accepted">
				<day>11</day>
				<month>06</month>
				<year>2020</year>
			</date>
		</history>
		<permissions>
			<copyright-statement>&#x00A9; 2020 KKITS All rights reserved</copyright-statement>
			<copyright-year>2020</copyright-year>
		</permissions>
		<abstract>
		<title>요약</title>
		<p>최근 전 세계적 환경오염이 심각해지고 있다. 대기오염으로 인한 경제적 손실과 초 미세먼지로 인한 위협은 사회적 문제가 되고 있다. 현재 목측과 광학적 측정 장비를 통해서 측정하고 있으나, 측정값과 인간의 시정감각이 일치하지 않는 문제점이 있다. 본 논문은 측정값과 인간의 시정감각을 일치시키기 위한 판단 알고리즘을 구현하는 것을 목표로 하였다. 건물의 IoT기반 카메라에서 측정된 사진 정보를 서버에 전송받아 시정거리 측정을 하고 텐서플로우를 통해 고속 연산 처리하여 신뢰도가 높은 시정 거리 분석이 필요하였다. 기존 시정거리 판단 알고리즘에 SVM 비선형 회귀모델 알고리즘로 보완함으로 사람이 직접 판단하는 것과 유사한 알고리즘으로 자동화 처리하는 알고리즘을 구현하였다. 본 연구에서는 SVM(Support Vector Machine) 비선형 회귀모델 알고리즘으로 텐서플로우를 이용하여 고속 연산처리를 하였고, 시정 판단 알고리즘 모델의 신뢰도를 향상시킬 수 있도록 시스템 구현 모델을 제안하였다.</p>
		</abstract>
		<trans-abstract xml:lang="en">
		<title>ABSTRACT</title>
		<p>Environmental pollution is getting serious around the world recently. Economic losses from air pollution and threats from ultra fine dust are becoming social problems. Currently, measurements are made through wood and optical measuring equipment, but there is a problem where measurements and human sense of corrective action do not match. This paper aimed to implement an algorithm of judgment to match the measured value with the human sense of visibility. Using IoT-based cameras in buildings, measured photo information is sent to the server to make corrective distance measurements, and real-time transmitted photos and existing measured photo information are processed in high-speed operation through Tensorflow, requiring high-reliability corrective distance. An algorithm that is supplemented with a SVM nonlinear regression model algorithm for existing corrective distance determination algorithms has been implemented to automate with algorithms similar to those that are directly judged by humans. In this study, a support vector machine (SVM) nonlinear regression model algorithm is used to perform high-speed computation using Tensorflow, and a system implementation model is proposed to improve reliability of the corrective judgment algorithm model.</p>
		</trans-abstract>
		<kwd-group kwd-group-type="author" xml:lang="en">
<title>K E Y W O R D S</title>
			<kwd>Tensorflow</kwd>
			<kwd>Smart factory</kwd>
			<kwd>IoT</kwd>
			<kwd>LwM2M(Lightweight M2M)</kwd>
			<kwd>Intelligent IoT systems</kwd>
			<kwd>Deep learning</kwd>
		</kwd-group>
	</article-meta>
</front>
<body>
<sec id="sec001" sec-type="intro">
	<title>1. 서 론</title>
	<p>최근 전 세계적으로 환경오염이 심각해지고 있으며 특히 안개, 황사 등 대기 오염으로 인한 경제적 손실과 초미세먼지로 부터의 위협은 사회적인 문제로 급격하게 부상하고 있다. 대기 오염으로 인한 피해를 최소화하기 위해서는 많은 지역에서 실시간으로 시정거리를 관측할 필요가 있다[<xref ref-type="bibr" rid="B001">1</xref>,<xref ref-type="bibr" rid="B003">3</xref>].</p>
	<p>현재까지 국내 시정관측은 목측(目測)을 통한 관측이나 광학 시정계를 사용하고 있다. 목측은 관측자의 주관적인 판단에 의해 시정거리를 관측하여 객관성에 문제가 있고 개개인에 따라 달라지는 문제가 있다. 광학 시정계는 목측과 비교적 추이는 비슷하나 사람이 공감하기 힘든 결과를 내놓는 경우가 많다[<xref ref-type="bibr" rid="B002">2</xref>,<xref ref-type="bibr" rid="B004">4</xref>].</p>
	<p>본 논문에서는 영상에서 혼탁도와 픽셀 값을 이용한 시정 관측법을 연구하였으며 사물인터넷(IoT)시스템에 이러한 관측법을 적용하고 IoT기반 카메라부터 얻은 빅데이터를 텐서플로우(Tensorflow)를 이용한 고속연산 처리하여 시정거리 판단에 대한 정확도 향상을 알고리즘 모델 방안을 제시하고 구현하였다</p>
	<p>본 논문의 구성은 다음과 같다. 2장에서는 시스템 구성, 시정거리 판단 알고리즘 구현, 시스템 구현 결과, 3장에서는 결론으로 기술한다.</p>
</sec>
<sec id="sec002">
	<title>2. 시정거리 관측을 위한 시스템 구성 및 알고리즘</title>
	<sec id="sec002-1">
		<title>2.1 시스템 구조</title>
		<p>&#x003C;<xref ref-type="fig" rid="f001">그림 1</xref>&#x003E;은 시정거리 관측향상을 위한 시스템 구성도이다. &#x003C;<xref ref-type="fig" rid="f001">그림 1</xref>&#x003E;의 Camera에서 받은 이미지를 Wifi를 통해 Server로 이미지(Image)를 전달하고 서버(Server)에서 이 이미지 데이터를 전처리(Pre-processing) 하여 Dataset을 생성한다. 생성된 Dataset을 Application 으로 보낸다. Application에서는 텐서플로우(Tensorflow)에서 만들어진 Model Graph를 Import 하여 해당 이미지의 보이는 유무를 처리하고 결과에 따라 거리를 산정한다. 산정된 결과 값은 Server의 DB에서 관리한다.</p>
		<fig id="f001" orientation="portrait" position="float">
			<label>그림 1.</label>
			<caption>
				<title>시스템 구성도</title>
				<p>Figure 1. System configuration</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f001.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
	</sec>
	<sec id="sec002-2">
		<title>2.2 시정거리 판단 알고리즘</title>
		<p>&#x003C;<xref ref-type="fig" rid="f002">그림 2</xref>&#x003E;의 Box1과 Box2처럼 일반적으로 사물은 거리가 멀수록 혼탁해지고 선명도는 낮아지고 밝기는 높아진다. 태양고도와 주변 광원의 영향을 받아서 선명도와 밝기는 달라진다. 그 이유는 빛이 사물에 반사되어 사람이 인식하는 과정에서 기체와 입자 등에 의해서 흡수 혹은 산란 때문이다.</p>
		<fig id="f002" orientation="portrait" position="float">
			<label>그림 2.</label>
			<caption>
				<title>거리에 따른 선명도와 혼탁도 변화</title>
				<p>Figure 2. Variation of sharpness and turbidity with distance</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f002.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>사물의 식별 능력의 정도는 거리에 따라 달라지는데 이러한 계수들의 합을 빛 소멸계수(β)라고 하며 시정계수는 빛 소멸계수와 밀접한 관련이 있다[<xref ref-type="bibr" rid="B005">5</xref>-<xref ref-type="bibr" rid="B009">9</xref>].</p>
<disp-formula-group>
<disp-formula id="d001">
	<label>(1)</label>
<mml:math id="dm01-1"><mml:mi>t</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>=</mml:mo><mml:mi>e</mml:mi><mml:mi>s</mml:mi><mml:mi>p</mml:mi><mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:msubsup><mml:mo>&#x222B;</mml:mo><mml:mn>0</mml:mn><mml:mrow><mml:mi>d</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced></mml:mrow></mml:msubsup><mml:mi>&#x3B2;</mml:mi><mml:mfenced><mml:mi>z</mml:mi></mml:mfenced><mml:mo>d</mml:mo><mml:mi>z</mml:mi></mml:mrow></mml:mfenced><mml:mspace linebreak="newline"/><mml:mo>=</mml:mo><mml:mi>e</mml:mi><mml:mi>s</mml:mi><mml:mi>p</mml:mi><mml:mfenced><mml:mrow><mml:mo>-</mml:mo><mml:mi>&#x3B2;</mml:mi><mml:mi>d</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced></mml:mrow></mml:mfenced><mml:mo>=</mml:mo><mml:mfrac><mml:mrow><mml:mi>I</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>-</mml:mo><mml:mi>A</mml:mi></mml:mrow><mml:mrow><mml:mi>J</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>-</mml:mo><mml:mi>A</mml:mi></mml:mrow></mml:mfrac></mml:math>
</disp-formula>
<disp-formula id="d002">
	<label>(2)</label>
<mml:math id="dm02-1"><mml:mi>I</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>=</mml:mo><mml:mi>J</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mi>t</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>+</mml:mo><mml:mi>A</mml:mi><mml:mfenced><mml:mrow><mml:mn>1</mml:mn><mml:mo>-</mml:mo><mml:mi>t</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced></mml:mrow></mml:mfenced></mml:math>
</disp-formula>
<disp-formula id="d003">
	<label>(3)</label>
<mml:math id="dm03-1"><mml:mi>&#x3B2;</mml:mi><mml:mo>=</mml:mo><mml:mo>-</mml:mo><mml:mfenced><mml:mrow><mml:mi>ln</mml:mi><mml:mfenced><mml:mfrac><mml:mrow><mml:mi>I</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>-</mml:mo><mml:mi>A</mml:mi></mml:mrow><mml:mrow><mml:mi>J</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced><mml:mo>-</mml:mo><mml:mi>A</mml:mi></mml:mrow></mml:mfrac></mml:mfenced><mml:mo>&#xD7;</mml:mo><mml:mfrac><mml:mn>1</mml:mn><mml:mrow><mml:mi>d</mml:mi><mml:mfenced><mml:mi>x</mml:mi></mml:mfenced></mml:mrow></mml:mfrac></mml:mrow></mml:mfenced></mml:math>
</disp-formula>
</disp-formula-group>
		<p><xref ref-type="disp-formula" rid="d001">식 (1)</xref>~<xref ref-type="disp-formula" rid="d003">(3)</xref>에서 <xref ref-type="disp-formula" rid="d001">(1)</xref>는 투과도에 대한 정의식이다. t(x)를 정리된 <xref ref-type="disp-formula" rid="d001">(1)식</xref>과 <xref ref-type="disp-formula" rid="d002">(2)식</xref>을 종합해서 정의하면 <xref ref-type="disp-formula" rid="d003">(3)식</xref>이 나온다.  t(x)는 이미지 영역의 투과도, i(x)는 측정 이미지 영역의 픽셀 벡터값, J(x)는 기준 이미지 영역의 픽셀 벡터값, A는 대기 산란광, d(x)는 관측물체와 사람 간의 거리이다[10,11].</p>
		<p><xref ref-type="disp-formula" rid="d003">식 (3)</xref>로부터 이미지의 영역 픽셀값과 그 혼탁도는 빛 소멸계수와 유의미한 관계가 있을 것이라 예측하였다. 따라서 이러한 관계의 알고리즘 공식을 적용한 결과 이미지 데이터를 빅데이터화하여 텐서플로우(Tensorflow)를 통해서 확인하였고, 정확도가 향상된 학습 모델을 제시한다.</p>
	</sec>
	<sec id="sec002-3">
		<title>2.3 시정거리 거리 측정</title>
		<fig id="f003" orientation="portrait" position="float">
			<label>그림 3.</label>
			<caption>
				<title>관측 P별 위치</title>
				<p>Figure 3. Location by observation P</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f003.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>2.2절에서 제시된 알고리즘 검증을 위해 &#x003C;<xref ref-type="fig" rid="f003">그림 3</xref>&#x003E;과 같은 위치에 &#x003C;<xref ref-type="fig" rid="f003">그림 3b</xref>&#x003E;의 P1, P2, P3와 같이 식별이 용이하고 거의 직선 위치를 지점으로 Point를 기준으로 데이터를 수집하였다. 해당 사진의 건물 중 식별이 용이하고 거의 직선에 있는 지점 3곳을 Point 정의하고 데이터를 수집하였다.</p>
		<p>식별이 용이하고 거의 직선에 있는 지점 3곳을 Point 정의하고 데이터를 수집하였다.</p>
		<fig id="f004" orientation="portrait" position="float">
			<label>그림 4.</label>
			<caption>
				<title>시간대별 촬영 사진</title>
				<p>Figure 4. Time photography</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f004.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>&#x003C;<xref ref-type="fig" rid="f004">그림 4</xref>&#x003E;는 1개월간 찍은 이미지에서 P1의 혼탁도와 픽셀 값을 보여준다. 1개월간 전체 수집된 데이터 분포도는 &#x003C;<xref ref-type="fig" rid="f005">그림 5</xref>&#x003E;와 같다.</p>
		<fig id="f005" orientation="portrait" position="float">
			<label>그림 5.</label>
			<caption>
				<title>전체 데이터 분포도</title>
				<p>Figure 5. Total Data Distribution Plot</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f005.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>&#x003C;<xref ref-type="fig" rid="f005">그림 5</xref>&#x003E;는 PASS와 FAIL 데이터가 모두 포함된 분포도이다. 이때 Pass 데이터만 선별하여 정리하면 &#x003C;<xref ref-type="fig" rid="f006">그림 6</xref>&#x003E;과 같다.</p>
		<fig id="f006" orientation="portrait" position="float">
			<label>그림 6.</label>
			<caption>
				<title>PASS 데이터 분포도</title>
				<p>Figure 6. PASS Data Distribution Plot</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f006.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>&#x003C;<xref ref-type="fig" rid="f006">그림 6</xref>&#x003E;는 P1이 PASS(보이는 상태)일 때 혼탁도와 픽셀 값에 대한 분포도이다. &#x003C;<xref ref-type="fig" rid="f005">그림 5</xref>&#x003E;에서 &#x003C;<xref ref-type="fig" rid="f006">그림 6</xref>&#x003E;을 선형적인 모델을 사용하여 분리할 수 있으나 중복되는 부분이 많아서 신뢰도가 떨어진다. 신뢰도를 향상하기 위하여 다음 식과 같이 일반적인 선형회귀 방법이 아닌 kernel을 사용한 SVM(Support Vector Machine) 비선형 회귀모델을 사용했다[<xref ref-type="bibr" rid="B012">12</xref>-<xref ref-type="bibr" rid="B014">14</xref>].</p>
<disp-formula-group>
<disp-formula id="d004">
	<label>(4)</label>
<mml:math id="dm04-1"><mml:mi>D</mml:mi><mml:mi>i</mml:mi><mml:mi>s</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mfenced><mml:mrow><mml:mi>A</mml:mi><mml:mo>-</mml:mo><mml:mfenced><mml:mrow><mml:mn>2</mml:mn><mml:mo>.</mml:mo><mml:mn>0</mml:mn><mml:mo>&#xD7;</mml:mo><mml:mi>X</mml:mi><mml:msup><mml:mi>T</mml:mi><mml:mi>T</mml:mi></mml:msup></mml:mrow></mml:mfenced></mml:mrow></mml:mfenced><mml:mo>+</mml:mo><mml:msup><mml:mi>B</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mspace linebreak="newline"/><mml:mi>K</mml:mi><mml:mi>e</mml:mi><mml:mi>r</mml:mi><mml:mi>n</mml:mi><mml:mi>e</mml:mi><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mi>&#x3B3;</mml:mi><mml:mfenced open="|" close="|"><mml:mrow><mml:mi>D</mml:mi><mml:mi>i</mml:mi><mml:mi>s</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub></mml:mrow></mml:mfenced></mml:mrow></mml:msup><mml:mspace linebreak="newline"/><mml:mi>O</mml:mi><mml:mi>u</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mo>=</mml:mo><mml:mfenced><mml:mrow><mml:msup><mml:mi>Y</mml:mi><mml:mi>T</mml:mi></mml:msup><mml:mo>&#xD7;</mml:mo><mml:mi>b</mml:mi></mml:mrow></mml:mfenced><mml:mi>K</mml:mi><mml:mi>e</mml:mi><mml:mi>r</mml:mi><mml:mi>n</mml:mi><mml:mi>e</mml:mi><mml:msub><mml:mi>l</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mspace linebreak="newline"/><mml:mi>D</mml:mi><mml:mo>=</mml:mo><mml:mi>O</mml:mi><mml:mi>u</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi></mml:mrow></mml:msub><mml:mo>-</mml:mo><mml:mfenced><mml:mrow><mml:mfrac><mml:mn>1</mml:mn><mml:mi>n</mml:mi></mml:mfrac><mml:munderover><mml:mo>&#x2211;</mml:mo><mml:mi>k</mml:mi><mml:mi>n</mml:mi></mml:munderover><mml:mi>O</mml:mi><mml:mi>u</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mrow><mml:mi>p</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mi>k</mml:mi></mml:msub></mml:mrow></mml:msub></mml:mrow></mml:mfenced><mml:mspace linebreak="newline"/><mml:mi>P</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:mi>t</mml:mi><mml:mo>=</mml:mo><mml:mfenced open="{" close=""><mml:mtable columnspacing="1.4ex" columnalign="left"><mml:mtr><mml:mtd><mml:mi>D</mml:mi><mml:mo>&lt;</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo></mml:mtd><mml:mtd><mml:mo>-</mml:mo><mml:mn>1</mml:mn></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mi>D</mml:mi><mml:mo>&gt;</mml:mo><mml:mn>0</mml:mn><mml:mo>,</mml:mo></mml:mtd><mml:mtd><mml:mn>1</mml:mn></mml:mtd></mml:mtr></mml:mtable></mml:mfenced></mml:math>
</disp-formula>
</disp-formula-group>
		<p><xref ref-type="disp-formula" rid="d004">식 (4)</xref> Predict의 결과를 통해서 testset이 속하는 영역을 알 수 있다.</p>
<disp-formula-group>
<disp-formula id="d005">
	<label>(5)</label>
<mml:math id="dm05-1"><mml:mi>A</mml:mi><mml:mi>c</mml:mi><mml:mi>c</mml:mi><mml:mi>u</mml:mi><mml:mi>r</mml:mi><mml:mi>a</mml:mi><mml:mi>c</mml:mi><mml:mi>y</mml:mi><mml:mo>=</mml:mo><mml:mfrac><mml:mn>1</mml:mn><mml:mi>n</mml:mi></mml:mfrac><mml:munderover><mml:mo>&#x2211;</mml:mo><mml:mi>k</mml:mi><mml:mi>n</mml:mi></mml:munderover><mml:mfenced open="{" close=""><mml:mtable columnspacing="1.4ex" columnalign="left"><mml:mtr><mml:mtd><mml:mfenced><mml:mrow><mml:mi>P</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mi>k</mml:mi></mml:msub><mml:mo>=</mml:mo><mml:msub><mml:mi>Y</mml:mi><mml:mi>k</mml:mi></mml:msub></mml:mrow></mml:mfenced><mml:mo>,</mml:mo></mml:mtd><mml:mtd><mml:mn>1</mml:mn></mml:mtd></mml:mtr><mml:mtr><mml:mtd><mml:mfenced><mml:mrow><mml:mi>P</mml:mi><mml:mi>r</mml:mi><mml:mi>e</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>c</mml:mi><mml:msub><mml:mi>t</mml:mi><mml:mi>k</mml:mi></mml:msub><mml:mo>&#x2260;</mml:mo><mml:msub><mml:mi>Y</mml:mi><mml:mi>k</mml:mi></mml:msub></mml:mrow></mml:mfenced><mml:mo>,</mml:mo></mml:mtd><mml:mtd><mml:mn>0</mml:mn></mml:mtd></mml:mtr></mml:mtable></mml:mfenced><mml:mspace linebreak="newline"/><mml:mi>G</mml:mi><mml:mi>r</mml:mi><mml:mi>a</mml:mi><mml:mi>d</mml:mi><mml:mi>i</mml:mi><mml:mi>a</mml:mi><mml:mi>n</mml:mi><mml:mi>t</mml:mi><mml:mo>&#xA0;</mml:mo><mml:mi>D</mml:mi><mml:mi>e</mml:mi><mml:mi>s</mml:mi><mml:mi>c</mml:mi><mml:mi>e</mml:mi><mml:mi>n</mml:mi><mml:mi>t</mml:mi><mml:mo>&#xA0;</mml:mo><mml:mi>O</mml:mi><mml:mi>t</mml:mi><mml:mi>i</mml:mi><mml:mi>m</mml:mi><mml:mi>i</mml:mi><mml:mi>z</mml:mi><mml:mi>e</mml:mi><mml:mi>r</mml:mi><mml:mo>=</mml:mo><mml:mn>0</mml:mn><mml:mo>.</mml:mo><mml:mn>01</mml:mn></mml:math>
</disp-formula>
</disp-formula-group>
		<p><xref ref-type="disp-formula" rid="d005">식 (5)</xref>은 testset의 원소들이 어떤 식으로 추론이 되었는지를 확인하고 SVM모델의 신뢰도를 계산하는 식이다[<xref ref-type="bibr" rid="B014">14</xref>,<xref ref-type="bibr" rid="B015">15</xref>].</p>
		<p>위의 알고리즘을 이용해 학습한 모델로 P1~P3의 PASS와 FAIL를 구분한다. 일반적인 목측 알고리즘을 적용하면 가장 먼 P1이 보이는 경우 나머지는 전부 보인다고 판단하기 때문에 &#x003C;<xref ref-type="table" rid="t001">표 1</xref>&#x003E;처럼 4가지 경우의 수가 나왔다.</p>
		<table-wrap id="t001">
			<label>표 1.</label>
			<caption>
				<title>타겟별 얻을 수 있는 경우의 수</title>
				<p>Table 1. Number of cases per target</p>
			</caption>
			<table frame="box" rules="all" width="100%">
<tbody align="center">
<tr>
<td>P1</td>
<td>P2</td>
<td>P3</td>
<td>거리판단 기준점</td>
</tr>
<tr>
<td>PASS</td>
<td>PASS</td>
<td>PASS</td>
<td>P1</td>
</tr>
<tr>
<td>FAIL</td>
<td>PASS</td>
<td>PASS</td>
<td>P2</td>
</tr>
<tr>
<td>FAIL</td>
<td>FAIL</td>
<td>PASS</td>
<td>P3</td>
</tr>
<tr>
<td>FAIL</td>
<td>FAIL</td>
<td>FAIL</td>
<td>-</td>
</tr>
				</tbody>
			</table>
		</table-wrap>
	</sec>
	<sec id="sec002-4">
		<title>2.4 시스템 구현 결과</title>
		<fig id="f007" orientation="portrait" position="float">
			<label>그림 7.</label>
			<caption>
				<title>Batch Accuracy</title>
				<p>Figure 7. Batch Accuracy</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f007.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<table-wrap id="t002">
			<label>표 2.</label>
			<caption>
				<title>학습 횟수 별 신뢰도 변화 추이</title>
				<p>Table 2. Confidence changes by number of studies</p>
			</caption>
			<table frame="box" rules="all" width="100%">
<tbody align="center">
<tr>
<td>Epoch(Generation)</td>
<td>Accuracy(True/Total Data)</td>
</tr>
<tr>
<td>50</td>
<td>약 72% (3545/4923)</td>
</tr>
<tr>
<td>100</td>
<td>약 85% (4202/4923)</td>
</tr>
<tr>
<td>200</td>
<td>약 91% (4510/4923)</td>
</tr>
<tr>
<td>300</td>
<td>약 96% (4736/4923)</td>
</tr>
				</tbody>
			</table>
		</table-wrap>
		<p>&#x003C;<xref ref-type="fig" rid="f007">그림 7</xref>&#x003E;은 1개월간 수집한 이미지 데이터를 직접 분류한 데이터와 본 알고리즘을 이용해 분류한 이미지의 결과값을 비교한 그래프이다. &#x003C;<xref ref-type="fig" rid="f007">그림 7</xref>&#x003E;에 따르면 시정거리 판단 알고리즘 모델의 신뢰도는 96% 이상이다.</p>
		<fig id="f008" orientation="portrait" position="float">
			<label>그림 8.</label>
			<caption>
				<title>Loss per Generation</title>
				<p>Figure 8. Loss per Generation</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f008.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>Loss는 &#x003C;<xref ref-type="fig" rid="f008">그림 8</xref>&#x003E;과 같이 Epoch를 반복할수록 일정하게 낮아진다.</p>
		<fig id="f009" orientation="portrait" position="float">
			<label>그림 9.</label>
			<caption>
				<title>Inference Result</title>
				<p>Figure 9. Inference Result</p>
			</caption>
			<graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f009.jpg" position="float" orientation="portrait" xlink:type="simple"></graphic>
		</fig>
		<p>&#x003C;<xref ref-type="fig" rid="f009">그림 9</xref>&#x003E;은 시정거리 판단 알고리즘 모델이 적용된 분류 그래프이다. gamma 값을 변경하더라도 &#x003C;<xref ref-type="fig" rid="f009">그림 9</xref>&#x003E;와 똑같은 출력이 나온다. &#x003C;<xref ref-type="fig" rid="f009">그림 9</xref>&#x003E;에서 See와 Non See가 겹치는 부분은 예외 값으로 해석해야 한다. gamma 값을 조정하면 분포에서 많이 벗어난 값을 해결 할 수 있어야 하나, Class 값을 학습하는 지도학습이므로 Dataset에 겹치는 값이나 혼란스러운 값이 있을 경우 발생한다. 따라서 학습 Dataset을 선별할 때에는 알고리즘의 X, Y의 값에 따라 결과 값에 큰 변화가 있을 수 있다는 점에 주의해서 선별해야 한다.</p>
		<p>결과적으로 시정거리 판단 알고리즘 모델을 사용하여 각각의 P위치가 PASS, FAIL 판단 여부를 결정하면 각각 P의 신뢰도는 90% 이상 신뢰도를 기대할 수 있다.</p>
	</sec>
</sec>
<sec id="sec003" sec-type="Conclusion">
	<title>3. 결 론</title>
	<p>본 논문에서는 텐서플로우를 이용해서 목측과 유사한 방법으로 시정거리를 산출하는 시스템 구현하는 시정 관측 알고리즘에 대해서 구현하였다. 알고리즘 구현을 통해서 높은 신뢰성, 그리고 비교적 간단한 알고리즘으로도 시정에 대한 평가가 가능하다는 점을 확인할 수 있었다. 기타 한계점으로는 계절 기상 변화에 따른 요인이 다양하게 영향을 주기 때문이다. 본 논문에서 구현한 시정거리 판단 알고리즘을 활용하여 계절 의 다양한 기상상황에 따른 후속 연구가 필요하다.</p>
</sec>
</body>
<back>
<ref-list>
<title>References</title>
<!--[1] S-K. Eun, H-G. Nam, B-G. Kim, J-S. Park, J-Y. Ahn, and S-J. Lee, Analysis of aerosol scatter coefficient in Seoul and Baengnyeong Island, Journal of Academic Conference of the Korean Journal of Atmospheric Environment, Vol. 29, No. 3 pp. 264-274, 2013.-->
<ref id="B001">
<label>[1]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Eun</surname><given-names>S-K.</given-names></name>
<name><surname>Nam</surname><given-names>H-G.</given-names></name>
<name><surname>Kim</surname><given-names>B-G.</given-names></name>
<name><surname>Park</surname><given-names>J-S.</given-names></name>
<name><surname>Ahn</surname><given-names>J-Y.</given-names></name>
<name><surname>Lee</surname><given-names>S-J.</given-names></name>
</person-group>
<year>2013</year>
<article-title>Analysis of aerosol scatter coefficient in Seoul and Baengnyeong Island</article-title>
<source>Journal of Academic Conference of the Korean Journal of Atmospheric Environment</source>
<volume>29</volume><issue>3</issue>
<fpage>264</fpage><lpage>274</lpage>
</element-citation>
</ref>
<!--[2] J-S. Park, T-S. Kim, and S-D. Kim, Analysis of correction characteristics by forward scattering meter and pastoral, Journal of Academic Conference of the Korean Society of Atmospheric Environment, pp. 13-14, 1997.-->
<ref id="B002">
<label>[2]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Park</surname><given-names>J-S.</given-names></name>
<name><surname>Kim</surname><given-names>T-S.</given-names></name>
<name><surname>Kim</surname><given-names>S-D.</given-names></name>
</person-group>
<year>1997</year>
<article-title>Analysis of correction characteristics by forward scattering meter and pastoral</article-title>
<source>Journal of Academic Conference of the Korean Society of Atmospheric Environment</source>
<fpage>13</fpage><lpage>14</lpage>
</element-citation>
</ref>
<!--[3] W-S. Choi, and J-J. Kim, Characteristics of nanoparticles (UFPs) observed in urban areas of the world, including Korea: Atmosphere. Korean Meteorological Society Vol. 28, No. 3, pp. 337-3555, 2018.-->
<ref id="B003">
<label>[3]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Choi</surname><given-names>W-S.</given-names></name>
<name><surname>Kim</surname><given-names>J-J.</given-names></name>
</person-group>
<year>2018</year>
<article-title>Characteristics of nanoparticles (UFPs) observed in urban areas of the world, including Korea: Atmosphere</article-title>
<source>Korean Meteorological Society</source>
<volume>28</volume><issue>3</issue>
<fpage>337</fpage><lpage>355</lpage>
<pub-id pub-id-type="doi">10.14191/Atmos.2018.28.3.337</pub-id>
</element-citation>
</ref>
<!--[4] J-S. Park, T-S. Kim, and S-D. Kim, Forward Scattering Meter, and the Pastoral Analysis, pp. 13-14, 1997.-->
<ref id="B004">
<label>[4]</label>
<element-citation publication-type="report">
<person-group>
<name><surname>Park</surname><given-names>J-S.</given-names></name>
<name><surname>Kim</surname><given-names>T-S.</given-names></name>
<name><surname>Kim</surname><given-names>S-D.</given-names></name>
</person-group>
<year>1997</year>
<source>Forward Scattering Meter and the Pastoral Analysis</source>
<fpage>13</fpage><lpage>14</lpage>
</element-citation>
</ref>
<!--[5] S. G. Narasimhan, and S. K. Nayar, Contrast restoration of weather degraded images, IEEE transactions on pattern analysis and machine intelligence, Vol. 25, No. 6, Jun. 2003.-->
<ref id="B005">
<label>[5]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Narasimhan</surname><given-names>S. G.</given-names></name>
<name><surname>Nayar</surname><given-names>S. K.</given-names></name>
</person-group>
<year>2003</year>
<month>Jun.</month>
<article-title>Contrast restoration of weather degraded images</article-title>
<source>IEEE transactions on pattern analysis and machine intelligence</source>
<volume>25</volume><issue>6</issue>
<pub-id pub-id-type="doi">10.1109/TPAMI.2003.1201821</pub-id>
</element-citation>
</ref>
<!--[6] S. K. Nayar, and S. G. Narasimhan, Vision in bad weather, Proc. Seventh IEEE Int’l Conf. Computer Vision, Vol. 2, pp. 820-827, 1999.-->
<ref id="B006">
<label>[6]</label>
<element-citation publication-type="paper">
<person-group>
<name><surname>Nayar</surname><given-names>S. K.</given-names></name>
<name><surname>Narasimhan</surname><given-names>S. G.</given-names></name>
</person-group>
<year>1999</year>
<article-title>Vision in bad weather</article-title>
<conf-name>Proc. Seventh IEEE Int’l Conf.</conf-name>
<comment>Computer Vision Vol. 2</comment>
<fpage>820</fpage><lpage>827</lpage>
<pub-id pub-id-type="doi">10.1109/ICCV.1999.790306</pub-id>
</element-citation>
</ref>
<!--[7] S. G. Narasimhan, and S. K. Nayar, Vision and the atmosphere, Int’l J. Computer Vision, Vol. 48, pp. 233-254, 2002.-->
<ref id="B007">
<label>[7]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Narasimhan</surname><given-names>S. G.</given-names></name>
<name><surname>Nayar</surname><given-names>S. K.</given-names></name>
</person-group>
<year>2002</year>
<article-title>Vision and the atmosphere</article-title>
<source>Int’l J. Computer Vision</source>
<volume>48</volume>
<fpage>233</fpage><lpage>254</lpage>
<pub-id pub-id-type="doi">10.1023/A:1016328200723</pub-id>
</element-citation>
</ref>
<!--[8] S. Y. Schechner, S. G. Narasimhan, and S. K. Nayar, Instant dehazing of images using polarization, Proc. Computer Vision & Pattern Recognition Vol. 1, pp. 325-332, 2001.-->
<ref id="B008">
<label>[8]</label>
<element-citation publication-type="paper">
<person-group>
<name><surname>Schechner</surname><given-names>S. Y.</given-names></name>
<name><surname>Narasimhan</surname><given-names>S. G.</given-names></name>
<name><surname>Nayar</surname><given-names>S. K.</given-names></name>
</person-group>
<year>2001</year>
<conf-name>Instant dehazing of images using polarization</conf-name>
<comment>Proc. Computer Vision &#x26; Pattern Recognition Vol.1</comment>
<fpage>325</fpage><lpage>332</lpage>
<pub-id pub-id-type="doi">10.1109/CVPR.2001.990493</pub-id>
</element-citation>
</ref>
<!--[9] S. G. Narasimhan, and S. K. Nayar, Chromatic framework for vision in bad weather, Proc. IEEE Conf. Computer Vision and Pattern Recognition, Vol. 1, pp. 598-605, Jun. 2000.-->
<ref id="B009">
<label>[9]</label>
<element-citation publication-type="paper">
<person-group>
<name><surname>Narasimhan</surname><given-names>S. G.</given-names></name>
<name><surname>Nayar</surname><given-names>S. K.</given-names></name>
</person-group>
<year>2000</year>
<month>Jun.</month>
<article-title>Chromatic framework for vision in bad weather</article-title>
<conf-name>Proc. IEEE Conf. Computer Vision and Pattern Recognition</conf-name>
<comment>Vol. 1</comment>
<fpage>598</fpage><lpage>605</lpage>
<pub-id pub-id-type="doi">10.1109/CVPR.2000.855874</pub-id>
</element-citation>
</ref>
<!--[10] R. Fattal, Single image dehazing, Proc. acm siggraph pp. 1-11, 2008.-->
<ref id="B010">
<label>[10]</label>
<element-citation publication-type="paper">
<person-group>
<name><surname>Fattal</surname><given-names>R.</given-names></name>
</person-group>
<year>2008</year>
<conf-name>Single image dehazing</conf-name>
<comment>Proc. acm siggraph</comment>
<fpage>1</fpage><lpage>11</lpage>
<pub-id pub-id-type="doi">10.1145/1360612.1360671</pub-id>
</element-citation>
</ref>
<!--[11] K. He, J. Sun, and X. Tang, Fellow, IEEE, Single image haze removal using dark channel prior, IEEE transactions on pattern analysis and machine intelligence, Vol. 33, No. 12, pp. 2341-2353, 2011.-->
<ref id="B011">
<label>[11]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>He</surname><given-names>K.</given-names></name>
<name><surname>Sun</surname><given-names>J.</given-names></name>
<name><surname>Tang</surname><given-names>X.</given-names></name>
</person-group>
<year>2011</year>
<article-title>Fellow, IEEE, Single image haze removal using dark channel prior</article-title>
<source>IEEE transactions on pattern analysis and machine intelligence</source>
<volume>33</volume><issue>12</issue>
<fpage>2341</fpage><lpage>2353</lpage>
<pub-id pub-id-type="doi">10.1109/TPAMI.2010.168</pub-id>
</element-citation>
</ref>
<!--[12] Z. Liu, and H. Xu, Kernel parameter selection for support vector machine classification, Journal of Algorithms & Computational Technology Vol. 8, No. 2, 2013.-->
<ref id="B012">
<label>[12]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Liu</surname><given-names>Z.</given-names></name>
<name><surname>Xu</surname><given-names>H.</given-names></name>
</person-group>
<year>2013</year>
<article-title>Kernel parameter selection for support vector machine classification</article-title>
<source>Journal of Algorithms &#x26; Computational Technology</source>
<volume>8</volume><issue>2</issue>
</element-citation>
</ref>
<!--[13] D. K. Srivastava, and L. Bhambhu, Data classification using support vector machine, journal of theoretical and applied information technology, Vol. 12, No. 1, 2010.-->
<ref id="B013">
<label>[13]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Srivastava</surname><given-names>D. K.</given-names></name>
<name><surname>Bhambhu</surname><given-names>L.</given-names></name>
</person-group>
<year>2010</year>
<article-title>Data classification using support vector machine</article-title>
<source>journal of theoretical and applied information technology</source>
<volume>12</volume><issue>1</issue>
</element-citation>
</ref>
<!--[14] SVM(Support Vector Machine) Nonlinear regression model algorithms, https://github.com/nfmcclure/tensorflow_cookbook/tree/master/04_Support_Vector_Machines/05_I mplementing_Nonlinear_SVMs, 28 Dec. 2019.-->
<ref id="B014">
<label>[14]</label>
<element-citation publication-type="webpage" publication-format="web">
<source>SVM(Support Vector Machine) Nonlinear regression model algorithms</source>
<comment><uri>https://github.com/nfmcclure/tensorflow_cookbook/tree/master/04_Support_Vector_Machines/05_Implementing_Nonlinear_SVMs</uri>, 28 Dec. 2019</comment>
</element-citation>
</ref>
<!--[15] J-O. Huh, J-H. Park, D-M. Shin, and Y-R. Choi, A hierarchical SVM based Behavior inference of human operators using a hybrid sequence kernel, Vol. 11, Issue 18, 2019.-->
<ref id="B015">
<label>[15]</label>
<element-citation publication-type="journal">
<person-group>
<name><surname>Huh</surname><given-names>J-O.</given-names></name>
<name><surname>Park</surname><given-names>J-H.</given-names></name>
<name><surname>Shin</surname><given-names>D-M.</given-names></name>
<name><surname>Choi</surname><given-names>Y-R.</given-names></name>
</person-group>
<year>2019</year>
<article-title>A hierarchical SVM based Behavior inference of human operators using a hybrid sequence kernel</article-title>
<volume>11</volume><issue>18</issue>
<pub-id pub-id-type="doi">10.3390/su11184836</pub-id>
</element-citation>
</ref>
</ref-list>
<bio>
	<p><graphic xlink:href="../ingestImageView?artiId=ART002597292&amp;imageName=jkits_2020_15_03_323_f010.jpg"></graphic><bold>Eig Seub Han</bold> received the B.S. degree and M.S. degree in the Department of Plasma Bio Display from KwangWoon University in 2011 and 2013, respectively. From 2006 to 2017, he was a CEO at WizCommunication. He has been a Visiting professor in the Department of Computer System at Koea Polytechic University since 2017. His current research interests include Tensorflow, smart factory, IoT, LwM2M(Lightweight M2M), Intelligent IoT systems, Deep Learning.</p>
	<p><italic>E-mail address</italic>: <email>eshan1627@kopo.ac.kr</email></p>
</bio>
</back>
</article>
