<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta charset="utf-8">
<meta name="generator" content="pdf2htmlEX">
<meta http-equiv="X-UA-Compatible" content="IE=edge,chrome=1">
<link rel="stylesheet" href="https://static.pudn.com/base/css/base.min.css">
<link rel="stylesheet" href="https://static.pudn.com/base/css/fancy.min.css">
<link rel="stylesheet" href="https://static.pudn.com/prod/directory_preview_static/6252987674bc5c0105d72e33/raw.css">
<script src="https://static.pudn.com/base/js/compatibility.min.js"></script>
<script src="https://static.pudn.com/base/js/pdf2htmlEX.min.js"></script>
<script>
try{
pdf2htmlEX.defaultViewer = new pdf2htmlEX.Viewer({});
}catch(e){}
</script>
<title></title>
</head>
<body>
<div id="sidebar" style="display: none">
<div id="outline">
</div>
</div>
<div id="pf1" class="pf w0 h0" data-page-no="1"><div class="pc pc1 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6252987674bc5c0105d72e33/bg1.jpg"><div class="t m0 x1 h2 y1 ff1 fs0 fc0 sc0 ls0 ws0">License Plate Location Recognition based on Multiagent System </div><div class="t m0 x2 h3 y2 ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x2 h3 y3 ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x3 h3 y4 ff2 fs1 fc0 sc0 ls2 ws2">M. Ebrahimi </div><div class="t m0 x4 h4 y5 ff3 fs1 fc0 sc0 ls2 ws2">Department of </div><div class="t m0 x5 h4 y6 ff3 fs1 fc0 sc0 ls3 ws1">Majlesi </div><div class="t m0 x6 h4 y7 ff3 fs1 fc0 sc0 ls2 ws2">Azad University </div><div class="t m0 x7 h4 y8 ff3 fs1 fc0 sc0 ls4 ws1">hoo_Ebrahimi@y</div><div class="t m0 x8 h4 y9 ff3 fs1 fc0 sc0 ls5 ws1">ahoo.com </div><div class="t m0 x9 h3 ya ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x9 h3 yb ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 xa h3 y4 ff2 fs1 fc0 sc0 ls6 ws3">S. Ildarabadi </div><div class="t m0 xb h4 y5 ff3 fs1 fc0 sc0 ls2 ws2">Department of </div><div class="t m0 xc h4 y6 ff3 fs1 fc0 sc0 ls4 ws1">Sepahan </div><div class="t m0 xd h4 y7 ff3 fs1 fc0 sc0 ls7 ws1">University </div><div class="t m0 xe h4 y8 ff3 fs1 fc0 sc0 ls8 ws1">sedigheh_ildarab</div><div class="t m0 xf h4 y9 ff3 fs1 fc0 sc0 ls4 ws1">adi@yahoo.com </div><div class="t m0 x10 h4 yc ff3 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x10 h4 yd ff3 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x11 h3 y4 ff2 fs1 fc0 sc0 ls4 ws4">R. Monsefi </div><div class="t m0 x12 h4 y5 ff3 fs1 fc0 sc0 ls6 ws3">Faculty of </div><div class="t m0 x13 h4 y6 ff3 fs1 fc0 sc0 ls8 ws1">engineering </div><div class="t m0 x14 h4 y7 ff3 fs1 fc0 sc0 ls9 ws1">Ferdowsi </div><div class="t m0 x12 h4 y8 ff3 fs1 fc0 sc0 ls7 ws1">University </div><div class="t m0 x15 h4 y9 ff3 fs1 fc0 sc0 ls2 ws1">rmonsefi@ferdow</div><div class="t m0 x16 h4 yc ff3 fs1 fc0 sc0 lsa ws1">si.um.ac.ir<span class="ff2 fs2 lsb"> </span></div><div class="t m0 x17 h5 ye ff2 fs2 fc0 sc0 ls1 ws1"> </div><div class="t m0 x18 h3 y4 ff2 fs1 fc0 sc0 ls9 ws5">M. R. Akbarzadeh<span class="fs2 ls1 ws1"> </span></div><div class="t m0 x19 h6 yf ff3 fs2 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1a h4 y5 ff3 fs1 fc0 sc0 ls6 ws3">Faculty of </div><div class="t m0 x1b h4 y6 ff3 fs1 fc0 sc0 ls8 ws1">engineering </div><div class="t m0 x1c h4 y7 ff3 fs1 fc0 sc0 ls9 ws1">Ferdowsi </div><div class="t m0 x19 h4 y8 ff3 fs1 fc0 sc0 ls7 ws1">University </div><div class="t m0 x18 h4 y9 ff3 fs1 fc0 sc0 ls8 ws1">akbarzadeh@ieee.</div><div class="t m0 x1d h4 yc ff3 fs1 fc0 sc0 lsa ws1">org</div><div class="t m0 x2 h3 y10 ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1e h7 y11 ff1 fs1 fc0 sc0 lsc ws1">Abstract </div><div class="t m0 x1e h8 y12 ff3 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1e h8 y13 ff3 fs3 fc0 sc0 lsd ws6">A novel approach is proposed for automatic license <span class="_ _0"></span><span class="lse ws7">plate location recognition <span class="lsf ws8">based on a fusion Gabor and </span></span></div><div class="t m0 x1e h8 y14 ff3 fs3 fc0 sc0 ls10 ws9">Multiple Interlacing in a multiagent <span class="ls11 wsa">configuration. License Plate Recogn<span class="_ _0"></span><span class="ls12 wsb">ition Systems based on machine vision </span></span></div><div class="t m0 x1e h8 y15 ff3 fs3 fc0 sc0 ls13 wsc">solely need to increasing speed and accuracy in real<span class="_ _0"></span> applications (i<span class="_ _0"></span>.e. atmospheric dark and glare)<span class="_ _0"></span>. </div><div class="t m0 x1e h8 y16 ff3 fs3 fc0 sc0 ls14 wsd">In this model, the images that Speedy-MI agent has not been able to detect<span class="_ _0"></span>, is delegated to the Accurate<span class="_ _0"></span>-Gabor </div><div class="t m0 x1e h8 y17 ff3 fs3 fc0 sc0 ls15 wse">agent by the Moderator agent. The proposed <span class="_ _1"></span>model is implemented and test<span class="_ _0"></span>ed on 200 images that is toke f<span class="_ _0"></span>rom </div><div class="t m0 x1e h8 y18 ff3 fs3 fc0 sc0 ls11 wsf">Tehran Control Traffic Apartment. The results show <span class="ls16 ws10">that overall speed and accu<span class="ls17 ws11">racy has been improved. </span></span></div><div class="t m0 x2 h3 y19 ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x2 h3 y1a ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1e h3 y1b ff1 fs1 fc0 sc0 ls6 ws1">Keywords: <span class="_"> </span><span class="ff2 ws12">License Plate Location (LPL), Multiagent<span class="ls18 ws13"> System<span class="_ _0"></span>s, Multiple Interlacing, Gabor </span></span></div><div class="t m0 x1e h3 y1c ff2 fs1 fc0 sc0 lsc ws1">Transformation. </div><div class="t m0 x1e h3 y1d ff2 fs1 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1f h9 y1e ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1f h7 y1f ff1 fs1 fc0 sc0 ls9 ws5">1. Introduction </div><div class="t m0 x1f h9 y20 ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x7 h9 y21 ff2 fs3 fc0 sc0 ls19 ws14">During the past few years, intelligen<span class="_ _1"></span>t transportation </div><div class="t m0 x1f h9 y22 ff2 fs3 fc0 sc0 ls12 ws15">systems (ITSs) have had a wide i<span class="_ _0"></span>mpact in people’s <span class="_ _0"></span>life as </div><div class="t m0 x1f h9 y23 ff2 fs3 fc0 sc0 ls14 ws16">their scope is to improve transportat<span class="_ _0"></span>ion safety and </div><div class="t m0 x1f h9 y24 ff2 fs3 fc0 sc0 ls16 ws17">mobility and to enhance productivity through the use of </div><div class="t m0 x1f h9 y25 ff2 fs3 fc0 sc0 ls10 ws18">advanced technologies. We will <span class="ws19">review 4 types of license </span></div><div class="t m0 x1f h9 y26 ff2 fs3 fc0 sc0 ls1a ws1a">plate location recogn<span class="_ _1"></span>ition algorithms in this sectio<span class="_ _1"></span>n. </div><div class="t m0 x7 h9 y27 ff2 fs3 fc0 sc0 ls14 ws1b">I) As far as extraction of the plat<span class="_ _0"></span>e region is <span class="_ _0"></span>concerned, </div><div class="t m0 x1f h9 y28 ff2 fs3 fc0 sc0 ls13 ws1c">techniques based upon combinat<span class="_ _0"></span>ions of edge stati<span class="_ _0"></span>stics </div><div class="t m0 x1f h9 y29 ff2 fs3 fc0 sc0 ls15 ws1d">and mathem<span class="_ _0"></span>atical m<span class="_ _0"></span>orphology [1-4] feat<span class="_ _0"></span>ured very good </div><div class="t m0 x1f h9 y2a ff2 fs3 fc0 sc0 ls1b ws1e">results. In these m<span class="_ _0"></span>ethods, gradient <span class="_ _0"></span>magnit<span class="_ _0"></span>ude and their </div><div class="t m0 x1f h9 y2b ff2 fs3 fc0 sc0 ls1c ws1f">local variance in an image ar<span class="ls16 ws20">e computed. They are based </span></div><div class="t m0 x1f h9 y2c ff2 fs3 fc0 sc0 ls17 ws21">on the property that the brightness change in the license </div><div class="t m0 x1f h9 y2d ff2 fs3 fc0 sc0 ls1b ws22">plate region is more rem<span class="_ _0"></span>arkable and more frequent t<span class="_ _0"></span>han </div><div class="t m0 x1f h9 y2e ff2 fs3 fc0 sc0 ls15 ws23">otherwise. Block-based processing i<span class="_ _0"></span>s also supported [5]<span class="_ _0"></span>. </div><div class="t m0 x1f h9 y2f ff2 fs3 fc0 sc0 ls1d ws24">Then, regions with a high edge magnitude and hi<span class="_ _0"></span>gh edge </div><div class="t m0 x1f h9 y30 ff2 fs3 fc0 sc0 ls10 ws25">variance are identified as po<span class="ls1 ws26">ssible license plate regions. </span></div><div class="t m0 x1f h9 y31 ff2 fs3 fc0 sc0 ls14 ws27">Since this m<span class="_ _0"></span>ethod does not depend on the edge of license </div><div class="t m0 x1f h9 y32 ff2 fs3 fc0 sc0 ls13 ws28">plate boundary, it can be applied to an im<span class="_ _0"></span>age with </div><div class="t m0 x20 h9 y33 ff2 fs3 fc0 sc0 ls13 ws29">unclear license plate boundary and can be im<span class="_ _0"></span>plem<span class="_ _0"></span>ented </div><div class="t m0 x20 h9 y34 ff2 fs3 fc0 sc0 ls1e ws2a">simply and fast<span class="_ _0"></span>. A disadvantage is that <span class="_ _0"></span>edge-based </div><div class="t m0 x20 h9 y35 ff2 fs3 fc0 sc0 ls14 ws2b">methods alone can hardl<span class="_ _0"></span>y be applied to com<span class="_ _0"></span>plex im<span class="_ _0"></span>ages, </div><div class="t m0 x20 h9 y36 ff2 fs3 fc0 sc0 ls1f ws2c">since they are too sensitiv<span class="_ _1"></span>e to unwanted edg<span class="_ _1"></span>es, which </div><div class="t m0 x20 h9 y37 ff2 fs3 fc0 sc0 ls20 ws2d">may al<span class="_ _0"></span>so show high edge magnit<span class="_ _0"></span>ude or variance (e.g., <span class="_ _0"></span>the </div><div class="t m0 x20 h9 y38 ff2 fs3 fc0 sc0 ls1b ws2e">radiator region in the front vi<span class="_ _0"></span>ew of the vehicle). <span class="_ _0"></span>In spite </div><div class="t m0 x20 h9 y39 ff2 fs3 fc0 sc0 ls13 ws2f">of this, when combined with m<span class="_ _0"></span>orphological steps that </div><div class="t m0 x20 h9 y3a ff2 fs3 fc0 sc0 ls10 ws30">eliminate unwanted edges in the processed images. </div><div class="t m0 x21 h9 y3b ff2 fs3 fc0 sc0 ls13 ws31">II) In [6], a method is developed to scan a vehicl<span class="_ _0"></span>e </div><div class="t m0 x20 h9 y3c ff2 fs3 fc0 sc0 ls13 ws32">image wit<span class="_ _0"></span>h N row distance and count the existent<span class="_ _0"></span> edges. </div><div class="t m0 x20 h9 y3d ff2 fs3 fc0 sc0 ls13 ws33">If the number of the edges is great<span class="_ _0"></span>er than a threshol<span class="_ _0"></span>d </div><div class="t m0 x20 h9 y3e ff2 fs3 fc0 sc0 ls21 ws34">value, this manifests the presen<span class="ls22 ws35">ce <span class="_ _1"></span>of a plate. If in the first </span></div><div class="t m0 x20 h9 y3f ff2 fs3 fc0 sc0 ls13 ws36">scanning process the plate is not found, t<span class="_ _0"></span>hen the </div><div class="t m0 x20 h9 y40 ff2 fs3 fc0 sc0 ls15 ws37">algorithm i<span class="_ _0"></span>s repeated, reducing the threshold <span class="_ _0"></span>for counting </div><div class="t m0 x20 h9 y41 ff2 fs3 fc0 sc0 ls17 ws11">edges. The method features very<span class="_ _0"></span> fast execution tim<span class="_ _0"></span>es as it </div><div class="t m0 x20 h9 y42 ff2 fs3 fc0 sc0 ls1e ws38">scans some rows of the im<span class="_ _0"></span>age. Nonetheless, <span class="_ _0"></span>this m<span class="_ _0"></span>ethod </div><div class="t m0 x20 h9 y43 ff2 fs3 fc0 sc0 ls23 ws39">is extremely simple to locat<span class="ls24 ws3a">e license plates in several </span></div><div class="t m0 x20 h9 y44 ff2 fs3 fc0 sc0 ls10 ws3b">scenarios, and moreover, it is not size or <span class="_ _0"></span>distance </div><div class="t m0 x20 h9 y45 ff2 fs3 fc0 sc0 ls20 ws3c">independent. </div><div class="t m0 x21 h9 y46 ff2 fs3 fc0 sc0 lsd ws3d">III) Fuzzy logic has been applied to the problem <span class="_ _0"></span>of </div><div class="t m0 x20 h9 y47 ff2 fs3 fc0 sc0 ls22 ws3e">locating license plates [7-9<span class="ls1e ws3f">]. The authors ma<span class="_ _0"></span>de some </span></div><div class="c x22 y48 w2 ha"><div class="t m0 x0 hb y49 ff4 fs4 fc0 sc0 ls1 ws1">5th Iranian Conference on Machine Vision and Image Processing, November 4-6, 2008</div></div></div><div class="pi" data-data='{"ctm":[1.568627,0.000000,0.000000,1.568627,0.000000,0.000000]}'></div></div>
</body>
</html>
<div id="pf2" class="pf w0 h0" data-page-no="2"><div class="pc pc2 w0 h0"><img class="bi x0 y0 w1 h1" alt="" src="https://static.pudn.com/prod/directory_preview_static/6252987674bc5c0105d72e33/bg2.jpg"><div class="t m0 x1f h9 y4a ff2 fs3 fc0 sc0 ls25 ws40">intuitive rules <span class="_ _1"></span>to describe th<span class="_ _1"></span>e license plate and gave so<span class="_ _1"></span>me </div><div class="t m0 x1f h9 y4b ff2 fs3 fc0 sc0 ls1b ws41">mem<span class="_ _0"></span>bership functions for the <span class="ls10 ws18">fuzzy sets “bright,” “dark,” </span></div><div class="t m0 x1f h9 y4c ff2 fs3 fc0 sc0 ls26 ws42">“bright and dark sequence,” “t<span class="ls11 ws43">exture,” and “yellowness” </span></div><div class="t m0 x1f h9 y4d ff2 fs3 fc0 sc0 ls14 ws44">to get the horizontal and vert<span class="_ _0"></span>ical plate <span class="_ _0"></span>positions, but<span class="_ _0"></span> these </div><div class="t m0 x1f h9 y4e ff2 fs3 fc0 sc0 ls22 ws45">methods are sensitive to the license plate color and </div><div class="t m0 x1f h9 y4f ff2 fs3 fc0 sc0 ls14 ws46">brightness and need longer processing tim<span class="_ _0"></span>e from<span class="_ _0"></span> the </div><div class="t m0 x1f h9 y50 ff2 fs3 fc0 sc0 ls14 ws47">conventional color-based m<span class="_ _0"></span>ethods. Consequently<span class="_ _0"></span>, in spite </div><div class="t m0 x1f h9 y51 ff2 fs3 fc0 sc0 ls21 ws48">of achieving<span class="_ _1"></span> better results, they still carry th<span class="_ _1"></span>e </div><div class="t m0 x1f h9 y52 ff2 fs3 fc0 sc0 lsd ws49">disadvantages of the color-based schemes. </div><div class="t m0 x7 h9 y53 ff2 fs3 fc0 sc0 ls10 ws4a">IV) Gabor filters have been one of the major tools for </div><div class="t m0 x1f h9 y54 ff2 fs3 fc0 sc0 ls1e ws4b">texture analysis. This technique has the advant<span class="_ _0"></span>age of </div><div class="t m0 x1f h9 y55 ff2 fs3 fc0 sc0 ls1d ws4c">analyzing texture in an unli<span class="_ _0"></span>mited num<span class="_ _0"></span>ber of directions </div><div class="t m0 x1f h9 y56 ff2 fs3 fc0 sc0 ls13 ws4d">and scales. A method for l<span class="_ _0"></span>icense plate location based on </div><div class="t m0 x1f h9 y57 ff2 fs3 fc0 sc0 ls24 ws4e">the Gabor transfor<span class="_ _1"></span>m is presented in [10]. The results </div><div class="t m0 x1f h9 y58 ff2 fs3 fc0 sc0 ls15 ws4f">were encouraging (98% for LP detection) <span class="_ _0"></span>when applied </div><div class="t m0 x1f h9 y59 ff2 fs3 fc0 sc0 ls26 ws50">to digital im<span class="_ _0"></span>ages acquired strict<span class="_ _0"></span>ly in a fixed and speci<span class="_ _0"></span>fic </div><div class="t m0 x1f h9 y5a ff2 fs3 fc0 sc0 ls1b ws51">angle, but the me<span class="_ _0"></span>thod is comput<span class="_ _0"></span>ationally expensi<span class="_ _0"></span>ve and </div><div class="t m0 x1f h9 y5b ff2 fs3 fc0 sc0 ls27 ws52">slow for images with la<span class="lse ws53">rge analysis. For a two-</span></div><div class="t m0 x1f h9 y5c ff2 fs3 fc0 sc0 ls1b ws2e">dimensional<span class="_ _0"></span> (2-D) input image of size N × N and a 2-D </div><div class="t m0 x1f h9 y5d ff2 fs3 fc0 sc0 ls21 ws54">Gabor filter o<span class="_ _1"></span>f size W × W, th<span class="_ _1"></span>e computational </div><div class="t m0 x1f h9 y5e ff2 fs3 fc0 sc0 ls28 ws55">complexity of 2-<span class="_ _1"></span>D Gabor filtering is in<span class="_ _1"></span> the order of </div><div class="t m0 x23 hc y5f ff2 fs5 fc0 sc0 ls1 ws1">2<span class="_ _2"></span>2</div><div class="t m0 x24 hd y60 ff3 fs6 fc0 sc0 ls1 ws1">N<span class="_ _3"></span>W<span class="_ _4"> </span><span class="ff2 fs3 ls26 ws56">, given that the im<span class="_ _0"></span>age orientation i<span class="_ _0"></span>s fixed at a </span></div><div class="t m0 x1f h9 y61 ff2 fs3 fc0 sc0 ls26 ws57">specific angle. Therefore, this m<span class="_ _0"></span>ethod was tested <span class="_ _0"></span>on </div><div class="t m0 x1f h9 y62 ff2 fs3 fc0 sc0 ls29 ws58">small sample im<span class="_ _0"></span>ages and it was reported that further </div><div class="t m0 x1f h9 y63 ff2 fs3 fc0 sc0 ls26 ws59">work remain to be done in order to alleviate t<span class="_ _0"></span>he </div><div class="t m0 x1f h9 y64 ff2 fs3 fc0 sc0 ls27 ws5a">limitations of 2-D Gabor filtering.<span class="_ _1"></span> </div><div class="t m0 x7 h9 y65 ff2 fs3 fc0 sc0 ls1e ws5b">There is a more com<span class="_ _0"></span>plete review of t<span class="_ _0"></span>hese met<span class="_ _0"></span>hods for </div><div class="t m0 x1f h9 y66 ff2 fs3 fc0 sc0 ls10 ws5c">recognizing the place of plate <span class="ls1e ws5d">in [11,15]. In section 2, </span></div><div class="t m0 x1f h9 y67 ff2 fs3 fc0 sc0 ls28 ws55">some concepts of Multi-Agent systems are described. </div><div class="t m0 x1f h9 y68 ff2 fs3 fc0 sc0 ls26 ws5e">Section 3 describes the proposed m<span class="_ _0"></span>odel and it incl<span class="_ _0"></span>udes </div><div class="t m0 x1f h9 y69 ff2 fs3 fc0 sc0 ls14 ws5f">agents’ descriptions. In section 4, we give t<span class="_ _0"></span>he results </div><div class="t m0 x1f h9 y6a ff2 fs3 fc0 sc0 ls20 ws60">obtained by applying the me<span class="_ _0"></span>thod to databases consist<span class="_ _0"></span>ing </div><div class="t m0 x1f h9 y6b ff2 fs3 fc0 sc0 ls20 ws3c">of 200 images from Tehran C<span class="_ _0"></span>ontrol Traffic Apartm<span class="_ _0"></span>ent. In </div><div class="t m0 x1f h9 y6c ff2 fs3 fc0 sc0 ls20 ws61">Section 5 we give a few concluding rem<span class="_ _0"></span>arks. </div><div class="t m0 x1f h9 y6d ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x1f h7 y6e ff1 fs1 fc0 sc0 ls18 ws62">2. Multi-Agent Systems of Artificial </div><div class="t m0 x1f h7 y6f ff1 fs1 fc0 sc0 ls7 ws63">Intelligence </div><div class="t m0 x1f h9 y70 ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x7 h9 y71 ff2 fs3 fc0 sc0 ls25 ws64">An agent is an in<span class="_ _1"></span>telligent computational structur<span class="_ _1"></span>e </div><div class="t m0 x1f h9 y72 ff2 fs3 fc0 sc0 ls2a ws65">which can be considered as a software program or a </div><div class="t m0 x1f h9 y73 ff2 fs3 fc0 sc0 lsf ws66">robot, having interacti<span class="_ _0"></span>ons with its envi<span class="_ _0"></span>ronment and </div><div class="t m0 x1f h9 y74 ff2 fs3 fc0 sc0 lsf ws67">independent behavior and working som<span class="_ _0"></span>ewhat based on </div><div class="t m0 x1f h9 y75 ff2 fs3 fc0 sc0 ls21 ws68">its experiences. By an<span class="_ _1"></span> intelligent structure we mean an </div><div class="t m0 x1f h9 y76 ff2 fs3 fc0 sc0 lsd ws69">agent that works reasonabl<span class="ls2b ws6a">y and flexibly in an </span></div><div class="t m0 x1f h9 y77 ff2 fs3 fc0 sc0 ls1b ws6b">environment, involvi<span class="_ _0"></span>ng various occurrences. There are </div><div class="t m0 x1f h9 y78 ff2 fs3 fc0 sc0 ls19 ws6c">two main reasons to consider multi-agent systems which </div><div class="t m0 x1f h9 y79 ff2 fs3 fc0 sc0 ls17 ws6d">could have been the basic factors of the development<span class="_ _0"></span>s in </div><div class="t m0 x1f h9 y7a ff2 fs3 fc0 sc0 ls2c ws6e">this scientific field in recent <span class="ls2d ws6f">years. The first reason is that </span></div><div class="t m0 x1f h9 y7b ff2 fs3 fc0 sc0 ls1f ws70">multi-agent systems are capable of playing the main role </div><div class="t m0 x1f h9 y7c ff2 fs3 fc0 sc0 ls13 ws71">in computer scie<span class="_ _0"></span>nce and its applications now <span class="_ _0"></span>and in the </div><div class="t m0 x1f h9 y7d ff2 fs3 fc0 sc0 ls17 ws72">future, because the growth of <span class="ls2e ws73">the complexity of com<span class="_ _0"></span>puter </span></div><div class="t m0 x1f h9 y7e ff2 fs3 fc0 sc0 ls2e ws74">systems and information system<span class="_ _0"></span><span class="ls22">s leads to the increase <span class="_ _0"></span>of </span></div><div class="t m0 x1f h9 y7f ff2 fs3 fc0 sc0 ls23 ws75">the complexity of<span class="_ _1"></span> applications, an<span class="_ _1"></span>d limitless growth of </div><div class="t m0 x1f h9 y80 ff2 fs3 fc0 sc0 ls11 ws43">calculations and centralizing them need processing m<span class="_ _0"></span>ore </div><div class="t m0 x20 h9 y81 ff2 fs3 fc0 sc0 ls15 ws76">various data obtained from<span class="_ _0"></span> different locat<span class="_ _0"></span>ions. In such </div><div class="t m0 x20 h9 y82 ff2 fs3 fc0 sc0 ls10 ws5c">scenarios computers are consid<span class="ls16 ws77">ered as agents and their </span></div><div class="t m0 x20 h9 y83 ff2 fs3 fc0 sc0 ls1b ws78">interactions need to be m<span class="_ _0"></span>anaged by technology<span class="_ _0"></span>. The </div><div class="t m0 x20 h9 y84 ff2 fs3 fc0 sc0 ls10 ws79">second reason is that multi-ag<span class="ls1c ws7a">ent systems are capable of </span></div><div class="t m0 x20 h9 y85 ff2 fs3 fc0 sc0 ls2f ws7b">playing an im<span class="_ _0"></span>portant role i<span class="_ _0"></span>n analyzing and directi<span class="_ _0"></span>ng </div><div class="t m0 x20 h9 y86 ff2 fs3 fc0 sc0 ls26 ws7c">models and theories of hum<span class="_ _0"></span>an society’s i<span class="_ _0"></span>nteractions. </div><div class="t m0 x21 h9 y87 ff2 fs3 fc0 sc0 ls25 ws7d">Being intelligent means that agen<span class="_ _1"></span>ts follow their goals </div><div class="t m0 x20 h9 y88 ff2 fs3 fc0 sc0 ls1b ws7e">and act through optimizi<span class="_ _0"></span>ng a set of factors of </div><div class="t m0 x20 h9 y89 ff2 fs3 fc0 sc0 ls30 ws7f">effectiveness. When<span class="_ _1"></span> it is said that agents are intelligen<span class="_ _1"></span>t it </div><div class="t m0 x20 h9 y8a ff2 fs3 fc0 sc0 ls1e ws80">doesn’t mean that they know every thing or t<span class="_ _0"></span>hey are </div><div class="t m0 x20 h9 y8b ff2 fs3 fc0 sc0 ls14 ws81">omnipotent<span class="_ _0"></span>, but that they never m<span class="_ _0"></span>ake mi<span class="_ _0"></span>stakes while </div><div class="t m0 x20 h9 y8c ff2 fs3 fc0 sc0 ls1e ws5d">functioning. It is bet<span class="_ _0"></span>ter to say that t<span class="_ _0"></span>hey act reasonably<span class="_ _0"></span> </div><div class="t m0 x20 h9 y8d ff2 fs3 fc0 sc0 ls1d ws82">and flexibly based on their i<span class="_ _0"></span>nformation, <span class="_ _0"></span>understanding, </div><div class="t m0 x20 h9 y8e ff2 fs3 fc0 sc0 ls1f ws83">and capabilities in vario<span class="_ _1"></span>us environmental situation<span class="_ _1"></span>s. </div><div class="t m0 x21 h9 y8f ff2 fs3 fc0 sc0 ls26 ws84">The reason for focusing on the processes like problem<span class="_ _0"></span> </div><div class="t m0 x20 h9 y90 ff2 fs3 fc0 sc0 ls1d ws85">solving, planning searching, decisi<span class="_ _0"></span>on making and </div><div class="t m0 x20 h9 y91 ff2 fs3 fc0 sc0 ls30 ws86">learning is that show<span class="_ _1"></span>ing the flexibility, reason<span class="_ _1"></span>ability and </div><div class="t m0 x20 h9 y92 ff2 fs3 fc0 sc0 ls21 ws87">manipulation of su<span class="_ _1"></span>ch processes is more possib<span class="_ _1"></span>le in multi-</div><div class="t m0 x20 h9 y93 ff2 fs3 fc0 sc0 ls24 ws88">agent systems. </div><div class="t m0 x21 h9 y94 ff2 fs3 fc0 sc0 ls11 ws89">Interacting shows that agents<span class="ls13 ws8a"> may be affected by<span class="_ _0"></span> other </span></div><div class="t m0 x20 h9 y95 ff2 fs3 fc0 sc0 ls31 ws8b">agents or human whi<span class="_ _0"></span>le following <span class="_ _0"></span>their goals<span class="_ _0"></span> or doing </div><div class="t m0 x20 h9 y96 ff2 fs3 fc0 sc0 ls26 ws8c">their duties. Essentiall<span class="_ _0"></span>y, congruency as a form of </div><div class="t m0 x20 h9 y97 ff2 fs3 fc0 sc0 ls31 ws8d">interaction i<span class="_ _0"></span>s import<span class="_ _0"></span>ant in order to be consta<span class="_ _0"></span>nt through </div><div class="t m0 x20 h9 y98 ff2 fs3 fc0 sc0 ls15 ws8e">the way of achieving goals and com<span class="_ _0"></span>pleting <span class="_ _0"></span>duties. The </div><div class="t m0 x20 h9 y99 ff2 fs3 fc0 sc0 ls12 ws8f">two concepts of competition and cooperation are som<span class="_ _0"></span>e </div><div class="t m0 x20 h9 y9a ff2 fs3 fc0 sc0 ls15 ws90">models of int<span class="_ _0"></span>eractions. In <span class="lsd ws91">cooperative sit<span class="_ _0"></span>uations several </span></div><div class="t m0 x20 h9 y9b ff2 fs3 fc0 sc0 ls22 ws92">agents with a wide set of<span class="lsd ws93"> knowledge and abilities word </span></div><div class="t m0 x20 h9 y9c ff2 fs3 fc0 sc0 ls30 ws94">together to <span class="_ _1"></span>achieve a shared goal. Bu<span class="_ _1"></span>t in competitive </div><div class="t m0 x20 h9 y9d ff2 fs3 fc0 sc0 ls11 ws95">situations agents work agai<span class="ls10 ws96">nst each other, because their </span></div><div class="t m0 x20 h9 y9e ff2 fs3 fc0 sc0 ls2a ws97">goals are contradictory. Cooperative agents try, l<span class="_ _0"></span>ike a </div><div class="t m0 x20 h9 y9f ff2 fs3 fc0 sc0 ls10 ws18">team, to accomplish a job that they can <span class="_ _0"></span>not do separately, </div><div class="t m0 x20 h9 ya0 ff2 fs3 fc0 sc0 ls2c ws98">therefore they eith<span class="ls1 ws99">er succeed or fail all together. While </span></div><div class="t m0 x20 h9 ya1 ff2 fs3 fc0 sc0 ls1f ws9a">competitive agents try <span class="_ _1"></span>to maximize their benefits at other </div><div class="t m0 x20 h9 ya2 ff2 fs3 fc0 sc0 ls1e ws9b">agents’ expenses, thus one’s success results in the others’ </div><div class="t m0 x20 h9 ya3 ff2 fs3 fc0 sc0 ls12 ws9c">failure. By the way it shoul<span class="_ _0"></span>d be emphasi<span class="_ _0"></span>zed that there i<span class="_ _0"></span>s </div><div class="t m0 x20 h9 ya4 ff2 fs3 fc0 sc0 ls13 ws9d">no general agreed upon defin<span class="ls1e ws9e">ition for agents and being </span></div><div class="t m0 x20 h9 ya5 ff2 fs3 fc0 sc0 ls25 ws9f">intelligent. [12] </div><div class="t m0 x21 h9 ya6 ff2 fs3 fc0 sc0 ls25 wsa0">A useful list of the theo<span class="_ _1"></span>ries of multi-agent systems in </div><div class="t m0 x20 h9 ya7 ff2 fs3 fc0 sc0 ls1d ws82">[13] and a famous te<span class="_ _0"></span>xt about agents in chapter t<span class="_ _0"></span>wo of </div><div class="t m0 x20 h9 ya8 ff2 fs3 fc0 sc0 ls1b ws47">[14], have been presented. </div><div class="t m0 x21 h9 ya9 ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x20 h7 yaa ff1 fs1 fc0 sc0 lsa wsa1">3. Proposed model </div><div class="t m0 x20 h9 yab ff2 fs3 fc0 sc0 ls1 ws1"> </div><div class="t m0 x21 h9 yac ff2 fs3 fc0 sc0 ls13 wsa2">We divide all license plat<span class="_ _0"></span>e location me<span class="_ _0"></span>thods to two </div><div class="t m0 x20 h9 yad ff2 fs3 fc0 sc0 ls1e wsa3">parts: </div><div class="t m0 x25 he yae ff2 fs3 fc0 sc0 ls32 ws1">1.<span class="ff5 ls1"> <span class="_ _5"> </span></span><span class="ls1d wsa4">Methods which focus on speed and <span class="_ _0"></span>have a high </span></div><div class="t m0 x26 h9 yaf ff2 fs3 fc0 sc0 ls21 wsa5">speed. Multiple In<span class="_ _1"></span>terlacing [6] is the highest o<span class="_ _1"></span>f </div><div class="t m0 x26 h9 yb0 ff2 fs3 fc0 sc0 ls2e wsa6">them thus is selected here. </div><div class="t m0 x25 he yb1 ff2 fs3 fc0 sc0 ls32 ws1">2.<span class="ff5 ls1"> <span class="_ _5"> </span></span><span class="ls2e wsa7">Methods which try to increase accuracy and lose </span></div><div class="t m0 x26 h9 yb2 ff2 fs3 fc0 sc0 ls1f wsa0">time to obtain it (i.e. utilizing Gabor filter [10] that </div><div class="t m0 x26 h9 yb3 ff2 fs3 fc0 sc0 ls2e wsa6">is selected here.). </div><div class="t m0 x21 h9 yb4 ff2 fs3 fc0 sc0 ls2d wsa8">Since we need accuracy only <span class="lsd wsa9">for some of the im<span class="_ _0"></span>ages </span></div><div class="t m0 x20 h9 yb5 ff2 fs3 fc0 sc0 ls13 wsaa">and for others imm<span class="_ _0"></span>ediate m<span class="_ _0"></span>ethods are adequat<span class="_ _0"></span>e, </div><div class="c x22 y48 w2 ha"><div class="t m0 x0 hb y49 ff4 fs4 fc0 sc0 ls1 ws1">5th Iranian Conference on Machine Vision and Image Processing, November 4-6, 2008</div></div></div><div class="pi" data-data='{"ctm":[1.568627,0.000000,0.000000,1.568627,0.000000,0.000000]}'></div></div>
mardi
