pr.javaanpr.1.2.5.source-code.config.xml Maven / Gradle / Ivy

Go to download
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE properties SYSTEM "http://java.sun.com/dtd/properties.dtd">
<properties>
    <comment>Global configuration file for the Automatic Number Plate
        Recognition System</comment>

	<!-- PHOTO -->

	<!-- thresholding mode 0 - plain thresholding N - adaptive thresholding 
		with radius N (must be greater or equal than 1) -->
    <entry key="photo_adaptivethresholdingradius">7</entry> <!-- thresholding mode -->

	<!-- SKEW DETECTION -->

	<!-- skew detection 0 - disable 1 - enable -->
    <entry key="intelligence_skewdetection">0</entry> <!-- skew detection -->


	<!-- PLATE CANDIDATES SEARCH -->

    <entry key="intelligence_numberOfBands">3</entry> <!-- how many bands from image should be extracted from image vertical graph -->
    <entry key="intelligence_numberOfPlates">3</entry> <!-- how many plates from band should be extracted from band horizontal graph -->
    <entry key="intelligence_numberOfChars">20</entry> <!-- maximum number of chars extracted from plate's horizontal graph -->

	<!-- PLATE HEURISTICS (DETERMINES CONSTRAINTS FOR PLATE ACCEPTANCE) -->

    <entry key="intelligence_minimumChars">5</entry> <!-- minimum number of detected characters -->
    <entry key="intelligence_maximumChars">15</entry> <!-- maximum number of detected characters -->
    <entry key="intelligence_maxCharWidthDispersion">0.5</entry> <!-- maximum character width dispersion -->
    <entry key="intelligence_minPlateWidthHeightRatio">0.5</entry> <!-- plate proportions: minimum plate width/height ratio -->
    <entry key="intelligence_maxPlateWidthHeightRatio">15.0</entry> <!-- plate proportions: maximum plate width/height ratio -->

	<!-- CHARACTER HEURISTICS (DETERMINES CONSTRAINTS FOR CHARACTERS ACCEPTANCE) -->

    <entry key="intelligence_minCharWidthHeightRatio">0.1</entry> <!-- char proportions: minimum char width/height ratio -->
    <entry key="intelligence_maxCharWidthHeightRatio">0.92</entry> <!-- char proportions: maximum char width/height ratio -->
    <entry key="intelligence_maxBrightnessCostDispersion">0.161</entry> <!-- maximum character brightness difference (from other chars) -->
    <entry key="intelligence_maxContrastCostDispersion">0.1</entry> <!-- maximum character contrast difference (from other chars) -->
    <entry key="intelligence_maxHueCostDispersion">0.145</entry> <!-- maximum character hue difference (from other chars) -->
    <entry key="intelligence_maxSaturationCostDispersion">0.24</entry> <!-- maximum character saturation difference (from other chars) -->
    <entry key="intelligence_maxHeightCostDispersion">0.2</entry> <!-- maximum character height difference (from other chars) -->
    <entry key="intelligence_maxSimilarityCostDispersion">100.0</entry> <!-- maximum character cost (recognition process) -->

	<!-- CHARACTER NORMALIZATION, FEATURE EXTRACTION AND RECOGNITION MODES -->

    <entry key="char_normalizeddimensions_x">8</entry> <!-- normalized character width (downsampled) -->
    <entry key="char_normalizeddimensions_y">13</entry> <!-- normalized character height (downsampled) -->

	<!-- path to directory containing already normalized characters. Dimensions 
		of these characters must match with normalized characters width and height -->
    <entry key="char_learnAlphabetPath">/alphabets/alphabet_8x13</entry>

	<!-- character downsampling methods 0 - linear resampling (good for preserving 
		edges (edge detection)) 1 - weighted average (good for direct pixel mapping) -->
    <entry key="char_resizeMethod">1</entry> <!-- character downsampling method -->

	<!-- feature extraction method 0 - direct pixel mapping (good for blurred 
		characters) 1 - edge detection (good for skewed/deformed characters) -->
    <entry key="char_featuresExtractionMethod">0</entry> <!-- feature extraction method. 0=map, 1=edge -->

	<!-- pattern classification methods 0 - euclidean distance pattern matching 
		1 - feedforward neural network -->
    <entry key="intelligence_classification_method">0</entry> <!-- classification method. 0=euclidean distance pattern mathing, 1=neural 
		network -->

	<!-- NEURAL NETWORK LEARNING PARAMETERS -->

    <entry key="char_neuralNetworkPath">neuralnetworks/network_avgres_813_map.xml</entry> <!-- neural network topology file (caution : dimensions must match with selected 
		extraction method) -->
    <entry key="neural_maxk">8000</entry> <!-- maximum number of iterations during learning process -->
    <entry key="neural_eps">0.07</entry> <!-- expected error ratio -->
    <entry key="neural_lambda">0.05</entry> <!-- lambda factor : speed of convergence -->
    <entry key="neural_micro">0.5</entry> <!-- micro factor : persistance ratio -->
    <entry key="neural_topology">20</entry> <!-- number of neurons in middle nn layer -->

	<!-- SYNTAX ANALYSIS OF RECOGNIZED PLATE -->

	<!-- syntax analysis mode : 0 - do not correct 1 - correct characters only 
		if character count matchs 2 - correct characters anyway (eliminate redundant 
		characters) -->
    <entry key="intelligence_syntaxanalysis">2</entry> <!-- syntax analysis mode -->
    <entry key="intelligence_syntaxDescriptionFile">syntax/syntax.xml</entry>

	<!-- CAR SNAPSHOT, BAND, PLATE GRAPH ANALYSIS -->

    <entry key="carsnapshot_graphrankfilter">9</entry>
    <entry key="carsnapshot_distributormargins">25</entry>
    <entry key="carsnapshotgraph_peakDiffMultiplicationConstant">0.1</entry>
    <entry key="carsnapshotgraph_peakfootconstant">0.55</entry>

    <entry key="bandgraph_peakDiffMultiplicationConstant">0.2</entry>
    <entry key="bandgraph_peakfootconstant">0.55</entry>

    <entry key="platehorizontalgraph_detectionType">1</entry> <!-- 1=edge detection 0=magnitude derivate -->
    <entry key="platehorizontalgraph_peakfootconstant">0.05</entry>
    <entry key="plateverticalgraph_peakfootconstant">0.42</entry>
    <entry key="plategraph_rel_minpeaksize">0.86</entry>
    <entry key="plategraph_peakfootconstant">0.7</entry>

</properties>