classvisionobject.cls

优秀的面部识别程序,用VB开发的

  Dim redImage As New classImageProcessing
  Dim greenImage As New classImageProcessing
  Dim blueImage As New classImageProcessing
  Const minStep = 10
  
  
  NoOfProbabilities = 0
  maxProbability = 0
  
  'grab the scene
  Set testScene = Nothing
  Set testScene = New classImageProcessing
  Call testScene.init(width, height)
  Call testScene.update(scene, topX, topY, width, height)
  
  'grab a red image
  Call redImage.init(width, height)
  redImage.processType = 1
  Call redImage.update(scene)
  Call redImage.applyThreshold(140)
  
  'grab a green image
  Call greenImage.init(width, height)
  greenImage.processType = 2
  Call greenImage.update(scene)
  Call greenImage.applyThreshold(140)
  
  'grab a blue image
  Call blueImage.init(width, height)
  redImage.processType = 3
  Call blueImage.update(scene)
  Call blueImage.applyThreshold(180)
  
  'subtract the blue image from the red to reveal possile locations of faces
  'possible areas are then further scrutinized
  For i = 0 To width - 1
    For j = 0 To height - 1
      p1 = redImage.getPoint(i, j)
      p2 = blueImage.getPoint(i, j)
      p2 = greenImage.getPoint(i, j)
      dp = (p1 * 2) - (p2 + p3)
      If (dp < 0) Or (p1 = 0) Or (p2 > 210) Then
        Call testScene.setPoint(i, j, 0)
      End If
    Next
  Next
  
  'kill
  Set redImage = Nothing
  Set blueImage = Nothing
  Set greenImage = Nothing
  
  For stretch = 0.6 To 1 Step 0.1
  
    iWidth = imageWidth + (imageTranslation * 2)
    iHeight = imageHeight + (imageTranslation * 2)
  
    subImageWidth = width * stretch
    subImageHeight = height
  
    While (subImageWidth > iWidth) And (subImageHeight > iHeight)
  
      w = Int((width / subImageWidth) * iWidth)
      h = Int((height / subImageHeight) * iHeight)
    
      'grab the scene and store it in an array
      For x = 0 To w - 1
        xx = ((x / iWidth) * subImageWidth)
        For y = 0 To h - 1
          yy = ((y / iHeight) * subImageHeight)
          vector(x, y) = testScene.getPoint(xx, yy) / 255#
        Next
      Next
    
      For x = 0 To w - iWidth - 1 Step LocatorStepSize
        For y = 0 To h - iHeight - 1 Step LocatorStepSize
        
          xx = ((x * width) / w) + (subImageWidth / 2)
          yy = ((y * height) / h) + (subImageHeight / 2)
          p1 = testScene.getPoint(xx, yy)
          
          If (p1 > 0) Then
        
            'set the network inputs
            k = 0
            For i = 0 To imageWidth - 1
              For j = 0 To imageHeight - 1
                p1 = VisionObjectTemplate.getPoint(i, j) / 255
                p2 = vector(x + i, y + j)
                dp = Abs(p1 - p2)
                Call NNet.setInput(k, dp)
                k = k + 1
              Next
            Next

            Call NNet.feedForward

            'get the decision
            If (NNet.getClassification = CLASSIFICATION_VisionObject) Then
          
              'show the current seek window
              'If (Not IsMissing(seekPicture)) Then
              '  Call showNNimage(seekPicture)
              'End If
          
              Confidence = NNet.ClassificationConfidence
          
              'confidence must be above a minimum level
              If (Confidence > 0.95) Then
           
                If (Confidence > maxProbability) Then
                  maxProbability = Confidence
                End If
          
                'store potential VisionObject locations
                If (NoOfProbabilities < 10000) Then
              
                  VisionObjectProbability(NoOfProbabilities, 0) = topX + ((x * width) / w)
                  VisionObjectProbability(NoOfProbabilities, 1) = topY + ((y * height) / h)
                  VisionObjectProbability(NoOfProbabilities, 2) = subImageWidth
                  VisionObjectProbability(NoOfProbabilities, 3) = subImageHeight
                  VisionObjectProbability(NoOfProbabilities, 4) = Confidence
            
                  NoOfProbabilities = NoOfProbabilities + 1
                End If
                      
              End If
                        
            End If
            
          End If
        
        Next
      Next
    
      'decrease size of subimage window
      subImageWidth = subImageWidth - Shrinkage
      subImageHeight = subImageHeight - Shrinkage
    
      If (subImageWidth < iWidth) And (subImageHeight > iHeight) Then
        subImageWidth = iWidth
      End If
    
      If (subImageHeight < iHeight) And (subImageWidth > iWidth) Then
        subImageHeight = iHeight
      End If
    
    Wend
  
  Next
    
  'calculate probability matrix
  Call calcProbabilityMatrix
   
End Function



Public Sub IdentifySkinWithinPicture(scene As PictureBox, topX As Integer, topY As Integer, width As Integer, height As Integer, Optional seekPicture As PictureBox)
'scans a picture and calculates the probability of a VisionObject
'existing at any given location
  
  Dim x As Integer
  Dim y As Integer
  Dim w As Integer
  Dim h As Integer
  Dim i As Integer
  Dim subImageWidth As Integer
  Dim subImageHeight As Integer
  Dim similarity As Single
  Dim skinFound As Boolean
  Dim minWidth As Integer
  Dim minHeight As Integer
  
  'grab the scene
  Set testScene = Nothing
  Set testScene = New classImageProcessing
  Call testScene.init(width, height)
  Call testScene.update(scene, topX, topY, width, height)
    
  minWidth = width / 10
  minHeight = height / 10
  subImageWidth = width / 4
  subImageHeight = height / 4
  maxProbability = 0
  NoOfProbabilities = 0
  
  While (subImageWidth > minWidth) And (subImageHeight > minHeight)
  
    w = width - subImageWidth
    h = height - subImageHeight
    For x = 0 To w - 1 Step LocatorStepSize
      For y = 0 To h - 1 Step LocatorStepSize
      
        'compare this area with the skin tones database
        skinFound = False
        i = 0
        While (Not skinFound) And (i < NoOfSkinColours)
          similarity = skinColourTemplate(i).colourSimilarity(scene, topX + x, topY + y, subImageWidth, subImageHeight)
          If (similarity > 0.9) Then
          
            If (similarity > maxProbability) Then
              maxProbability = similarity
            End If
          
            'store potential VisionObject locations
            If (NoOfProbabilities < 10000) Then
             
              VisionObjectProbability(NoOfProbabilities, 0) = topX + x
              VisionObjectProbability(NoOfProbabilities, 1) = topY + y
              VisionObjectProbability(NoOfProbabilities, 2) = subImageWidth
              VisionObjectProbability(NoOfProbabilities, 3) = subImageHeight
              VisionObjectProbability(NoOfProbabilities, 4) = similarity
            
              NoOfProbabilities = NoOfProbabilities + 1
            End If
          
            skinFound = True
          End If
          i = i + 1
        Wend
        
      Next
    Next
  
    'decrease size of subimage window
    subImageWidth = subImageWidth - 5 'LocatorStepSize
    subImageHeight = subImageHeight - 5 'LocatorStepSize
    
  Wend
   
End Sub



Public Function IdentifyWithinPictureLinear(scene As PictureBox, topX As Integer, topY As Integer, width As Integer, height As Integer, Shrinkage As Integer, Optional seekPicture As PictureBox) As String
'scans a picture and calculates the probability of a VisionObject
'existing at any given location
  
  Dim x As Long
  Dim y As Long
  Dim subImageWidth As Long
  Dim subImageHeight As Long
  Dim w As Integer
  Dim h As Integer
  Dim i As Integer
  Dim j As Integer
  Dim vector(1000, 1000) As Single
  Dim iWidth As Integer
  Dim iHeight As Integer
  Dim Confidence As Single
  Dim xx As Integer
  Dim yy As Integer
  Dim p1 As Single
  Dim p2 As Single
  Dim dp As Single
  Dim edist As Double
  Dim stretch As Single
  Const minStep = 10
  
  NoOfProbabilities = 0
  maxProbability = 0
  
  'grab the scene
  Set testScene = Nothing
  Set testScene = New classImageProcessing
  Call testScene.init(width, height)
  Call testScene.update(scene, topX, topY, width, height)
  
  For stretch = 0.8 To 1.2 Step 0.1
  
    iWidth = imageWidth + (imageTranslation * 2)
    iHeight = imageHeight + (imageTranslation * 2)
  
    subImageWidth = width * stretch
    subImageHeight = height
  
    While (subImageWidth > iWidth) And (subImageHeight > iHeight)
  
      w = Int((width / subImageWidth) * iWidth)
      h = Int((height / subImageHeight) * iHeight)
    
      'grab the scene and store it in an array
      For x = 0 To w - 1
        xx = ((x / iWidth) * subImageWidth)
        For y = 0 To h - 1
          yy = ((y / iHeight) * subImageHeight)
          vector(x, y) = testScene.getPoint(xx, yy) / 255#
        Next
      Next
    
      For x = 0 To w - iWidth - 1 Step LocatorStepSize
        For y = 0 To h - iHeight - 1 Step LocatorStepSize
          
          edist = 0
          For i = 0 To imageWidth - 1
            For j = 0 To imageHeight - 1
              p1 = VisionObjectTemplate.getPoint(i, j) / 255
              p2 = vector(x + i, y + j)
              dp = Abs(p1 - p2)
              edist = edist + dp
            Next
          Next
          edist = edist / (imageWidth * imageHeight)
                
          'get the decision
          If (edist < 0.2) Then
           
            'show the current seek window
            'If (Not IsMissing(seekPicture)) Then
            '  Call showNNimage(seekPicture)
            'End If
           
            Confidence = 1 - edist
          
            If (Confidence > maxProbability) Then
              maxProbability = Confidence
            End If
          
            'store potential VisionObject locations
            If (NoOfProbabilities < 10000) Then
             
              VisionObjectProbability(NoOfProbabilities, 0) = topX + ((x * width) / w)
              VisionObjectProbability(NoOfProbabilities, 1) = topY + ((y * height) / h)
              VisionObjectProbability(NoOfProbabilities, 2) = subImageWidth
              VisionObjectProbability(NoOfProbabilities, 3) = subImageHeight
              VisionObjectProbability(NoOfProbabilities, 4) = Confidence
            
              NoOfProbabilities = NoOfProbabilities + 1
            End If
                        
          End If
        
        Next
      Next
    
      'decrease size of subimage window
      subImageWidth = subImageWidth - Shrinkage
      subImageHeight = subImageHeight - Shrinkage
    
    Wend
    
  Next
   
End Function





Public Sub showProbabilityMatrix(canvas As PictureBox)
  Call ProbabilityMatrix.show(canvas)
End Sub