HALCON示例程序check_blister_mixed.hedv藥品膠囊缺陷檢測

示例程序源碼（加注釋）

• 讀入圖片與顯示相關設置
  dev_close_window ()
  read_image (Image, ‘blister/blister_mixed_reference’)
  dev_open_window_fit_image (Image, 0, 0, -1, -1, WindowHandle)
  set_display_font (WindowHandle, 14, ‘mono’, ‘true’, ‘false’)
  dev_update_off ()
  dev_display (Image)
  dev_set_draw (‘margin’)
• 首先提取每個膠囊
  disp_message (WindowHandle, ‘Train gmm classifier on pill types’, ‘window’, -1, -1, ‘black’, ‘true’)
• 注意注意敲黑板，劃重點。下邊的這個函數是halcon函數的封裝，里邊代碼有好多。內容是上一篇文章介紹過，大同小異，所以不理解的話看一下上一篇例子。下邊的代碼塊是extract_pill_types里邊的內容。
  extract_pill_types (Image, Chambers, ChambersUnion, Classes, PhiRef, RowRef, ColumnRef, PillTypeCount)

threshold (Image, Region, 90, 255)
connection (Region, ConnectedRegions)
select_shape (ConnectedRegions, SelectedRegions, 'area', 'and', 5000, 999999)
shape_trans (SelectedRegions, Blister, 'convex')
gen_empty_region (Chambers)
for I := 0 to 4 by 1Row := 107 + I * 70for J := 0 to 2 by 1Column := 177 + J * 150gen_rectangle2 (Rectangle, Row, Column, 0, 64, 30)concat_obj (Chambers, Rectangle, Chambers)endfor
endfor
difference (Blister, Chambers, Pattern)
union1 (Chambers, ChambersUnion)
orientation_region (Blister, PhiRef)
PhiRef := rad(180) + PhiRef
area_center (Blister, Area, RowRef, ColumnRef)*  +++ Extract pattern for classification +++
select_shape (Chambers, PillType1, 'row', 'and', 1, 145)
union1 (PillType1, PillType1)
select_shape (Chambers, PillType2, 'row', 'and', 145, 270)
union1 (PillType2, PillType2)
select_shape (Chambers, PillType3, 'row', 'and', 270, 390)
union1 (PillType3, PillType3)* Extract color space of yellow pills
reduce_domain (Image, PillType1, ImageReduced)
decompose3 (ImageReduced, ImageR, ImageG, ImageB)
threshold (ImageB, Region, 60, 95)* Extract color space of red pills
reduce_domain (Image, PillType2, ImageReduced)
decompose3 (ImageReduced, ImageR, ImageG, ImageB)
* invert_image反轉圖像的灰度值
invert_image (ImageB, ImageInvert)
* hysteresis_threshold使用遲滯閾值操作
* 函數原型：hysteresis_threshold(Image : RegionHysteresis : Low, High, MaxLength : )
* Image ：輸入圖像；RegionHysteresis ：輸出區域；Low：低于這個灰度不會被選中；High：高于這個灰度立刻被選中；
* MaxLength ：在Low與High之間的根據MaxLength 對比度進行選擇，和carry邊緣類似。
hysteresis_threshold (ImageInvert, RegionHysteresis2, 190, 200, 5)* Extract color space of green pills
reduce_domain (Image, PillType3, ImageReduced)
decompose3 (ImageReduced, ImageR, ImageG, ImageB)
invert_image (ImageB, ImageInvert)
hysteresis_threshold (ImageInvert, RegionHysteresis3, 180, 200, 10)* 求兩個區域的交集
intersection (Region, PillType1, PillType1)
intersection (RegionHysteresis2, PillType2, PillType2)
PillTypeCount := [3,6,6]
intersection (RegionHysteresis3, PillType3, PillType3)
concat_obj (PillType1, PillType2, Classes)
concat_obj (Classes, PillType3, Classes)
return ()

NumClasses := |PillTypeCount|

• 創建創建高斯混合模型

• 輸入參數： NumDim：幾個特征；NumClasses:樣本分類個數；NumCenters:類中心的個數；CovarType:

• 協方差矩陣（‘Spherical’,‘diag’,‘full’）;Preprocessing:預處理是否使用，及特征向量轉換類型；

• NumComponents:預處理參數，轉換特征數；RandSend:迭代次數，初始化GMM；

• 輸出參數：GMMHandle:高斯混合模型
  create_class_gmm (3, 3, [1,5], ‘spherical’, ‘normalization’, 10, 42, GMMHandle)

• 含義：增加訓練樣本（圖片區），放入GMM 進行訓練數據

• 輸入參數：Image:訓練的圖像（多通道）;ClassRegions:圖像中用于訓練的類區域;

• GMMHandle:高斯混合模型;Randomize:高斯噪聲標準差
  add_samples_image_class_gmm (Image, Classes, GMMHandle, 0)

• 含義：訓練高斯混合模型

• 輸入參數：GMMHandle：高斯混合模型;MixIter: 期望最大值算法中迭代最大數;

• Threshold:期望誤差閾值，超過此值，計算取消;

• ClassPriors:類的先驗概率類型;（‘training’,‘uniform’）'uniform’表示使用相同的權重，1/NumClasses;'training’表示類發生的概率是依據樣本中的概率推斷。

• Regularize:調整值，為了防止計算中出現奇異協方差矩陣。增加到協方差矩陣的對角線上的一個很小的值

• 輸出參數：Centers:每個類的中心個數

• Iter:每個類的迭代次數
  train_class_gmm (GMMHandle, 100, 0.001, ‘training’, 0.0001, Centers, Iter)

• 檢測膠囊
  Count := 12
  for FileIndex := 1 to Count by 1

  • 讀入圖片
    read_image (Image, ‘blister/blister_mixed_’ + FileIndex$‘02’)
  • 閾值分割
    threshold (Image, Region, 90, 255)
  • 分割連通域
    connection (Region, ConnectedRegions)
  • 面積對區域篩選
    select_shape (ConnectedRegions, SelectedRegions, ‘area’, ‘and’, 5000, 9999999)
  • 求取最大區域邊緣
    shape_trans (SelectedRegions, RegionTrans, ‘convex’)
  • 利用橢圓變換，得出區域角度
    orientation_region (RegionTrans, Phi)
    if (abs(Phi) > rad(90))
    Phi := rad(180) + Phi
    endif
  • 求取區域面積與中心坐標
    area_center (RegionTrans, Area1, Row, Column)
  • 創建平移旋轉變換矩陣
    vector_angle_to_rigid (Row, Column, Phi, RowRef, ColumnRef, PhiRef, HomMat2D)
  • 使用仿射變換矩陣進行任意仿射變換
    affine_trans_image (Image, ImageAffinTrans, HomMat2D, ‘constant’, ‘false’)
  • 減少圖像定義域
    reduce_domain (ImageAffinTrans, ChambersUnion, ImageReduced)
  • 將圖片分為rgb三通道圖像
    decompose3 (ImageAffinTrans, ImageR, ImageG, ImageB)
  • 對每個膠囊進行識別檢測
  • 含義：使用GMM分類一個圖像
  • 輸入參數：Image:訓練的圖像（多通道）；GMMHandle:GMM模型；
  • RejectionThreshold:分類拒絕閾值，如果圖像所有的像素概率低于此值將不會分類任何類。
  • 輸出參數：ClassRegions:分類的各個區域
    classify_image_class_gmm (ImageReduced, ClassRegions, GMMHandle, 0.005)
  • 對ClassRegions進行計數
    count_obj (ClassRegions, Number)
  • 生成空區域
    gen_empty_obj (FinalClasses)
  • 分割連通域
    connection (Chambers, ChambersRemaining)
    for Index := Number to 1 by -1
    dev_clear_window ()
    • 選擇ClassRegions區域中的第Index個對象
      select_obj (ClassRegions, Region, Index)
    • 求交集
      intersection (ChambersRemaining, Region, Region)
    • 使用面積與寬度進行區域篩選
      select_shape (Region, PillsOfOneType, [‘area’,‘width’], ‘and’, [200,40], [3000,68])
    • 求取兩個區域的差集
      difference (ChambersUnion, PillsOfOneType, RegionDifference)
    • 分割連通域
      connection (RegionDifference, ConnectedRegions)
    • 使用面積進行區域篩選
      select_shape (ConnectedRegions, SelectedRegions, ‘area’, ‘and’, 0, 7868)
    • 求取區域的最大邊緣
      shape_trans (SelectedRegions, SelectedRegions, ‘convex’)
    • 聯合區域
      union1 (SelectedRegions, SelectedRegions)
    • 求取區域差集
      difference (ChambersRemaining, SelectedRegions, ChambersRemaining)
    • 聯合兩個區域
      concat_obj (SelectedRegions, FinalClasses, FinalClasses)
      endfor
  • 得出結果進行處理得出結論
    gen_empty_obj (MissingPills)
    gen_empty_obj (WrongPills)
    gen_empty_obj (WrongNumberOfPills)
    difference (ChambersUnion, FinalClasses, LeftOvers)
    area_center (LeftOvers, Area, Row1, Column1)
    if (Area > 0)
    connection (LeftOvers, LeftOvers)
    count_obj (LeftOvers, Number)
    for Index := 1 to Number by 1
    select_obj (LeftOvers, ObjectSelected, Index)
    intensity (ObjectSelected, ImageB, Mean, Deviation)
    if (Deviation > 40)
    concat_obj (WrongPills, ObjectSelected, WrongPills)
    else
    concat_obj (MissingPills, ObjectSelected, MissingPills)
    endif
    endfor
    endif
  • Compute histogram
    CountFinalClass := []
    for Index := 1 to NumClasses by 1
    select_obj (FinalClasses, ObjectSelected, Index)
    connection (ObjectSelected, ObjectSelected)
    count_obj (ObjectSelected, Size)
    CountFinalClass := [CountFinalClass,Size]
    endfor
  • Display classification results and output allover statistic
    display_results (ImageAffinTrans, LeftOvers, FinalClasses, WrongPills, CountFinalClass, PillTypeCount, WindowHandle)
    if (FileIndex < Count)
    disp_continue_message (WindowHandle, ‘black’, ‘true’)
    endif
    stop ()
    endfor

• Clear classifier handle
  clear_class_gmm (GMMHandle)

處理思路

此例子使用GMM高斯分類器進行膠囊識別識別，因為一版藥上有多個種類的藥片。

后記

大家有什么問題可以向我提問哈，我看到了第一時間回復，希望在學習的路上多多結交良師益友。