Halcon classification——（二）MLP classifier based on shape feature

目录：百家号

Step 1: Create classifier创建分类器，例程使用基于MLP的分类器

Step 2: Add training samples to the classifier添加样例

Step 3: Train the classifier训练

Step 4: Classify new objects分类未知的物 ，分割图像，获取图像的形状特征，使用分类器分类图像，达到所属的类的编号。

?分类

Step 1: Create classifier创建分类器，例程使用基于MLP的分类器

 create_class_mlp (6, 5, 3, 'softmax', 'normalization', 3, 42,MLPHandle)

Step 2: Add training samples to the classifier添加样例

添加已知类别的图像作为每个类的样例图，其中，每幅增加的图像都要进行处理：segment、add_samples

procedure segment (Image, Regions)简单的blob分析
*二值化
binary_threshold (Image, Region, 'max_separability', 'dark', UsedThreshold) 
connection (Region, ConnectedRegions) 
fill_up (ConnectedRegions, Regions) 
return ()add_samples获取形状特征，并训练分类器。
procedure add_samples (Regions, MLPHandle, Class)
count_obj (Regions, Number)
for J := 1 to Number by 1select_obj (Regions, Region, J)get_features (Region, Features)add_sample_class_mlp (MLPHandle, Features, Class)
endfor
return ()get_features获取每个region的形状特征：circular、roudness、PSI并存入Features
procedure get_features (Region, Features)
select_obj (Region, SingleRegion, 1)
circularity (SingleRegion, Circularity)
roundness (SingleRegion, Distance, Sigma, Roundness, Sides)
moments_region_central_invar (SingleRegion, PSI1, PSI2, PSI3, PSI4)
Features := [Circularity,Roundness,PSI1,PSI2,PSI3,PSI4]
return ()

Step 3: Train the classifier训练

train_class_mlp (MLPHandle, 200, 1, 0.01, Error, ErrorLog)

Step 4: Classify new objects分类未知的物 ，分割图像，获取图像的形状特征，使用分类器分类图像，达到所属的类的编号。

分类

procedure classify (Regions, MLPHandle, Classes)

count_obj (Regions, Number)

Classes := []

for J := 1 to Number by 1

select_obj (Regions, Region, J)

get_features(Region, Features)

classify_class_mlp (MLPHandle, Features, 1, Class, Confidence)

Classes := [Classes,Class]

endfor

return ()