分类器——高斯混合模型之水果分类

* 此示例程序演示如何应用通用 GMM
* 使用“面积”和“圆度”特征区分柑橘类水果的分类。此外，还对提取的水果的 2D 特征空间进行了可视化。
* 
read_image (Image, 'color/citrus_fruits_01')
get_image_pointer1 (Image, Pointer, Type, Width, Height)
dev_close_window ()
dev_open_window (0, 0, Width, Height, 'white', WindowHandle)
set_display_font (WindowHandle, 12, 'mono', 'true', 'false')
dev_set_draw ('margin')
dev_set_line_width (2)
dev_display (Image)
dev_update_window ('off')
dev_update_pc ('off')
dev_update_var ('off')
* 
FeaturesArea := []
FeaturesCircularity := []
ClassName := ['橘子','柠檬']
* 
* 创建一个高斯混合模型分类器
create_class_gmm (2, 2, 1, 'spherical', 'normalization', 10, 42, GMMHandle)
* 
* 增加训练样本
for I := 1 to 4 by 1
    read_image (Image, 'color/citrus_fruits_' + I$'.2d')
    dev_display (Image)
    * 获取样本区域
    get_regions (Image, SelectedRegions)
    dev_display (SelectedRegions)
    count_obj (SelectedRegions, NumberObjects)
    for J := 1 to NumberObjects by 1
        select_obj (SelectedRegions, ObjectSelected, J)
        get_features (ObjectSelected, WindowHandle, Circularity, Area, RowRegionCenter, ColumnRegionCenter)
        FeaturesArea := [FeaturesArea,Area]
        FeaturesCircularity := [FeaturesCircularity,Circularity]
        *样本圆度和面积特征用于，训练GMM模型
        FeatureVector := real([Circularity,Area])
        if (I <= 2)
            *前两张图像为橘子的样本
            add_sample_class_gmm (GMMHandle, FeatureVector, 0, 0)
            disp_message (WindowHandle, '至样本增加:' + ClassName[0], 'window', RowRegionCenter, ColumnRegionCenter - 100, 'white', 'false')
        else
            *从第三张图像为柠檬的样本
            add_sample_class_gmm (GMMHandle, FeatureVector, 1, 0)
            disp_message (WindowHandle, '至样本增加:' + ClassName[1], 'window', RowRegionCenter, ColumnRegionCenter - 100, 'white', 'false')
        endif
    endfor
    disp_continue_message (WindowHandle, 'white', 'false')
    stop ()
endfor
dev_clear_window ()
* 
*可视化样本特征空间
visualize_2D_feature_space (Cross, Height, Width, WindowHandle, FeaturesArea[0:5], FeaturesCircularity[0:5], 'orange', 18)
* 'oranges', 40, 440
visualize_2D_feature_space (Cross, Height, Width, WindowHandle, FeaturesArea[6:11], FeaturesCircularity[6:11], 'green', 18)
* 'lemons', 70, 440
disp_continue_message (WindowHandle, 'white', 'false')
stop ()
* 
* 训练分类器
train_class_gmm (GMMHandle, 100, 0.001, 'training', 0.0001, Centers, Iter)
* 
* 开始分类
for I := 1 to 15 by 1
    read_image (Image, 'color/citrus_fruits_' + I$'.2d')
    dev_display (Image)
    * 'Classify Image', 10, 10
    get_regions (Image, SelectedRegions)
    dev_display (SelectedRegions)
    count_obj (SelectedRegions, NumberObjects)
    for J := 1 to NumberObjects by 1
        select_obj (SelectedRegions, ObjectSelected, J)
        get_features (ObjectSelected, WindowHandle, Circularity, Area, RowRegionCenter, ColumnRegionCenter)
        FeaturesArea := [FeaturesArea,Area]
        FeaturesCircularity := [FeaturesCircularity,Circularity]
        FeatureVector := real([Circularity,Area])
        classify_class_gmm (GMMHandle, FeatureVector, 1, ClassID, ClassProb, Density, KSigmaProb)
        disp_message (WindowHandle, '得分最高类: ' + ClassName[ClassID], 'window', RowRegionCenter, ColumnRegionCenter - 100,'white', 'false')
        disp_message (WindowHandle, '得分: ' + KSigmaProb, 'window', RowRegionCenter + 30, ColumnRegionCenter - 100, 'white', 'false')
    endfor
    if (I != 15)
        disp_continue_message (WindowHandle,'white', 'false')
    endif
    stop ()
endfor
* 
* 释放分类器内存
clear_class_gmm (GMMHandle)

1.样本训练

 2.样本空间特征统计

 3.分类结果