分类器——高斯混合模型之缺陷检测（纹理缺陷检测）

合集 - Halcon(58)

50.分类器——高斯混合模型之缺陷检测（纹理缺陷检测）2024-03-07

51.基本操作之——图像纹理处理texture_laws2024-04-19 52.分类器——高斯混合模型之查找表分类器（保险丝分类）2024-03-08 53.模版匹配——set_shape_model_param2024-03-13 54.模板匹配——set_shape_model_clutter2024-03-13 55.模板匹配——金字塔图像计算gen_gauss_pyramid2024-03-13 56.模板匹配——determine_shape_model_params2024-03-14 57.模板匹配——create_shape_model2024-03-14 58.模版匹配——inspect_shape_model2024-03-15

* This example program shows you how to use the GMM classifier for novelty
* detection to perform a web inspection task.  To perform the novelty detection,
* all pixels belonging to the single trained class are computed, and are then
* subtracted from the classification ROI to extract the erroneous pixels.  For
* the web inspection task, the GMM can consequently be used to detect
* textures that do not correspond to the texture of the trained good objects.

* 此示例程序向您展示如何使用 GMM 分类器进行新颖性检测以执行 网状检测任务。 
*为了执行新颖性检测，将计算属于单个训练类的所有像素，然后从分类 ROI 中减去以提取错误像素。 
*因此，对于卷材检测任务，GMM 可用于检测与训练好的物体的纹理不对应的纹理。
* 
dev_update_off ()
* 
ReadPretrainedClassifier := false
* 取消注释以下行以从磁盘读取预训练的分类器。训练可能持续长达半分钟。
* ReadPretrainedClassifier := true
SaveClassifier := false
* 取消注释将 GMM训练样本后将分类器保存到磁盘
* SaveClassifier := true
* 
read_image (Image, 'plastic_mesh/plastic_mesh_01')
get_image_size (Image, Width, Height)
dev_close_window ()
dev_open_window (0, 0, Width, Height, 'black', WindowHandle)
dev_set_color ('red')
set_display_font (WindowHandle, 16, 'mono', 'true', 'false')
get_system ('example_dir', HalconExamples)
* The texture filters used for the classification will return artifacts at the image
* borders because the images of the plastic mesh to be inspected do not
* contain an integer number of mesh cells.  Because this would lead to wrongly
* detected errors at the image borders, we must exclude the area close to the
* image border from the training and classification.  This is done with the following
* rectangle.  Note that the image is later scaled down by a factor of two.
* 用于分类的纹理过滤器将在图像边框处返回伪影，因为要检查的塑料网格的图像不包含整数个网格单元。
*由于这会导致在图像边界处错误检测到错误，因此我们必须从训练和分类中排除靠近图像边界的区域。 
*这是通过以下矩形完成的。 请注意，图像稍后会缩小两倍。
gen_rectangle1 (Rectangle, 10, 10, Height / 2 - 11, Width / 2 - 11)
if (ReadPretrainedClassifier)
    * 从磁盘读取先前训练的GMM模型
    dev_display (Image)
    disp_message (WindowHandle, 'Reading classifier from disk...', 'window', 10, 10, 'cyan', 'false')
    read_class_gmm (HalconExamples + '/hdevelop/Segmentation/Classification/novelty_detection.gmm', GMMHandle)
    wait_seconds (1.5)
else
    * 创建GMM模型分类器
    create_class_gmm (5, 1, [1,5], 'spherical', 'normalization', 5, 42, GMMHandle)
    * 以5件完好样品来训练模型
    for J := 1 to 5 by 1
        read_image (Image, 'plastic_mesh/plastic_mesh_' + J$'02')
        * The images are zoomed down because the resolution of the mesh is very
        * high.  This saves a large amount of processing time.
        *由于网格的分辨率非常高，因此图像被缩小。 这节省了大量的处理时间。
        zoom_image_factor (Image, ImageZoomed, 0.5, 0.5, 'constant')
        dev_display (ImageZoomed)
        disp_message (WindowHandle, '训练样本添加中...', 'window', 10, 10, 'cyan', 'false')
        * Generate the texture image.
        *生成纹理图像
        gen_texture_image (ImageZoomed, ImageTexture)
        * 纹理图像添加至分类器
        add_samples_image_class_gmm (ImageTexture, Rectangle, GMMHandle, 2.0)
    endfor
    dev_display (ImageZoomed)
    disp_message (WindowHandle, 'Training GMM...', 'window', 10, 10, 'cyan', 'false')
    * 训练模型.
    train_class_gmm (GMMHandle, 100, 0.1, 'training', 0.0001, Centers, Iter)
    if (SaveClassifier)
        write_class_gmm (GMMHandle, HalconExamples + '/hdevelop/Segmentation/Classification/novelty_detection.gmm')
    endif
endif
* 检测塑料网中的异常
dev_set_draw ('margin')
dev_set_line_width (3)
for J := 1 to 14 by 1
    read_image (Image, 'plastic_mesh/plastic_mesh_' + J$'02')
    zoom_image_factor (Image, ImageZoomed, 0.5, 0.5, 'constant')
    dev_display (ImageZoomed)
    dev_set_color ('white')
    dev_display (Rectangle)
    gen_texture_image (ImageZoomed, ImageTexture)
    reduce_domain (ImageTexture, Rectangle, ImageTextureReduced)
    * Classify samples belonging to the trained class with the GMM.
    *使用 GMM 对属于训练类的样本进行分类。
    classify_image_class_gmm (ImageTextureReduced, Correct, GMMHandle, 0.000002)
    * Subtract them from the ROI to obtain the texture errors.
    *从 ROI 中减去它们以获得纹理误差。
    difference (Rectangle, Correct, Errors)
    * Postprocess the returned raw errors to remove insignificant parts of the
    * detected errors.
    * 对返回的原始错误进行后处理，以删除检测到的错误中无关紧要的部分。
    opening_circle (Errors, ErrorsOpening, 3.5)
    closing_circle (ErrorsOpening, ErrorsClosing, 10.5)
    connection (ErrorsClosing, ErrorsConnected)
    select_shape (ErrorsConnected, FinalErrors, 'area', 'and', 300, 1000000)
    count_obj (FinalErrors, NumErrors)
    dev_set_color ('red')
    dev_display (FinalErrors)
    if (NumErrors > 0)
        disp_message (WindowHandle, 'Mesh not OK', 'window', 10, 10, 'red', 'false')
    else
        disp_message (WindowHandle, 'Mesh OK', 'window', 10, 10, 'forest green', 'false')
    endif
    if (J < 14)
        disp_continue_message (WindowHandle,'cyan', 'false')
    endif
    stop ()
endfor