特别注意事项常规模板匹配仅支持刚性变换平移、旋转、均匀缩放。无法处理透视形变若目标存在透视倾斜必须先通过标定和图像校正消除形变才能匹配。且仅返回目标的位置、旋转角、缩放系数无法描述透视形变。若是需要匹配发生倾斜形变的目标参考Halcon案例create_planar_uncalib_deformable_model_xld.hdev该案例展示直接从XLD亚像素轮廓生成模型无需原始模板图像支持平面透视投影形变Homography变换。可应对倾斜拍摄带来的轮廓透视拉伸完全无需相机标定直接在像素坐标系下完成匹配直接返回包含透视变换的2D投影映射参数。常见的模板匹配有三种基于灰度值由于该模板是基于灰度值进行匹配导致使用时要求图像的光照稳定有较高的要求在实际中使用的不多。基于互相关NCC可以解决光照不均不足光照变化的场景以及焦点变化的场景图像模糊的模板匹配。*特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度。场景1焦距变化图片模糊。*本例的目的是检查ncc检测的偏差 *由其参考模型在镜头焦点变化时得到,即在图片模糊时也可很好查找到目标 *迭代遍历序列。这里使用的匹配方案是 * NCC模型。被检测模型在x和y方向上的总体运动 *最终被描绘出来以像素为单位。 * dev_close_window () read_image (ImageRef, pcb_focus/pcb_focus_telecentric_061) get_image_size (ImageRef, Width, Height) dev_open_window_fit_image (ImageRef, 0, 0, -1, -1, WindowHandle) set_display_font (WindowHandle, 14, mono, true, false) dev_update_off () dev_set_draw (margin) dev_set_color (green) dev_set_line_width (1) * * create ncc model gen_rectangle1 (ModelRegion, 81.5, 148.5, 419.5, 633.5) reduce_domain (ImageRef, ModelRegion, TemplateImage) dev_clear_window () dev_display (TemplateImage) disp_message (WindowHandle, Creating the ncc model may take a few seconds ... , image, -1, -1, black, true) *特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度 *例如模板区域旋转角度是20度使用create_ncc_model创建模板后该区域旋转角度变为0度。 create_ncc_model (TemplateImage, auto, -rad(5), rad(10), auto, use_polarity, ModelID) * * find ncc model in image area_center (ModelRegion, Area, ModelRow, ModelColumn) Rows : [] Columns : [] for Index : 1 to 121 by 1 read_image (Image, pcb_focus/pcb_focus_telecentric_ Index$03 .png) find_ncc_model (Image, ModelID, -rad(5), rad(10), 0.5, 1, 0.5, true, 0, Row, Column, Angle, Score) Rows : [Rows,Row] Columns : [Columns,Column] *特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度所以这里是0 vector_angle_to_rigid (ModelRow, ModelColumn, 0, Row, Column, Angle, HomMat2D) affine_trans_region (ModelRegion, RegionAffineTrans, HomMat2D, nearest_neighbor) set_system (flush_graphic, false) dev_display (Image) dev_display (RegionAffineTrans) gen_cross_contour_xld (Cross, Row, Column, 12, rad(0)) dev_display (Cross) disp_message (WindowHandle, Finding ncc model in image: Index$03, image, -1, -1, white, false) set_system (flush_graphic, true) disp_line (WindowHandle, -101, -101, -99, -99) endfor * * display results dev_display (Image) wait_seconds (1) gen_contour_polygon_xld (Contour, Rows, Columns) fit_line_contour_xld (Contour, tukey, -1, 0, 5, 2, RowBegin, ColBegin, RowEnd, ColEnd, Nr, Nc, Dist) gen_contour_polygon_xld (Regression, [RowBegin,RowEnd], [ColBegin,ColEnd]) * resize window so that pixels appear square Ratio : (394 - 389 1) / real(252 - 250 1) dev_resize_window_fit_size (0, 0, Height * Ratio, Height, 640, 480) dev_set_part (250, 389, 252, 394) dev_display (ImageRef) dev_set_color (green) dev_display (Contour) dev_set_color (yellow) dev_display (Regression) disp_message (WindowHandle, [Deviation of ncc matches in (x,y)-direction,while defocusing lens], image, 249.75, 389, white, false) disp_message (WindowHandle, Deviation (green), Tukey\s robust regression (yellow) in pixel, image, 251.75, 389.25, white, false) * * close all handles clear_ncc_model (ModelID)模板图像模糊的图像也能完成模板查找场景2亮度变化* This example program shows how to use HALCONs correlation-based * matching. In particular it demonstrates the robustness of this method against * linear illumination changes. The training is performed in an image with good * illumination. The matching is applied in images where the exposure time varies * extremely from very short to very long. read_image (Image, cap_exposure/cap_exposure_03) get_image_size (Image, Width, Height) dev_close_window () dev_open_window (0, 0, Width, Height, black, WindowHandle) dev_update_window (off) gen_circle (Circle, 246, 336, 150) area_center (Circle, Area, RowRef, ColumnRef) reduce_domain (Image, Circle, ImageReduced) create_ncc_model (ImageReduced, auto, 0, 0, auto, use_polarity, ModelID) dev_set_draw (margin) dev_display (Image) dev_display (Circle) stop () Rows : [] Cols : [] for J : 1 to 10 by 1 read_image (Image, cap_exposure/cap_exposure_ J$02) find_ncc_model (Image, ModelID, 0, 0, 0.5, 1, 0.5, true, 0, Row, Column, Angle, Score) vector_angle_to_rigid (RowRef, ColumnRef, 0, Row, Column, 0, HomMat2D) affine_trans_region (Circle, RegionAffineTrans, HomMat2D, nearest_neighbor) Rows : [Rows,Row] Cols : [Cols,Column] dev_display (Image) dev_display (RegionAffineTrans) stop () endfor * Compute the standard deviation of the found positions. If the individual * positions in Rows and Cols are examined, it can be seen that the standard * deviation is caused mainly by the last four images, which are severely * overexposed. StdDevRows : deviation(Rows) StdDevCols : deviation(Cols) clear_ncc_model (ModelID)模板图片不同亮度下均可匹配成功。样例图片基于形状的匹配多模板匹配参考Halcon样例位于HDevelop示例-方法-模板匹配基于形状-pm_multiple_models.hdev*多模板匹配案例 dev_close_window () dev_open_window (0, 0, 512, 512, black, WindowHandle) * dev_update_pc (off) * dev_update_window (off) * dev_update_var (off) *模型句柄数组 ModelIdS:[] *模型对应轮廓的开始索引数组 IndexS:[] *模型对应轮廓结束索引数组 IndexE:[] *多模型轮廓集合 gen_empty_obj (Models) *定义显示轮廓的颜色 Colors:[red,green,blue] *加载模板图片 for Index : 1 to 3 by 1 read_image (Image, metal-parts/metal-part-model-Index$02) get_image_size (Image, Width, Height) dev_resize_window_fit_size (0, 0, Width, Height, -1, -1) binary_threshold (Image, Region, max_separability, light, UsedThreshold) smallest_rectangle1 (Region, Row1, Column1, Row2, Column2) gen_rectangle1 (Rectangle, Row1, Column1, Row2, Column2) *获取模板图像 dilation_rectangle1 (Rectangle, RegionDilation, 10, 10) reduce_domain (Image, RegionDilation, ImageReduced) *创建模板 create_shape_model (ImageReduced, auto, rad(0), rad(360), auto, auto, use_polarity, auto, auto, ModelID) *获取模板轮廓方法1 *connection (Region, ConnectedRegions) *获取其轮廓 * gen_contour_region_xld (ConnectedRegions, Contours, border_holes) *获取模板轮廓方法2,注意ModelRegions是边界而不是区域 * inspect_shape_model (ImageReduced, ModelImages, ModelRegions, 1, 30) * connection (ModelRegions, ConnectedRegions) * select_shape (ConnectedRegions, SelectedRegions, area, and, 27.98, 500) * union1 (SelectedRegions, RegionUnion) * gen_contour_region_xld (RegionUnion, Contours, border) *获取模板轮廓方法3 get_shape_model_contours (ModelContours, ModelID, 1) select_contours_xld (ModelContours, SelectedContours, contour_length, 20, 2000, -0.5, 0.5) count_obj (SelectedContours, ContoursNumber) count_obj (Models, ModelsNumber) *将模型轮廓添加到集合中 concat_obj (Models, SelectedContours,Models) *记录该模型轮廓的起始索引与结束索引 IndexS:[IndexS,ModelsNumber1] IndexE:[IndexE,ModelsNumberContoursNumber] *将轮廓添加到集合中 *concat_obj (Models,SelectedContours, Models) *将模板句柄添加到数组 ModelIdS:[ModelIdS,ModelID] disp_message (WindowHandle, 创建模板Index, window, 20, 20, blue, true) wait_seconds (0.5) endfor stop () dev_set_line_width (2) *进行模板匹配 for i : 1 to 15 by 1 read_image (Image, metal-parts/metal-parts-i$02) dev_display (Image) *进行多模板查找。贪婪系数可影响查找数量为0时精确查找尽量查找到所有速度慢为1时快速查找可能有遗漏 find_shape_models (Image, ModelIdS, rad(0), rad(360), 0.5, 0, 0.5, least_squares, 0, 0.6, Row, Column, Angle, Score, Model) *计算匹配到的数量 num:|Score| *进行匹配成功的模板轮廓的查找仿射变换 for j : 0 to num-1 by 1 *轮廓集合中获取匹配成功的模型轮廓. *IndexS[Model[j]]:该模板轮廓在轮廓集合中的起始索引 *IndexE[Model[j]]-IndexS[Model[j]]1构成该模板轮廓的轮廓数量 copy_obj (Models, ObjectsSelected,IndexS[Model[j]] , IndexE[Model[j]]-IndexS[Model[j]]1) *进行反射变换 vector_angle_to_rigid (0, 0, 0, Row[j], Column[j], Angle[j], HomMat2D) *设置显示的颜色 dev_set_color (Colors[Model[j]]) affine_trans_contour_xld (ObjectsSelected, ContoursAffinTrans, HomMat2D) dev_display (ContoursAffinTrans) endfor set_display_font (WindowHandle, 20, mono, true, false) disp_message (WindowHandle, 查找到num个, window, 20, 20, white, false) stop () endfor for t : 0 to 2 by 1 clear_shape_model (ModelIdS[t]) endfor如果在运行时出现如下显示在图形窗口右键-更新窗口-取消“在运行模式”选项左上角轮廓将不再显示。
特别注意事项常规模板匹配仅支持刚性变换平移、旋转、均匀缩放。无法处理透视形变若目标存在透视倾斜必须先通过标定和图像校正消除形变才能匹配。且仅返回目标的位置、旋转角、缩放系数无法描述透视形变。若是需要匹配发生倾斜形变的目标参考Halcon案例create_planar_uncalib_deformable_model_xld.hdev该案例展示直接从XLD亚像素轮廓生成模型无需原始模板图像支持平面透视投影形变Homography变换。可应对倾斜拍摄带来的轮廓透视拉伸完全无需相机标定直接在像素坐标系下完成匹配直接返回包含透视变换的2D投影映射参数。常见的模板匹配有三种基于灰度值由于该模板是基于灰度值进行匹配导致使用时要求图像的光照稳定有较高的要求在实际中使用的不多。基于互相关NCC可以解决光照不均不足光照变化的场景以及焦点变化的场景图像模糊的模板匹配。*特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度。场景1焦距变化图片模糊。*本例的目的是检查ncc检测的偏差 *由其参考模型在镜头焦点变化时得到,即在图片模糊时也可很好查找到目标 *迭代遍历序列。这里使用的匹配方案是 * NCC模型。被检测模型在x和y方向上的总体运动 *最终被描绘出来以像素为单位。 * dev_close_window () read_image (ImageRef, pcb_focus/pcb_focus_telecentric_061) get_image_size (ImageRef, Width, Height) dev_open_window_fit_image (ImageRef, 0, 0, -1, -1, WindowHandle) set_display_font (WindowHandle, 14, mono, true, false) dev_update_off () dev_set_draw (margin) dev_set_color (green) dev_set_line_width (1) * * create ncc model gen_rectangle1 (ModelRegion, 81.5, 148.5, 419.5, 633.5) reduce_domain (ImageRef, ModelRegion, TemplateImage) dev_clear_window () dev_display (TemplateImage) disp_message (WindowHandle, Creating the ncc model may take a few seconds ... , image, -1, -1, black, true) *特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度 *例如模板区域旋转角度是20度使用create_ncc_model创建模板后该区域旋转角度变为0度。 create_ncc_model (TemplateImage, auto, -rad(5), rad(10), auto, use_polarity, ModelID) * * find ncc model in image area_center (ModelRegion, Area, ModelRow, ModelColumn) Rows : [] Columns : [] for Index : 1 to 121 by 1 read_image (Image, pcb_focus/pcb_focus_telecentric_ Index$03 .png) find_ncc_model (Image, ModelID, -rad(5), rad(10), 0.5, 1, 0.5, true, 0, Row, Column, Angle, Score) Rows : [Rows,Row] Columns : [Columns,Column] *特别需要注意使用create_ncc_model创建模板后模板区域角度将自动旋转到0度所以这里是0 vector_angle_to_rigid (ModelRow, ModelColumn, 0, Row, Column, Angle, HomMat2D) affine_trans_region (ModelRegion, RegionAffineTrans, HomMat2D, nearest_neighbor) set_system (flush_graphic, false) dev_display (Image) dev_display (RegionAffineTrans) gen_cross_contour_xld (Cross, Row, Column, 12, rad(0)) dev_display (Cross) disp_message (WindowHandle, Finding ncc model in image: Index$03, image, -1, -1, white, false) set_system (flush_graphic, true) disp_line (WindowHandle, -101, -101, -99, -99) endfor * * display results dev_display (Image) wait_seconds (1) gen_contour_polygon_xld (Contour, Rows, Columns) fit_line_contour_xld (Contour, tukey, -1, 0, 5, 2, RowBegin, ColBegin, RowEnd, ColEnd, Nr, Nc, Dist) gen_contour_polygon_xld (Regression, [RowBegin,RowEnd], [ColBegin,ColEnd]) * resize window so that pixels appear square Ratio : (394 - 389 1) / real(252 - 250 1) dev_resize_window_fit_size (0, 0, Height * Ratio, Height, 640, 480) dev_set_part (250, 389, 252, 394) dev_display (ImageRef) dev_set_color (green) dev_display (Contour) dev_set_color (yellow) dev_display (Regression) disp_message (WindowHandle, [Deviation of ncc matches in (x,y)-direction,while defocusing lens], image, 249.75, 389, white, false) disp_message (WindowHandle, Deviation (green), Tukey\s robust regression (yellow) in pixel, image, 251.75, 389.25, white, false) * * close all handles clear_ncc_model (ModelID)模板图像模糊的图像也能完成模板查找场景2亮度变化* This example program shows how to use HALCONs correlation-based * matching. In particular it demonstrates the robustness of this method against * linear illumination changes. The training is performed in an image with good * illumination. The matching is applied in images where the exposure time varies * extremely from very short to very long. read_image (Image, cap_exposure/cap_exposure_03) get_image_size (Image, Width, Height) dev_close_window () dev_open_window (0, 0, Width, Height, black, WindowHandle) dev_update_window (off) gen_circle (Circle, 246, 336, 150) area_center (Circle, Area, RowRef, ColumnRef) reduce_domain (Image, Circle, ImageReduced) create_ncc_model (ImageReduced, auto, 0, 0, auto, use_polarity, ModelID) dev_set_draw (margin) dev_display (Image) dev_display (Circle) stop () Rows : [] Cols : [] for J : 1 to 10 by 1 read_image (Image, cap_exposure/cap_exposure_ J$02) find_ncc_model (Image, ModelID, 0, 0, 0.5, 1, 0.5, true, 0, Row, Column, Angle, Score) vector_angle_to_rigid (RowRef, ColumnRef, 0, Row, Column, 0, HomMat2D) affine_trans_region (Circle, RegionAffineTrans, HomMat2D, nearest_neighbor) Rows : [Rows,Row] Cols : [Cols,Column] dev_display (Image) dev_display (RegionAffineTrans) stop () endfor * Compute the standard deviation of the found positions. If the individual * positions in Rows and Cols are examined, it can be seen that the standard * deviation is caused mainly by the last four images, which are severely * overexposed. StdDevRows : deviation(Rows) StdDevCols : deviation(Cols) clear_ncc_model (ModelID)模板图片不同亮度下均可匹配成功。样例图片基于形状的匹配多模板匹配参考Halcon样例位于HDevelop示例-方法-模板匹配基于形状-pm_multiple_models.hdev*多模板匹配案例 dev_close_window () dev_open_window (0, 0, 512, 512, black, WindowHandle) * dev_update_pc (off) * dev_update_window (off) * dev_update_var (off) *模型句柄数组 ModelIdS:[] *模型对应轮廓的开始索引数组 IndexS:[] *模型对应轮廓结束索引数组 IndexE:[] *多模型轮廓集合 gen_empty_obj (Models) *定义显示轮廓的颜色 Colors:[red,green,blue] *加载模板图片 for Index : 1 to 3 by 1 read_image (Image, metal-parts/metal-part-model-Index$02) get_image_size (Image, Width, Height) dev_resize_window_fit_size (0, 0, Width, Height, -1, -1) binary_threshold (Image, Region, max_separability, light, UsedThreshold) smallest_rectangle1 (Region, Row1, Column1, Row2, Column2) gen_rectangle1 (Rectangle, Row1, Column1, Row2, Column2) *获取模板图像 dilation_rectangle1 (Rectangle, RegionDilation, 10, 10) reduce_domain (Image, RegionDilation, ImageReduced) *创建模板 create_shape_model (ImageReduced, auto, rad(0), rad(360), auto, auto, use_polarity, auto, auto, ModelID) *获取模板轮廓方法1 *connection (Region, ConnectedRegions) *获取其轮廓 * gen_contour_region_xld (ConnectedRegions, Contours, border_holes) *获取模板轮廓方法2,注意ModelRegions是边界而不是区域 * inspect_shape_model (ImageReduced, ModelImages, ModelRegions, 1, 30) * connection (ModelRegions, ConnectedRegions) * select_shape (ConnectedRegions, SelectedRegions, area, and, 27.98, 500) * union1 (SelectedRegions, RegionUnion) * gen_contour_region_xld (RegionUnion, Contours, border) *获取模板轮廓方法3 get_shape_model_contours (ModelContours, ModelID, 1) select_contours_xld (ModelContours, SelectedContours, contour_length, 20, 2000, -0.5, 0.5) count_obj (SelectedContours, ContoursNumber) count_obj (Models, ModelsNumber) *将模型轮廓添加到集合中 concat_obj (Models, SelectedContours,Models) *记录该模型轮廓的起始索引与结束索引 IndexS:[IndexS,ModelsNumber1] IndexE:[IndexE,ModelsNumberContoursNumber] *将轮廓添加到集合中 *concat_obj (Models,SelectedContours, Models) *将模板句柄添加到数组 ModelIdS:[ModelIdS,ModelID] disp_message (WindowHandle, 创建模板Index, window, 20, 20, blue, true) wait_seconds (0.5) endfor stop () dev_set_line_width (2) *进行模板匹配 for i : 1 to 15 by 1 read_image (Image, metal-parts/metal-parts-i$02) dev_display (Image) *进行多模板查找。贪婪系数可影响查找数量为0时精确查找尽量查找到所有速度慢为1时快速查找可能有遗漏 find_shape_models (Image, ModelIdS, rad(0), rad(360), 0.5, 0, 0.5, least_squares, 0, 0.6, Row, Column, Angle, Score, Model) *计算匹配到的数量 num:|Score| *进行匹配成功的模板轮廓的查找仿射变换 for j : 0 to num-1 by 1 *轮廓集合中获取匹配成功的模型轮廓. *IndexS[Model[j]]:该模板轮廓在轮廓集合中的起始索引 *IndexE[Model[j]]-IndexS[Model[j]]1构成该模板轮廓的轮廓数量 copy_obj (Models, ObjectsSelected,IndexS[Model[j]] , IndexE[Model[j]]-IndexS[Model[j]]1) *进行反射变换 vector_angle_to_rigid (0, 0, 0, Row[j], Column[j], Angle[j], HomMat2D) *设置显示的颜色 dev_set_color (Colors[Model[j]]) affine_trans_contour_xld (ObjectsSelected, ContoursAffinTrans, HomMat2D) dev_display (ContoursAffinTrans) endfor set_display_font (WindowHandle, 20, mono, true, false) disp_message (WindowHandle, 查找到num个, window, 20, 20, white, false) stop () endfor for t : 0 to 2 by 1 clear_shape_model (ModelIdS[t]) endfor如果在运行时出现如下显示在图形窗口右键-更新窗口-取消“在运行模式”选项左上角轮廓将不再显示。
——字符串格式化)
![Retrospective: [Sprint N / Milestone Name]](http://pic.xiahunao.cn/yaotu/Retrospective: [Sprint N / Milestone Name])
)
)
