Halcon template matching actual code (I)

Template image ： The goal is to get the number of the upper left corner of the image

Direct thinking , Directly use a box to frame the number , Then template match the image （ Is not workable , Because the number in the image is not fixed ）

So we need to choose the fixed area in the image as the template , Then find our target area according to the template area , The trademark name in the upper left corner of the case is used as the template area

 

 

Code ： The case picture is in C:\Users\HJ\AppData\Roaming\MVTec\HALCON-21.05-Progress\examples\images\blister（ Install it on your own halcon To find ）

* Read template image , Understand the target requirements ： Get the number in the upper left corner of the picture and recognize it 
read_image (TemplateImage, 'C:/Users/HJ/Desktop/test_image/cd_cover/cd_cover_01.png')
get_image_size (TemplateImage, Width, Height)
dev_close_window ()
dev_open_window (0, 0, Width, Height, 'black', WindowHandle)
dev_display(TemplateImage)

* The first thought must be ： Since you are looking for numbers , Just match the template directly and find the number , But template matching looks for areas similar to the template in the image , But the numbers will obviously change .
* So we need to find all fixed areas on the image that will not change as templates , And find the center of this region , In this example, the trademark name in the upper left corner is used as the template area 
draw_rectangle1(WindowHandle, Row1, Column1, Row2, Column2)
gen_rectangle1(NumberRectangle, Row1, Column1, Row2, Column2)
reduce_domain(TemplateImage, NumberRectangle, TemplateImageReduced)
area_center(NumberRectangle, Area, CenterModelROIRow, CenterModelROIColumn)

* After selecting the template area , Our target area （ Numbers ） Below the template area , So use a box to frame the target area 
* In this way, we can know the relationship between the target area and the template area , In the subsequent template matching , After matching to the template area , The target region can be found by affine transformation 
gen_rectangle1 (NumberROI, Row2, Column1, Row2 + 30, Column2)

* Create a shape Model , Angle range 0-rad(360)
create_shape_model(TemplateImageReduced, 'auto', 0, rad(360), 'auto', 'auto', 'use_polarity', 'auto', 'auto', ModelID)
get_shape_model_contours(ShapeModel, ModelID, 1)
ImageFiles := 'cd_cover/cd_cover_'
for I := 1 to 4 by 1
    read_image(SearchImage, ImageFiles + I$'.2d')
    find_shape_model (SearchImage, ModelID, 0, rad(360), 0.7, 1, 0.7, 'least_squares', 0, 1, RowMatch, ColumnMatch, AngleMatch, Score)
    if (|Score| > 0)
        * First, affine transform the template area , Find the template area in the picture 
        vector_angle_to_rigid (0, 0, 0, RowMatch, ColumnMatch, AngleMatch, MovementOfModel)
        affine_trans_contour_xld (ShapeModel, ModelAtNewPosition, MovementOfModel)
        
        * Make an affine transformation , Obtain the affine transformation matrix of the template area , Because the result of template matching defaults to （0,0） As the origin , So you need to map to the position in the actual picture  
        vector_angle_to_rigid (CenterModelROIRow, CenterModelROIColumn, 0, RowMatch, ColumnMatch, AngleMatch, MovementOfObject)
        affine_trans_region(NumberROI, RegionAffineTrans, MovementOfObject, 'nearest_neighbor')
        
        * After obtaining the target area , Find the target area （ matrix ） The inverse of , Then do matrix operation with the inverse matrix and the original graph , Generate new images 
        hom_mat2d_invert(MovementOfObject, HomMat2DInvert)
        affine_trans_image(SearchImage, RectifiedSearchImage, HomMat2DInvert, 'constant', 'false')
        * The generated image is rotated correctly , The location of the area is the same as that of the template image , So you can get the front out target area 
        reduce_domain(RectifiedSearchImage, NumberROI, RectifiedNumberROIImage)    
    endif

endfor
clear_shape_model(ModelID)