One 、 brief introduction

Image mosaic is a technology we often use , Personally, I once naively thought that the longest used in the project is to use halcon The kind of puzzle in the case , But it's not like this when you really do a project , This is a real case in the project .

Two 、 Project background

1、 Yes On 、 in 、 Next three pictures , And on the right of each picture is a row of calibration circles

2、 The calibration circle in the figure above can be photographed 9 A round hole , In the middle of the picture, all of the above picture and the part of the following picture are guaranteed to overlap , The following figure and the middle figure must also overlap , Here's the picture ：

3、 ... and 、 Puzzle principle

As shown in the previous two steps ：

First step ： We need to find the calibration circle on the right side of the picture , When looking for a circle ,a、 Through the first blob Analyze and find out the contour of the desired round hole , Then fit the circle , Another method is to calculate the value of the circular hole area_center, Then use a two-dimensional measuring caliper to fit the circle

*  Calculate the radius by area 
Radius:=pow(mean(Area)/rad(180),0.5)

*  Create caliper 
create_metrology_model (MetrologyHandle)
get_image_size (ImageReduced, Width, Height)
set_metrology_model_image_size (MetrologyHandle, Width, Height)

tuple_gen_const (|Row|, Radius, Radiuses)
add_metrology_object_circle_measure (MetrologyHandle, Row, Column, Radiuses, 5, 10, 1, 30, ['num_measures','measure_transition','measure_select','min_score'], [16,'positive','all',0.1], Index_circle)

apply_metrology_model (ImageReduced, MetrologyHandle)


get_metrology_object_result (MetrologyHandle, Index_circle, 'all', 'result_type', 'all_param', Parameter_circle)

The second step ： In the first step, we get the row and column coordinates of circular holes of each picture , This step is only to calculate the difference of overlapping positions , The distribution is shown in the above figure and the middle figure Of offsetx1 offsety1, And in the middle figure and the following figure offsetx2,offsety2 . There are many ways to calculate , Not here .

The third step ： adopt tile_images_offset Operator to puzzle

Four 、 According to the effect

Core code ：

Find a circle

    *  obtain 9 The radius of a circle and    Row column coordinates 
    Albert_Carib_Circle (ImageResult, ContCircleT, 'TOP', tilParams[6], [tilParams[7],tilParams[9],tilParams[10]], [tilParams[8],1,tilParams[11]], [], ProcessObjOut, ProcessObjOut, RowT, ColumnT)
    SetDictObject (ContCircleT, ProcessObjOut, ' The circle finding result in the figure above ')
    
     *  Calculate the distance between two adjacent circles  
    distance_pp (RowT[0:|RowT|-2], ColumnT[0:|RowT|-2], RowT[1:|RowT|-1], ColumnT[1:|RowT|-1], Distance)
    a := max(Distance)-min(Distance)

  Puzzle ：

 OffsetYX := [0,0,OffsetY11,OffsetX11,OffsetY11+OffsetY22,OffsetX11+OffsetX22]
   * set_dict_tuple (DictHandle_tile, ' Calibration circle extraction ', ProcessObjOut)
    * Puzzle 
    gen_empty_obj (TiledImageLight)
    gen_empty_obj (TiledImageDark)
    get_image_size (ObjectSelected, Width, Height)
    tile_images_offset (ImageLights, TiledImageLight, [OffsetYX[0],OffsetYX[2],OffsetYX[4]], [OffsetYX[1],OffsetYX[3],OffsetYX[5]], [0,0,0], [0,0,0], [Height,Height,Height], [Width,Width,Width], Width, Height*3-2000)

Final effect ：