Geometric image transforms

 The sample code and data for this tutorial can be downloaded from here. The result files are here. The accompanying slide deck for this tutorial is here. After downloading the code, run the following commands: 


acv@acv-vm:~/vision/practice/bin$ cmake ..; make; ./homework transformDemo
-- The C compiler identification is GNU 7.5.0
-- The CXX compiler identification is GNU 7.5.0
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Detecting C compile features
-- Detecting C compile features - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- Found PkgConfig: /usr/bin/pkg-config (found version "0.29.1") 
-- Found OpenCV: /usr/local (found version "4.4.0") 
-- Configuring done
-- Generating done
-- Build files have been written to: /home/acv/vision/practice/bin
Scanning dependencies of target homework
[ 14%] Building CXX object CMakeFiles/homework.dir/src/CorrelationDemo.cpp.o
[ 28%] Building CXX object CMakeFiles/homework.dir/src/GeometricTransform.cpp.o
[ 42%] Building CXX object CMakeFiles/homework.dir/src/ImageUtilities.cpp.o
[ 57%] Building CXX object CMakeFiles/homework.dir/src/Morphology.cpp.o
[ 71%] Building CXX object CMakeFiles/homework.dir/src/homework.cpp.o
[ 85%] Building CXX object CMakeFiles/homework.dir/src/util/LogManager.cpp.o
[100%] Linking CXX executable homework
[100%] Built target homework
acv@acv-vm:~/vision/practice/bin$
The code in function GeometricTransform::similarityTransformDemo of GeometricTransform.cpp shows how to perform rotation and translation of an image. The result from this code will be in the following location: /tmp/similiarityTransform_rotate.png

The code in GeometricTransform::demoAffineNonAffine shows how to match different points in two images and correct for affine and non-affine distortion. Code between lines 101 and 121 demonstrates how to match two printed circuit boards using affine transform. Using the matching points, one board can be transformed to look like the other. The matching process is shown visually in the image below. This process is also discussed in the slide deck.  


Code between lines 123 to 148 of GeometricTransform::demoAffineNonAffine demonstrates how to correct an image for perspective skew. Another example is given between lines 150 to 177. 

Comments

Post a Comment

Popular posts from this blog

Discrete Wavelet Transform

Code and sample files for this demo can be downloaded from here . Code uses only libraries from the ACV-VirtualMachine.No extra libraries are needed. You can unzip the code to any folder on your system. All paths used in the code are relative. We are using the folder "~/acv/code/dwt_demo/" Compiling and running the code is fairly straight forward. You only need to run the following two commands: acv@acv-vm : ~/acv/code/dwt_demo/bin $ cmake ../src/wavelet_demo_others/ -- The C compiler identification is GNU 7.5.0 -- The CXX compiler identification is GNU 7.5.0 -- Check for working C compiler: /usr/bin/cc -- Check for working C compiler: /usr/bin/cc -- works -- Detecting C compiler ABI info -- Detecting C compiler ABI info - done -- Detecting C compile features -- Detecting C compile features - done -- Check for working CXX compiler: /usr/bin/c++ -- Check for working CXX compiler: /usr/bin/c++ -- works -- Detecting CXX compiler ABI info -- Detecting CXX compiler ABI info - ...
Read more

Testing image capture with high speed cameras

Image
In industrial systems, objects may be moving along conveyor belts or carried on pick and place machines. A camera based inspection solution for such situations needs to take into accoun that the object in question may not be stationery, or at a fixed distance from the camera. Ergo, the captured image may not contain the object in focus, or not have the object in view at all. One option would be to stop the conveyor belt to take a photo of the object:   However, stopping a conveyor belt is not always possible. You would need cameras that can capture a clear image even when objects are moving. Enthusiasts naiively assume that one can take a video and perform analysis on individual frames of the video. Unless the video has the correct frame rate and good resolution, the solution would not work. Assume that you need to implement an industrial imaging solution to monitor items moving on a conveyor belt. You should consider specifications such as these: Distance from camera t...
Read more