In this repository you can find notebooks of applications of computer vision algorithums using python or C++ with OpenCV.
tracking.ipynb implements a histogramm based approach with CAMSHIFT.
- First we select a patch of the image.
- Use one or more changells to creat a histogramm: Check each color and put in the appropriate bin.
- Store the histogramm.
- For each image of the video feed:
- Get the color at each location.
- Find the appropriate bucket.
- The value of the bucket is the weight of the pixel.
- Use mean shift to find the center.
tracking_with_homography C++ project Findes keypoints in the selected region and tracks them with optic flow. Based on the motion of the points computes a homography and transforms the tracking window. However the computation of the homography has an unkown problem. The window is shifted to the upper left corner.
TrackingObjectFlow.py Its a tweaked version of the above in python. It adds feature matching and resets the tracking window, if match is possible.