1 Meanshift and Camshift {#tutorial_meanshift}
4 @prev_tutorial{tutorial_background_subtraction}
5 @next_tutorial{tutorial_optical_flow}
12 - We will learn about the Meanshift and Camshift algorithms to track objects in videos.
17 The intuition behind the meanshift is simple. Consider you have a set of points. (It can be a pixel
18 distribution like histogram backprojection). You are given a small window (may be a circle) and you
19 have to move that window to the area of maximum pixel density (or maximum number of points). It is
20 illustrated in the simple image given below:
22 ![image](images/meanshift_basics.jpg)
24 The initial window is shown in blue circle with the name "C1". Its original center is marked in blue
25 rectangle, named "C1_o". But if you find the centroid of the points inside that window, you will
26 get the point "C1_r" (marked in small blue circle) which is the real centroid of the window. Surely
27 they don't match. So move your window such that the circle of the new window matches with the previous
28 centroid. Again find the new centroid. Most probably, it won't match. So move it again, and continue
29 the iterations such that the center of window and its centroid falls on the same location (or within a
30 small desired error). So finally what you obtain is a window with maximum pixel distribution. It is
31 marked with a green circle, named "C2". As you can see in the image, it has maximum number of points. The
32 whole process is demonstrated on a static image below:
34 ![image](images/meanshift_face.gif)
36 So we normally pass the histogram backprojected image and initial target location. When the object
37 moves, obviously the movement is reflected in the histogram backprojected image. As a result, the meanshift
38 algorithm moves our window to the new location with maximum density.
40 ### Meanshift in OpenCV
42 To use meanshift in OpenCV, first we need to setup the target, find its histogram so that we can
43 backproject the target on each frame for calculation of meanshift. We also need to provide an initial
44 location of window. For histogram, only Hue is considered here. Also, to avoid false values due to
45 low light, low light values are discarded using **cv.inRange()** function.
48 - **Downloadable code**: Click
49 [here](https://github.com/opencv/opencv/tree/3.4/samples/cpp/tutorial_code/video/meanshift/meanshift.cpp)
52 @include samples/cpp/tutorial_code/video/meanshift/meanshift.cpp
56 - **Downloadable code**: Click
57 [here](https://github.com/opencv/opencv/tree/3.4/samples/python/tutorial_code/video/meanshift/meanshift.py)
60 @include samples/python/tutorial_code/video/meanshift/meanshift.py
64 - **Downloadable code**: Click
65 [here](https://github.com/opencv/opencv/tree/3.4/samples/java/tutorial_code/video/meanshift/MeanshiftDemo.java)
68 @include samples/java/tutorial_code/video/meanshift/MeanshiftDemo.java
71 Three frames in a video I used is given below:
73 ![image](images/meanshift_result.jpg)
78 Did you closely watch the last result? There is a problem. Our window always has the same size whether
79 the car is very far or very close to the camera. That is not good. We need to adapt the window
80 size with size and rotation of the target. Once again, the solution came from "OpenCV Labs" and it
81 is called CAMshift (Continuously Adaptive Meanshift) published by Gary Bradsky in his paper
82 "Computer Vision Face Tracking for Use in a Perceptual User Interface" in 1998 @cite Bradski98 .
84 It applies meanshift first. Once meanshift converges, it updates the size of the window as,
85 \f$s = 2 \times \sqrt{\frac{M_{00}}{256}}\f$. It also calculates the orientation of the best fitting ellipse
86 to it. Again it applies the meanshift with new scaled search window and previous window location.
87 The process continues until the required accuracy is met.
89 ![image](images/camshift_face.gif)
91 ### Camshift in OpenCV
93 It is similar to meanshift, but returns a rotated rectangle (that is our result) and box
94 parameters (used to be passed as search window in next iteration). See the code below:
97 - **Downloadable code**: Click
98 [here](https://github.com/opencv/opencv/tree/3.4/samples/cpp/tutorial_code/video/meanshift/camshift.cpp)
100 - **Code at glance:**
101 @include samples/cpp/tutorial_code/video/meanshift/camshift.cpp
105 - **Downloadable code**: Click
106 [here](https://github.com/opencv/opencv/tree/3.4/samples/python/tutorial_code/video/meanshift/camshift.py)
108 - **Code at glance:**
109 @include samples/python/tutorial_code/video/meanshift/camshift.py
113 - **Downloadable code**: Click
114 [here](https://github.com/opencv/opencv/tree/3.4/samples/java/tutorial_code/video/meanshift/CamshiftDemo.java)
116 - **Code at glance:**
117 @include samples/java/tutorial_code/video/meanshift/CamshiftDemo.java
120 Three frames of the result is shown below:
122 ![image](images/camshift_result.jpg)
127 -# French Wikipedia page on [Camshift](http://fr.wikipedia.org/wiki/Camshift). (The two animations
128 are taken from there)
129 2. Bradski, G.R., "Real time face and object tracking as a component of a perceptual user
130 interface," Applications of Computer Vision, 1998. WACV '98. Proceedings., Fourth IEEE Workshop
131 on , vol., no., pp.214,219, 19-21 Oct 1998
136 -# OpenCV comes with a Python [sample](https://github.com/opencv/opencv/blob/3.4/samples/python/camshift.py) for an interactive demo of camshift. Use it, hack it, understand