doc/tutorials/video/background_subtraction/background_subtraction.markdown

   1 How to Use Background Subtraction Methods {#tutorial_background_subtraction}
   2 =========================================
   3
   4 -   Background subtraction (BS) is a common and widely used technique for generating a foreground
   5     mask (namely, a binary image containing the pixels belonging to moving objects in the scene) by
   6     using static cameras.
   7 -   As the name suggests, BS calculates the foreground mask performing a subtraction between the
   8     current frame and a background model, containing the static part of the scene or, more in
   9     general, everything that can be considered as background given the characteristics of the
  10     observed scene.
  11
  12     ![](images/Background_Subtraction_Tutorial_Scheme.png)
  13
  14 -   Background modeling consists of two main steps:
  15
  16     -#  Background Initialization;
  17     -#  Background Update.
  18
  19     In the first step, an initial model of the background is computed, while in the second step that
  20     model is updated in order to adapt to possible changes in the scene.
  21
  22 -   In this tutorial we will learn how to perform BS by using OpenCV.
  23
  24 Goals
  25 -----
  26
  27 In this tutorial you will learn how to:
  28
  29 -#  Read data from videos or image sequences by using @ref cv::VideoCapture ;
  30 -#  Create and update the background model by using @ref cv::BackgroundSubtractor class;
  31 -#  Get and show the foreground mask by using @ref cv::imshow ;
  32
  33 Code
  34 ----
  35
  36 In the following you can find the source code. We will let the user choose to process either a video
  37 file or a sequence of images.
  38
  39 We will use @ref cv::BackgroundSubtractorMOG2 in this sample, to generate the foreground mask.
  40
  41 The results as well as the input data are shown on the screen.
  42
  43 @add_toggle_cpp
  44 -   **Downloadable code**: Click
  45     [here](https://github.com/opencv/opencv/tree/3.4/samples/cpp/tutorial_code/video/bg_sub.cpp)
  46
  47 -   **Code at glance:**
  48     @include samples/cpp/tutorial_code/video/bg_sub.cpp
  49 @end_toggle
  50
  51 @add_toggle_java
  52 -   **Downloadable code**: Click
  53     [here](https://github.com/opencv/opencv/tree/3.4/samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java)
  54
  55 -   **Code at glance:**
  56     @include samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java
  57 @end_toggle
  58
  59 @add_toggle_python
  60 -   **Downloadable code**: Click
  61     [here](https://github.com/opencv/opencv/tree/3.4/samples/python/tutorial_code/video/background_subtraction/bg_sub.py)
  62
  63 -   **Code at glance:**
  64     @include samples/python/tutorial_code/video/background_subtraction/bg_sub.py
  65 @end_toggle
  66
  67 Explanation
  68 -----------
  69
  70 We discuss the main parts of the code above:
  71
  72 -   A @ref cv::BackgroundSubtractor object will be used to generate the foreground mask. In this
  73     example, default parameters are used, but it is also possible to declare specific parameters in
  74     the create function.
  75
  76 @add_toggle_cpp
  77 @snippet samples/cpp/tutorial_code/video/bg_sub.cpp create
  78 @end_toggle
  79
  80 @add_toggle_java
  81 @snippet samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java create
  82 @end_toggle
  83
  84 @add_toggle_python
  85 @snippet samples/python/tutorial_code/video/background_subtraction/bg_sub.py create
  86 @end_toggle
  87
  88 -   A @ref cv::VideoCapture object is used to read the input video or input images sequence.
  89
  90 @add_toggle_cpp
  91 @snippet samples/cpp/tutorial_code/video/bg_sub.cpp capture
  92 @end_toggle
  93
  94 @add_toggle_java
  95 @snippet samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java capture
  96 @end_toggle
  97
  98 @add_toggle_python
  99 @snippet samples/python/tutorial_code/video/background_subtraction/bg_sub.py capture
 100 @end_toggle
 101
 102 -   Every frame is used both for calculating the foreground mask and for updating the background. If
 103     you want to change the learning rate used for updating the background model, it is possible to
 104     set a specific learning rate by passing a parameter to the `apply` method.
 105
 106 @add_toggle_cpp
 107 @snippet samples/cpp/tutorial_code/video/bg_sub.cpp apply
 108 @end_toggle
 109
 110 @add_toggle_java
 111 @snippet samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java apply
 112 @end_toggle
 113
 114 @add_toggle_python
 115 @snippet samples/python/tutorial_code/video/background_subtraction/bg_sub.py apply
 116 @end_toggle
 117
 118 -   The current frame number can be extracted from the @ref cv::VideoCapture object and stamped in
 119     the top left corner of the current frame. A white rectangle is used to highlight the black
 120     colored frame number.
 121
 122 @add_toggle_cpp
 123 @snippet samples/cpp/tutorial_code/video/bg_sub.cpp display_frame_number
 124 @end_toggle
 125
 126 @add_toggle_java
 127 @snippet samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java display_frame_number
 128 @end_toggle
 129
 130 @add_toggle_python
 131 @snippet samples/python/tutorial_code/video/background_subtraction/bg_sub.py display_frame_number
 132 @end_toggle
 133
 134 -   We are ready to show the current input frame and the results.
 135
 136 @add_toggle_cpp
 137 @snippet samples/cpp/tutorial_code/video/bg_sub.cpp show
 138 @end_toggle
 139
 140 @add_toggle_java
 141 @snippet samples/java/tutorial_code/video/background_subtraction/BackgroundSubtractionDemo.java show
 142 @end_toggle
 143
 144 @add_toggle_python
 145 @snippet samples/python/tutorial_code/video/background_subtraction/bg_sub.py show
 146 @end_toggle
 147
 148 Results
 149 -------
 150
 151 -   With the `vtest.avi` video, for the following frame:
 152
 153     ![](images/Background_Subtraction_Tutorial_frame.jpg)
 154
 155     The output of the program will look as the following for MOG2 method (gray areas are detected shadows):
 156
 157     ![](images/Background_Subtraction_Tutorial_result_MOG2.jpg)
 158
 159     The output of the program will look as the following for the KNN method (gray areas are detected shadows):
 160
 161     ![](images/Background_Subtraction_Tutorial_result_KNN.jpg)
 162
 163 References
 164 ----------
 165
 166 -   [Background Models Challenge (BMC) website](https://web.archive.org/web/20140418093037/http://bmc.univ-bpclermont.fr/)
 167 -   A Benchmark Dataset for Foreground/Background Extraction @cite vacavant2013benchmark