doc/tutorials/video/background_subtraction/background_subtraction.markdown

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