Traffic Density Estimation Using OpenCV

Analysis of runtime-utility trade offs for different methods to estimate the dynamic and static traffic density using OpenCV libraries.

Installation and Building (for Ubuntu 20.04 )

$ sudo apt install libboost-program-options-dev
$ sudo apt install python3.8 python3.8-dev python3-matplotlib python3-numpy
$ mkdir -p bin
$ make

*** For Older versions of Ubuntu, replace with appropriate versions of python3 in installation and in makefile

Executing

$ bin/main -f trafficvideo.mp4 -m0

More execution options are given below.

Clean Workspace

$ make clean

Allowed options:

Flag	Usage
-h, --help	produce help message
-f, --file < filepath >	video filepath, Default : input/trafficvideo.mp4
-s, --save < filepath >	path to image of saved graph, Default : output/output.png
-m, --method< method >	choose method to apply, Default : 0
-t, --thread < num_thread >	number of threads (for method 3,4), Default: 4
-r, --frameskip < frameskip >	to process every rth frame (for method 1), Default : 1
-v, --verbose	enable logging
-x, --resolution	provide resolution at which to process image (for method 2)

Argument --method

• -m0 - Normal(baseline) method for queue density calculation
• -m1 - Subsampling (frameskip) method for queue density calculation ( Method 1)
• -m2 - Reduce resolution method for queue density calculation (Method 2)
• -m3 - Dividing work spatially for queue density calculation (Method 3)
• -m4 - Dividing work frame-wise for queue density calculation (Method 4)
• -m5 - Sparse optical flow method for moving density calculation (Method 5)
• -m6 - Dense optical flow method (baseline) for moving density calculation

Folders' description

• src - Contains all the cpp and header files
• input - Default location for input video file
• output - Contains baseline data (density.csv) and is the default output location for the output graph of traffic density
• bin - Contains executable

### Libraries Used : - **OpenCV** - For image and video operations and plotting graph in realtime - **Boost** - For argument parser - **matplotlib** - For plotting final graph - **matplotlib-cpp** - C++ bindings for matplotlib [link](https://github.com/lava/matplotlib-cpp)