This project implements some texture and color descriptors for real time background subtraction using cuda and opencv.
This projects aims to create a surveillance system that can be used to detect moving objects in a video stream.
We implemented two color descriptors: rgb and ycbcr and gray scale and three local binary pattern descriptors: lsbp, olbp and xcslbp.
For further details check our report Parallel_implementation_of_Local_SVD_Binary_Pattern_for_real_time_Background_Subtraction_using_CUDA.pdf
This is an example of how you may give instructions on setting up your project locally.
You need opencv4 installed with cuda support.
- Install the prerequesites
- Clone the repo
git clone https://github.com/texsmv/lsbp_background_subtraction
- Compile the project
make
You can change some configurations on the config.cfg
file.
- Place a video in data/video_input, like
video1.mp4
. - Run the executable exe: with parameters: (height, width, inputName, outputName, 1, 0)
./exe 200 400 video1.mp4 video1_out 1 0
- Check the output in data/video output.
Synthetic video:
Real video:
Alexis Mendoza - amendozavil@unsa.edu.pe - alexis@paulonia.dev
Project Link: https://github.com/texsmv/lsbp_background_subtraction