An algorithm that detects and outlines sidewalks or driving lanes using python. These outlines can be used to help the blind navigate. The programs gives real time voice feed backs to the blind.
P.S - This algorithm was originally made to detect sidewalks. It just happened to work on driving lanes too.
Both the python files have to be ran simultaneously. The detection.py file will append it's results into a text file. The output.py file will read the results and say it out loud for the blind.
- Opencv (image processing module)
- Numpy (support for large multi-dimensional arrays and matrices)
- Math
- Time
- Scipy (polynomial regression)
- pyttsx3 (text to speech)
All the wheel files for python 3 can be downloaded from https://www.lfd.uci.edu/~gohlke/pythonlibs/ To install, run
pip install WheelFileName.whl
Alternative -
pip install opencv-python
pip install numpy
pip install pyttsx3
pip install scipy
Both the files detection.py and out.py have to be run simultaneously. The file detection.py is running its algorithm over a mp4 video called challenge.mp4. However, the same algorithm can be performed over a live video feed from a camera.
The algorithm uses the results from hough lines probabilistic (an opencv algorithm) and creates 2 linear lines to outline driving lanes. Experiments suggests that the 2 linear lines will intersect if the driving lanes curve or turn. \
Perform edge detection over the frames from challenge.mp4
Cut out reigon of interest.
Perform hough lines probabilistic over ROI frames.
Average out 2 linear lines using the results from hough lines probabilistic.
Perform quadratic regression on hough line results to outline the lanes.
Final results -
- Python3
- A ton of pizza at a hackathon ;)