Spottr

Navigate with ease, locate open spots quickly, and get notified about busy lots. Find parking stress-free, save time, and reduce carbon emissions with our parking lot tracker. Join us in creating a greener future, one parking spot at a time.

How to run it

git clone https://github.com/sunami09/parkingprocessing.git
pip3 install -r dependencies.txt
python3 main_.py

Walkthrough

• Starting Page

  • The program starts and a pop-up window appears.
  
  
  

• Learn More

  • It gives you a brief description of what Spottr does.
  
  
  

• Try Out!

  • It gives you option to select from:
  
  
  

• Parking Spot Detection

  • It detects all the spot that are available and marks them as green while others red.
  
  
  

How we built it

• Data Collection

  • Our system requires video footage of parking lots from a top/bird's-eye view.
  • We conducted an internet search to find the necessary videos for our project.

• Reading The Videos

  • We utilized the OpenCV library to read, write, and display the videos.
  • We created two Python files, 'ParkingSpaceMarker.py' and 'main.py'.

• Area of Interest(ParkingSpaceMaker.py)

  • As running a for-loop was inconvenient due to irregular spacing in the parking lots, we manually marked the spots using rectangular boxes.
  • We then stored these coordinates in a binary file for 'main.py' to utilize and mark the spots on the video.

• Image Processing(Main.py)

  • We read the binary file and marked the parking spots as empty or occupied on the video.
  • To increase focus and reduce distractions, we converted the RGB image to grayscale.
  • We applied Gaussian blur to reduce visual noise such as shadows and reflections and to detect the edges of the parking spots more accurately.
  • We then used adaptive thresholding to segment the foreground and background and create a binary image based on pixel intensity due to poor lighting conditions in the footage.
  • We applied median blur to further smooth the image and reduce salt-and-pepper noise.
  • We used dilation to fill small gaps and reduce the risk of false negatives.
  • Finally, we counted the number of non-zero pixels in the binary image to determine which parking spots were occupied or available.

• User Experience

  • We utilized the tkinter library to create a more user-friendly interface instead of command-line interaction.

Authors

• Sunami Dasgupta
• Alessandro Sisniegas