Automatic parking slots detection using Python

[falyusuf]

TL;DR Summary: How to automatically detect the parking slots using Python?

I have a Deep Machine Learning course and would like to do my course project about "Available Parking Slots Detection using Deep Machine Learning Based on Image Processing".
My plan is to split the project into two parts: selector part and detector part.

For the selector part; how to automatically select the parking slots in the parking area? using what algorithms?

 

[berkeman]

TL;DR Summary: How to automatically detect the parking slots using Python?

I have a Deep Machine Learning course and would like to do my course project about "Available Parking Slots Detection using Deep Machine Learning Based on Image Processing".
My plan is to split the project into two parts: selector part and detector part.

For the selector part; how to automatically select the parking slots in the parking area? using what algorithms?

Since this is for a schoolwork project, please tell us what you have found so far. Also, how are you going to be given the input data? Will it be some video file that is recorded while driving around a parking lot or something? What are the constraints on that video input? Time of day/night, lighting, shadows, parking stall markings, backgrounds, etc.?

What is your experience so far with Python?

 

[falyusuf]

Several related works have manually selected parking slots using different methods, such as selecting the region of interest (ROI) through mouse clicks or determining the coordinates of each parking slot through trials and errors (which I used in my previous course project). However, these methods are not practical for large-scale parking areas and are challenging to use when selecting parallel parking spaces. Therefore, I am searching for automatic parking slot detection and updating my previous project.

I have prior experience with Python in a project that involved detecting parking slot availability through image processing. In that project, I took a picture of the parking area and manually selected the slots, as previously mentioned. I then segmented the image into five images, each representing a parking slot. The next step involved converting each image from RGB to grayscale, using the Canny function to obtain edges, counting the white pixels in each edge image, and determining if the number of white pixels was large enough to classify the parking slot as occupied (if there was a car) or free.

Challenges: Manually selection of parking slots. The challenge is that rain spots on the parking slots themselves may be detected as white pixels in the edge image, leading to incorrect detection.

My new idea is: Record a video from a high vantage point of the parking area using a laptop webcam. Use algorithms to automatically detect the parking slots and load a large dataset of images of free and occupied parking slots in Python to train a predictive model. The model will determine whether the parking slot is free or busy, and the border color of the slot will change to green or red, respectively, to indicate its status.

 

[berkeman]

I understand that this is a school project, but is there a projected application in mind? For example, for fixed cameras in an airport parking lot to be able to identify available empty slots to direct airport customers to? Or fixed cameras in a parking structure to direct customers to?

Any example pictures or videos that you can post will help this discussion.

 

[falyusuf]

but is there a projected application in mind?

Not yet.
I searched for projects in this topic but still did not get any.

 

[berkeman]

Not yet.
I searched for projects in this topic but still did not get any.

I have a Deep Machine Learning course and would like to do my course project about "Available Parking Slots Detection using Deep Machine Learning Based on Image Processing".

So it sounds like you have some flexibility in your project definition then? In that case, it would probably help the success of your project (and your grade) if you could pick a practical application to apply your program to.

I'd suggest doing some more research on practical parking space applications, and then chosing one of those or focusing on the airport parking lot application with fixed security cameras. Airport parking lots will typically have lots of cameras for security monitoring, so piggybacking on that infrastructure to add a big customer convenience application like available parking spot identification would be a good thing.

Maybe check out typical airport parking lot video monitoring setups (if you can get security clearance for that) to see what the coverage and video artifacts look like.

 

[falyusuf]

So it sounds like you have some flexibility in your project definition then?

Yes, we can choose any topics related to Deep Machine Learning. I can redo my previous project in a different way; by training a model using deep ML. However, I want to improve it even more by selecting the slots automatically and that's what I am looking for.
Here's some of the results from my previous project.