Documente Academic
Documente Profesional
Documente Cultură
Austen Roberson
Intern/Mentor G/T
Title: The Future of Parking Assist: Machine Learning Using Optical Passive Sensing
Overview of Research:
This project will attempt to mitigate the problem of finding a parking space in a crowded
lot. We will be creating a machine learning program that takes input data from a camera, and
uses it to learn what a open parking space looks like. Then, we will create some type of
human-machine interface that transmits the machines findings so that a human user can know
where the spaces are. This interface may be a drone that leads a car to the parking space, or an
app that allows users to see a real time map of the parking space. This product will make it easier
to find open parking spaces more quickly and drastically reduce the time it takes to park. This
will require research of machine learning programs as well as image processing software.
Rationale:
The use of this program could impact all people globally. The reduction of the time it
takes to park can lead to less missed appointments, less stress, and less money spent on gas.
Having a program that can quickly identify parking spaces outside of a traditional parking garage
could be adapted to many different uses. People could use the program to find parking spaces in
busy cities, where otherwise they’d have to search countless minutes.
Research Methodology:
Research Design:
I will be conducting quasi-experimental research. I plan on gathering test data from a
parking lot at the Applied Physics Lab by taking a recording of the parking lot during a high
traffic time. Then using that recording, I will begin to label each individual frame with the
correct data in order to provide a baseline from which the computer can learn from and begin to
create the program. Next, we will run the remaining test data into the program to test the validity
and allow the program to continue learning. We will continue to gather test data until the
program reaches a sufficient accuracy. After the accuracy has reached an acceptable level, then a
human machine interface can be created to allow humans to access the data. This may be in the
form of an app or a drone valet.
Operational Definitions:
Artificial Intelligence- An area of computer science that supports the creation of
machines that work and react like humans
Machine Learning- A subset of artificial intelligence in which systems learn from data,
identify patterns and make decisions.
Image Recognition- The ability of software to identify objects in images.
Optical Passive Sensing- a visual sensing device that detects and responds to some type
of input from the physical environment, i.e. a camera
Human-Machine Interface-The component of the program where interactions between
humans and machines occur.
Product Overview-
By the end of the calendar year, we hope to have the testing data collected and beginning
the labeling process. By the end of January we hope to have finished labelling every frame of
testing data required. By the end of the ASPIRE program, we hope to have fully developed the
machine learning program that identifies the empty parking spaces.
Our product will first be implemented at the Applied Physics Laboratory, where it will be
rigorously tested. If the product is successful, and is allowed to continue outside of the
laboratory, then it may be implemented in schools across Howard County.
Logistical Considerations-
The major logistical concerns will involve the immense materials required to successfully
gather the data required. First, we will need a high quality camera in order to take the test data.
Then, we may have to purchase an image processing program in order to successfully label the
data. If we decide to use a drone, we will have to acquire a drone capable of processing and
receiving the data. We will also need a server with enough processing power in order to run the
code and relay the instructions to the drone. Finally, we will need power stations for the drone so
it will not die and fall on an unsuspecting person’s car.