אפקה - פרויקטי גמר 2020

Smart Parking lot System

Avia Maza, Shay Rashinsky, Aviv Lazar

Advisor: Mr. Amit Stekel

Software Engineering

Car lots today do not leverage technology in any way. Our

goal is to embed algorithms into existing parking lots for the

purpose of tracking cars and managing the flow of traffic

inside the lots. Our system can be installed in any lot using

just one camera and can decrease the hassle of parking for

both parking lot owners and customers.

Initialize map of spaces by

the deployment utility

The program track all cars in the

parking lot

The program detect the

when car enter and direct

it where to park

Initialize routes map by

navigator utility

Detector:

Car Detection is used for detection from images or video, using

pre-trained deep neural networks using a combination of VGG and custom

networks for detecting cars. The Detector gives the ability to detect if in each

crop a car is present or not using a custom made NN.

Tracker:

Tracker that keeps track of all cars that move within the lot.

Car Tracking finds movement between the frames of the video.

Car Tracking does it by employing background subtraction between the

current frame and the next frame. Then it finds contours, and bounding

boxes can be formed around moving blobs of contours.

Recognizer:

Recognition giving the ability to recognize cars that were lost.

Recognition is used when the tracking system fails. By storing all detected

cars, meaning feature vector of the car in memory, cars that get lost in the

tracking process can be rediscovered using their distance from the last

known position/predicted position, as well as by deploying feature matching

and template matching between known and unknown cars.

Examples of results:

System Architecture:

The Algorithm: