Student Projects

Time-limited Parking Spot Detection and Timeouts Notification

Project Video

Team Members

Team Members:

Lao Danning, Chang Yiding, Chen Caiwei, Zhu Wentao, Jiang Lingyuan

Instructors:

Chong Han

Project Description

  • Problem

    Nowadays, time-limited parking spots are becoming more and more popular in cities. However, car owners often forget about the time limit, leading to fines and tickets. This project aims to develop a system that can recognize the time limit on the parking spot and convey the information to car owners to notify them of the imminent limit.


    Fig. 1 A time-limited parking spot [1]


  • Concept Generation

    For this project, we expect a fully-functional time detection and reminder system with three parts: the data-fetching part, the data-processing part, and the data-visualizing part. These three parts are connected to be a whole system with data interaction and collaboration.

    Fig. 2 General function structure


    Fig. 3 Concept Diagram



  • Design Description

    We place a Raspberry PI and a fisheye camera on a remote control car. Images fetched by the camera will be displayed on the LED monitor and sent to the computerfromtheRaspberryPI through socket connections. Using parking spot detection and OCR algorithms, we will have the time-range text. The information will be sent to our server which connects with our WeChat mini-app. When the time is out, the app will send a notification to the car owners.

    Fig. 4 The hardware system



  • Validation

    Validation Process:      

    For accuracy and missing rate, repetitive experiments are conducted.

    For text rotation degrees, the parking slots with different rotated times are used to be tested for recognition.

    For image processing speed, a timer can be used to measure.

    Validation Results:

    According to the validation part, most specifications can be met.

    Use fixed camera
     Max text length
     Visualization software
     Environment lightness>=30lux
     Processing rate<=100ms per image
     Max text rotation degree 180°
    Time range recognition accuracy>=90%
    Max parking slot detection missing rate<=10%

    √ means having been verified and · means to be determined.


  • Modeling and Analysis

    Data processing includes image preprocessing, parking slot detection and optical character recognition (OCR).

    First, images received from the server undergo gray scale conversion and fisheye distortion correction. 


    Fig. 5 Image preprocessing effect: Left: before, right: after 


    Second, contour detection extracts vertices coordinates of target parking slot and images are cropped into regions of interest.


    Fig. 6 Parking slot detection 

    Third, text obtained from Paddle OCR is formatted and sent to the server. 


  • Conclusion

    The time-limited parking spot detection and timeouts notification system can achieve end-to-end service that assist car owners to have a better control of their parking status.


  • Acknowledgement

    Faculty Advisor: Chong Han, Jigang Wu, Chengbin Ma from UM-SJTU Joint Institute

    Company Mentor: Huan Lu, Jony Xu, Zihan Chen from General Motors


  • Reference

    [1] https://wap.zzrbnews.com/view/74094