How to drive motors through image processing using Resberry pi

Discussion in 'Embedded Systems and Microcontrollers' started by Alex Grame, Aug 4, 2018.

  1. Alex Grame

    Thread Starter New Member

    Jul 6, 2018
    6
    0
    Hi experts !
    I have chosen self drive car as my final year project. In this project i will take video from front end camera, process it using Resberry pi (mean lane detection) and then drive the motors to steers the car . But I am little bit confused. How i will give the signal to motors from resberry pi after image prpcessing.? Please guide me . Thank you so much
     
  2. danadak

    Well-Known Member

    Mar 10, 2018
    1,885
    394
    Your challenge is not trivial, especially the image recognition aspect. Road
    conditions, lane markers faded, road edge, other traffic, speed.....

    The coding for this is man years kind of effort unless you can piggy back
    on a tried and tested solution.

    I suspect the RPI is not enough horsepower to do all the recognition and
    control in real time, unless the vehicle's velocity quite low.

    Good luck, that's a challenging project for sure.

    Regards, Dana.
     
    atferrari, Alex Grame and absf like this.
  3. Alex Grame

    Thread Starter New Member

    Jul 6, 2018
    6
    0
    let suppose i have successfully detected the road lane. Now i want to steer the car according to the lanes. Mean i wish that car must remain between the road lines. but how i will guide the car motors to stay in between the road lines according to the image. please guide me
     
  4. DNA Robotics

    Member

    Jun 13, 2014
    398
    146
    The tricky part will be determining your position on the road. But since you said after image processing, it will be the same as a line following robot.
    You should be able to find lots of software and hardware examples for line following robots.
     
    Alex Grame likes this.
  5. danadak

    Well-Known Member

    Mar 10, 2018
    1,885
    394
    Line following robots not normally image recognition, they are simple
    reflective measurement opto/led/laser illuminated approaches.

    The IEEE did a lot of papers and paper collection books on image recognition,
    on issues like edge detection, and more complex objects. I am sure there must
    be more lit on the topic on the web by searching.

    Regards, Dana.
     
    Alex Grame likes this.
  6. Sensacell

    Moderator

    Jun 19, 2012
    2,016
    632
    Simplistically, if you can detect the road lanes, you can compare the current position to a theoretical desired center position.
    This could be as simple as a subtraction: (current position - desired position = error signal) the error output would be a positive or negative number. (too far left, or too far right) This error signal would be sent to a motor actuator via serial port, DAC, TCPIP or ANY OTHER output mechanism.

    The motor controller must then convert the error signal into a command position for the motor / actuator, to move the steering to the desired angle.

    When it's working, the system corrects the error to zero - i.e. car too far left, then steer right.

    When the error is zero, the car steers straight ahead, in the center of the lane.

    This is a horrendous oversimplification, but at the core, it's really what happens.
     
    Alex Grame likes this.
  7. Alex Grame

    Thread Starter New Member

    Jul 6, 2018
    6
    0
    sorry for one more thing.
    can i use Robot operating system (ROS) for self drive car? is it feasible?
    and one more question can i use laptop inside the car for controlling the over all operations.? please suggest. thank you
     
Loading...