2
$\begingroup$

I am working on an idea for my capstone project. I am trying to get an idea of a device, sensor or reader that can detect the front traffic light situation. so that it can tell when the light is green or red. I am still not sure what is the best way to detect the light? Is there any possibilities that the sensor can tell the light color without using any corresponding sensors or devices in each single traffic light? I was trying to find how driverless cars can see the traffic lights but still no ideas yet.

I appreciate your participation.

$\endgroup$
1
  • $\begingroup$ This is really too broad. Are you designing a system? Hardware? Software? All of the above? $\endgroup$
    – DLS3141
    Commented Jul 18, 2016 at 12:51

2 Answers 2

3
$\begingroup$

This will be done with a video camera. Writing software to process images is easier than it used to be. People are writing software on Raspberry pi (and I think, impressively, even Arduino) that do this sort of thing. And given that your application is a driverless car, it would need similar capabilities anyway. MATLAB software is actually capable of virtually writing the code for you. The difficult bits anyway. You'll need to buy various add-ons. While this is the easiest solution, bare in mind that this is still a hard problem. The tools are not for the faint hearted or the impatient. See here:

I can't imagine it having been done any way besides processing images from a camera.

Does that answers your question?

$\endgroup$
3
  • $\begingroup$ Yep, the really hard part is identifying a traffic light assembly and distinguishing from all the other crapola in the scene. (Just imagine Times Square, for example). $\endgroup$ Commented Jul 18, 2016 at 12:24
  • $\begingroup$ I am fairly positive that you cannot use Arduino for production for this application in question $\endgroup$ Commented Jul 19, 2016 at 2:33
  • $\begingroup$ @MahendraGunawardena good point - although I was just trying to say that I am pretty sure I have seen high end Arduino models used for basic computer vision in robotics. It's not impossible with suitably low frame rates/resolution, simple tasks and minimal neural network complexity, if that is the Arduino's sole task within a larger system, and if it is efficiently written in C etc (or assembler). $\endgroup$
    – Jodes
    Commented Jul 19, 2016 at 6:00
1
$\begingroup$

The state of any traffic lights is just one of the many elements that are extracted from the scene in front of the vehicle using video cameras and machine vision.

Software identifies objects such as other vehicles, pedestrians and other potential obstacles, plus information such as road edges, lane markings, signs and traffic signals.

$\endgroup$
4
  • $\begingroup$ This is not an answer. $\endgroup$ Commented Jul 18, 2016 at 12:23
  • $\begingroup$ @CarlWitthoft: I don't understand how this is any less of an answer than the one Jodes gave. I'm saying exactly the same thing, using different words. $\endgroup$
    – Dave Tweed
    Commented Jul 18, 2016 at 12:31
  • $\begingroup$ Just saying "the software does it" is not really true -- it's not like you can buy the MATLAB IdentifyStopLight toolbox. You also don't address the question of video vs. any other metric. $\endgroup$ Commented Jul 18, 2016 at 12:36
  • $\begingroup$ @CarlWitthoft: But it IS true. The only hardware involved is the video camera. The rest is entirely software. Nobody is suggesting that all of the software is commerically available -- this is a very active research topic, and anyone working on it is going to have to write some custom code. Look, this is a very broad question, because the OP is just looking for clues to get him started. Jodes and I have provided the same kind of clues. $\endgroup$
    – Dave Tweed
    Commented Jul 18, 2016 at 12:45

Not the answer you're looking for? Browse other questions tagged or ask your own question.