Printable View
Ibam wanting to make a lane keeper. I have been wanting to do this ffor a while with an arduino or FB but i am unsure what type of sencer to use. The only thing i can ome up with is something hiw a line following robot works. But that type of light sencer has problems when its not right on tom of the line its trying to follow.
So any ideas??
Note: a lane keeper is a system that displays how centered you are in the lane. Not all that useful but it would be cool.
If you have an android phone check out the application called iOnTheRoad .
They use the camera on the phone to do this...
I have tride that app but sinse i have a in car pc it seems a little silly to keep my phone on the windshield.
I am thinking something like a lens on a phototransistor in a tube.... complex and i dont know how to make it work. But how else could o focus on a spot smaller then a dime from two foot away?
There has to be a way to get a development board to "see" the dotted line on the road.
Well one thing you could do is look at the color of the pavement... Anything different than that is either a line or a vehicle.
Note Visual clues like this will only work on dry clear pavement.
I have actually considered putting Cameras in each of my Headlight fixtures and building a 3D camera system to figure out where things are in front of me while also using radar. Haven't done it yet but sounds cool.
how could i check the colour of something with a phototransistor?Quote:
Originally Posted by redheadedrod
a ccd seems like it would be a a heavy tax on my cpu. but on that though; how could i take one frame thats 800x600, then crop it to 800x1, then find which pixals are the brightest, then put which pixals they are in to a int so i can use the value to figure out where the car is in relevance to the line?Quote:
Originally Posted by redheadedrod
this is not hard to build if you get IR or next to IR Camera ex Wattec 802 or 803 dont remember they are 100-300 on ebay, then use computer vision object recognition to track target, use aircraft tracker (will work perfectly day and night)
If you check out the OpenCV project there are many ways to accomplish this in software once you have something in memory using one of the open source libraries you can use things like edge detection and much more.
You can look at snapshots from the video and look at these snapshots. A picture will normally reside in memory like it does on the screen.. In one long memory string that starts at the top left of the screen and ends at the bottom right. But using an open source library you really shouldn't need to worry about coding them directly.
Edit: Said OpenCL, meant OpenCV (Which stands for Computer Vision).
Hmm, very interesting...Quote:
Originally Posted by scott_fx
I was going to build something like this for a project I am working on.. No need to with this... ;)
I am designing a security application that makes use of Multiple cameras and using something like this would be nice to preserve battery power and the need for a big processor.
Rodney
chip looks good :) OpenCV is probably easier to integrate or even bettor Mitov video lab (you need 4 lines of code for tracking)