@Surinder Yes, the camera model is pinhole. We didn't use Haar features because they are patented, instead HOG can be used in combination with shadows, symmetry, etc. The distance to the vehicle can be computed with the homography projecting the lowest pixels of the vehicle to the road plane, and defining the coordinate system to be at the front of the vehicle. Google inverse perspective mapping for more details. Regards!
@Marcos，greate works, excellent! Could you share your codes ? I want to fix bugs in my similar program , only for learning;and share my similar project later after fixing; thank you very much in advance! libo22@gmail.com
Hi! Well, actually if it has an ARM Cortex A9, you can build OpenCV and most of the work is done : ) Some code optimization (NEON, TBB) and possibly you can run it real-time!
Hi! Thanks a lot. I am using Hough transform to get the vanishing point, but only on startup. After that I use bump intensity detectors for lane markings detection, and road segmentation to guess the number of lanes and the presence of vehicles.
It's all done with a single camera and image processing. We had the calibration of the camera, and compute distances from the camera to the position of the vehicles in the road plane.
Could you share your code , please? I am a student and try to do a task on detection of cars and road lines of my code has some trouble keeping the car recognition
Sí! Aunque este trabajo es un poco antiguo ya utilizaba OpenCV 1.1. Ahora tenemos cosas mucho mejores con OpenCV 2.4.6, y pronto saldrá OpenCV 3.0.0. Un saludo!
@marcos I used HOG+SVM for car detection using car.xml. But It is very slow taking 4-6 minutes to detect car and it also have false detections. Pls help.
Hi Marcos, I would like to know, how do you handle the situations, when there is no lane on the road ? There are times, when vehicle moves from a road having lane to a road having no lane.
@marcos Nice work you did. I need your some help. did you used pinhole camera model to get camera matrix? did you used haar cascade classifier for object detection? Which formula you have used to calculate distance of car from camera using homography ?
hi, I really appreciate the good result. I know it may be inconvenient for you to share the code. Could you tell some key points? How do you detect the lanes without other lines? Do you use hough or other to detect lines? How do you improve the accuracy of detection? Best regards, Neal
@Macros Is the height of object is known? In my case object height needs to be calculate and also distance to camera. Can you tell me any way to find object height? Also, now I have bird eye view of Object, but how to find the height of Object in Bird eye view. I am only left with Object Height. Please help.
You can use flat world assumption to get linear reprojection from image plane to road plane (assuming undistortion applied) and metric calibration available. For better accuracy, tracking the vehicle in the bird's-eye view can also help fix the homography. Finally, if you have a good detector, you can also obtain 3D estimates of the vehicle from its 2D bounding box, which then you can use to estimate it's distance to the camera if the vehicle dimensions are approximately known.
se puede utilizar en robotica, m mi carrera es industrias alimentarias, podria aplicarlo en procesos industriales. te agradeceria que fueras mas especifico en el programa o la idea del codigo fuente.
hola, soy estudiante de ingenieria y estoy realizando un trabajo similar en opencv, me gustaria saber si me podrias enviar algun tipo de tutorial de como haces para determinar ese "porcentaje" que sale en que supongo que es el grado de desviacion de la linea amarilla marcada con respecto a que tanto se acerca a las azules dentro del cuadro verde.. necesito hacer algo parecido.. espero me puedas ayudar, de ante mano gracias,, genial el video
Hola! Gracias por escribir. Lo mejor que puedo hacer es dirigirte a los links que he compartido anteriormente. En mi blog (marcosnietoblog.wordpress.com) puedes encontrar prácticamente todo mi material, incluido mi tesis doctoral, donde viene todo explicado. Espero que tengas suerte! Un saludo, Marcos
Do a Google search for Self Driving Cars. There's an article in the NY Daily News about Audi and BMW pushing forward with this technology. Google has authorization from the state of California to operate thier autonomous vehicle as long as a person is supervising. Applications similar to this that are connected to various sensors around the vehicle extend the autonomous technology. In about 20 years we'll just punch in our destination on the GPS, and the car will do the rest.
Hi! first of all, thank you for sharing your work with us. :D I am working one vehicle tracking algorithms. what you have done is great. and I would like to know what did you used or serouce code if it's public plz :)
Porschen Hund, The road is approximated as a plane in the local neighborhood so we can compute an homography between image and road plane. The distance is then computed assuming the cars are on the plane.
Great work there, am doing similar thing in my fyp.. But using matlab instead.. I am now in dire need of code or at least know how to reach this point.......
it could be ported to Android, I am sure, but I am not focusing on that now. Actually, porting such security systems to mobile platforms would be only for entertainment, not security. BR
This is awesome... Imagine if you can make a game out of this, something like if you can get pass any number of vehicle in a certain time(My weird thoughts) :)
Yes, we use a single camera, and indeed the processing was done in grayscale, although visualization was kept in color.
@Surinder
Yes, the camera model is pinhole. We didn't use Haar features because they are patented, instead HOG can be used in combination with shadows, symmetry, etc.
The distance to the vehicle can be computed with the homography projecting the lowest pixels of the vehicle to the road plane, and defining the coordinate system to be at the front of the vehicle. Google inverse perspective mapping for more details.
Regards!
@Marcos，greate works, excellent!
Could you share your codes ? I want to fix bugs in my similar program , only for learning;and share my similar project later after fixing;
thank you very much in advance!
libo22@gmail.com
Marcos Nieto Awesome work, its like an airplane to avoid terrains. Very similar to Rolls Royce program for parking.
Do youhave the code？
The distance is measured wit computer vision only. The camera is calibrated and the distance can be computed assuming that the road is locally planar.
Hi! Well, actually if it has an ARM Cortex A9, you can build OpenCV and most of the work is done : )
Some code optimization (NEON, TBB) and possibly you can run it real-time!
Hi! Thanks a lot.
I am using Hough transform to get the vanishing point, but only on startup. After that I use bump intensity detectors for lane markings detection, and road segmentation to guess the number of lanes and the presence of vehicles.
Hi Marcos, how do you calculate bump intensity ? Could you please explain? I appreciate it in advance
Hi Marcos, I am also curious about how you segment the roads ? Could you please give a short brief ?
We assume the road is flat in the close distance (
The distance is computed since we know the calibration matrix of the camera, and compute the homography between the image plane and the road plane.
It's all done with a single camera and image processing. We had the calibration of the camera, and compute distances from the camera to the position of the vehicles in the road plane.
Yes, mostly data structures, image handlers, and some drawing functions.
Wondering what platform you are using to achieve this frame rate.
Could you share your code , please? I am a student and try to do a task on detection of cars and road lines of my code has some trouble keeping the car recognition
Check this you may find what you're looking for
ua-cam.com/video/821ZnewFLcY/v-deo.html
I'm working on a project about taking referential spatial reference to the observer using a camera.
Sí! Aunque este trabajo es un poco antiguo ya utilizaba OpenCV 1.1. Ahora tenemos cosas mucho mejores con OpenCV 2.4.6, y pronto saldrá OpenCV 3.0.0. Un saludo!
Hello, can you say how you determine angle of rotetion for the car?
@marcos I used HOG+SVM for car detection using car.xml. But It is very slow taking 4-6 minutes to detect car and it also have false detections. Pls help.
Hi Marcos, I would like to know, how do you handle the situations, when there is no lane on the road ? There are times, when vehicle moves from a road having lane to a road having no lane.
@marcos Nice work you did.
I need your some help.
did you used pinhole camera model to get camera matrix?
did you used haar cascade classifier for object detection?
Which formula you have used to calculate distance of car from camera using homography ?
hi,
I really appreciate the good result. I know it may be inconvenient for you to share the code. Could you tell some key points? How do you detect the lanes without other lines? Do you use hough or other to detect lines? How do you improve the accuracy of detection?
Best regards,
Neal
I want to share my driving video after completing my code from your github sources soon. Thanks
@Macros Is the height of object is known?
In my case object height needs to be calculate and also distance to camera.
Can you tell me any way to find object height?
Also, now I have bird eye view of Object, but how to find the height of Object in Bird eye view. I am only left with Object Height.
Please help.
Very nice work! May I ask the model of the camera?
how do you get the distance. Is there some sensor or are you using some computer vision functionality?
Hello Marcos! How do you recognize the cars? blob? SURF? And How do you measure distance to the cars? Do you use 2 cameras?
Hello!!!! Impressive work!!!!!! Did you just use hough transform or other algorithm to detect the lane markings?
How did you measure the distance of cars? Would you recommend me some resources? Thank you.
You can use flat world assumption to get linear reprojection from image plane to road plane (assuming undistortion applied) and metric calibration available. For better accuracy, tracking the vehicle in the bird's-eye view can also help fix the homography. Finally, if you have a good detector, you can also obtain 3D estimates of the vehicle from its 2D bounding box, which then you can use to estimate it's distance to the camera if the vehicle dimensions are approximately known.
Hii sir your vihical dirve him slefe?
Hi Marcus
Could you tell me which method you used to segment the cars?
Regars
very nice! did you use imaging to obtain the distance of other cars? or radar?
how do you make the reference point so you know that car is right/left ?
So basically its Mobileye but your own version?
can you tell me the method or function that you use to build this?
No lo sé, esta secuencia es Madrid, la M30 si no recuerdo mak.
se puede utilizar en robotica, m mi carrera es industrias alimentarias, podria aplicarlo en procesos industriales. te agradeceria que fueras mas especifico en el programa o la idea del codigo fuente.
Single Camera? How you have the distance?
hola, soy estudiante de ingenieria y estoy realizando un trabajo similar en opencv, me gustaria saber si me podrias enviar algun tipo de tutorial de como haces para determinar ese "porcentaje" que sale en que supongo que es el grado de desviacion de la linea amarilla marcada con respecto a que tanto se acerca a las azules dentro del cuadro verde.. necesito hacer algo parecido.. espero me puedas ayudar, de ante mano gracias,, genial el video
Hola!
Gracias por escribir. Lo mejor que puedo hacer es dirigirte a los links que he compartido anteriormente. En mi blog (marcosnietoblog.wordpress.com) puedes encontrar prácticamente todo mi material, incluido mi tesis doctoral, donde viene todo explicado.
Espero que tengas suerte!
Un saludo,
Marcos
Do you have the code？
Excelente trabajo, felicitaciones ¡¡¡, usaste OpenCV? Saludos
I want to order 1 unit of this thing please
Do a Google search for Self Driving Cars. There's an article in the NY Daily News about Audi and BMW pushing forward with this technology. Google has authorization from the state of California to operate thier autonomous vehicle as long as a person is supervising. Applications similar to this that are connected to various sensors around the vehicle extend the autonomous technology. In about 20 years we'll just punch in our destination on the GPS, and the car will do the rest.
awesome job!!! Now I got inspired, I wanna learn it and to research about these stuff!!! thanks
how does this approximate distance?
is this for public use or just testing? if its for the public can you get it for android? thanks in advance!
using openCV?
Hi! first of all, thank you for sharing your work with us. :D I am working one vehicle tracking algorithms. what you have done is great. and I would like to know what did you used or serouce code if it's public plz :)
can i follow your work？
hi, I'm working for my final year project. This is really useful. Can i get a source code?
Hello, can you build a terminator thx
How do you calculate distance with images?
Porschen Hund, The road is approximated as a plane in the local neighborhood so we can compute an homography between image and road plane. The distance is then computed assuming the cars are on the plane.
Oh, I see. That's brilliant. What is the longest accurate range when it comes to safety (not hitting a pedestrian) ?
Great work there, am doing similar thing in my fyp..
But using matlab instead..
I am now in dire need of code or at least know how to reach this point.......
@Marcos Thanks a lot.
wow excellent..would you make some lesson for this? Or please share the codes. Thanks
Se parece a la salida de Santiago de Chile...¿?
Bčne kamere in še zadnja kamara fali zraven vse štiri strani bi morale delovati ko se voziš če želiš da to pomaga v prometu.
it could be ported to Android, I am sure, but I am not focusing on that now. Actually, porting such security systems to mobile platforms would be only for entertainment, not security.
BR
كاميرا إستشعار الإشارات الطريق ومسافة المركبات الأمامية وإنحاء مسار الطريق
an awesome work , I love this
now this is really nice. thanks.
This is awesome...
Imagine if you can make a game out of this, something like if you can get pass any number of vehicle in a certain time(My weird thoughts) :)
comment ta fé
hard work for devloper
Will love to make it with my pandaboard
wonderful!
Excelente
hi Marcos, great work,can you help me do similar project, I can pay you!
Hey great work. I’m working on a similar project and need some help. Can you give me your email or something and help me please?
very nice :)
Wow!! Don't believe this!!!!!!!!!!!
amazing :)
Niiiiiiiceeee!!!!!
Dudu