How can i detect car using tail light color and shape of the car?
2 visualizaciones (últimos 30 días)
Mostrar comentarios más antiguos
Nahid Islam
el 30 de Oct. de 2014
Respondida: Muhammad Talib
el 27 de Feb. de 2016
I m working on car detection from a given image. I m thinking about using tail light color and shape of the rear side of car as my feature. I used color segmentation and edge detection. now i m stuck with classification. I can't use SVM as my supervisor won't allow SVM. What can i do? I attached my color segmentation and edge detection code. Also an image of car.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/145901/image.jpeg)
0 comentarios
Respuesta aceptada
Dima Lisin
el 31 de Oct. de 2014
Out of curiosity, is there any reason your supervisor does not let you use SVM?
You can try using the trainCascadeObjectDetector function in the Computer Vision System Toolbox to train a boosted cascade detector. It is not an SVM, and it does not use color information at all. Since you only need to detect the rear side of cars, it may be feasible.
3 comentarios
Image Analyst
el 1 de Nov. de 2014
I don't know of anyway to figure out what release a feature was first introduced other than manually going through the release notes looking for it. See http://www.mathworks.com/help/relnotes/index.html
Dima Lisin
el 5 de Nov. de 2014
Yes, trainCascadeObjectDetector is available in 2013a. There is also a GUI tool for labeling ground truth on the File Exchange. If you get a more recent version, then you can use the Training Image Labeler app.
Más respuestas (2)
Image Analyst
el 30 de Oct. de 2014
That's a very challenging problem. You should look up CBIR since the tools used in that will help you. Warning though - it's not trivial. You'll use things like SURF and SIFT and HOG and lots of other things.
As far as color goes, that will probably be a fruitless attempt. There are so many ways the color in your digital image could change and the change from one lighting condition to the next will most likely be way more than the color change between two different cars under the same lighting. In other words, if your mean RGB of a tail light is [160, 60, 70], there's no way to know what model of car it is from just that color. It could be a Toyota in daylight, or a BMW at dusk, or a Ford in the morning.
Muhammad Talib
el 27 de Feb. de 2016
further i am also working on car make and model recognition system. for that i used these feature as well out of others ....
0 comentarios
Ver también
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!