Ask Your Question

How to extract specific feature in pattern recognition algorithm in image processing for object detection?

asked 2016-07-13 01:31:08 -0500

astronaut gravatar image

would like to detect the object (airplane door) from short distance. The algorithm should be very robust so can be implemented by any plane (with lots of different paintings, logos ) and by any weather condition(sun, rain, day and night).

I search in OpenCV and implemented some of them feature extracting algorithms such as SURF , SIFT but the results is not so good.

I would like to improve the algorithm’s robustness with respect to the correct matching of features. So, more reliable feature matching for the purpose of object recognition and detection to be more robust and reliable. Like for example, can include the distance between the window and door frame (which in every airplane model is fix), then the thickness of the door frame etc.

I would like to extract some customs features so that the algorithm work for any planes with any paintings and logos. Means the algorithm should be robust to detect the door by any type of plane.Features like Logos of some airlines and paintings should not be keypoints/features.

So because of that I like to extract features that can be general like distance between the window and the door frame(as this feature is always the same for given airplane model). Like for example the minimal distance between the door frame and the nearest window in Airbus A350 is let we say 1m. So I would like to use this feature in my algorithm. Any advice how to extract such features?

Should I use in this case pattern recognition and machine learning techniques such and Deep Neural Networks or KNN?

edit retag flag offensive close merge delete

1 answer

Sort by » oldest newest most voted

answered 2016-07-13 08:34:22 -0500

MRDaniel gravatar image

updated 2016-07-13 08:37:00 -0500


I saw the other post.

You should edit your question and link it so that people can see the images associated with this question.

My personal opinion is that you are trying to be too bold. You want to detect the door in one step. Sure, this would be simple if the door was the only object in the scene, but as you've experienced, applied computer vision is never so clean and simple.

You have to do two things. Locate the parts of the image that you are interested in, and remove parts of the image that you are not interested in.

Instead of a one step detector, you may want to detect the fuselage, or the cockput window to begin with. This would give you a rough estimate as to where the door should be located, and where there may be artwork or details.

Then feature analysis or a cascade classifier may be useful at this point to narrow down and postiviely detect the door.

edit flag offensive delete link more


Also, to answer your question, deep learning could probably do this task but it is complete overkill in terms of preparation and output.

MRDaniel gravatar imageMRDaniel ( 2016-07-13 08:35:15 -0500 )edit

Remove the features found in common airline logos. i.e. have a database of logos a reject any features that score highly when matched against those reference features.

MRDaniel gravatar imageMRDaniel ( 2016-07-13 08:36:06 -0500 )edit

The problem in that case is that some airlines like one from NEW Zealand have painted the whole airplane in black. And I have to do image processing from short distance like 1-2 m from the door. So now much can see like cockpit windows. Other important issue is that algorithm should work by heavy rain and strong sun. So how should I proceed in these case?

astronaut gravatar imageastronaut ( 2016-07-14 01:27:51 -0500 )edit

So I was thinking to have an approach where first remove parts of the image that you Im not interested in.I this case can be rain or sun reflection. So for this I was thinking to use Background Subtraction Methods. Any other advice. Then , second part of the algorithm is detect and locate part of the image that Im interest like part of the door frame and the corners edge of the door, or door thickness. Third par would be feature analysis techniques and machine learning algorithm or cascade classifier can be applied. So its a three stage process algorithm and approach. Any suggestions or opinion?

astronaut gravatar imageastronaut ( 2016-07-14 02:05:07 -0500 )edit
Login/Signup to Answer

Question Tools

1 follower


Asked: 2016-07-13 01:31:08 -0500

Seen: 753 times

Last updated: Jul 13 '16