Which method would be better for my purpose?

asked 2017-07-18 23:05:17 -0500

Aldur gravatar image

updated 2017-07-19 00:13:44 -0500

berak gravatar image

Hello,

Im a civil engineering student, and I will do my thesis about recognizing structural failures on concrete walls. Im searching for information about which method/algorithm would be the better for this purpose.

I spect to have 200+- images for training like this: image description.

I read about LBP, Haar-like, HOG+SVM, SURF and SIFT, but I really dont know which one would be better.

If some1 can help me Ill very very thankful.

ps: sorry for my english, im not native.

edit retag flag offensive close merge delete

Comments

i'm afraid, that none of those ideas will work reliably, since your cracks may come in any size, shape or orientation

berak gravatar imageberak ( 2017-07-19 07:39:48 -0500 )edit

Thanks for your answer berak,

You are right, these cracks/failures can come in any size or shape, but they tend to have a determinated orientation or angle depending of the reason of the failure (shear, flexure, etc), and obviously the crack comes always in gray/grey.

Here is an ilustration of how the shear failures comes, and a picture of a real shear failure.

http://ijripublishers.com/IJRI-Publishers/pdf/CIVIL/VOLUME%201/IJRI-CCE-01-003/3.png

http://alert.air-worldwide.com/Alert/alertdata/Protect/Data/Chile__Earthquake/L3/Fig9_Shear.jpg

As you can see the shear failure tend to come in a 45° angle

Aldur gravatar imageAldur ( 2017-07-19 19:29:46 -0500 )edit