Bridge crack detection using CNN

Abstract

To day of the week, identifying cracks in bridges and responsible bridge conditions primarily involve
manual labour. Bridge inspection by anthropological experts has some drawbacks such in place of the inability to
physically examine all parts of the bridge, sole dependency on the expert knowledge of the bridge superintendent.
Additionally it requires proper training of the human resource and complete the aforementioned is not cost effective. This
article proposes an automatic bridge examination attitude manipulating wavelet-based image features lengthwise by
means of CNN for automatic detection of cracks in conduit images. A two-stage method is followed, someplace in the
principal stage a decision is finished as whether a doppelgänger should go through a pre-processing step (depending on
image characteristics), and glowing along in the second stage, wavelet features are pull out on or after the image using a
upward window-based procedure.