ABSTRACT: Identifying bridge cracks and assessing bridge conditions has historically relied on labor. Bridge scrutiny by human consultants has bound disadvantages, like the failure to in person examine all sections of the bridge and sole reliance on the bridge inspector's specialist experience. Moreover, it necessitates adequate human resource coming up with, and it's not cost effective within the long-standing time. This article proposes an automatic bridge scrutiny technique that uses wavelet-based image options in conjunction with CNN to find cracks in bridge pictures mechanically. A two-stage approach is employed, with the primary stage determinative whether or not an picture are often pre-processed (based on image characteristics), and therefore the second stage extracting wavelet features from the image using a sliding window-based technique. Also on noisy and sophisticated bridge pictures, that they had been capable of achieving an overall accuracy of 92.11%.

Keywords: Convolution Neural Network, Image Processing, Machine Learning, Python

| DOI: 10.17148/IARJSET.2021.85103