Bridge Inspection Using Image Enhancement and Machine Learning

Bridges are the essential part of our transportation in daily life. Infrastructure safety and integrity are crucial to prevent accidents and ensure smooth traffic flow. Regular inspections require careful observation and analysis of various components, including decks, beams, piers, abutments, and cables. It can consume a lot of time, be costly, be prone to errors, and not always provide accurate and reliable results. Machine Learning techniques with drone technology is going to address efficient results. In this work, image processing and Machine Learning are combined for better performance. Brightness based image enhancement technique is used in this paper. Decision trees and Support Vector Machine are explored to identify various bridge defects. Further, crack length is estimated using image processing. Initially, edges are highlighted in the image. Later, these edges are joined using morphological operations. Eight connected neighborhood operations are applied after morphological operations. Finally, length of the crack is computed by counting the number of ones in the mask. Image processing using the most advanced algorithms can automate and enhance the images of bridges. It is expected that the proposed techniques would play a key role in identifying potential issues quickly and accurately, reduce inspection time and costs, and improve the safety and longevity of bridges. Performance of the algorithms is assessed using accuracy, precision, and recall.