Research On Crack Detection Of Concrete Bridge Based On Image Processing

Bridges play an irreplaceable role in China's transportation industry and effectively safeguards China's economic development.Most modern bridges are made of reinforced concrete.Due to the factor of natural disasters,aging of building materials,overload,and man-made impact,most bridges in service now have more or less damage.As one of the most common diseases of bridges,cracks not only affect the normal use of bridges,but also induce other diseases.Therefore,it is very necessary to monitor the health of bridges regularly.Some traditional detection methods using manpower can no longer meet the detection requirements of bridge damage and defects in China.With the rapid development of computer technology and artificial intelligence,more and more attention has been paid on bridge crack detection technology based on image processing.Due to the complex geography of the bridge and the slightly rough surface of the concrete,the captured images often have non-uniform illumination,weaker crack information,and much noise.Traditional crack detection algorithms cannot solve these problems well.This paper designs a crack detection system suitable for the surface of concrete bridges.The system is consisted of energy supply module,image acquisition and processing module,furthermore,the image acquisition and processing module is the most important parts of the system.The industrial camera is used to acquire the image in real time and median filter is applied to the captured images.On the basis of background subtraction,an adaptive threshold selection method based on gray-level estimation is proposed so that complete extraction of cracks is achieved accurately.According to the image local dissimilarity feature,the gray anisotropic filtering method is used to filter out a large number of pseudo crack pixels.A multilevel filter algorithm based on morphology and connected domain is designed to remove large-area noise points and small pieces of noise.At the same time,the fracture cracks are connected to ensure the integrity of the crack.In order to effectively distinguish the remaining crack fragments and non-crack fragments such as water stains,grooves,and holes in images,a crack recognition algorithm base on SVDD is designed in this paper.The method combining gray-level difference ratio with four different shape features,such as circularity,area ratio,eccentricity,and filling degree to construct the crack identification eigenvector,is proposed.The non-crack sample database is also constructed to train the crack recognition model.Finally,using the actual images of the concrete surface for experiments,the recognition accuracy of this algorithm is 94.92%,which meets the actual needs of engineering applications,and has high practical value and application prospect.