Design And Implementation Of Bridge Surface Defect Analysis System

Bridges play an important role in the transportation system.At present,most of the bridge surface defect detection in China still relies on manual detection.The conditions are difficult and inefficient.The visual dead angle of the naked eye recognition will lead to large detection errors,and regular inspection will consume.A lot of manpower,material resources and financial resources also affect traffic efficiency.The artificial surface-based bridge surface defect analysis system aims to solve the above problems by replacing human resources with machines.The artificial surface-based bridge surface defect analysis system uses image processing,convolutional neural network(CNN)and other related technologies to manage video files or real-time video streams of drones,identify corresponding cracks,and pass training.A good crack classifier to determine the grade of the crack.The main work of this thesis includes: firstly collected lots of crack pictures,construct the crack image data set;secondly,extract the feature data needed by each level through training,and construct a strong classifier based on the weak classifier cascade;then,construct and train CNN To achieve the classification of cracks;finally,based on the expertise of bridge crack classification,a part of the bridge crack image was selected for testing.The system not only realizes the general identification of bridge cracks,but also proposes that the bridge cracks are divided into three grades,corresponding to the three crack standards in actual production.Finally,a bridge crack analysis model based on artificial intelligence is established,and a convolutional neural network algorithm is used to enhance the feature expression of cracks.In this paper,the bridge crack images taken in Kunming,Hangzhou,Suzhou and other places are used as source data,and the AdaBoost algorithm is used to implement and train the classifier,and the classification model of convolutional neural network is constructed.And using a part of the data for testing,according to the industry recommendations of the bridge inspection department,the accuracy rate should be at least 92%.,which can prove the effectiveness and superiority of the system mentioned in this paper.