Identification And Measurement Of Bridge Cracks Based On Image Processing

Due to the effects of loads,temperature variation and other environmental factors,it is easy for bridges to produce cracks during the long-term use,which will cause security risks.Therefore,it is of great significance to carry out fast and efficient inspection as well as targeted maintenance and repair work in order to ensure the normal use of bridges.At present,the construction of bridges in China is changing from "mainly new construction" to " build and maintain simultaneously ".However,the manual detection method,which is used currently,has low inspection efficiency and large resource consumption,and also has difficulties to form inspection data systematically,making it difficult to guarantee the effectiveness of subsequent maintenance measures.Therefore,it is imminent to conduct research on the intelligent detection technologies for bridge cracks.Based on the summary of previous research on crack image acquisition and recognition,an image processing-based assessment scheme for bridge cracks is proposed,including image acquisition based on image acquisition systems,identification of crack images based on convolutional neural network technology and extraction of crack parameters based on connected domain analysis technology.First,an image acquisition system was designed independently and a system prototype was used to complete the image acquisition experiment.Then,a convolutional neural network model was established.After model training and parameter optimization,the crack recognition experiments were carried out on the collected bridge images.Finally,the panoramic image of cracks was generated using the collected images,and the three main parameters of crack length,average width and maximum width were collected using connected domain analysis technology.The results show that the proposed scheme can realize the bridge image acquisition,crack identification and parameter extraction.The accuracy rate of crack identification is over 98%,and the error rate of extracted parameters is less than 10%,which can provide the basis for bridge inspection and maintenance.