Design And Implementation Of Concrete Surface Damage Identification System Based On The CNN

Affected by material,construction,maintenance and environment,the surface damage of concrete structure may occur in terms of direction,width,depth and degree.If this kind of damage can not be controlled,the effective load of the structure will be reduced and even cause the structure failure.As such,concrete surface damage is one of the important monitoring targets of concrete structure health.Often the manual detection method is accompanied by many problems such as poor universality,low efficiency and low recognition accuracy in massive surface image processing.Firstly,this paper aims to summarize the relevant researches at home and abroad,which concludes the existing results and analyzes the improvement.And then,the following works are finished: First of all,this paper analyzes the requirements of surface damage identification system from the perspective of engineering application,and designs a reasonable classification system according to the causes and characteristics of surface damage of concrete structures.Secondly,the target image data is processed by graying,piecewise linear transformation,image denoising and binarization,which effectively reduces the noise in the image and decrease the feature dimension.Thirdly,in the process of system design and implementation,this paper adopts the model training method combining transfer learning and parameter adjustment to train the target model,and obtains a model with shorter training time and higher recognizing accuracy through experiments,and verifies the generalization ability of the model based on the damage data of the concrete bridge and its attached structure in service provided by the construction safety and quality agency.Finally,this paper designs the logical structure,functional modules and system database of the surface damage detection system,and realizes the Web detection system for traffic engineering monitoring that can be accessed by mobile devices based on the ResNet-50 model with parameter learning has been completed.Besides,the functional test and performance test of the system are also completed.Based on the significant advantages of Convolutional Neural Network(CNN)in the field of image recognition,surface damage identification of concrete structures is carried out in this paper.CNN,one of the excellent schemes to copy with surface damage identification of concrete structures.Compared with the traditional identification method,this way can provide high-precision and efficient damage target identification,and can effectively reduce the influence of various complex interference information in image data in practical application on classification results.This paper also realizes the practical application of CNN in the field of traffic engineering detection.The system can directly access the shooting data and complete image recognition on the Unmanned Aerial Vehicle(UAV)flight control platform,which improves the work efficiency to a greater extent.The functions of project creation,identification result management and data statistical analysis in the system meet the actual requirements,which also simplifies the work of engineers.The solutions mentioned in this paper can be widely used in the identification,analysis and monitoring of surface damage of concrete structures such as roads,bridges and tunnel linings in traffic engineering.