Research On Bridge Crack Image Recognition Technology Based On Big Data Analysis

With the continuous development of artificial intelligence and the arrival of the era of big data,the application of advanced information technology to practical engineering has achieved great success,such as the current face recognition technology,intelligent traffic monitoring system,but there is still a lot of room for technology application in civil engineering.In the aspect of bridge and tunnel,bridge crack detection is a time-consuming and laborious work,and it also has certain dangers.To solve this problem,this paper will use big data,combined with massive crack images,and use deep learning to recognize for crack images.Now research in this field is in full swing,and has achieved fruitful research results.This paper is also devoted to the research of image crack recognition based on previous research.This paper combines big data information and uses deep learning technology to identify concrete cracks.The specific research focuses are as follows:(1)Collect crack images and build data sets.In this paper,through the detection of on-site collection and laboratory acquisition,a total of more than 1500 crack images were obtained.Through image segmentation,7000 sheets were selected as sample data.After image processing technology,the crack images were pre-processed.5000 sheets were used as the training set,2000 sheets were used as the verification set,and the training iteration was 200,000 times,and the final crack model was generated.(2)In view of the diversity of crack images,this paper identifies the cracks of four different shapes in the horizontal,vertical,oblique and intersecting directions respectively.The accuracy rate is 95%,the recognition effect is good,and the crack orientation can be well identified.The recognition speed of the algorithm is tested and the recognition speed reaches 3.5 FPS.At the same time,this paper also analyzes the factors affecting the crack identification effect,and analyzes the batch size,sample size,training times and thresholds respectively,and determines the optimal recognition result under a single variable.(3)In order to better apply the model to the actual work of bridge inspection,this paper also migrates the crack model to the mobile intelligent to realize real-time identification of cracks.It is seen from the research of this paper that in view of the rapid requirements of realtime recognition,the crack image is not segmented and recognized,resulting in a real-time recognition effect.The UAV is an important helper for the heavy detection work in the future.This paper makes a simple analysis of the feasibility of real-time identification of the UAVequipped crack model.It is considered that the UAV is equipped with a simplified model for bridge crack identification.There are still many research points in this paper that have not been resolved,and there are still many shortcomings.At the end of this paper,the research direction is prospected,and I hope that the next research will make a breakthrough.