Research On Concrete Crack Identification In Complex Environment Based On Improved SSD Algorithm

In recent years,the development of artificial intelligence has made more and more disciplines which seek to expand their research scope in the field of artificial intelligence.There are also more and more researchers who apply artificial intelligence to actual social situations,such as face recognition,unmanned driving and intelligent transportation,and there are still many problems that need to be overcome in the field of civil engineering.In the detection of highway diseases such as bridges,most of the previous methods of manual detection were used,which was time-consuming and labor-intensive,and the safety of inspectors could not be guaranteed.Therefore,this paper makes corresponding contributions to the final realization of unmanned intelligent detection: improve the identification accuracy and noise resistance of unmanned crack detection,so that the detection algorithm can adapt to any complex environment.In this paper,a large number of crack pictures are used to identify concrete cracks based on neural network knowledge.The specific research focuses are as follows:(1)The SSD algorithm is used to train and identify the crack dataset,and the recognition accuracy rate reaches more than 96%.Firstly,a large number of cracks were collected in the actual scene and laboratory,and the crack pictures were pre-processed by segmentation,amplification and filtering,and a total of 8,000 sample data were formed,of which 5,000 were used as training sets and 3,000 were used as verification sets.It is then manually annotated and fed into the algorithm model for continuous training and parameter adjustment.The final shape is formed into an initial crack recognition algorithm model,and the accuracy of the image recognition of noise-free cracks reaches more than 96%.(2)The crack image is noise-canceled to simulate the crack in a complex environment,and an improved SSD algorithm based on the pruning prior frame is proposed according to the characteristics of the SSD algorithm network,and the accuracy of crack image recognition of 20% of the noise is more than 90%.First of all,in order to simulate the crack pictures in various complex environments,the collected crack pictures are noise-canceled,and the pepper salt algorithm is selected to be more suitable for the crack pictures in the actual complex environments.The level of noise reinforcement is also divided into five levels: 0%,5%,10%,15%,20%,and then use the initial algorithm to identify these crack pictures after noise,but the recognition accuracy rate is not high.Therefore,a pruning improved SSD algorithm based on a priori candidate box is proposed,and the number of transcendental candidate boxes of each resolution and shape is first counted,and the group of a priori candidate boxes with the highest accuracy is studied by variable analysis,forming an improved crack identification algorithm model.Finally,the noise cracks of each level are identified,and finally the accuracy of the crack picture recognition of 20% of the noise is more than90%,and the algorithm before the improvement is compared and analyzed,indicating that the improved algorithm has good anti-noise ability.(3)The improved algorithm is applied to the actual test for identification,and the feasibility analysis of intelligent disease detection is carried out.Firstly,crack images were collected for the test of studying the bending resistance of printed concrete slabs.Then the image is processed in three different ways,and it is transmitted to the server terminal for recognition,which achieves good results,and finally the feasibility analysis of unmanned intelligent disease detection based on the algorithm is carried out.It is proved that the improved algorithm can be used for crack identification in some complex environments.Based on the above work,this paper conducts an in-depth study on the final realization of unmanned intelligent detection,summarizes and analyzes the research content and results,and lays the foundation for the application in this field and also looks forward to it accordingly.