Research And Application Of Tower Hidden Danger Detection Based On Deep Learning

In the field of communication,as the basic building carrying communication equipment,the security of the communication tower plays an important role in ensuring communication and personal safety.Therefore,in the communication industry,tower inspection exists as daily work.At present,the tower inspection work still adopts manual inspection and manual verification.At the same time,the inspection process has no sound system management,and some work contents are still managed in the form of a standing book.On the one hand,this inspection method increases the labor cost,on the other hand,it is not conducive to subsequent data management and analysis.This paper studies the automatic detection of hidden dangers and the closed-loop inspection process and solves the key problems such as the small sample and small target of the detection model and the system design and implementation of the closed-loop process.The main research results of this paper are as follows:(1)The common algorithms in deep learning are studied in detail,and the SSD algorithm with balanced speed and accuracy is selected as the basic model.Take the actual patrol image as the basic data,expand the data set through image enhancement,and train the model with the mixed data set.The conclusion has practical value.Finally,the generated hidden danger detection model map reaches 0.756.(2)Aiming at the small target problem of hidden danger image of the communication tower,the selection optimization is carried out based on the SSD algorithm.Through comparative experiments ts,the FSSD model can effectively improve the accuracy and detect small targets more accurately.Compared with the SSD model,the map is improved by 0.035;Aiming at the small sample problem in the hidden danger data set of the iron tower,an image fusion method is proposed to continuously optimize the data domain.The samples generated by image fusion can achieve a threatening effect similar to the original image.Under the FSSD algorithm model,the map reaches 0.815,which is 3%higher than the original data set.(3)Aiming at the problems of decentralized process and non-closed loop system in tower inspection,based on the generated hidden danger detection model,a tower hidden danger detection and alarm system is designed and implemented.The front end is completed by bootstrap,and the back end is implemented by spring boot.MySQL and Redis are used for data storage.The system is divided into three parts:Tower image transmission and detection module,mail transmission and alarm module,and data query module.It realizes the reasonable series of the inspection process excess and solves the problem of low closed-loop inspection of tower.The model and system completed in this project are highly integrated with the actual scene,and have certain practical significance.