Road Damage Detection Based On Intelligent Inspection Terminal

With the development of deep learning theory and the performance for computers and mobile devices,deep learning has brought profound changes to the research fields of many disciplines.In the field of civil engineering,deep learning has brought tremendous impetus to many directions.Deep learning has become one of the most important research topics for the realization of intelligent civil engineering in the future.Roads are an important part of infrastructure.Road damage detection plays an important role in maintaining traffic safety.The current mainstream road damage detection method is manual inspection.This method requires a high labor cost,and the damage cannot be discovered when the damage occurs.In order to overcome these shortcomings,this paper proposes a road damage detection method based on intelligent inspection terminal.(1)A road damage data set based on deep learning theory has been built,the training and evaluating of deep learning models has been completed.This article listed four common types of road damage,which are long strip cracks,reticulation cracks,road potholes and road rutting.(2)Using the Raspberry Pi,camera module,acceleration sensor,GPS module,Intel neural network computing stick and other hardware to develop intelligent inspection terminal.The intelligent inspection terminal has a built-in wireless transmission module,which can transmit the collected data to the server in real time to realize data cloud storage and analysis.Building a deep learning environment on the intelligent inspection terminal and completed the transplantation of the deep learning model,so as to realize automatic inspection.(3)Developed a mobile APP to realize road damage detection based on smartphone,the APP has a built-in deep learning model.The APP can meet the detection needs of engineers in various scenarios.After the actual engineering inspection experiment,the road damage detection method based on the intelligent inspection terminal proposed in this paper can accurately identify and classify the road damage.The object detection process has extremely low latency.The test results show that for different types of damage,the average accuracy(AP)level of detection is 80.87%,which proves the effectiveness of the method.