Research On The Recognition And Location Of Street Lamp In The Process Of Street Lamp Cleaning

Street lamps is a lamp that provides lighting function for the road.It can improve traffic conditions,reduce driver fatigue,improve road capacity and ensure traffic safety,beautify the city appearance,and widely used in urban roads and highways.With the increase of service life,Industrial tail gas and natural pollutants will attach to the surface of street lamps,which will reduce the brightness of street lamps and affect traffic safety.Due to the high hanging height of street lamps,manual cleaning method is time-consuming and laborious,there are personnel safety problems.Therefore,it is urgent to develop a mechanized street lamps cleaning device,identify and locate street lamps at the same time,obtain the position information of street lamps,and then realize the automatic cleaning of street lamps,which is of great significance to prolong the service life of street lamps and save the maintenance cost.This paper analyzes the working condition requirements of street lamps cleaning,and straight arm cleaning scheme of street lamps is put forward.Aiming at the key problem of accurate recognition and location of street lamps,the parameter index of street lamps detection is determined,and the detection method of street lamps recognition and location based on camera and Lidar is proposed.In order to realize the problem of street lamps recognition,a street lamps recognition system is built.The data sets of street lamps under different lighting and environmental conditions are established by collecting data sets manually.According to the characteristics of street lamps,a detection method based on camera and deep learning is proposed.80% and 20% of the sample set are used as training set and test set respectively,two models,yolov2 and faster R-CNN,are used for migration learning,the comparison results show that yolov2 model has the highest accuracy,and the model recognition accuracy indexes F1,precision and recall reach 89%,96.2% and 91% respectively,which can realize the street lamps detection within the camera range and meet the detection requirements of cleaning vehicle.In order to verify the effectiveness of the proposed model,the simulation environment under different background,weather and light conditions is built on the platform of PRESCAN,and the recognition program of street lamps is built in MATLAB / Simulink.The simulation experiment shows that proposed deep learning model can effectively identify street lamps in overcast days and different background under different experimental conditions.In order to realize the accurate recognition of the angle and orientation of the street lamps relative to the cleaning vehicle,Lidar sensor is used to detect the position of the street lamps.In PRESCAN software,Lidar sensor is used to simulate the distance and orientation of street lamps relative to the vehicle,and the simulation model is used to verify the results.The results show that Lidar sensor can effectively obtain the angle and distance information of street lamps.