Image Processing Method Based On Depth Learning In Intelligent Road Lamp System Of Vehicle Network

With the continuous progress of automobile industry and Internet of Things Technology at home and abroad,intelligentized traffic management technology is a breakthrough point of urban intelligent management.Any intelligentized application management is based on information collection and processing.Intelligent image processing technology is a research focus under the background of AI.In this master thesis,street lamp is used as an important part of the Internet of Vehicles.Based on the intelligent street lamp system under the background of vehicle networking,we explore the deep learning based image processing technology.The main contents are as follows:First of all,before studying and analysis of the specific algorithm application,we first introduce the basic concepts and tasks of image processing briefly,and understand the common problems and typical operators in the field of image processing.In order to achieve intelligent image processing,we introduce the concept of deep learning,introduce the basic concepts of deep learning and neural network,and the basic concept of convolution neural network applied in image processing area.Secondly,we introduce the CBCL StreetScenes data set which is suitable for road traffic applications under the background of the Internet of Vehicles.On the basis of this,the basic ideas and training steps of the classic Fast-RCNN algorithm and YOLO algorithm based on depth learning are further explored.Then the CBCL StreetScenes data set is used to train it to verify its detection effect on the common targets in the road traffic.Then,aiming at the shortcomings of the classic algorithm in the background of vehicle networking,we propose an improved CBCL StreetScenes data set and an improved YOLO algorithm model.The original CBCL StreetScenes data set is tailored,and the detection target is divided into two categories: dynamic and static.Aiming at the more important dynamic class targets in detection algorithm,an improved YOLO algorithm is further proposed.The detection results of the improved CBCL StreetScenes data set and the improved YOLO algorithm are verified.For the difference between them and the classical algorithms,the detection results are visualized and visualized by drawing the method of confusion matrix and ROC curve.Finally,the feasibility of the algorithm in the intelligent street lamp system under the background of vehicle networking is discussed,and two reliable system models which can be applied to the actual scene are put forward,which are traffic lights intelligent control system and vehicle fire accident monitoring system.The specific function design and detailed module design are given to judge the application of the algorithm in the actual scene.