Research On Traffic Scene Element Detection Based On Deep Learning

If you want to make the application of self-driving vehicles truly,it is mainly to solve the three problems of perception,decision-making and control,and the perception of the environment is an important basis for achieving other aspects.Various sensors installed on the car can sense the surrounding environment at any time during the driving process of the car,collect data,perform identification,monitoring and tracking of static and dynamic objects,and combine the map data in the navigator to perform system calculations and Analysis,so that drivers can be aware of potential hazards in advance,in order to increase the comfort and safety of car driving.Therefore,in terms of sensor perception,a method for driving assistance combining multiple sensors is constructed.If the vehicle "sees" the condition of the vehicle ahead at all times,traffic accidents will be reduced accordingly,and road safety will be guaranteed.Modern cars are equipped with various advanced visual driving assistance systems,which provide the most basic visual capabilities for vehicles.Among the car cameras,the forward-looking camera is the most frequently used camera,and its importance cannot be underestimated.Therefore,in the visual elements of self-driving vehicles,this paper uses the Paddle Paddle deep learning development platform to do vehicle target detection based on YOLOV3 on the UA-DETRAC data set,at the same time,the lane line recognition is performed,and finally the m AP index is used for evaluation.Vehicles not only have to "see",but they also have to "listen" like human drivers.With the improvement of computer performance,audio processing technology has also made great progress.In recent years,deep learning technology has been used in the research of audio signal separation,and has become an increasingly popular topic in the field of audio signal processing.Promote the development of sound source separation based on deep learning technology.In this paper,the mixed speech audio composed of human speech and background music in the simulated vehicle is selected as the research object,and the audio signal analysis and neural network algorithm are used for analysis and research.Analyze the time-domain and frequency-domain features of various audio,and select the frequency-domain information of audio as the features to preprocess and extract the samples.A sound source separation model based on deep neural network was implemented,and the Spleeter sound source separation framework was tested.In summary,this paper has done some research on the detection of traffic elements based on various sensors.Among them,visual information can provide road and direction guidance for driving,and audio information can be used to perceive human activities to establish a friendly interior atmos phere and collect emergency.Case handling measures to provide services.In order to build a high-quality driving environment,both are indispensable.