Research On Driving Decision Of Night Unmanned Vehicle Based On Infrared And Radar

Research on driving decisions has always been an important research link in the navigation of unmanned vehicles.In many cases,unmanned vehicles need to be driven at night,and in the dark,even at night,the infrared imager and radar are used to obtain driving directions.The information needed is crucial.Infrared images obtained by infrared imagers usually have defects such as low contrast and low signal-to-noise ratio.Therefore,obtaining reliable information and models from night-time infrared images and radar data is a key research project in the field of nocturnal machine vision,which is obtained by radar.The distance information is an important basis for the decision-making of unmanned vehicle navigation,and its research results can be applied practically in civil,industrial,and military fields.The angle and speed are the key to the driving of the unmanned vehicle at night.Therefore,the angle and distance information of the unmanned vehicle at night is studied,and the corresponding driving decisions are obtained,making it possible for the unmanned vehicle to travel smoothly in a dark environment.Significance.This paper applies deep learning technology and gives a driving decision model for unmanned vehicles at night.The thesis is divided into three parts: The first part introduces the background and research status of the driving decision of the unmanned vehicle,and elaborates the deep learning framework and its role in machine learning.The second part is the unmanned vehicle based on deep learning classification network.The directional decision model transforms directional decision research into a classification problem.In the third part,a depth estimation model is added on the basis of the previous section,and a combination of them is proposed to propose a vehicle speed decision model based on depth information.The innovations of this paper are as follows:(1)Applying the idea of deep learning classification to decision-making of the unmanned vehicle direction,optimizing the number of traditional classification network output categories,enabling the classification network to identify road direction information and giving corresponding driving advice.Secondly,the network that was originally applied to color images was applied to infrared images to make it possible to give more reasonable suggestions for night scenes.(2)Based on the direction decision model,a depth estimation network was added to better integrate the depth model and the classification model,and a speed classification decision model was obtained,which not only improved the prediction accuracy,but also obtained additional infrared images.Distance information in the scene.Comprehensive speed model and direction model,and finally get a complete driving decision model.