Research On Key Technologies Of Lamp Language Intention Recognition In Mixed Traffic Flow Of Connected Automated Vehicles

In recent years,the development of artificial intelligence technology is in full swing.Connected and Automated Vehicles(CAV)have emerged.However,there are a considerable number of manual vehicles(MV)in the traffic flow at this stage.CAV and MV formed the mixed traffic flow gradually.The interactive problem of driving intention between CAV and MV in mixed traffic flow environment has become a research hotspot of domestic and foreign scholars in the field of Intelligent Transport System(ITS).However,how CAV can quickly and accurately recognize the intention of the MV sent by the MV is a difficult point in ITS research.At present,researchers lack a systematic method for identifying the intention of the MV lamp language.The recognition of the lamp language is an important basis for CAV automatic driving decision-making.Recognizing the intent of the lamp language of the surrounding MV can enable the CAV automatic driving system to correctly select the driving path and correctly control the vehicle operating state.Therefore,for the light information in the mixed traffic flow,a CAV lamp language intention recognition model is established.The main research work of the thesis is as follows:(1)Aiming at the detection requirements of CAV for vehicle lights in a mixed traffic environment.Firstly,the image de-halation algorithm based on the dark channel prior principle,the image segmentation algorithm based on ABT clustering and HSV space,and the lamp pair matching algorithm based on edge shape features preprocess the car light images obtained by CAV.Secondly,the feature extraction is based on the Histogram of Gradients(HOG),the car lights are detected based on the Convolutional Neural Network(CNN),and the corner tracking algorithm(Kanade-Lucas-Tomasi,KLT)is used to perform the feature extraction following the MV with the lamp language.Finally,the lamp detection and lamp pair tracking experiments show that the average detection rate of the lamp is 96.4%,and the angle error of the tracking and positioning is less than 1°,which basically meets the CAV’s detection and recognition requirements for MV lamp.(2)Aiming at the requirements of CAV for LED light data processing in a mixed traffic environment.First,the light radiation flux algorithm is established based on the vehicle light projection model and the road reflection model.Secondly,considering the influence of various light noises in the mixed traffic flow,a multi-input and multi-output model of LED car lights was established.Finally,the lamp language luminous flux detection and the lamp language error rate experiment show that the closer the MV and CAV are,the better the calculation effect of the algorithm on the luminous flux of the lamp language,and due to the increase of the distance between the vehicles,the LED lamp The higher the bit error rate,the more optical noise received.In the five typical lamp language scenarios,the characteristics of the light radiant flux received by CAV are obvious,which conforms to the lamp language rules in different scenarios,and reflects the high feasibility and reliability of the algorithm.(3)Aiming at meeting the needs of CAV lamp language intention recognition in mixed traffic environment.First,a classification algorithm based on improved SVM is proposed to classify the recognized LED lights.Secondly,a lamp language intention recognition model is established based on the double-layer Hidden Markov Model(HMM),which is divided into the driving behavior layer HMM model and the lamp language intention layer HMM.Finally,five typical lamp language experiments are used to verify the feasibility of the identification of prompting the intent of the driving lamp language,the intent of changing lanes to overtake the lamp language,the intent of accelerating and passing the lamp language,the intent of keeping the distance between the lights,and the intent of prompting the turn signal,and the results are realized.It shows that the average recognition rate reaches 94.8%,which verifies the accuracy and applicability of lamp language intention recognition.In summary,this article proposes a trinity methodology of "vehicle lamp detection-light data processing-lamp language intent recognition" for CAV’s recognition of MV lamp language intentions,which makes up for the communication defects between CAV and MV in mixed traffic flow.V2V communication complements each other and further expands the vehicle-to-vehicle communication technology in the mixed traffic environment.