The Research And Implementation Of Detecting Techniques Of Driver’s Illegal Behaviors Based On Road Monitoring Images

In recent years,our country has entered the fast lane of development and the standard of living of the people has been increasing.At the same time,it also drove the rapid development of the automobile industry in China.Cars have become an indispensable means of transportation for us to travel.However,it brings conveniences and frequent traffic accidents.Traffic safety problems have become increasingly prominent.Of course,there are many factors that cause traffic accidents.The simple subjective behavior of the driver to answer the phone while driving is one of the important factors causing traffic accidents.At present,there are few people who research and detect driver’s violations and lack of technical means.It is impossible to identify violations that occur within the vehicle.In order to make the traffic control system more effective for supervision and to ensure the safety of people’s lives and property,government departments at all levels are building large-scale intelligent transportation systems.The system currently has a lot of research on red light,retrograde,and not wearing safety belts.However,the detection of illegal phone calls during the driving process is not mature.This article mainly studies the behavior of the driver based on the road-surveillance image detection system during the driving process and provides an effective solution.The article focuses on the following aspects of research,the first image pre-processing.Retinex algorithm and histogram equalization(CLAHE)method are combined to improve the image preprocessing effect for the problem of uneven light and low amount of original image information.After the experimental test,it is verified that the algorithm can enhance the picture quality problems caused by illumination,which lays a good foundation for the road monitoring of face detection and the recognition rate of violations.Secondly,the behavior of the image is identified.The traditional Adaboost algorithm adds weight to the wrong classification samples during the training process,which leads to the increase of computational complexity and prolongation of the algorithm.The algorithm is improved to make it effective in the actual scene..The main improvement of the algorithm is to define a threshold HWt in the training process to determine the update weight of the new round of the cycle,and finally obtain the detection result.At the same time,after the head-circle border is expanded,the color space Ycbcr with better clustering is used to perform skin color separation..Experimental test results show that the skin color segmentation results are more accurate and lay a certain foundation for the detection of late behavior.Finally,it is judged whether there is any violation.After the skin color segmentation result is completed,a combination of HOG feature extraction and SVM classifier is used to determine whether the driver has irregular behavior.After researching and improving the above algorithms,the paper proposes a technical program to detect the driver’s illegal behavior recognition on the telephone and on the phone.The data sets provided by Xinjiang Uyghur Autonomous Region Traffic Police Corps are tested on the Matlab2015b platform.The experimental test and result analysis verify the effectiveness and practicability of the improved algorithm.