| With the rapid development of economy and the increasing of car ownership,traffic safety problems become severe increasingly.According to relevant statistics,the driver illegal behavior is the main cause of traffic accidents.Therefore,the traditional transportation system has been unable to meet the needs of traffic management.There is an urgent need for an efficient intelligent traffic system to recognition the driver illegal behavior and avoid the traffic accidents.With the advantages of non-contact,real-time and high-efficiency,driver illegal behavior recognition system based on computer vision has been used widely.This paper mainly researched the recognition algorithm of driver illegal behavior in intelligent transportation system and designed a detection system of driver illegal behavior which realized windows detection,seatbelt detection,and phone use recognition.In this paper,the research mainly focused on the following aspects:(1)On the basis of the existing technology,this paper combined the integral projection and MBI feature to improve the algorithm of window detection with accurate detection.(2)Researching on seatbelt recognition,this paper firstly used Hough transform to detect the straight lines on the image of driver,detected the seatbelt according to the decision condition secondly.Based on HOG feature and SVM,the new algorithm was redesigned and got better performance.(3)Researching on phone use recognition,this paper firstly designed a CNN to detect the face base on the structure of AlexNet.Secondly,the hand was located by the face area and a CNN which can detect the hand using phone was built.But considering the driver’s special action,this paper use the driver area to train a new CNN.Finally,three CNNs were used to realize the phone use recognition with high accuracy.(4)This paper designed a set of ITS which include human-computer interaction interface and module of behavior detection.Among the system,the main functions of UI include image and parameter display,detection result display and system control.the main functions of behavior detection module include windows detection,seatbelt detection,and phone use recognition. |