Design And Development Of Safe Driving Assistance System Based On Neural Network

In order to ensure driving safety and reduce traffic accidents,a fast,reliable and high-performance safe driving assistance system based on neural network is developed in this paper.It can detect the state of drivers in real time and give timely warning to unsafe driving behaviors.Unsafe driving behaviors detected by this system include fatigue driving and smoking behavior.System functions are mainly divided into face detection and ROI extraction module,target classification module,unsafe driving behavior recognition and early warning module.The main research contents of this paper are as follows:(1)Face detection and ROI extraction moduleThe environmental requirements for driver behavior detection are analyzed,and the multi-task cascade convolutional neural network MTCNN framework is used as the theoretical basis for face detection and key point positioning in this paper.The superiority of the model is proved by experiments with the parameter adjustment of ONet subnetwork.Combined with the relative position of the facial features of "three chambers,five eyes and two points",the ROI extraction of the driver’s eyes and mouth is realized.(2)Target classification moduleIn this paper,EM-Caps Net model based on capsule neural network was established to classify drivers’ eye and mouth conditions and smoking behaviors.In order to enhance the capability of feature extraction and optimize the real-time performance of the model,the CAPSNET network structure and dynamic routing algorithm were improved.Through experiments,it is proved that the multi-state classification model designed in this paper is reliable.(3)Unsafe driving behavior recognition and warning modulePerclos criterion,blink frequency and POM index are combined as the criterion to determine the fatigue state of drivers in this paper.The smoking behavior of the driver is judged by analyzing the cigarette in the mouth area.In this experiment,Python language and Tensor Flow framework are used to complete the development of the safe driving assistance system.According to the unsafe driving behaviors of drivers,the system makes corresponding early warning.The validity and practicability of the system are proved by the test.