Smart Mobile Phone Based Vehicle Detection And Distance Detection

In recent years,with the increase of car ownership,traffic accidents have become a topic of great importance.ADAS(Advanced driver assistance system),especially the vehicle collision warning system and lane keeping assistance system are becoming more and more important.Preceding vehicle detection and vehicle distance measurement are the key modules of vehicle collision warning system.In the preceding vehicle detection system,the common sensors are radar(millimeter wave),infrared laser radar,camera and so on,different kinds of sensors are applicable to different scenes,in order to adapt to a variety of scenes,manufacturers generally collect information using a variety of sensors.Due to the low cost and the rapid development of computer vision in recent years,vision based vehicle detection system has attract more and more attention.Computer vision based vehicle detection system is mainly based on the surface characteristics of vehicle,such as symmetry,color,texture,shadow,geometric feature,lights,followed by template matching or image detection method of machine learning.Vision based vehicle distance measurement methods mainly have two major categories which are binocular vision ranging method and monocular vision ranging method.In view of the high cost of binocular vision and high complexity,the monocular vision ranging method is still the mainstream.The current ADAS manufacturers needs customized hardware equipment to achieve vehicle detection,the cost is very high,so the popularity is low.In this paper,a vehicle detection and distance measurement system based on mobile phone platform has been implement,including Android and IOS two mainstream platform,greatly reduces the cost of the vehicle driving assistance system,improved the vehicle detection method so that it can adapt to multiple environments,and reduced the false positive rate effectively,so it is expected to gain popularity.In order to meet the property of real time and accuracy,the system uses two steps preceding vehicle detection technique which are finding the vehicle hypothesis area and verifying the vehicle hypothesis area.By using two steps threshold extracting method based on lane prior to find the shadows,By using lane prior to select the shadows to find the hypothetical vehicle region,and using Adaboost method based on haar-like features to verify vehicle hypothesis.After that the video stream characteristic was used to remove the false positive detection and to track the vehicles.This paper also combines the camera perspective geometry relationship to measurement the distance of front vehicles and the users just need to provide the height of the camera.It can be used as an effective means of risk warning.The characteristic of real time is also an important demand of driving assistance system,while keeping high accuracy,by considering the characteristics of intelligent mobile phone,the system is optimized to ensure the characteristic of real time.The experimental results show that the system has high accuracy and low false alarm rate,and can perform well in a variety of environment,like city road,high-speed road in both sunny and rainy day.Therefore it can be effectively applied to vehicle detection and distance measurement module in ADAS.