Abstract：The study set up a platform for gathering CAN data from the vehicles independently. The system proposed 3 accessing index for drivers’ comprehensive quality based on driving safety mainly, supplemented by riding comfortably and fuel consumption of vehicles, which reflects the driving safety to a large extent.The model analyze these index from 7 factors the entropy of steering wheel angle, the entropy of vehicle speed,steering angle rate and so on. Firstly, the analytic hierarchy process(AHP) was used to establish a single factor weight vector. Then the single factor fuzzy judgment matrix was calculated by membership function. Finally, we got the result of the driver′s comprehensive quality using maximum membership. The practice show that this model provides a simple and accurate data analysis method for driving safety.

Abstract：Lane level traffic violation refers to the behavior of the moving target violates the traffic laws whose relative position moves within one lane level relative to the ground lane or other moving targets. Due to the inadequate data accuracy, this kind of behavior has always been a difficulty for traffic management law enforcement and evidence collection. The realization of high precision BDS/GPS positioning relies on wide area real-time precision positioning technology. The satellite positioning accuracy has improved from an average of about 4-20 meters to a public level of sub-meter through the national/global BDS ground-based augmentation system. From the point of view of high precision positioning technology, we have designed a terminal with lane level positioning ability for law enforcement and evidence collection in this paper, and we have carried out technical validation in Wuxi and other places. The experiments showed that these technologies could provide support for the IOV system such as vehicle behavior analysis and vehicle active safety.

Abstract：Real-time lane detection and tracing is achieved on TMS320DM6437 platform. Lane detection is achieved with the method of two-stage Hough transform, and the correlation of frams is used to trace lane. After the algorithm is transplanted, the algorithm will be optimized further based on the character of DSP. Experimental results show that the system achieved real-time lane detection and tracing and it has some noise immunity.

Abstract：Aiming at the simplification of the driver′s state detection, feature extraction, and the high cost of detecting equipment, the implementation of driver′s state detection based on multiple feature fusion isproposed. The system adopts STM32L4 DSP chip as the core control. Firstly, characteristics such as pulse, acceleration, angular velocity, attitude angle and human body temperature are acquired and processed by SON1303, MPU6050. Secondly, for pulse, DSP library is called to implement FFT and Chebyshev filter is used to extract the frequency spectrum. Finally, a method via analyzing the frequency spectrum of the driver′s state, tiredness, distraction and tension , defining the first main peak B and spectral ratio K, and fusing these characteristics as B, K, attitude angle, acceleration and angular velocity can judge driver′s states effectively. Experiments show that the system has the characteristics of strong anti-interference, low cost, etc. It can be widely used in the state detecting of divers, and it can promote the popularization of ADAS.

Abstract：A robust lane detection algorithm is presented for ADAS applications. Utilizing a distance based image multiscale resampling method, horizontal dark-bright-dark feature, parabola lane model and Kalman filter based tracking, the algorithm extracts and tracks road lane in realtime. A fast pitch angle alignment method is put forward for realtime detection on Android mobile devices. Experiments show that the algorithm runs realtime on an Android car dashcam.

Abstract：Advanced Driving Assistant System(ADAS) are important to increase driving safety. Nowadays, the Open Computing Language(OpenCL) framework and mobile devices integrated with Graphics Processing Unit(GPU) has made it possible to launch ADAS applications on mobile devices efficiently. In this paper, a typical ADAS lane detection application based on OpenCL framework is implemented and launched on three mobile devices. By evaluating the accuracy, processing frame speed and energy efficiency of this application, it's concluded that the mobile devices are suitable for ADAS applications.