A Design Of Seat-belt Warning System Based On Machine Vision

The premise of seat belt protect passengers effectively is that they have worn seat belt already and in a proper way.So far,most studies on seat belt intelligent detection were focused on whether the passengers had worn seat belt or not,while whether in a proper way(such as: Wear style of seat belt,Flabby extent of seat belt,High or low wear position of seat belt,etc.)had not attracted sufficient attention yet.In order to promote the application of three-point seat belt correctly and effectively,this paper presents a novel seat belt wearing warning system based on machine vision,with the help of infrared sensitive markers printed on the seat belt.The main contents are as follows:(1)Building the experimental platform for image data collecting under different seat belt wear patterns.Designing and implementing different lighting conditions and different colors of passengers' clothes under the vehicle driving test.The purpose is to obtain the video information of driver's seat belt wear patterns(such as: not wearing seat belts,wearing only diagonal belts,wearing only lap belts,diagonal belts and lap belts wear loose,wear too high,wear too low and wear proper,etc.)after the vehicle starts.After the data processing,a total of 72,000 images were obtained,which provided image data support for the design of the seat belt detection algorithm and system verification.(2)Augmented reality technology was used for rapid identifying and locating seat belt markers.Then some total threshold method,fuzzy clustering threshold method,curve fitting relaxation method and the shoulder's outer contour intersection location method were designed respectively,aiming to realize the purposes of detecting weather the seat belt was worn or not,and particularly weather the worn style,flabby extent as well as height status were proper,to achieve the detection propose of whether the occupant seat belt has been worn in a proper way.(3)To allow the seat belt detection algorithm can be used in vehicle,two kinds of hardware schemes are used,such as: develop the DSP hardware systems independent and use raspberry pi to do application integration.Infrared camera and sound and light alarm module are designed,combining the principle of infrared light imaging and the hardware features of Raspberry Pi.The “seat belt wear proper detection software based on mark” was designed,based on the seat belt detection algorithm.Besides,the software and hardware integration and code transplant are carried out.According to the hardware features of Raspberry GPIO,the control algorithm of sound and light alarm module is designed based on WiringPi library function.(4)Test the function of each module of software and hardware system developed by this research and seatbelt detection algorithm validation.After testing,the hardware and software system modules work properly,and each detection modes can achieve the desired results.The results of seat belt detection algorithm validation showed that the method invented in this paper had high accuracy rates as 95.4% when detecting whether the seat belt was worn,93.1% when detecting whether worn style of the seat belt was right,79.5% when detecting whether the seat belt was too flabby,and 85.3% when detecting whether the seat belt was worn in a exceeding high or low position,under different kinds of image-forming conditions and background interferences.Finally,integrated accuracy rate of detecting whether the passengers had worn seat belt in proper way came up to 89.5% in total,which verified the accuracy,effect and adaptability of the method proposed.