Real-time Monitor Driver Vigilance Using Infrared Images

Public traffc accident, a serious social issue, is always threating the security ofpeople'slifeandproperty. Thus,howtoreducetraffcaccidentattractstheattentionanddedication of academy and industry. Statistics and analysis show that fatigue driving isa most important cause for traffc accident. Therefore, to monitor the vigilance of thedriver can effectively decrease the fatigue driving, and then reduce the happening oftraffc accident.The traditional driving vigilance evaluation systems do multi-features analysis inthe human face images under light or double ring infrared illumination, and then evalu-atethedriver'svigilance. Thesemethodshaveseveraldefects. Forexample, (1)imagesunder the light or double ring infrared illumination image have their own defects. Therobustness of the light condition is very poor. It may not work in the overcast or light-less condition. The double ring infrared camera need a lot of control circuit.(2) Thetraditional methods do multi-features analysis using eye, mouth as seperate part. Thesemethods run very slow and are hard to process in common ways.In order to ffx the mentioned defects. This paper mainly studies the evaluationmethods of driving vigilance using infrared images and establishes the real-time driv-ing vigilance evaluation system using infrared images. The main innovation worksdetailed as follows:1. After the improvement of the double ring infrared camera, this paper gets themodiffed single ring infrared camera. It not only lower the cost of the hardwareof the system but also improves the analysis effciency and algorithm stability.2. In the analysis of vigilance expression, this paper treats all the features of humanexpression as a united organism and uses the Active Shape Model to analyze and modelhumanexpression, thuspeoplecanquicklyextractexpressioninformationfrom the face images.3. Asforthecomplicatedhumanfaceexpressioninformation, thispaperofferssim-ple methods to extract facial expression. It can effectively extract the feature in-formation which is closely related with the vigilance, like wink frequency, theopen-and-close state of the mouth, and deffection angle of the head, etc. further-more, this paper saves the instantaneous characteristics and temporal character-istics in a sextuple vigilance parameter vector, which can facilitate the vigilancemodel building. In the choice of vigilance model, this paper gets and appliesan effective vigilance evaluation model based on support vector machine afterexperimenting and comparing in various model experiments.4. In order to facilitate the simulation experiment, this paper provides specializedvigilance evaluation software CIVigilance.5. We builds specialized experiment methods of image vigilance evaluation andkeep improving the methods in the experiment. This paper also participates invigilance driving simulation experiment by the combination of EEG and images,and veriffes the conclusion of this paper in the experiment.The innovations and improvements of this paper enable this paper to build relativesophisticated vigilance evaluation algorithm and experiment platform using infraredimages. It can lay a solid foundation for latter study and analysis.