Based On Driver’s Visual Characters Lane Change Intention Recognition

Lane change behavior is often happened during driving. When changing lanes, if thedriver didn’t fully observe surrounding environment of vehicles or judged with deviation orjudged wrongly, traffic accident will be caused extremely easily. If the driver’s operationintention can be predicted before lane change, then we can do safety assessment to theoperation behaviors in advance, and give the alarm before danger happens, finally avoidingthe traffic accidents. Vision is the most important approach to obtain traffic information forthe driver. The paper talks about the identification of lane change intention, based on thecharacteristics of visual behavior of the driver.In the paper, based on the study of relevant research at home and aboard, thelane-changing behavior was simply analyzed, and the lane-changing process was divided intotwo phase: intention phase and implementation phase. Combined with the experiment testequipment used, the starting point of the implementation phase was confirmed. Then themethod of watching the rearview mirror in order to assure the time of intention phase ofdrivers was put forward. The test video and test data of lane-changing behavior gotten formthe testing car provide data support for the study on lane-changing intention identification.Based on the test video and test data, the paper deeply analyzed the changing regularityof fixation and saccade parameter of drivers in the lane-keeping phase and lane-changingintention phase. The fixation areas were divided into seven parts, and eye movement ofdrivers was analyzed with statistical analysis. We use Markov chain theory and calculateone-step transition probabilities of drivers in different fixation areas to solve its stationarydistribution. Finally through the comprehensive analysis, some visual behavior indexes wasput forward that can be used to characterize the driver’s lane-changing intentions, and theLogistic regression model was established that can identify lane-changing intentions.The results of the study show that:1. When changing lanes, the time of intention phase of drivers is5s;2. Compared to the lane-keeping phase, the attention of drivers of the rearview mirrorsignificantly increased in the intention phase; both in lane-keeping phase and intention phasedrivers are very concerned about the traffic information in current lane, but the performance is more obvious in the lane-keeping phase; in lane-keeping phase, the visual search breadth inhorizontal direction, the standard deviation of head turning angle and the scanning range aregenerally lower than those in intention phase.3. The Logistic regression model was established that can identify driver’s lane-changingintentions effectively.