道路交通事故多状态识别方法研究

Recognition Method Study on Road Traffic State

研究道路交通事故多状态准确识别的方法。交通事故的现场情况较为复杂,不能依靠单一的特征进行智能判断,当前的道路交通状态监测系统对意外状况的检测中,针对小范围内事故的集聚,受到车辆重叠、交错等视觉像素重叠的干扰,可能无法准确地捕捉事故发生异常路段的边界,仅仅依靠车辆的长时间聚集判断车祸的发生,检测的准确性较低。提出了一种车辆运动目标状态解析的道路交通事故多状态识别方法。引入车祸多态识别模型,根据道路交通中的车辆图像,对车辆轻微刮碰事件、碰撞事件、严重碰撞事件等车辆状态进行识别,从而完成道路交通事故多状态识别。实验结果表明,运用该方法进行交通道路事故多状态识别,能够对各种不同的交通状态进行有效识别,从而为智能交通管理提供依据。

The accurate recognition method for multi - state of road traffic accidents was researched in this paper. This paper presented a moving target state analysis based on the multi - state recognition in vehicle＇s road accidents. It introduced multi - state recognition model of car accident. According to the road traffic in the vehicle images, the slight scratching, collision, vehicle condition and other serious collision events in the vehicle incident were identi- fied, thus the multi - state road accidents identification was completed. Experimental results show that the method for road traffic accidents state recognition is effective, and it can provide the basis for intelligent traffic management.