中文摘要：

灯语是无人驾驶车与周围车辆的主要交互手段，但目前国内尚未展开专门针对灯语的研究工作。对灯语的检测和识别有助于实现无人驾驶车的拟人驾驶能力。在对车辆进行检测与识别的基础上，提出了一种具有环境自适应特征的灯语状态转换检测方法，并借鉴数据场思想构建灰度势场对车灯进行精确定位和状态分类，最后根据交通法规对常用灯语建模，构建二值化波形模板对车灯状态序列进行分类，仿真显示该模型能正确理解出目标车辆的灯语含义。本研究对无人驾驶车通过图灵测试，提高其驾驶智能性和安全性具有重要意义。

英文摘要：

Light language is the main interactive means between unmanned vehicle and other vehi- cles around, yet there isn＇t any research about unmanned vehicle especially on light language. Detec- ting and recognizing vehicle＇s light language can help to realize the impersonation drive ability of the unmanned vehicle. A state conversion detecting method of light language is proposed on the basis of the detecting and recognizing vehicle, which is able to adapt the surroundings. Accurate locating and state classifing of the lamp is achieved by gray-potential-energy-field which borrows the idea of data- field. Finally, common used light language is modeled according to the traffic rules of law. Two-value-waveform templet is used to classify the serial lamp states. Simulation shows this model can understand target vehicle＇s light language exactly. This research may help unmanned vehicle to pass Turing test, and improve its driving intelligence and safety.