Automatic driving scene semantic prediction method

The invention discloses an automatic driving scene semantic prediction method. The method comprises the following steps: performing word vector expansion on context vocabularies rich in semantics by using a word embedding unit; extracting local semantic information by using a local semantic convolution unit to obtain a semantic feature combination among several objects; extracting global semantic information by using a global semantic loop unit to obtain all forward and reverse semantic information features; and performing feature fusion by using a semantic fusion unit to realize semantic prediction of the scene. According to the method, secondary processing is carried out on the sensor data in the automatic driving scene, so that missing, errors and conflicts possibly existing among the sensor data from different sources can be solved to a certain extent, and automatic driving scene semantics can be predicted more accurately. 本发明公开了一种自动驾驶场景语义预测方法，包括：使用词嵌入单元将富含语义的上下文词汇进行词向量扩充；使用局部语义卷积单元对局部语义信息进行提取，得到几个物体间的语义特