90％。Road temperature depends on road condition (dry, wet, snowy, icy) and solar radiation (mapped to season, geographical location, time, air temperature and air humidity), and there is the nonlinear causality among them, thus, road condition can be detected indirectly by road temperature and solar radiation with BP neural network. During the experiment to detect road condition (dry, wet), BP neural network was trained with 1344 group data and validated by 96 group data, the road condition detection accuracy reached 90%.