Road target identification method for automatic driving system

The invention discloses a road target identification method for an automatic driving system. The road target identification method comprises the following steps of adjusting the size of an input picture; extracting the picture features through a trunk network; downsampling through an expansion convolution module; constructing a feature pyramid structure based on the partial feature map; generatingan anchor frame on the characteristic pyramid structure and characteristic graphs with different scales obtained by downsampling; obtaining an offset value of the boundary frame relative to the anchor frame from the picture through a network, and decoding the offset value to obtain the boundary frame; performing descending sort and non-maximum suppression on the bounding boxes to obtain a final detection and identification result; in the training part, training the network by using the smoothed absolute value loss function and the weighted average loss function of the class of the backgroundsamples before balance. Ac