The invention relates to a road intersection detection method and device based on improved YOLOv3. The method mainly comprises the following steps: firstly, acquiring a road image; then, performing network training, and constructing an improved YOLOv3 network model; wherein the improved YOLOv3 network model comprises a feature extraction end and a feature detection end, the feature detection end comprises a plurality of channels, and in each channel, transversely broadening a corresponding convolution module to generate different feature maps, and then performing longitudinal aggregation; andadopting the improved YOLOv3 network model to identify a road image to be detected, and outputting a result. According to the invention, the convolution module in each channel of the improved YOLOv3 feature detection terminal is broadened transversely. different feature maps are generated according to the feature map set, and then longitudinal aggregation is carried out, so that the network widthof the convolution module of each channel can be wider, the expression capability of the network is enhanced, the detection difficulty of small-size road intersections in a complex remote sensing scene is reduced, and the detection precision is improved.