Wood board small target surface defect identification and detection method based on machine vision

The invention discloses a wood board surface defect small target identification and detection method based on machine vision. A backbone network, an SE module, a GSConv module, a VoV-GSCSP module, a CPCA module and an NWD module are included. The backbone network extracts multi-scale features of a to-be-detected image containing small target defects; the SE module multiplies attention feature maps obtained by channels and spatial dimensions and inputs the attention feature maps into a feature map, so that adaptive feature refinement is carried out, and background feature training weights are reduced; the GSConv module enables common convolution and depth separable convolution to be combined, the calculation cost of the model is reduced, and better feature expression ability and calculation efficiency are achieved; the VoV-GSCSP module uses a one-time aggregation mode to design a cross-level part network, has the advantage that multiple receptive fields represent multiple features, and reduces the complexity o