The invention discloses an ore
grinding hard rock target detection method and
system based on a HardRockNet
network structure, and the method comprises the steps: constructing an ore
grinding hard rock target detection model which comprises an input layer, a multi-scale
feature extraction layer, a DFPN
feature fusion layer and an output layer which are sequentially connected in series; the multi-scale
feature extraction layer comprises a plurality of
feature extraction sub-
layers which are sequentially connected in series and have different feature extraction scales, the output ends of the plurality of feature extraction sub-
layers are connected with the input end of the DFPN
feature fusion layer, and the plurality of feature extraction sub-
layers perform feature extraction by adopting a DPBlock module; and detecting hard rock on the ore
grinding picture by using the trained target detection model. According to the invention, the DPBlock module and the DFPN module are combined, and a HardRockNet
network structure is constructed, so that the
small target detection precision is improved, the robustness of complex backgrounds and target overlapping problems is enhanced, and accurate detection of hard rocks is realized.