Photovoltaic cell defect detection bionic model based on mimicry vision

The invention relates to the technical field of photovoltaic cells, in particular to a photovoltaic cell defect detection bionic model based on mimicry vision. The method comprises the following three parts: (1) a backbone network: the backbone network is used for extracting effective information in an image and mapping original data to a required space, and features extracted by the backbone network are transmitted to a neck network; the neck network is used for fusing features of different scales and levels and enriching the expression ability of the features, and the features fused by the neck network can be transmitted to a detection head; and (3) a detection head which is mainly responsible for predicting features, generating a bounding box and predicting categories. A backbone network based on a human receptive field and a visual mechanism has stronger abstract ability for features, and effective features can be efficiently extracted from the electroluminescent image of the photovoltaic cell with single