High-robustness target detection model and target detection method

The invention relates to a high-robustness target detection model and a target detection method. The system comprises a backbone feature extraction network, a feature fusion network and a head network, the backbone feature extraction network comprises a feature extraction basic network and an efficient feature extraction module, and the efficient feature extraction module is connected in series in the feature extraction basic network; feature fusion is carried out on different scale features extracted by the backbone feature extraction network by using a feature fusion network, the feature fusion network carries out feature fusion based on a feature fusion method of a bidirectional feature pyramid network, reasoning is carried out through an attention enhancing module after feature fusion, and reasoning information output by the attention enhancing module is subjected to feature fusion based on the feature fusion method of the bidirectional feature pyramid network. And performing network output processing on