Small sample target detection method based on semantic enhancement feature generation and predictive optimization

The invention discloses a small sample target detection method based on semantic enhancement feature generation and predictive optimization, and the method comprises the steps: constructing a small sample target detection model which comprises a query branch, a support branch, a class-independent aggregation module, a detection head and a detection result optimization module, and constructing a feature generator to train the small sample target detection model. The constructed feature generator integrates semantic and visual information and allows the generator to enhance category-centered representation through cross-modal constraints, thereby defining boundaries of different categories while ensuring improvement of data diversity. Besides, the prediction optimization module disclosed by the invention can accurately filter potential false alarms caused by boundary box offset, and ensures that only the most reliable detection result is left. Compared with other methods, the method provided by the invention ha