Deep learning-based biomedical image accurate detection processing system and method

Abstract

The invention discloses a biomedical image accurate detection processing system and method based on deep learning in the biomedical field, and the system comprises a dynamic anchor frame generation module which is used for generating a self-adaptive anchor frame through employing an improved K-means + + algorithm based on the prior of biomedical image target size statistics; and the multi-modal similarity measurement module comprises a multi-scale feature pyramid fusion unit and a mixed distance measurement unit. Through the dynamic anchor frame generation algorithm and the optimized similarity measurement model, the small target detection AP is significantly improved compared with the prior art, missing detection and misjudgment in image detection are effectively reduced, and more accurate data support is provided for biomedical research; through accurate image detection and processing, a technical guarantee is provided for establishment of a scientific research integrity system, and scientific research integrity risks caused by image problems are avoided.

Images