High-precision tiny target detection and size measurement method based on deep learning

Abstract

The invention belongs to the technical field of computer vision and deep learning, and provides a high-precision tiny target detection and size measurement method based on improved YOLOv8, which realizes intelligent upgrading of industrial quality inspection and military reconnaissance in combination with a high-resolution image sensor (such as a CMOS / infrared sensor). Aiming at small target missing detection, complex scene interference and real-time bottleneck, according to the scheme, the shallow feature detail capturing capability is enhanced through a small target detection branch (SMA), and multi-scale features are dynamically fused in combination with two-stage routing attention (BRA), so that the detection robustness of a small target under a complex background is remarkably improved; and a lightweight RepGhost module and C2f structure optimization are introduced, so that high reasoning efficiency is maintained while the calculation complexity of the model is reduced, and the real-time requirements of industrial edge equipment and military embedded terminals (such as an unmanned aerial vehicle-mounted infrared imaging system) are met.

Images