Fragment hole image segmentation algorithm of target plate using PIDNet with multi-scale fusion dual attention

To address the challenges of ineffective segmentation of perforations on target plates in static explosion damage experiments, which lead to the loss of edge characteristic information and low segmentation accuracy, this paper introduces fragment hole image segmentation algorithm of target plate using PIDNet with multi-scale fusion dual attention. The algorithm is based on the PID triple-branch network architecture and adopts the dual-attention module to enhance the hole edge feature information; and it uses the multi-scale feature fusion MSFF-Bag module to improve the segmentation accuracy of the hole contour. The experimental results show that the algorithm in this paper achieves better results in the mean intersection and concurrency ratio and pixel accuracy compared with the current state-of-the-art methods, which provides a guarantee for analyzing the dispersion characteristics of the fragments and the destructive effect and has important military theoretical value and practical application significance.