IAE-Unet: Pork Ribs Segmentation Method Integrating Improved Dual Attention Mechanism with Edge Feature Enhancement Module

Autonomous segmentation of carcasses represents a crucial advancement in the meat processing industry, with accurate segmentation of carcass images being a pivotal technology in achieving this objective. However, the presence of fascia, blood stains, and other interferences on the surface of pork ribs poses challenges for image segmentation networks, hindering accurate segmentation of pork ribs. To address these issues, this paper proposes a pork ribs segmentation method integrating improved dual attention mechanism with edge feature enhancement module. Firstly, the paper enhances the network’s focus on the target region by constructing an improved dual-attention mechanism. Secondly, through the edge feature enhancement module, it augments the edge features of pork ribs and compensates for missing edge features. Ultimately, the method achieves accurate segmentation of pork ribs by fusing the U-net backbone network with the improved dual attention mechanism and the edge feature enhancement module. Experimental results demonstrate a $1.58 \%$ improvement in the Dice coefficient over existing algorithms, indicating superior performance.