“The latest research results on concrete surface crack segmentation technology have been unveiled. The research team proposed an innovative crack segmentation model that combines linear guidance and grid optimization to address issues such as low segmentation accuracy, missing fine cracks, and background interference. By introducing a multi branch linear guidance module, this model significantly improves the network's ability to express the linear structure of cracks, strengthens the connection between cracks in different regions, improves the ability to perceive global context information, and thus improves the segmentation accuracy. Meanwhile, the introduction of the grid detail optimization module has successfully prevented the omission of subtle cracks, bringing new breakthroughs to the field of crack segmentation. In addition, the embedding of hybrid attention modules further highlights crack features, reduces background interference, and further improves the performance of the model. The experimental results on Deepcrack537, Crack500, and CFD crack datasets show that the IoU and F1 score values of this model are significantly better than most existing methods, demonstrating higher segmentation accuracy. This research achievement undoubtedly provides a new direction for the development of concrete surface crack segmentation technology and provides strong support for solving related practical problems.”