The purpose of this article is to upgrade the vertical roller damage alarm of steel bar transmission lines from manual inspection to artificial intelligence recognition. The original manual inspection method required production shutdown to achieve the inspection of the quality of the vertical rollers. The discovery of faults has a delay and can affect the quality of steel bar production. Moreover, with the construction of digitalization, the construction of unmanned workshops at Xianggang has been greatly expanded, and the research on unmanned alarm for vertical rollers is imperative. The main content of this article includes scheme design and demonstration, design and optimization of machine vision algorithms, analysis and troubleshooting of various faults encountered during on-site installation and debugging. This article is the first in the domestic steel industry to achieve automatic recognition and alarm of vertical rollers, freeing up manpower, ensuring personal safety, and improving the production quality of steel.