Against the backdrop of rapid technological development in China, intelligent technologies such as big data, artificial intelligence (AI), and robotics have been widely applied, driving society into the era of intelligence. In this transformation, the application of industrial robots (IR) is becoming increasingly widespread, and their interaction with humans is becoming closer. However, the sudden failures of IR and the safety issues it causes are gradually becoming more prominent. With the continuous progress and widespread application of computer vision (CV) technology, integrating it into IR action intelligent recognition is expected to significantly improve recognition rate. This article proposes an IR action recognition algorithm based on CV and deep learning (DL). This algorithm captures dynamic images of IR, extracts key action features using DL models, and achieves accurate action classification and anomaly detection. This article aims to improve the security and stability of IR, reduce potential security risks, and provide strong support for the further development of IR technology in the future. The research results show that the algorithm designed in this study shows excellent performance. Compared with many traditional methods, this algorithm not only has higher recognition accuracy, but also has faster processing speed.