For the problem that it was difficult to recognize the low-quality DataMatrix code formed by the image changes at the surface of metal forgings under the complex environment of thermal processing, a set of three-dimensional visual identification technology was proposed, which relied on the “depression” feature of the DM two-dimensional code and obtained the original point cloud data by scanning of vision sensor. Then, removing interference information points and screening effective identification points were conducted by data layering analysis and local projection algorithm to finally realize the DM two-dimensional code identification function. In addition, in order to verify the identification ability of this technology, DM two-dimensional code marking test was carried out. Pneumatic dot-needle marking technology was used for marking, and the DM codes of four typical metal materials with low surface quality were obtained after heat treatment. The test results show that the technology can realize 100% identification of low quality DM two-dimensional codes marking at the metal surface, which provides technical guidance for the metal forging industry and other manufacturing industries to carry out quality control and information traceability.