This paper proposes a Data Matrix code recognition technology, which includes initial position, rotation correction and final accurate locating based on corner detection and cluster analysis. The accurate Data Matrix code region can be obtained from the complex background. By detection method proposed in this paper, the success rates of various types of Data Matrix barcode recognition are improved, especially in the test with the brightness difference, and the effect is very obvious, which embodies the characteristics of Data Matrix code used for Internet of Things. It can also more effectively find the area of uneven illumination. The improved algorithm is more stable and adaptive, so as to improve the success rate of recognition.