The work aims to propose a new method of surface defect detection based on improved gray level co-occurrence matrix to solve the problem of poor real-time calculation and low defect type recognition rate in the detection of printing defects. Firstly, the mainstream defect detection process was optimized to determine the presence of shape defects on the printing surface to be tested by pre-processing and differential operation of the image. Then, a feature parameter extraction algorithm for the defective region was designed to address the problems of high dimensionality, easy loss of information and poor rotational invariance of the traditional gray level co-occurrence matrix. Finally, combined with the obtained feature parameters, the defects were classified and recognized by the classifier based on support vector machine. The experimental results showed that the feature parameters extracted by the improved algorithm designed in this paper can more accurately characterize the features of defect areas. Meanwhile, the extraction time of feature parameters and the defect classification and recognition rate were more advantageous than the traditional detection methods. On the premise of ensuring real-time computation, the detection method designed in this paper can effectively identify the texture features of the defect areas on the surface of printing and has a high rate of classification and recognition.