In recent years, thanks to the continuous development of concrete 3D printing technology, the quality of printed products has attracted more and more attention. The geometric characteristics of printed products have become an important index to evaluate the printing quality. Aiming at solving a series of problems existing in the current 3D printing concrete geometric feature extraction methods e.g. complex process, limited precision and lack of evaluation indicators, this paper designs an automatic extraction method of 3D printing concrete geometric features based on machine vision. Firstly, the camera array deployment method is used to obtain the original image which is orthodontic to improve the image quality. Then, the geometric features of printed products are extracted by semantic segmentation model U-Net, after which the results are optimized. Finally, through the quantitative index value, it is visualized and the print result is evaluated. Experiments have shown that the geometric feature automatic extraction method proposed in this paper has significant advantages such as fast speed and high accuracy. In addition, systematic evaluation indicators can provide reference for the quality evaluation of concrete 3D printing.