The invention provides a depth camera automatic calibration algorithm based on three-dimensional feature points. According to the calibration algorithm, calibration work of a depth camera can be completed at a time by using two depth images and corresponding three-dimensional feature pixel points. The method comprises the following steps: firstly, combining a depth camera measurement model and a classical camera calibration model according to known coordinates of characteristic pixel points in a three-dimensional space, coordinates of corresponding points in a depth image and an original erroroffset to obtain a basic calibration model of a depth camera; secondly, obtaining an initial value of an internal parameter of the camera through an internal parameter constraint condition of the depth camera, and obtaining an external parameter according to the initial value of the internal parameter; and finally, optimizing the established depth image error control function to obtain an optimalparameter of the depth camera. According to the calibration algorithm, a traditional calibration model of the depth camera is simplified, the defect that classic calibration is affected by illumination conditions, scene uncertainty and the like is overcome, internal and external parameters of the camera are accurately obtained, a good restoration effect on edge distortion of the depth image is achieved, and the calibration algorithm has the advantages of being easy to operate, high in calibration precision, good in practicability and robustness and the like.