The invention discloses a circular target pose detection method based on binocular vision three-dimensional reconstruction. The method comprises the following steps: adopting a Zhang Zhengyou calibration method to calibrate internal and external parameters and distortion parameters of a binocular camera, using the binocular camera to photograph images of a same circular target, and performing distortion correction on left and right images according to the calibrated distortion parameters to obtain undistorted left and right images; using a direct least squares ellipse detection method to identify and extract the undistorted left and right images to obtain an ellipse; traversing the left and right images separately, and finding left and right reference image dot pairs on the ellipse in theimages; processing to obtain left and right first image intersection pairs according to left and right camera optical center coordinates under a left and right camera coordinate system and in combination with left and right reference image dot pairs; and constructing a left and right reference coordinate system, and using left and right transformation space dot pairs and the left and right reference coordinate system to calculate and obtain the center of the circular target in the space. According to the circular target pose detection method based on the binocular vision three-dimensional reconstruction provided by the invention, the radius dimension of the circular target is finally obtained without giving the radius dimension of the circular target in advance, and subsequent normal vector screening is not needed to be performed.