The invention discloses a method for constructing a mathematical model for calibrating a 3D rotation difference and a method and a device for calibrating the 3D rotation difference. The method for constructing the mathematical model comprises the following steps of acquiring first detection data of a motion detection device and second detection data of a 3D measurement device; presetting a transfer function model, and converting the first detection data into corresponding first conversion data through the transfer function model; establishing a loss function, and carrying out optimization solution on a coefficient qa and a coefficient qb by using a nonlinear least square method; and introducing coefficient qa and the coefficient qb which are obtained by optimization solution to the transfer function model, and establishing the mathematical model for calibrating the 3D rotation difference. According to the constructed mathematical model, a coordinate direction of the motion detection device can be rotated to be consistent with a coordinate direction of the 3D measurement device, and zero reference positions and the coordinate directions of the motion detection device and the 3D measurement device are kept consistent, so that a condition is provided for calibrating the 3D rotation difference between the motion detection device and the 3D measurement device.