Satellite attitude vision measurement algorithm of nonlinear residual self-attention mechanism

The invention discloses a satellite attitude vision measurement algorithm of a nonlinear residual self-attention mechanism. The measurement method comprises the following steps: constructing a satellite attitude rotation data set, obtaining an image single-channel matrix, extracting local features through a convolutional neural network, carrying out position encoding on the local features by a position encoder, and learning the image features after position encoding by a nonlinear residual self-attention mechanism. And outputting attitude measurement information through the full connection layer. By using the characteristic that a single-channel image matching self-attention mechanism can only process a two-dimensional matrix, the image position information and the size of a compression model can be reserved. Aiming at the defects of a self-attention mechanism, a nonlinear activation function and a residual network are introduced, a nonlinear residual self-attention mechanism is provided, and the model expres