Neural network for calculating attention weight for laser point cloud and training method

The invention relates to a neural network for calculating attention weight for laser point cloud, which comprises an input layer, a one-dimensional convolution layer, an activation function layer and an output layer, and is characterized in that the input layer is a 1*N-dimensional laser point cloud sequence or information capable of representing the contour of the point cloud, the sequence length is a positive integer, the convolution kernel of the one-dimensional convolution layer is strip-shaped and is used for processing the sequence, the activation function layer is used for enhancing the nonlinear representation capability of the network and improving the fitting capability of the network to complex data, and the output layer is a point cloud matching process of substituting attention weight sequences of 1*N dimensions into Scan-to-Scan or Scan-to-Map, and is used for reducing the interference of features without distinction degree on positioning and map construction. The neural network does not have a