Machine learning-based polyurethane self-flowing floor flatness detection and optimization method

The invention discloses a machine learning-based polyurethane self-flowing floor flatness detection and optimization method. The method comprises the following steps of 1, multi-source data acquisition and fusion; 2, applying and optimizing a feature extraction depth prediction network; 2.1) constructing a coding-decoding convolutional network based on the fused data in the step 1, and performing depth residual prediction by adopting a residual structure; 2.2) processing the original depth map Dorig by using a ReLU activation function and a convolutional layer weight Wk; high-precision depth residual error information is output through adjustment of an activation function and a learning rate alpha k, and the original depth map is optimized by utilizing the predicted depth residual error information, so that the accuracy of the depth information is improved; step 3, performing accurate flatness evaluation and iteration; 3.1) the optimized depth information is used for three-dimensional reconstruction, and a hi