What is the process of convolution operation? How does the matrix of convolution calculation come from, as follows, how to understand this convolution operation?

First of all, the convolution kernel is the same, the input is the same, and the characteristics of the output are the same. Only the matrix form of the output becomes the form of the column vector.

In essence, there is no difference between general convolution operation and convolution operation in matrix, the only difference is that matrix elements are rearranged into row vectors or column vectors.

Kernel matrix is often selected according to experience or obtained through learning.