This paper proposes an online measurement method for cable diameter based on amplitude-frequency characteristics. Firstly, the correlation mechanism between cable impedance and diameter is analyzed through modeling based on transmission line theory. Then, to reduce the loss of computational accuracy of the neural network and accelerate its convergence speed, a feature processing module based on grouped aggregation and partial least squares is designed to extract parameter features from the dataset. Finally, a measured dataset of amplitude-frequency characteristics of 8 cables with different diameters is constructed based on the FPGA hardware platform. The measurement accuracy was analyzed, which verifies the effectiveness of the method proposed in this paper.