Aiming at the problem of amplitude and phase errors in full-waveform light detection and ranging (LiDAR), a method based on neural network and pulse feature parameters is proposed. The feature parameters of waveform, including shape information, energy information, gradient information, symmetry information, and distance information were extracted from non-cooperative target detection information and graded based on the Pearson correlation coefficient. A neural network model combining multi-echo characteristic information with amplitude-phase error correction was established to correct the amplitude and phase errors of each channel in full-waveform laser measurement. The experiment used 5%, 20%, 60% and 80% standard reflectorsand laser acquisition modules to collect pulse data at 7 different distances. The processed data was also compared with the traditional full waveform measurement method. The results show that this method could effectively reduce the influence of amplitude and phase errors in the full-waveform laser measurement, and the measurement accuracy is improved by more than 51.2%.