Feature Extraction Method Of Pulsed Lidar Echo Based On Deep Learning

In this paper,a method of pulse lidar echo feature extraction based on deep learn ing is proposed,which improves the dynamic range of pulse lidar and ensures the ran ging accuracy without increasing the hardware complexity.In order to improve the dy namic range of pulse lidar,a saturation echo recovery method based on LSTM model is proposed in this paper.The structure of the LSTM model is obtained by using the simulation echo model,and the parameters of the LSTM model are adjusted by the ac tual echo data of the pulse laser radar.Because the unsaturation echo and the saturati on recovery echo need to carry on the distance characteristic extraction,so this paper presents a pulse lidar echo feature extraction method based on LSTM-CNN model to extract the range feature of lidar.The laser radar echo signals are grouped and labeled according to the 0.15 m resolution.The pulse laser radar echo data is classified by L STM-CNN model,and the range characteristic information under this classification is obtained.The average ranging error of the LSTM-CNN model is 0.063 m and the stan dard error is 0.061 m,and the average error of the range feature extraction is 0.070 m,the standard deviation is 0.073 m,which meets the requirement of laser radar ranging accuracy of 0.15 m.The results show that the proposed method can recover the saturat ed echo signal of lidar and extract the range feature with 0.15 m resolution.