Abstract: Based on the wideband radar echo data measured in a microwave anechoic chamber. A study on the recognition method of micro drone models is conducted based on HRRP, and proposes the residual attention pyramid pooling net (RAPPNet) is proposed, which verifies the feasibility of using HRRP for drone recognition. Comparative experiments on echo data of different bandwidths show that larger bandwidths can effectively improve the accuracy of UAV recognition based on high resolution range profile. Under the 6 GHz bandwidth, the proposed method can identify drones with an accuracy of 90. 63%.