A multi-rotor UAV target recognition method based on deep normalized network

Abstract

The invention belongs to the technical field of target recognition and neural network, and in particular relates to a multi-rotor unmanned aerial vehicle target recognition method based on a deep normalized network. The present invention preprocesses the echo data of the multi-rotor UAV as the input of the depth normalization network. The depth normalization network is composed of multiple normalization sub-networks, followed by a softmax classification Layer, the input of each normalized sub-network in the present invention is spliced ​​by the output and input of the previous normalized sub-network, so that the learned network parameters depend on the feature information of this sub-network and the front sub-network at the same time , so as to better describe the feature information in the original radar echo data. Moreover, by introducing normalization processing to the input of the sub-network, the data distribution is optimized, so that the new distribution is more in line with the real distribution of the data, and the network model is further guaranteed. nonlinear expression ability.

Description

technical field [0001] The invention belongs to the technical field of target recognition and neural network, and in particular relates to a multi-rotor unmanned aerial vehicle target recognition method based on a deep normalized network. Background technique [0002] With the rapid development of UAV technology, UAVs have been widely used in military and civilian fields, but they have also brought security problems caused by illegal intrusion into private areas, collisions with aircraft, terrorist attacks, etc. , flight safety, etc. have brought great trouble. Therefore, it is very important to accurately identify the type of UAV. [0003] At present, the method of identifying drones mainly uses the rotating parts of multi-rotor drones to form micro-Doppler spectra with obvious differences, and then uses conventional machine learning methods for classification and recognition. However, conventional machine learning methods must be artificially set Identification features,...

AI Technical Summary

Problems solved by technology

[0003] At present, the method of identifying drones mainly uses the rotating parts of multi-rotor drones to form micro-Doppler spectra with obvious differences, and then uses conventional machine learning methods for classification and recognition. However, conventional machine learning methods must be artificially set Identification features, and for multi-rotor UAV targets, its identification features are not easy to determine manually

Images