|
|
|
| FREQUENCY-DOMAIN RECOGNITION METHOD FOR WIDEBAND RADAR OPTICAL REGION TARGET |
| Li Ying①; Ren Yong①; Shan Xiuming①; Xiao Zhihe②; Chao Zengming② |
| ①Department of Electronic Engineering Tsinghua University Beijing 100084 China;②Beijing Institute of Environmental Features Beijing 100854 China |
|
|
|
|
Abstract Meeting the application requirements of wideband radar optical region target recognition, this paper presents a simple and effective frequency-domain recognition method. First, two kinds of waves called backscattering amplitude wave and phase feature wave are constructed directly from frequency measured data sets, which keep invariant on the shift of target in the radial direction. Based on these waves, generalized frequency and length of wave are extracted as recognition features insensitive to target azimuth. With the aid of the idea of "ruler transform" in time-frequency analysis, the feature sets are further completed. Aiming at lessening the effect of measuring noise, the paper then designs a specific preprocessing method. FMM neural network is chosen as the classifier with modified training algorithm. The recog-
nition results show that this target recognition algorithm can obtain high correct classification rate.
|
|
Received: 29 June 2000
|
|
|
|
|
|
|
|
|
2001, Vol. 23 
