Research On Weak Target Detection Algorithm Of FOD Radar

FOD is the abbreviation of Foreign Object Debris,which refers to any object located in the airport environment that may cause damage to the aircraft.Timely detection and removal of FOD is an important part of ensuring the safety of aircraft taking off and landing.Automatic FOD detection system can detect FOD targets timely and efficiently,but the existing detection systems have poor effect on detection small-sized FOD targets.Based on this,this paper studies the small FOD target detection algorithm,then applies the classification idea to the target detection problem,and proposes several classification algorithms suitable for small FOD target detection.The main contents of this paper are as follows:1.Two classical CFAR detection algorithms are studied,namely CA-CFAR and CACMCFAR.Firstly,the principle of two CFAR algorithms is introduced in detail,and then the detection effect of them on small FOD targets is verified based on the actual measurement data of the airport.2.For the first time,the classification idea is applied to the FOD target detection,and two small FOD target detection algorithms based on SVDD and NSVDD classifiers are proposed.First,the detection principle of two methods for FOD targets is introduced.And then the design process of SVDD and NSVDD models is described in detail,including the extraction of echo power spectrum features and the optimization of classifier parameters using PSO algorithm.Finally,the detection performance of two methods is verified and compared with two CFAR algorithms.3.The application of classification to FOD detection is further discussed and a small FOD target detection algorithm based on CNN is proposed.First,a CNN model for detecting FOD targets is designed,and then an image data set containing only ground clutter and both ground clutter and FOD target is constructed to train and test the network.The classification results of the test set verify the effectiveness of the CNN method in detecting small FOD targets.Finally,the performance of the algorithms mentioned in this paper is compared.