Research On Moving Target Recognition In The Airport Based On Sensor Network

Due to the incresion of air traffic and weather influence, runway incursion has gradually become a serious risk of airport security. Target identification as a basic part of the the runway intrusion prevention system, can provide imformaiton support for target tracking and invasion control. So for the target types in the airport, a new feature exatration method and a data fusion method which based on AMR sensor network are proposed in this paper.The first one is target detection. Build the magnetic-dipole model of the moving target according to detection principle of AMR sensor. And several groups of sensor are deployed in the airport and made up a sensor network. Using a dynamic threshold method detecte target and obtain target signal.The second one is target feature exatration.A combination of time and frequency domain feature extraction algorithm is put forward in the paper. Firstly, the magnetic signals caused by surface moving target are sensed by sensors placed at the center line of runway, which are used to calculate the length of target. After regularizing time domain signal, peak features can be extracted based on the target length. Secondly, the regularized signal is decomposed using DB wavelet. The obtained wavelet coefficients with the energies of normalizing approximate signal and every layer signal are realized as frequency domain feature of the signal. The time domain features and the frequency features consitute all features of the target. Using a method of target features differentiation measurement to demonstrat that signal features provided by the proposed method are more recognizable than the ones provided by time domain or frequency domain feature method individually.The last one is target recognition of multi-sensors. The algorithm of two-level data fusion of multisensors network based on D-S evidence theory is put forword in the paper. Using the fuzzy recognition method obtain the recognition result of single sensor in one axial. Because of the conrelation of evidences and support level of evidences and reliability of evidences, using algorithm of the first level of data fusion based on similarity coefficient of evidences obtain the recognition result of single sensor. And allocate the weights according to the signal energy, using the second level of data fusion get recognition result of multi-sensors. The finally regonition results is gain by decision rules. Compard with three other algorithms, the fusion algorithm proposed in the paper has more superiority in the simulation.