【正文】 echoes (a) DFT (b) CZT Fig. 7 (a) and (b) result from the I/Q data obtained by the CZT method. For parison, 北京科技大學(xué)本科生畢業(yè)設(shè)計（論文） 27 Fig. 7 (c) and (d) show the imaging results of the I/Q data obtained by the DFT method, the imaging method are the same as Fig. 7 (a) and (b), respectively. Fig. 7 (c) shows that even with accurately motion pensation, the true 1D crossrange image could not be obtained using the I/Q data obtained by the DFT method. Therefore, it is necessary to use CZT method to get the range gates39。 flight tracks are knownexactly. Using the known tracks to pensate the phases of the radar returns, we can get the ideal 1D crossrange images, shown in Fig. 7 (a). During the observation time, the formation targets rotate approximately 176。 I/Q data are obtained with DFT method, the amplitude and phase errors badly worsen the WVD results, and target number can hardly be determined. On the other hand, the distances between adjacent targets are decreased, which are beyond the resolution of the timefrequency transform. In addition,the interference of crossterms makes the target number couldn39。 echoes (a) DFT (b)CZT Suppose the radar parameters are the same as in Section . In addition to the two targets discussed in Section , a third target is included, which start from (x50,Y39。 RWT method can greatly reduce the affection of crossterms and obtain good target number detection results. These two figures also show that, because of the reduced amplitude and phase errors, the WVD and RWT maps of the I/Q data obtained by CZT method are clearer than those of the data obtained by DFT method. 北京科技大學(xué)本科生畢業(yè)設(shè)計（論文） 25 Fig. 4. The WVD of two formation flight targets39。 by applying CZT to the echoes, the peak value can be more accurately obtained, which assures the reliability of further signal processing. . CZT vs. DFT processing (corresponding to Fig. 1 (b)) 北京科技大學(xué)本科生畢業(yè)設(shè)計（論文） 24 . Fourscatterer model of the targets III. SIMULATION RESULTS OF TARGET NUMBER DETECTION A. Target number detection based on Doppler analysis Suppose a radar operates at 2GHz and radiates LFM signal with a bandwidth of . Its PRF is radar locates at the origin of the XYZ coordinate system. Two targets fly with 250m/s along the X direction from (X0, Y0, Zo ) and (X0100,Yo100,Zo) respectively, where x0= 57321 Yo = 138380 m, andzo= 8000 m. Both the targets are simulated by a fourscatterer model, shown in Fig. 3. The distance between the two targets is less than the range resolution, so the targets couldn39。 2) getting 1D crossrange image of targetinformation and resolving targets from the azimuth dimension 北京科技大學(xué)本科生畢業(yè)設(shè)計（論文） 22 [JIA, 01][WAN, 93]. Both methods need to obtain the I/Q data of the range gate containing the formation targets. However, the range gates are discretely sampled and the sample point may not be the peak value, which will bring extra amplitude and phase errors [JIA, 01]. Fig. 1 illustrates the range gates of a radar, where the amplitudes are normalized. The targets are between the 30 1st and 302nd sample points. Whether the 301st or the 302nd sample point is used for further processing, there will be amplitude and phase errors. In [JIA, 01], the author proposed to use conic interpolation to get the peak values. In this paper, we propose to apply chirpz transform (CZT) to getting the range gate peak value, which equivalent to more detailedly describe the range gate. The method has more distinct physical meanings than the conic interpolation method. II. CZT METHOD IN GETTING THE I/Q DATA OF RANGE GATE DFT calculates the equalinterval samples on the entire unit circle of the ztransform. However, CZT can calculate the ztransform on any interested curves, which is more flexible than the DFT. By choosing the start point, we can analyze the input data with narrowband highresolution. Given a discrete time signal x (n), 0 n N 1, its CZT is defined as X（ zk） =CZT[x(n)]=???1N0nnkznx ）（ =???1N0nnkn WAnx ）（ （ 1） where， zk=AWk， k=0,1,….,M 1, （ 2） Fig. 1. Range gates of a radar (a) the whole range gates (b) local magnification 北京科技大學(xué)本科生畢業(yè)設(shè)計（論文） 23 where A=A0ejθ0, s the start sample point in the zplane, W=W0ejφ0， where W0, the amplitude of W, controls the way the curve twist, and 0, the phase of W, indicates the angular frequency difference between adjacent samples, and M is the number of sample points to be analyzed,which may not equal to N. In a narrowband linearfrequency modulated (LFM) radar system, correlation pulse pression technique isusually used. And the correlation processor is oftenperformed digitally using FFT. Therefore, CZT can beused in place of FFT to finely describe the range gate containing targets. Thus, the peak value can be accurately obtained and its amplitude and phase errors can be greatly reduced. In this situation, the amplitudes of A and W are set to be 1, which indicate the analysis is on an arc of the unit circle. The phases of A and W, together with the parameter M, jointly control the start and ending points of the arc. The parameters should satisfy that the angular frequency of the range gate containing the targets is approximately in the middle of the arc, ., the angular M0 + 9m frequency should approximate 2 Fig. 2 shows the CZT results of the echo data in the range gates of that shown in Fig. 1 (b). The amplitudes are normalized. It can be seen from Fig. 2 that, the sample point of 301st range gate isn39。30 (for example 20), which corresponds to several kilometers39。Nanjing Research Institute of Electronics Technology Nanjing, Jiangsu 210013, P. R. China 2Nanjing University of Science amp。應(yīng)嚴(yán)格區(qū)分自己的成果與他人（特別是導(dǎo)師的）科研成果的界限。不要簡單重復(fù)、羅列實驗結(jié)果，要認(rèn)真闡明本人在結(jié)業(yè)工作