Research On Near-field 3-D Synthetic Aperture Radar Imaging And Target Extraction Technology

Recently years,near-field 3--D synthetic aperture radar(SAR)has been used in se-curity inspection,radar cross section measurement and other fields.These applications bring higher requirements for the resolution,imaging efficiency and target extraction ac-curacy.The linear frequency modulated(LFM)signal is a traditional transmit signal of SAR imaging system,however,which is easily limited by the performance of radio fre-quency system.Its range resolution is easily limited by the bandwidth of transmit signals.Some scholars have studied the stepped frequency linear frequency modulated(SFLFM)signal composed of multiple wide bandwidth sub-pulses to achieve high range resolution,and all sub-pulse echo data are fused by bandwidth synthesis method.However,the large bandwidth sub-pulses result in a high sampling rate,which causes a large amount of echo data.The bandwidth synthesis algorithm further increases the imaging time consumption.Hence,this dissertation studies the fast high range resolution 3--D SAR imaging algorithm.The background in 3--D SAR imaging result is complex,which is composed of noise,side lobes,multipath interference and other factors,and seriously affects the quality of imag-ing results and the measurement accuracy of target scattering characteristics.Therefore,it is needed to study how to strip the background and extract the target area from the 3--D SAR imaging results.The target extraction is usually achieved by the image segmen-tation.Combined with the characteristics of 3--D SAR images and image segmentation algorithms in other fields,this dissertation conducts in-depth research on the technology of 3--D SAR image target extraction.The main contributions and innovations include:(1)A 3--D VCFBP algorithm based on IAFHLFM is proposed which realizes fast high resolution 3--D SAR imaging.For the high requirement of receiver sampling rate,an inter-array hopping LFM signal model is proposed.The proposed signal transmits only one narrow bandwidth sub-pulse with hopping carrier frequency in each array element.Narrow bandwidth allows the receiver to sample the echo data at a low frequency.Lower sub-pulse number and sampling frequency make the amount of echo data significantly reduced.To improve the imaging efficiency,a 3--D VCFBP algorithm is proposed.The BP algorithm can perform accurate phase compensation and accumulation on the echo data.The proposed algorithm not only completes the image focusing of the along-track direction and cross-track direction,but also fuses the low range resolution imaging results of sub-pulses with different carrier frequencies into the high range resolution imaging re-sults.Hence,the proposed algorithm completes bandwidth synthesis and 3--D SAR imag-ing simultaneously,the computational efficiency is improved.The experimental results demonstrate that the proposed algorithm achieves high range resolution imaging,and its imaging time is only 2.35% ? 2.74% of that of the comparison algorithm.(2)A global seed region growing(GSRG)algorithm is proposed,which realizes the multi-region 3--D SAR image target extraction.When the seed region growing(SRG)al-gorithm in other fields is applied in the SAR field,the problems such as difficulty in initial seed selection and the spatial discreteness of the target region lead to serious target region missing and over-segmentation in the extraction results.Therefore,this dissertation pro-poses a GSRG algorithm.Firstly,the proposed algorithm uses the weighted Otsu method to automatically generate multiple initial seeds.The seeds grow in and cover different target areas,so the difficulty of selecting initial seed and the problem of missing target areas are solved.Then,the seed state matrix,growing rate function,and growing state matrix are introduced to refine the growing process.The threshold is adjusted reasonably to control the growing process.Thereby,the over-segmentation is suppressed and the target extraction accurate is improved.The experimental results indicate that the relative foreground region error and intersection of union of the proposed algorithm are 0.2290 and 0.4983,which are 31.66% and 29.93% higher than the SRG algorithm.(3)A region adaptive morphological reconstruction fuzzy C-means(RAMRFCM)algorithm is proposed,which realizes the 3--D SAR image target extraction with anti-multipath interference ability.Multipath interference is usually difficult to eliminate,and it is easy to cause over-segmentation in the 3--D SAR image target extraction.This disser-tation researsches that the edge gradient of multipath interference is weaker than that of the target in the SAR imaging results,and a strategy is proposed to first strip the interfer-ence and then accurately extract the target.The RAMRFCM algorithm is proposed based on this strategy.Firstly,the anisotropic diffusion algorithm is used to smooth the image to suppress noise,and the target edge is extracted while removing the interference edge by the 3--D Kirsch operator and the hysteresis threshold algorithm.Thereby,a rough target area is extracted,and the separation of target and multipath interference is realized.Then,the target area is accurately extracted by the adaptive morphological reconstruction fuzzy C-means algorithm with strong anti-noise ability and high image segmentation accuracy.The proposed algorithm not only overcomes the problem of multipath interference,but also reduces the image size processed by the image segmentation algorithm and improves the computational efficiency.The experimental results demonstrate that the accuracy,pre-cision and Dice similarity coefficient of the proposed algorithm reach 0.9993,0.8198 and0.7978,which are 10.86%,123.87% and 213.97% higher than GSRG.Moreover,the time consumption is only 10.85% of the comparison algorithm.(4)A region of interest extraction with adaptive threshold(REAT)algorithm is pro-posed,which not only has high-precision multipath interference stripping ability,but also can flexibly introduce multiple image segmentation algorithms in the other fields to achieve high-precision target extraction.Based on the advanced strategy of stripping the interference first and then accurately extracting the target,accurately separating the target and the interference has become the key problem of 3--D SAR image target extrac-tion.In order to achieve high-precision interference stripping in the face of the images obtained from different SAR systems and more complex imaging scenarios,a REAT al-gorithm is proposed in this dissertation.Firstly,the proposed algorithm increases the amplitude difference between the target and the background through the saliency detec-tion algorithm.Then,the thresholds required to process images are obtained through the adaptive threshold calculation method based on image features,which makes the edge detection algorithm to extract the target edge more accurately.Based on the proposed algorithm,a framework for 3--D SAR image target extraction is established,which can flexibly introduce image segmentation algorithms from other fields to complete accurate target region extraction.The experimental results demonstrate that the interference strip-ping ability and computational efficiency of the proposed algorithm are better than those of the RAMRFCM algorithm.After using the proposed framework to enhance three image segmentation algorithms,the target extraction accuracy and computational efficiency are greatly improved.The experiments not only verify the high-precision interference strip-ping ability of the proposed algorithm but also the flexibility of the proposed framework.In summary,this dissertation proposes a fast high range resolution 3--D SAR imaging algorithm which decreases the time consumption of high range resolution imaging.This dissertation also proposes two 3--D SAR image target extraction algorithms and a target extraction framework to strip the background in 3--D SAR images.