哈尔滨工程大学学报
哈爾濱工程大學學報
합이빈공정대학학보
JOURNAL OF HARBIN ENGINEERING UNIVERSITY
2015年
5期
616-622
,共7页
沉底目标%时频滤波%脊波变换%Hough变换%目标亮点
沉底目標%時頻濾波%脊波變換%Hough變換%目標亮點
침저목표%시빈려파%척파변환%Hough변환%목표량점
bottom target%time-frequent filtering%ridgelet transform%Hough transform%target highlight
主动声呐探测沉底目标时,由于存在严重的海底混响干扰,使得拷贝相关方法不能有效地提取目标回波亮点特征,大大降低了主动声呐的目标识别性能。针对该问题,研究了时频域滤波Hough变换的目标回波亮点特征提取方法。该方法采用互魏格纳变换将接收信号与拷贝的发射信号变换到时频域,依据目标回波与混响及噪声能量的时频聚集特性不同,采用时频域脊波变换滤波滤除混响及噪声,最后利用Hough变换提取目标回波中各亮点峰,并进行投影计算形成目标回波的亮点特征。同时研究了Hough变换域中的峰值位置与目标亮点对应的数学关系,给出了基于时频滤波提取沉底目标亮点特征的算法步骤,并采用支持向量机对比分析了该方法与拷贝相关方法提取亮点特征的识别效果。结果表明,该方法能够在较低的信混比下有效抑制混响干扰提取亮点特征,从而提高沉底目标的识别性能。
主動聲吶探測沉底目標時,由于存在嚴重的海底混響榦擾,使得拷貝相關方法不能有效地提取目標迴波亮點特徵,大大降低瞭主動聲吶的目標識彆性能。針對該問題,研究瞭時頻域濾波Hough變換的目標迴波亮點特徵提取方法。該方法採用互魏格納變換將接收信號與拷貝的髮射信號變換到時頻域,依據目標迴波與混響及譟聲能量的時頻聚集特性不同,採用時頻域脊波變換濾波濾除混響及譟聲,最後利用Hough變換提取目標迴波中各亮點峰,併進行投影計算形成目標迴波的亮點特徵。同時研究瞭Hough變換域中的峰值位置與目標亮點對應的數學關繫,給齣瞭基于時頻濾波提取沉底目標亮點特徵的算法步驟,併採用支持嚮量機對比分析瞭該方法與拷貝相關方法提取亮點特徵的識彆效果。結果錶明,該方法能夠在較低的信混比下有效抑製混響榦擾提取亮點特徵,從而提高沉底目標的識彆性能。
주동성눌탐측침저목표시,유우존재엄중적해저혼향간우,사득고패상관방법불능유효지제취목표회파량점특정,대대강저료주동성눌적목표식별성능。침대해문제,연구료시빈역려파Hough변환적목표회파량점특정제취방법。해방법채용호위격납변환장접수신호여고패적발사신호변환도시빈역,의거목표회파여혼향급조성능량적시빈취집특성불동,채용시빈역척파변환려파려제혼향급조성,최후이용Hough변환제취목표회파중각량점봉,병진행투영계산형성목표회파적량점특정。동시연구료Hough변환역중적봉치위치여목표량점대응적수학관계,급출료기우시빈려파제취침저목표량점특정적산법보취,병채용지지향량궤대비분석료해방법여고패상관방법제취량점특정적식별효과。결과표명,해방법능구재교저적신혼비하유효억제혼향간우제취량점특정,종이제고침저목표적식별성능。
When active sonar detects a bottom target, because of the disturbance from serious bottom reverberation, the copy?correlation method cannot effectively extract the highlight feature of the bottom target, which reduces the performance of target recognition. In this paper, the method for extracting the highlight feature of bottom target based on time?frequency domain filtering and Hough transforming is researched. In this method, firstly the received signal and copy of transmitted signal were transformed into time?frequency domain by means of cross Wigner?Ville transform. Next, the reverberation and noise were filtered by Ridgelet transform filter according to the different char?acters of energy distribution in time?frequency domain between the target echo and reverberation. Finally, the peaks of highlight were obtained using Hough transform and the feature vectors were extracted by projecting it intoρaxis. The mathematical relation between the target highlight and the position of peak value in Hough transform domain was researched at the same time. The steps for extracting the bottom target highlight features based on time?frequen?cy domain were given. The performance of recognition was compared between the new method and copy?correlation method using support vectors machine in simulation and real experiments. The results showed that the new method can reduce the reverberation and noise in time?frequency domain and thereby improve recognition performance with a low signal reverberation ratio.