Connect public, paid and private patent data with Google Patents Public Datasets

Super broad band land radar automatic target identification method based on information fusion

Info

Publication number
CN1332220C
CN1332220C CN 200410025217 CN200410025217A CN1332220C CN 1332220 C CN1332220 C CN 1332220C CN 200410025217 CN200410025217 CN 200410025217 CN 200410025217 A CN200410025217 A CN 200410025217A CN 1332220 C CN1332220 C CN 1332220C
Authority
CN
Grant status
Grant
Patent type
Prior art keywords
target
broad
fusion
band
information
Prior art date
Application number
CN 200410025217
Other languages
Chinese (zh)
Other versions
CN1595195A (en )
Inventor
李建勋
郑军庭
Original Assignee
上海交通大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Grant date

Links

Abstract

一种基于信息融合的超宽带探地雷达自动目标识别方法,首先利用探地雷达回波信号中的直达波相对目标信号有一个较大的时间差,进行直达波的剔除,利用宽相关处理进行滤波和典型数据提取,提高信号的信噪比,提取纵向和横向典型数据用于目标形状识别,提取典型回波道数据进行Welch功率谱估计,并利用RBF网络进行目标材质分类,最后把目标形状识别和材质识别的结果进行信息融合,达到对不同形状,不同材质的地下目标的全面有效的自动识别。 An ultra-wideband ground penetrating radar automatic target recognition method based on information fusion, firstly GPR echo signals relative to the direct wave signal has a certain time a large difference, for the direct wave removed, with a wide filter correlation processing and typical data extraction, to improve the signal to noise ratio, extracting data for representative longitudinal and transverse shape recognition target, the echo extraction channel data Welch typical power spectrum estimates, and using the target material classification RBF network, the final target shape recognition and the results of the identification of material information integration, to achieve comprehensive and effective automatic identification of different shapes, different materials of underground target. 本发明实现了超宽带探地雷达目标的全面自动识别,对于实际的应用系统,特别是手持机具有重要意义和实用价值。 The invention realizes the full automatic identification of ultra-wideband ground penetrating radar targets, for practical applications, in particular handset has important significance and practical value.

Description

基于信息融合的超宽带探地雷达自动目标识别方法 Automatic target recognition method based on information fusion ultra wideband Penetrating

技术领域 FIELD

本发明涉及一种基于信息融合的超宽带探地雷达自动目标识别方法----基于宽相关处理、韦尔奇(Welch)功率谱分析、径向基函数(RBF)神经网络以及形状特征数据对目标进行全面自动识别,可广泛应用于地下金属/非金属管道探测、考古遗址定位、地质剖面勘探、高速公路质量检查以及安全检查等国家安全和经济领域中。 The present invention relates to an ultra-wideband ground penetrating radar automatic target recognition information fusion method based on broad-based correlation process ---- Welch (Welch) power spectrum analysis, radial basis function (RBF) neural network, and shape feature data full automatic target recognition, can be widely used in underground metal / nonmetal pipeline detection, locating archaeological sites, geological exploration profile, highway quality inspection and safety inspection of national security and economic fields.

背景技术 Background technique

探地雷达作为非破坏性探测手段正被广泛应用于地下目标(如空洞、管道、地雷等)的探测,如何对雷达回波信号进行处理以识别地下埋设的目标始终是困扰探地雷达应用的难题。 GPR as a non-destructive means of detecting the target is being widely used in underground (such as voids, pipelines, mines, etc.) probe, radar echo signals how processed to identify a target buried underground always plagued the GPR problem. 目前主要的处理手段包括成像识别和特征变量识别。 At present, the processing means includes an imaging recognition and feature identification variables.

成像处理通过对探地雷达回波信号的处理,获取了埋藏物体几何特征,从而可以根据几何特征(主要是外形)对目标加以判别,主要以合成孔径雷达(SAR)成像为主。 Imaging processing by the processing of the ground penetrating radar echo signals, acquired buried object geometry features can be (mainly shape) to be determined according to the geometric characteristics of the target, mainly synthetic aperture radar (SAR) imaging based. 实现的方法包括三维距(Stanislav Vitebskiy,Lawrence Carin and MarcA.Ressler,Ultra-wideband,short-pulse Ground-penetrating radar:simulation andmeasurement.IEEE Trans.On geoscience and remote sensing.35(3),1997,762-772)和相位处理(Sai,B.;Ligthart,LP;GPR Phase-Based Techniques forProfiling Rough Surfaces and Detecting Small,Low-Contrast Landmines Under FlatGround Geoscience and Remote Sensing,IEEE Transactions on,Volume:42,Issue:2,Feb.2004 Pages:318-326)。 Implemented method comprising a three-dimensional distance (Stanislav Vitebskiy, Lawrence Carin and MarcA.Ressler, Ultra-wideband, short-pulse Ground-penetrating radar: simulation andmeasurement.IEEE Trans.On geoscience and remote sensing.35 (3), 1997,762- 772) and a phase processing (Sai, B; Ligthart, LP; GPR phase-Based Techniques forProfiling Rough Surfaces and Detecting Small, Low-Contrast Landmines Under FlatGround Geoscience and Remote Sensing, IEEE Transactions on, Volume:. 42, Issue: 2, Feb.2004 Pages: 318-326). 由于大地的衰减和色散特性,使得探地雷达回波相互间具有不一致性,获取清晰的图像相对比较困难,从而造成很高的虚警率。 Due to attenuation and dispersion characteristics of the earth, so that the Ground Penetrating Radar another by an inconsistency, acquiring a clear image is relatively difficult, resulting in a high false alarm rate. 同时成像识别忽略了信号中原有的其它特征信息,尤其是比较难于区分形状相似的目标。 Identification signal while the imaging ignoring any original information other features, especially the more difficult to distinguish similar shape object. 同时成像处理对实验设备要求高、计算复杂,不易实时处理。 While high image processing requirements of the experiment apparatus, the computational complexity, real-time processing difficult. 处理结果一般由人工加以解释,含有较多的主观因素。 Processing results are generally explained by manual, it contains more subjective factors.

基于特征变量识别主要是利用探地雷达的回波信号进行特征变量的提取,借助神经网络完成自动目标识别。 Based on the identification characteristic variable is the use of ground penetrating radar echo signals are extracted characteristic variable, the neural network is completed by the automatic target recognition. 已有的相关探地雷达特征提取方法包括:连续子波变换(T.Le-Tien,H.Talhami and DTNguyen,“Target SignatureExtraction Based on the Continuous Wavelet Transform in Ultra-WidebandRadar,”IEE Electronics Letters,Vol.33,Issue 1,January 1997),和时频分析(Guillermo C.Gaunaurd,Hans C.Strifors,Applications of(Wigner-Type)Time-Frequency Distributions to Sonar and Radar SignalAnalysis,7th.International Wigner Symposium held in College park,MD USA,2001)等。 Related GPR conventional feature extraction methods include: continuous wavelet transform (T.Le-Tien, H.Talhami and DTNguyen, "Target SignatureExtraction Based on the Continuous Wavelet Transform in Ultra-WidebandRadar," IEE Electronics Letters, Vol. 33, Issue 1, January 1997), and frequency analysis (Guillermo C.Gaunaurd, Hans C.Strifors, Applications of (Wigner-Type) time-frequency Distributions to Sonar and Radar SignalAnalysis time, 7th.International Wigner Symposium held in College park , MD USA, 2001) and so on. 已有的方法主要是依据两维功率谱进行识别,特征量复杂不便于识别的工程识别,同时由于特征变量识别主要强调回波信号的特性,对于不同形状目标的识别却无能为力。 Conventional method mainly based on two-dimensional power spectrum recognition, the feature amount is not complicated construction facilitates recognition of identification, and because the main emphasis characteristic variable identifying characteristics of the echo signals, for identifying the different shape of the target can not do anything.

发明内容 SUMMARY

本发明的目的在于针对现有技术存在的不足,提供一种新的基于信息融合的超宽带探地雷达自动目标识别方法,即克服成像技术的设备要求高,不能区分形状相似目标的缺点,也克服了现有特征变量识别技术的复杂不易实现和对于不同形状目标的识别无能为力的不足,可以对不同形状、不同材质的地下目标进行有效的自动识别,达到工程化的实用效果。 Object of the present invention is for deficiencies of the prior art, to provide a new ultra-wideband ground penetrating radar automatic target recognition method based on information fusion, i.e. to overcome the high equipment requirements imaging techniques can not distinguish certain disadvantages similar shape, but also overcomes the prior art to identify characteristic variable and complex to implement for lack of identification of different shapes powerless object, can be effectively automatic identification of different shapes, different materials subsurface target, to achieve practical results engineered.

为实现这样的目的,本发明的技术方案中,首先对超宽带探地雷达回波信号进行直达波的剔除,利用宽相关处理进行信号滤波和典型数据提取。 To achieve this object, the technical solution of the present invention, the first UWB ground penetrating radar echo signals excluding the direct wave, and the filtered signal is typically associated with a wide data extraction process. 提取纵向和横向典型数据用于目标形状识别;提取典型回波道数据并进行韦尔奇(Welch)功率谱分析,并利用RBF神经网络对目标材质进行分类,最后把目标形状识别和材质识别的结果进行信息融合,从而实现目标的全面自动识别。 Typical extraction longitudinal and transverse shape recognition for the target data; typical echo extraction channel data and Welch (Welch) power spectrum analysis, and the target material using the RBF neural network classification, and finally the material of the target shape recognition and recognition information fusion results in order to achieve full automatic target recognition.

本发明的基于信息融合的超宽带探地雷达自动目标识别方法包括如下具体步骤:1.数据处理数据处理主要包括直达波剔除和信号滤波,用于提取典型纵向和横向切面数据和典型道数据。 Ultra-wideband ground penetrating radar automatic target recognition method based on information fusion of the present invention includes the following specific steps: 1. processing data including data processing and signal filtering excluding direct waves, typically for extracting data and vertical and horizontal section a typical data track. 将探地雷达的三维回波数据进行横向和纵向方向的平均,获取垂直方向的平均回波数据,从中选择第二和第三个回波的连接点作为截断点进行数据截断,抑制直达波,剔除前面的回波数据部分,将余下的回波数据作为含信号的数据进行后续处理,对截断后的探地雷达回波数据进行宽相关处理,得到三个典型切面和三个回波信号的最大值点处的X、Y、Z坐标值,此坐标值是以探地雷达为原点建立的三维坐标系中的坐标值。 The GPR three-dimensional echo data for transverse and longitudinal directions of the average, the average in the vertical direction acquired echo data, select a connection point of the second and third echoes as a cutoff point data truncation, suppressing the direct wave, excluding the front portion of the echo data, the remaining echo signals containing data as data for subsequent processing, on ground Penetrating radar data related to a wide-cut process, and cut to give three three typical echo signal at the maximum point X, Y, Z coordinate values, the coordinate value is a coordinate value of GPR origin establishing a three-dimensional coordinate system.

由于探地雷达回波信号由收发天线间直接耦合波、地面反射波、地下介质不连续产生的后向散射波、随机干扰等构成。 Since GPR echo signals directly from the coupling between the transmitting and receiving antennas wave, ground reflected wave, after the subsurface discontinuities constituting backscattered waves generated random disturbance. 由直接耦合波和地面反射波组成的直达波直接影响回波目标信号。 By the direct coupling of the reflected wave and ground wave component of the direct wave directly affect a target echo signal. 由于直达波相对目标信号有一个较大的时间差,因此本发明通过数据时间轴截断抑制直达波。 Since the direct wave signal has a relatively large target time difference, thus the present invention is by inhibition of the direct wave data truncation axis.

信号滤波采用宽相关处理方法实现。 Signal filtering broad correlation processing methods. 对截断后的探地雷达回波数据进行宽相关处理,可以提高回波信号的信噪比。 GPR echo data related to a wide-cut process, can improve the signal to noise ratio of the echo signal. 宽相关处理的主要思想就是通过引入伸缩因子,所得的回波信号与伸缩的母波具有匹配关系。 The main idea is that correlation processing width by introducing the factor of stretching, the resulting echo signals and having a telescoping mother wavelet matching relationship. 经过宽相关处理后,可以得到三个典型切面和三个回波信号的最大值点(X,Y,Z)。 After the correlation processing width can be obtained maximum point (X, Y, Z) and three cut three typical echo signal.

2.特征提取特征提取主要包括两部分:用于目标形状识别的纵向和横向典型数据的提取和用于目标材质识别的典型道数据的提取。 2. The feature extraction feature extraction consists of two parts: a typical longitudinal and transverse for extracting data of the target shape recognition and for extracting data of the target material typical channel identification. 根据宽相关处理后得到的回波信号最大值点处的X、Y值,得到对应的纵向切面和横向切面数据,再取切面图最大值附近的各道数据对应的最大值,得到两个切面的轮廓点,得到用于形状识别的特征数据,确定不同的X、Y值,得到对应的纵向切面和横向切面交点的典型道数据,然后经Welch功率谱处理后,可以得到用于材质识别的数据。 The X-echo signal at a point of maximum width of the obtained correlation processing, Y-values, to obtain a corresponding longitudinal section and transverse section data, and then take the maximum value in each section of FIG channel data corresponding to the vicinity of the maximum value, to obtain two cut contour points, to obtain characteristic data for shape recognition, determining different X, Y value, obtain a representative channel data corresponding to the longitudinal section and transverse section of the intersection, and then after treatment Welch power spectrum can be obtained for material recognition data.

经过宽相关处理后,可以得到三个典型切面和三个回波信号最大值点处的X,Y,Z值。 After the correlation processing width can be obtained at the three X, and typical three section maximum point of the echo signal, Y, Z value. 其中一个是水平切面,显示目标反射面的形状信息,一个纵向切面和一个横向切面,纵向切面的典型数据和横向切面的典型数据相结合用于目标形状的识别;最大值X、Y对应的宽相关处理数据代表回波的典型数据,用于目标材质的识别。 Where a is a horizontal section, the reflecting surface shape of the display of the information of the target, a longitudinal section and a transverse section, longitudinal section and transverse typical typical data combination section data identifying a target shape; maximum X, Y corresponding to the width a typical representative of data related to echo data processing for the recognition target material.

根据宽相关处理后得到的回波信号最大值点处的X、Y值,得到对应的纵向切面和横向切面数据,再取切面图最大值附近的各道数据对应的最大值,得到两个切面的轮廓点,得到用于形状识别的特征数据。 The X-echo signal at a point of maximum width of the obtained correlation processing, Y-values, to obtain a corresponding longitudinal section and transverse section data, and then take the maximum value in each section of FIG channel data corresponding to the vicinity of the maximum value, to obtain two cut contour points, to obtain characteristic data for shape recognition. 根据两道数据的相似性进行目标形状的识别。 Identifying similarity of the target shape data according to two.

确定不同的X、Y值,得到对应的纵向切面和横向切面交点的典型道数据,然后经Welch功率谱处理后,可以得到用于材质识别的数据。 Determine the different X, Y value, obtained longitudinal section corresponding to data channel and typical transverse section of the intersection, and then Welch power spectrum after processing, the data can be used to identify the material.

基于宽相关处理所得到的最大值X、Y以及宽相关处理的三维结果,提取对应于(X,Y)的单道宽相关处理数据形成典型道回波数据。 Correlation processing based on the maximum width X of the obtained three-dimensional correlation processing result Y and the width of the extracts corresponding to the (X, Y) of the width of a single channel data forming correlation processing echo data representative track. 由于探地雷达回波信号的非平稳性,尤其是对于超宽带瞬态电磁散射信号,传统的基于傅立叶变换的谱估计方法都将不能使用。 Since the non-stationary nature of the ground penetrating radar echo signals, in particular for ultra-wideband signal transient electromagnetic scattering, conventional spectral estimation methods based on Fourier transform will not be used. 考虑部分扫描的Welch平均重叠周期谱可以较好的用于非平稳信号的处理和一维的数据量,可以较好的用于目标特征的提取。 And Welch considering partial scans overlap period profile may be preferred for processing and the data amount of non-stationary one-dimensional signals, can be well used for extracting the target feature.

将提取的典型道数据经Welch功率谱处理即可得到一维的功率谱,进而用于材质的识别。 Typical channel data extracted by Welch power spectrum processing may be one-dimensional power spectrum, and further to identify the material.

3.分类识别将得到的形状识别特征数据进行曲线拟合,比较不同曲线对应的平方差,来确定拟合结果,利用不同形状目标回波信号对应不同的拟合曲线,并结合切面图显示,实现目标形状的识别;利用径向基函数RBF神经网络对目标材质进行分类,将与不同材质对应的典型道数据经Welch功率谱估计,得到用于材质识别的样本数据,送入径向基函数RBF神经网络进行训练建立特征量与目标值的函数关系,将上一步特征提取得到的用于材质识别的数据作为特征量输入神经网络,实现目标材质的自动识别;最后把目标形状识别和材质识别的结果进行信息融合,实现对不同材质,不同形状目标的全面自动识别。 3. The shape recognition feature data classification was subjected to curve fitting, curve corresponding to the square of the difference compare to determine the fitting results, the use of different shapes corresponding to the target echo signal curve fitting, in conjunction with FIG display section, achieve recognition target shape; using the radial basis function RBF neural network to classify certain material, different materials corresponding to the channel data by Welch typical power spectrum estimates, to obtain the sample data for identifying the material, into the radial basis function RBF neural network is trained with the feature quantity established as a function of the target value, the feature extraction step to obtain the data for identifying the material characteristic input as the neural network, automatically identifying a target material; the final target shape recognition and recognition material the result of information integration, to achieve full automatic identification of different materials, different shapes goals.

利用特征提取得到的轮廓点的数据进行一次曲线和二次曲线拟合,比较两次拟合曲线的平方差,来确定拟合结果是直线还是二次曲线。 Using the feature points extracted contour data obtained once a quadratic curve fit curve and comparing the two squared difference curve fitting, to determine fitting result is a line or quadratic curve. 并结合三维显示中的纵向和横向典型切面结果,不同形状物体的两个典型切面的典型道数据的分布形状的不同。 Different shapes of typical distribution channel data and combined vertical and horizontal section a typical result of the three-dimensional display, typically two objects of different shapes of the cut surface. 如果两个切面数据拟合都是二次的,显示为两个高峰,对应为球;如果一个为一次的,一个为二次的,显示一个为高峰分布,一个为不连续极值分布,则对应为管。 If the two are quadratic fit data slice, shown as two peaks, corresponding to a ball; if one of the primary, a secondary, and a display for the peak distribution, a discontinuous extreme value distribution, the corresponding to a tube. 这样可以实现目标形状识别。 This allows the target shape recognition.

利用径向基函数RBF神经网络对目标材质进行分类,首先将与不同材质对应的典型道数据经Welch功率谱估计,得到用于材质识别的样本数据,送入径向基函数RBF神经网络进行训练建立特征量与目标值的函数关系,将上一步特征提取得到的用于材质识别的数据作为特征量输入神经网络,实现目标材质的自动识别。 Using a radial basis function RBF neural network to classify the target material, the first material is different from the data channel corresponding to a typical power spectrum estimation by Welch, sample data obtained for material recognition, into radial basis function RBF neural network was trained establishing a function of the target value of the feature amount, the feature extraction step to obtain the data for identifying the material characteristic input as the neural network, automatically identifying a target material.

针对得到的典型道特征数据,利用径向基函数RBF神经网络对目标材质进行分类。 A typical channel characteristic data obtained using the radial basis function RBF neural network to classify the target material. 首先分别从测量数据选取典型的土壤、铁和PVC数据,分别通过直达波剔除、Welch功率谱估计得到典型特征用于神经网络训练的输入,同时将对应的目标信息——土壤、铁和PVC分别用不同的值表示形成训练的期望输出。 Typical soils were selected first, PVC and iron data from the measured data, are removed by the direct wave, the estimated input power spectrum Welch typical characteristic for the neural network training, while the information corresponding to the target - soil, PVC and iron were It represents the desired output of the training formed with different values. 当网络训练收敛以后的网络权值即代表了特征量与目标信息的映射关系。 When the network weights after training convergence network, which represents a mapping between feature quantity and target information. 针对特征提取的典型道数据的功率谱,通过训练收敛的神经网络即可进行目标材质的自动分类识别。 Power spectrum for a typical feature extraction channel data by training a neural network to converge for automatic classification of the target material.

最后把目标形状识别和材质识别的结果进行信息融合,可以实现对不同材质,不同形状目标的全面自动识别。 Finally, the target shape recognition and material identification result information fusion, can achieve full automatic identification of different materials, different shapes goals.

本发明的方法中,利用了探地雷达回波信号中的直达波相对目标信号有一个较大的时间差,进行直达波的剔除,并利用宽相关处理进行信号滤波和典型数据提取,提高了信号的信噪比。 The method of the present invention, use is made of the direct wave signal relative to the target ground penetrating radar echo signal has a large time difference, for the direct wave removed, and signal filtering and extracting with a wide typical data correlation process, improves signal the signal to noise ratio. 方法中提取纵向和横向典型数据用于目标形状识别,提取典型回波道数据并进行Welch功率谱分析,并利用RBF神经网络对目标材质进行分类,最后把目标形状识别和材质识别的结果进行信息融合,实现对不同材质,不同形状目标的自动识别。 The method of extracting data for representative longitudinal and transverse shape recognition target, typical echo extraction channel data and Welch power spectral analysis, and classification of the target material using the RBF neural network, and finally the material of the target shape recognition and recognition result information fusion, automatic identification of different materials, different shape of the target. 本发明的方法易于实现,即克服现有成像技术的设备要求高,不能区分形状相似目标的缺点,也克服特征变量识别技术的对于不同形状目标的识别无能为力的不足,为探地雷达的工程化提供了一个有效的技术实现方法。 The method of the present invention is easy to implement, i.e., high equipment requirements to overcome the prior art imaging, can not distinguish certain disadvantages similar shape, but also to overcome the disadvantages characteristic variable identification technology for identifying a target inability of different shapes, is engineered GPR It provides an effective technology implementation. 本发明对于实际的应用系统,特别是手持机具有重要意义和实用价值。 The present invention is useful for practical applications, in particular handset has important significance and practical value.

附图说明 BRIEF DESCRIPTION

图1为本发明基于信息融合的超宽带探地雷达自动目标识别的原理框图。 Figure 1 a block diagram of the principle of information fusion UWB ground penetrating radar automatic target recognition based on the present disclosure.

图2为不同形状物体的识别效果对照图。 FIG 2 is a different shape of the object recognition results with reference to FIGS.

其中,图2(a),(b),(c)为针对两根铁管的处理与显示对照图,图2(a)为原始数据显示,图2(b)为宽相关处理结果显示,图2(c)为三维显示;图2(d),(e),(f)为铝立方体的处理与显示对照图,图2(d)为原始数据显示,图2(e)为宽相关处理结果显示,图2(f)为三维显示。 Wherein FIG. 2 (a), (b), (c) is a processing for the two display control iron pipe, and FIG. 2 (a) is the original data, FIG. 2 (b) is a broad correlation processing result display, FIG. 2 (c) is a three-dimensional display; FIG. 2 (d), (e), (f) for the processing and display control in FIG aluminum cube, FIG. 2 (d) to the original data, FIG. 2 (e) is the width-related processing result display, FIG. 2 (f) three-dimensional display.

图3为不同材质的典型道数据的Welch功率谱对照图。 FIG 3 is a control diagram of a typical power spectrum Welch data channel different materials.

其中,图3(a)为典型道数据的Welch功率谱,图3(b)为PVC的典型道数据的Welch功率谱,图3(c)为土壤的典型道数据的Welch功率谱。 Wherein, the power Welch FIG 3 (a) is a typical data spectrum channel, the data channel is typically power Welch FIG. 3 (b) is the spectrum of PVC, FIG. 3 (c) is a typical channel data Welch power spectrum of soil.

具体实施方式 detailed description

为了更好地理解本发明的技术方案,以下结合附图对本发明的实施方式作进一步描述。 To better understand the technical solutions of the present invention, the following embodiment in conjunction with the accompanying drawings of embodiments of the present invention will be further described.

本发明基于信息融合的超宽带探地雷达自动目标识别的原理框图如图1所示,总共包括三个主要部分,即数据处理、特征提取和分类识别。 The present invention is based on the principle of information fusion UWB ground penetrating radar automatic target recognition diagram shown in Figure 1, a total of three main parts, i.e. the data processing, feature extraction and classification. 其中数据处理部分主要包括直达波剔除和采用宽相关处理方法实现信号滤波,用于提取典型横向和纵向切面数据和典型道数据。 Wherein the data processing section including the direct wave and the culling processing method implemented using the wide-correlation signal filtering, typically for extracting horizontal and vertical section and typical data channel data. 特征提取部分包括用于目标形状识别的横向和纵向典型数据的提取和用于目标材质识别的典型道数据的提取及提取后的功率谱估计。 Feature extraction section includes a typical power spectrum after extraction channel data of the target material identifying and extracting target for extracting horizontal and vertical shape recognition and for estimating the typical data. 分类识别部分利用横向和纵向两个典型数据完成目标形状的识别和分类,对得到的目标材质识别特征数据利用RBF神经网络对目标材质进行识别和分类。 Classification using both horizontal and vertical portions of a typical data complete identification and classification of the target shape, the target material recognition feature data obtained using the RBF neural network to identify and classify the target material. 最后把目标形状识别和材质识别的结果进行信息融合从而获得目标识别结果。 Finally, the material of the target shape recognition and recognition result information to obtain the fusion target recognition results.

各部分具体实施细节如下:1.数据处理针对每一道测试数据,可建立超宽带探地雷达回波模型如下:探地雷达超宽带天线发射的探测脉冲为r1(t)=x(t),则回波信号为:S(t)=S0(t)+Σj=1m+1ΣφKi,jx(si,j(t-τi,j))+Σj=1m+1Σφ-Ki,jx(si,j(t-τi,j))+n(t)]]>其中:S0(t)为直达波,i表示第i次反射波,j表示第j层反射波。 Specific implementation details of each part are as follows: 1. Data processing test data for each channel, may be established ultra-wideband Ground Penetrating Radar model is as follows: The Detection of transmitted UWB antenna is r1 (t) = x (t), the echo signals: S (t) = S0 (t) + & Sigma; j = 1m + 1 & Sigma; & phi; Ki, jx (si, j (t- & tau; i, j)) + & Sigma; j = 1m + 1 & Sigma; & phi; -Ki, jx (si, j (t- & tau; i, j)) + n (t)]]> where: S0 (t) is a direct wave, i denotes the i-th reflected wave, j represents layer j of the reflected wave. m表示地面距埋藏目标可分的层数。 m represents a certain ground separable from the buried layers. φ={i|τi,j∈目标回波信号宽度内}, {| Inner τi, j∈ target echo signal width i}, φ = 为φ补集。 Φ is the complement. n(t)为高斯噪声。 n (t) is Gaussian noise. Ki,j为衰减常数(对应反射系数),s1,m+1和τ1,m+1是待估计的未知参数,代表目标的时延、频谱展宽。 Ki, j is the attenuation constant (corresponding to the reflection coefficient), s1, m + 1 and τ1, m + 1 are unknown parameters to be estimated, on behalf of the target delay, spectral broadening.

经过直达波剔除后的回波信号可描述为:S′(t)=Σj=1m+1ΣφKi,jx(si,j(t-τi,j))+Σj=1m+1Σφ-Ki,jx(si,j(t-τi,j))+n(t)]]>在均匀介质条件下,忽略介质和多次反射波的影响,则用于目标检测和参数估计的有效回波信号可近似描述为:r2(t)=ΣiKi,Tx(si,T(t-τi,T))+n(t)]]>宽带相关处理器的输出为:WC(s,τ)=s∫r*1(s(t-τ))r2(t)dt]]>在非均匀介质情况下,通过多道数据纵向或横向平均,以纵向或横向分辨率的降低为代价换取正确的匹配和参数得稳健估计。 After the direct wave echo signal removed can be described as: S & prime; (t) = & Sigma; j = 1m + 1 & Sigma; & phi; Ki, jx (si, j (t- & tau; i, j)) + & Sigma; j = 1m + 1 & Sigma; & phi; -Ki, jx (si, j (t- & tau; i, j)) + n (t)]]> under uniform medium conditions, ignoring the medium and multiple influence of the reflected wave, the effective echo signal for target detection and parameter estimation may be described approximately as: r2 (t) = & Sigma; iKi, Tx (si, T (t- & tau; i, T)) + n (t)]]> broadband output correlation processor are: WC (s, tau &;) = s & Integral; r * 1 (s (t- & tau;)) r2 (t) dt]]> in the inhomogeneous medium case, the multi-channel data longitudinal or the average lateral, longitudinal or transverse direction in order to reduce the resolution in exchange for the right to match the cost and obtain robust estimation parameters.

经过宽相关处理后,可以得到三个典型切面和三个回波信号的最大值点(X,Y,Z)。 After the correlation processing width can be obtained maximum point (X, Y, Z) and three cut three typical echo signal. 一个是水平切面,显示目标反射面的形状信息,一个纵向切面和一个横向切面,纵向切面的典型数据和横向切面的典型数据相结合用于目标形状的识别。 It is a horizontal section, the reflecting surface shape of the display of the information of the target, a longitudinal section and a transverse section, longitudinal section of a typical data and typical transverse section of the combined data identifying a target shape. 两个切面交点的道数据代表回波的典型数据,用于目标材质的识别。 Road data representing the two echo data representative of an intersection section, for identifying the target material.

2. 2. 特征提取特征提取主要包括两部分:用于目标形状识别的纵向和横向典型数据的提取和用于目标材质识别的典型道数据的提取。 Feature extraction feature extraction consists of two parts: a typical longitudinal and transverse for extracting data of the target shape recognition and for extracting data of the target material typical channel identification.

经过宽相关处理后,可以得到回波信号最大值点处的X、Y、Z值,分别取X、Y值,可以得到对应的纵向切面和横向切面数据,再取切面图最大值附近的各道数据对应的最大值,得到两个切面的轮廓点,这样就得到了用于形状识别的典型道数据。 After the correlation processing width can be obtained by X, Y, the Z value of the echo signal at a maximum point, were taken X, Y values ​​can be obtained corresponding to the longitudinal section and transverse section data, and then take the vicinity of each cut maxima maximum channel data corresponding to the contour points obtained two facets, thus obtaining the channel data for the typical shape recognition.

部分扫描Welch功率谱被证明可以用于目标材质的有效识别,Welch法谱估计采取数据分段加窗处理再求平均的办法,先分别求出每段的谱估计,然后进行总平均。 Welch partial scan may be used to power spectrum proved effective recognition target material, the spectral estimation method taking Welch windowed data segments and then averaging approach, the first spectral estimate of each segment are determined, then the overall average. 根据概率统计理论证明,若将原长度为N的数据分成K段,每段长度取M=N/K,如各段数据互为独立,则估计的方差将只有原来不分段的1/K,达到一致估计的目的。 According to the theory of probability and statistics shown that if the original data length of N is divided into K segments, each segment lengths are M = N / K, such as the mutually independent segments of data, then the variance of the estimates will not only the original segment 1 / K , reach consensus estimate of purpose. 但若K增加、M减小,则分辨率下降。 However, if K increases, M is reduced, the resolution is reduced. 相反,若K减小、M增加,虽偏差减小,但估计方差增大。 Conversely, if K is reduced, increasing M, although the deviation decreases, but the estimation variance increases. 所以在实际中必须兼顾分辨率与方差的要求适当选取K与M的值。 Therefore, the resolution must be balanced with the requirements of the variance in practice appropriately selected value of K to M.

Welch功率谱估计的计算过程如下:设信号s(n)的长度为512,将其分成K=7段,每段长度为N=128,重叠50%。 Welch estimation of the power spectrum is calculated as follows: the length of a reset signal s (n) 512, K = 7 will be divided into segments, each of length N = 128, 50% overlap. 并对每个子集加上一个hanmin窗w(n)(n=128)。 And each subset plus hanmin a window w (n) (n = 128).

Welch功率谱估计按下式计算:Pw=1UKΣi=1kSi(w)Si*(w)]]>Si(w)=Si(n)w(n)e-2πmwn]]>U=1mΣn=0m-1w2(n)]]>图3为不同材质的典型道数据的Welch功率谱对照图,对比可以看到三者之间存在着较大的差别,因此可以用来作为目标的材质识别和目标的检测。 Welch power spectrum estimation is calculated as follows: Pw = 1UK & Sigma; i = 1kSi (w) Si * (w)]]> Si (w) = Si (n) w (n) e-2 & pi; mwn]]> U = 1m & Sigma; n = 0m-1w2 (n)]]> FIG. 3 is a comparison of FIG Welch power spectrum of a typical channel data of different materials, the contrast can be seen that there are large differences between the three, and therefore can be used as a target detection and identification of target materials. 确定不同的X,Y值,得到对应的纵向切面和横向切面交点的典型道数据,然后经Welch功率谱处理后,可以得到用于材质识别的数据. Determine the different X, Y value, obtained longitudinal section corresponding to data channel and typical transverse section of the intersection, and then Welch power spectrum after processing, the data can be used to identify the material.

3.分类识别本发明目标形状识别的试验采用的数据分别为针对球和管的测量数据。 3. The test data using a target shape recognition classification of the present invention were measured data for the ball and the tube. 实验的方法是首先针对测量的数据进行宽相关信号处理,获得水平切面图、横向切面图和纵向切面图。 The method of experiment was first performed broad correlation signal processing for the measured data, obtained horizontal section view and a transverse sectional view through a longitudinal section of FIG. 结合纵向和横向切面中的典型数据进行目标形状识别。 Combined longitudinal and transverse section of a typical data for target shape recognition.

经过宽相关处理,回波信号的信噪比得到了增强。 After broad correlation processing, the echo signal to noise ratio is enhanced. 利用特征提取得到的轮廓点的数据进行一次曲线和二次曲线拟合,比较两次拟合曲线的平方差,来确定拟合结果是直线还是二次曲线。 Using the feature points extracted contour data obtained once a quadratic curve fit curve and comparing the two squared difference curve fitting, to determine fitting result is a line or quadratic curve. 并结合三维显示中的纵向和横向典型切面结果,不同形状物体的两个典型切面的典型道数据的分布形状的不同。 Different shapes of typical distribution channel data and combined vertical and horizontal section a typical result of the three-dimensional display, typically two objects of different shapes of the cut surface. 如图2所示,如果两个切面数据拟合都是二次的,显示为两个高峰,对应为球;如果一个为一次的,一个为二次的,显示一个为高峰分布,一个为不连续极值分布,则对应为管。 As shown, if the two are quadratic fitting section data, the display 2 as two peaks, corresponding to a ball; if one of the primary, a secondary, showing a peak for the distribution, is not a continuous extreme value distribution, corresponds to the pipe. 这样可以实现目标形状识别。 This allows the target shape recognition.

本发明采用RBF径向基函数神经网络进行目标识别。 The present invention uses radial basis function RBF neural network target recognition. RBF选取具有单隐层的三层前馈网络,包括输入层、中间层和输出层。 RBF Select the previous single hidden layer having a three-layer feedforward network, comprising an input layer, an intermediate layer and output layer. 输入层个数的选取依据选取的特征向量的采样点数。 Select the number of the input layer according to the selection of the sampling points of the feature vector. 考虑回波信号中有用信息的长度,本采样点数取为128。 Regardless of the length of the useful information in the echo signal, the sampling points is taken as 128. 中间层个数的选取原则为2倍的输入层个数减去输出层个数。 The number of input layer principle of selecting the number of intermediate layer 2 by subtracting the number of times the output layer. 输出层个数为1,根据不同的应用分别用0,1,2代表待识别物体的种类---土壤、铁和PVC。 The number of output layer of 1, 0,1,2 respectively representative of the type of object to be recognized --- soil, iron and PVC depending on the application.

针对实际数据的宽相关处理结果,分别取土壤和目标上不同的X,Y值,将对应的不同的典型道数据经Welch功率谱估计,得到用于材质识别的样本数据,对比可以看到三者之间存在着较大的差别,因此可以用来作为目标的材质识别和目标的检测。 For wide data related to the actual processing result, were taken a different X, Y value and the target soil, corresponding to the different data channels by typical power spectrum estimation Welch, to give a sample of the material identifying data, comparison can see three there are large differences between persons, it can be used to detect the target material and the target recognition. 将功率谱特征量送入径向基函数RBF神经网络进行训练。 The power spectrum feature amount into the radial basis function RBF neural network is trained. 同时针对待识别的测量数据通过宽相关处理得到的回波信号最大值。 While measured data for the echo signal to be recognized by the maximum value obtained by the correlation processing width. 对应(X,Y)的典型道信号经过Welch功率谱估计,进而通过神经网络进行分类识别。 Typical channel signals corresponding to (X, Y) through Welch power spectrum estimation, and further classified by the neural network recognition. 根据网络的输出值的范围进行目标材质的自动识别。 Automatic recognition target range based on the output value of the material of network. 当输出值∈(-0.5,0.5),判定为土壤;当输出值∈(0.5,1.5),判定为铁;当输出值∈(1.5,2.5),判定为PVC;其它输出值,判定其它。 When the output value of ∈ (-0.5,0.5), it is determined that the soil; when the output value ∈ (0.5,1.5), it is determined that iron; when the output value ∈ (1.5,2.5), it is determined that of PVC; other output value, determining other.

如图3所示。 As shown in Figure 3. 对于伪铁管和PVC管的神经网络的训练与识别,输出结果为表1,反映Welch功率谱可以有效的借助神经网络完成对地下目标材质的识别。 For the training and recognition neural network pseudo iron pipe and the PVC pipe, output results in Table 1, reflecting the Welch power spectrum by means of the neural network can effectively complete the identification of the subsurface target material.

表1 Table 1

对比现有成像识别技术和特征变量识别,本发明可以有效地对不同形状,不同材质的地下目标进行有效的自动识别,能够达到工程化的实用效果。 Compare the prior art image recognition and identification characteristic variables, the present invention can be effectively effective automatic identification of different shapes, different materials subsurface target, it is possible to achieve practical results engineered. 同时从整个实现步骤可知,本发明的方法易于实现,从而为探地雷达的工程化提供了一个技术实现方法。 Meanwhile seen from the entire implementation step, the method of the present invention is easy to implement, providing a method for the technology engineered GPR.

Claims (1)

1.一种基于信息融合的超宽带探地雷达自动目标识别方法,其特征在于包括如下具体步骤:1)数据处理:包括直达波剔除和信号滤波,将探地雷达的三维回波数据进行横向和纵向方向的平均,获取垂直方向的平均回波数据,从中选择第二和第三个回波的连接点作为截断点进行数据截断,抑制直达波,剔除前面的回波数据部分,将余下的回波数据作为含信号的数据进行后续处理,对截断后的探地雷达回波数据进行宽相关处理,得到三个典型切面和三个回波信号的最大值点处的X、Y、Z坐标值,此坐标值是以探地雷达为原点建立的三维坐标系中的坐标值;2)特征提取:包括用于目标形状识别的纵向和横向典型数据的提取和用于目标材质识别的典型道数据的提取,根据宽相关处理后得到的回波信号最大值点处的X、Y值,得到对应的纵向切面和横向切面数据, CLAIMS 1. A method for ultra wideband automatic target recognition GPR based on information fusion, comprising the following specific steps: 1) Data processing: comprising a direct wave signal and reject filtering, the GPR transverse dimensional echo data and the mean longitudinal direction, obtaining an average of the echo data in the vertical direction, the second select data truncation and the connection point of the third cut-off point as echo suppressing the direct wave, excluding the front portion of the echo data, the remaining echo data for subsequent processing as a data signal containing, on ground Penetrating radar data related to a wide-cut process, to obtain the maximum point X at three typical three section and an echo signal, Y, Z coordinate value, the coordinate values ​​are coordinate values ​​GPR origin establishing a three-dimensional coordinate system; 2) feature extraction: extracting comprises means for longitudinal and transverse typical shape identification data of the target material and the target track identifying typical extracting data, the X-echo signal at a point of maximum width of the obtained correlation processing, Y-values, to obtain a corresponding longitudinal section and transverse section data, 取切面图最大值附近的各道数据对应的最大值,得到两个切面的轮廓点,得到用于形状识别的特征数据,确定不同的X、Y值,得到对应的纵向切面和横向切面交点的典型道数据,然后经Welch功率谱处理后,可以得到用于材质识别的数据;3)分类识别:将得到的形状识别特征数据进行曲线拟合,比较不同曲线对应的平方差,来确定拟合结果,利用不同形状目标回波信号对应不同的拟合曲线,并结合切面图显示,实现目标形状的识别;利用径向基函数RBF神经网络对目标材质进行分类,将与不同材质对应的典型道数据经Welch功率谱估计,得到用于材质识别的样本数据,送入径向基函数RBF神经网络进行训练建立特征量与目标值的函数关系,将上一步特征提取得到的用于材质识别的数据作为特征量输入神经网络,实现目标材质的自动识别;最后把目标形状识别 FIG section taken near the maximum data for each channel corresponding to a maximum value, to obtain contour points of the two facets to obtain characteristic data for shape recognition, determining different X, Y value, to obtain a corresponding longitudinal section and transverse section of the intersection typical channel data, and then Welch power spectrum after treatment, the material can be used to identify the data; 3) classification: the resultant shape recognition feature data curve fitting, comparing the squared difference corresponding to different curves, to determine fitting As a result, the use of different shapes corresponding to the target echo signal curve fitting, in conjunction with FIG display section, to achieve the target shape recognition; using radial basis function RBF neural network to classify the target material, typically the channel corresponding to the different materials Welch data through power spectrum estimation, obtained for the sample material identifying data, into the radial basis function RBF neural network is trained with the feature quantity established as a function of the target value, the feature extraction step on the data obtained for material recognition wherein the neural network as input, automatically identifying a target material; the final target shape recognition 材质识别的结果进行信息融合,实现对不同材质,不同形状目标的全面自动识别。 The results of material identifying information, and realize the full automatic identification of different materials, different shapes goals.
CN 200410025217 2004-06-17 2004-06-17 Super broad band land radar automatic target identification method based on information fusion CN1332220C (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN 200410025217 CN1332220C (en) 2004-06-17 2004-06-17 Super broad band land radar automatic target identification method based on information fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN 200410025217 CN1332220C (en) 2004-06-17 2004-06-17 Super broad band land radar automatic target identification method based on information fusion

Publications (2)

Publication Number Publication Date
CN1595195A true CN1595195A (en) 2005-03-16
CN1332220C true CN1332220C (en) 2007-08-15

Family

ID=34663608

Family Applications (1)

Application Number Title Priority Date Filing Date
CN 200410025217 CN1332220C (en) 2004-06-17 2004-06-17 Super broad band land radar automatic target identification method based on information fusion

Country Status (1)

Country Link
CN (1) CN1332220C (en)

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
DE102005024716A1 (en) 2005-05-30 2006-12-07 Robert Bosch Gmbh Method and apparatus for detection and classification of objects
CN102012509B (en) * 2010-09-28 2012-11-21 中国水电顾问集团华东勘测设计研究院 Radar detection method for occurrence of geological structure surface
CN102298160B (en) * 2011-05-18 2013-05-22 西安交通大学 Method for tracking data layer position of ground penetrating radar
CN102571230A (en) * 2011-12-22 2012-07-11 中国人民解放军总参谋部第六十三研究所 Distributed collaborative signal identification method based on blind estimation of higher order statistics and signal to noise ratio
CN103308899A (en) * 2013-05-23 2013-09-18 中国人民解放军第四军医大学 Biological radar human body target identification method based on zero crossing point technology
CN105319229A (en) * 2014-06-24 2016-02-10 国家电网公司 Defect detecting method and system of high voltage buried cables
CN105137498A (en) * 2015-09-17 2015-12-09 鲁东大学 Underground target detection and recognition system and method based on feature fusion

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5384715A (en) * 1993-08-27 1995-01-24 The Texas A&M Univeristy System System identification and analysis of subsurface radar signals
US5644314A (en) * 1996-03-29 1997-07-01 The United States Of America As Represented By The Secretary Of The Army Portable geophysical system using an inverse collocation-type metehodology
US5912639A (en) * 1997-05-23 1999-06-15 Power Spectra, Inc. Ground penetrating radar with synthesized end-fire array
US6222481B1 (en) * 1996-07-05 2001-04-24 Forsvarets Forskningsanstalt Method of detecting and classifying objects by means of radar

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US5384715A (en) * 1993-08-27 1995-01-24 The Texas A&M Univeristy System System identification and analysis of subsurface radar signals
US5644314A (en) * 1996-03-29 1997-07-01 The United States Of America As Represented By The Secretary Of The Army Portable geophysical system using an inverse collocation-type metehodology
US6222481B1 (en) * 1996-07-05 2001-04-24 Forsvarets Forskningsanstalt Method of detecting and classifying objects by means of radar
US5912639A (en) * 1997-05-23 1999-06-15 Power Spectra, Inc. Ground penetrating radar with synthesized end-fire array

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
基于模糊神经网络和D-S推理的智能特征信息融合研究 敬忠良,杨永胜,李建勋,戴冠中,信息与控制,第26卷第2期 1997;基于小波能量特征的探雷方法研究 王群,倪宏伟,徐毅刚,数据采集与处理,第18卷第2期 2003;利用复合特征进行模式识别的探雷研究 王群,倪宏伟,徐毅刚,应用科学学报,第21卷第1期 2003 *

Also Published As

Publication number Publication date Type
CN1595195A (en) 2005-03-16 application

Similar Documents

Publication Publication Date Title
Carin et al. Ultra-wide-band synthetic-aperture radar for mine-field detection
Brunzell Detection of shallowly buried objects using impulse radar
Verma et al. Analysis of clutter reduction techniques for through wall imaging in UWB range
US6222481B1 (en) Method of detecting and classifying objects by means of radar
US4896116A (en) Pulse radar method and apparatus for detecting an object
Al-Nuaimy et al. Automatic detection of buried utilities and solid objects with GPR using neural networks and pattern recognition
Ho et al. Discrimination mode processing for EMI and GPR sensors for hand-held land mine detection
Lahouar et al. Automatic detection of multiple pavement layers from GPR data
Savelyev et al. Investigation of time–frequency features for GPR landmine discrimination
Potin et al. Landmines ground-penetrating radar signal enhancement by digital filtering
Lopera et al. Filtering soil surface and antenna effects from GPR data to enhance landmine detection
Huang et al. A novel method for speckle noise reduction and ship target detection in SAR images
Baili et al. GPR signal de-noising by discrete wavelet transform
Torrione et al. Application of the LMC algorithm to anomaly detection using the Wichmann/NIITEK ground-penetrating radar
Soldovieri et al. Combination of advanced inversion techniques for an accurate target localization via GPR for demining applications
Stanley et al. Feature and decision level sensor fusion of electromagnetic induction and ground penetrating radar sensors for landmine detection with hand-held units
Sassen et al. 3D polarimetric GPR coherency attributes and full-waveform inversion of transmission data for characterizing fractured rock
Benedetto et al. Reliability of signal processing technique for pavement damages detection and classification using ground penetrating radar
He et al. Review of GPR rebar detection
Brandt et al. Detection of buried ice and sediment layers in permafrost using multi-frequency ground penetrating radar: A case examination on Svalbard
Eppelbaum et al. Application of informational and wavelet approaches for integrated processing of geophysical data in complex environments
Eyuboglu et al. Detection of water leaks using ground penetrating radar
Sato et al. Automatic signal processing of front monitor radar for tunneling machines
Schultz et al. Controlled GPR grave research: comparison of reflection profiles between 500 and 250 MHz antennae
Milisavljević et al. Improving mine recognition through processing and Dempster–Shafer fusion of ground-penetrating radar data

Legal Events

Date Code Title Description
C06 Publication
C10 Request of examination as to substance
C14 Granted
C17 Cessation of patent right