阅读说明：本技术 一种基于乘幂法的低复杂度最小方差超声成像方法 (A kind of low complex degree minimum variance ultrasonic imaging method based on power method ) 是由 王平 孔露 王林泓 杜婷婷 孔美娅 李锡涛 柳学功 田训 石轶哲 于 2019-08-15 设计创作，主要内容包括：本发明涉及一种基于乘幂法的低复杂度最小方差超声成像方法,属于超声成像技术领域；该方法首先利用离散余弦变化,将超声回波信号由阵元域转换到波束域,提取部分波束域数据,降低样本协方差矩阵维数；其次,利用乘幂法对样本协方差矩阵求解最大的特征值及其对应的特征向量,忽略部分低能量回波信号数据,对协方差矩阵的求逆进行简化,构造新的最优加权矢量,达到降低算法复杂度的目的。本发明提出的算法成像效果优于传统的延时叠加算法(DAS),最小方差算法(MV),运行效率远高于最小方差算法(MV)和基于特征空间的最小方差算法(ESBMV),有效克服了传统自适应算法由于运行时间较长而限制了其应用范围的问题,具有较好的应用前景和价值。(The present invention relates to a kind of low complex degree minimum variance ultrasonic imaging method based on power method, belongs to ultrasonic imaging technique field；This method is transformed into Beam Domain by Element space first with long-lost cosine code, by ultrasound echo signal, extracts subwave beam numeric field data, reduces sample covariance matrix dimension；Secondly, maximum characteristic value and its corresponding feature vector are solved to sample covariance matrix using power method, ignores part low energy echo signal data, inverting for covariance matrix is simplified, new optimal weighting vector is constructed, achievees the purpose that reduce algorithm complexity.Algorithm imaging effect proposed by the present invention is better than traditional delay superposition algorithm (DAS), minimum variation algorithm (MV), operational efficiency is much higher than minimum variation algorithm (MV) and the minimum variation algorithm (ESBMV) based on feature space, effectively overcome traditional adaptive algorithm due to runing time is longer and the problem of limit its scope of application, with good application prospect and value.)

1. a kind of low complex degree minimum variance ultrasonic imaging method based on power method, it is characterised in that: this method includes following Step:

S1: filtering processing and AD conversion are amplified to the received echo-signal of ultrasound element, obtain Element space echo data；

S2: carrying out discrete cosine processing to Element space echo data, determines dimensionality reduction parameter by comparing, extracts subwave beam Domain echo data reduces covariance matrix dimension；

S3: being handled covariance matrix by power method, is obtained maximum characteristic value and its corresponding feature vector, is obtained High-intensity echo signal；

S4: ignore part low energy echo data, invert to covariance matrix and carry out abbreviation, the inversion operation of matrix is reduced to The multiplying of vector reduces the complexity level of matrix inversion；

S5: by the simplification Beam Domain covariance inverse matrix and Beam Domain direction vector of acquisition, the calculation of Beam Domain minimum variance is obtained The weight vector of method；

S6: the weight vector is projected to the corresponding feature vector of maximum eigenvalue, obtains the low complex degree based on power method The optimal weight vector of minimum variation algorithm；

S7: the optimal weight vector using the low complex degree minimum variation algorithm based on power method adds Beam Domain echo-signal Power summation, obtains the output signal of the low complex degree minimum variation algorithm based on power method.

2. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In step s 2, it specifically includes:

S21: constructing the transition matrix of (1+p) × L dimension using discrete cosine transform, and concrete form is as follows:

Matrix T meets TT^H=I, wherein T^HFor the associate matrix of T, I is unit battle array, and L is the array number of subarray, parameter p For dimensionality reduction parameter, the selection of p value is needed to meet (p+1) < L, is enabled dimensionality reduction parameter p=8 with the minimum principle of amount distortion；

S22: Element space echo-signal is obtained after carrying out space smoothing to echo-signal, multiplied by the available wave of transition matrix T The echo-signal in beam domain, extracting the sample covariance matrix that subwave beam data obtains becomes (p+ from original L × L dimension matrix 1) × (p+1) dimension matrix is converted as the following formula by taking first of subarray as an example:

Wherein k indicates k-th of sampled point,Indicate first of received echo-signal of Element space subarray, vector dimension is L × 1, T_(1+p)×LIndicate the transition matrix of (1+p) × L dimension,Indicate Beam Domain subarray echo-signal；

S23: after obtaining Beam Domain subarray echo-signal by S22, sample covariance matrix accordingly becomes:

WhereinFor the Beam Domain sample covariance matrix of (p+1) × (p+1) dimension, N indicates probe array number, T_(1+p)×LTable Show the transition matrix of (1+p) × L dimension,For the Element space sample covariance matrix of L × L dimension, E [] is indicated Covariance operation, x_nIndicate Element space echo-signal, []^HIndicate conjugate transposition operation.

3. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In step s3, it specifically includes:

S31: the simplified operation of Eigenvalues Decomposition is carried out to Beam Domain covariance matrix by improving power method, seeks covariance matrix Maximum eigenvalue feature vector corresponding with its, i.e. λ_maxAnd e_max, set e₀=[1,1 ..., 1], is iterated according to the following formula It solves:

Wherein, e_i+1It is the feature vector after iteration i+1 times, λ_i+1For its corresponding characteristic value, max () is greatest member solution Operation after each interative computation, is normalized into next iteration obtained vector, by taking i-th iteration as an example, Obtain vector e_iIt is normalized afterwards, i.e. e_i=e_i/max(e_i), interative computation is executed again after normalized, Interative computation terminate Rule of judgment be | λ_i+1-λ_i|/|λ_i+1| < ε selectes ε value according to required precision, takes ε=0.001.

4. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In step s 4, it specifically includes:

S41: the energy of the characteristic value signal of covariance matrix in ultrasonic signal, the corresponding signal energy of the bigger characterization of characteristic value Amount is stronger, and the corresponding feature vector of low energy part constitutes noise subspace and it is low to ignore part for simplification matrix inversion operation Energy signal, the corresponding characteristic value of noise subspace take identical value in the case where covariance matrix mark is constant, guarantee that ultrasound is returned The energy constant of wave signal, it may be assumed that

Wherein, q is the number of feature vector in signal subspace, and trace () asks mark operation for matrix；

S42: it enablesThen Beam Domain sample association side Poor inverse of a matrix matrix can be with abbreviation are as follows:

Wherein, I is unit matrix, e_iFor ith feature vector, λ_iFor its corresponding characteristic value；

S43: the solution procedure of further abbreviation S42 ignores part beam signal, q=1 is taken to be converted to the inversion operation of matrix The multiplying of vector, it may be assumed that

Above formula only gets maximum characteristic value feature vector corresponding with its, i.e. λ₁=λ_max, e₁=e_max,I is unit matrix, by matrix inversion operation It is reduced to the multiplying of vector, computational complexity is by Ο (L³) it is reduced to Ο ((p+1)²), wherein L is Element space subarray Element number of array, p be Beam Domain dimensionality reduction parameter.

5. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In step s 5, it specifically includes:

S51: the Beam Domain transition matrix obtained using S21 obtains the direction vector of Beam Domain:

a_b=Ta

Wherein, [1,1 ..., 1] a=^TIt is the Element space direction vector that N × 1 is tieed up, T is transition matrix, a_bFor Beam Domain direction to Amount；

S52: the Beam Domain direction vector that the Beam Domain sample covariance matrix inverse matrix and S51 acquired using S43 is acquired obtains The weight vector of low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibIt is the weight vectors of the low complex degree minimum variance ultrasonic imaging method based on power method,It is simplified Beam Domain covariance matrix inverse matrix, a_bFor Beam Domain direction vector.

6. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In step s 6, it specifically includes:

S61: the optimal weight vector expression formula of the low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibmvFor the optimal weighting vector of the low complex degree minimum variance ultrasonic imaging method based on power method, e_maxFor most The corresponding feature vector of big characteristic value, []^HIndicate conjugate transposition operation.

7. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: In the step s 7, adaptive beamformer output signal is calculated as follows:

In formula,For weight vector w_ibmvConjugate transposition, x_bIt (k) is the wave after the echo-signal space smoothing of sub-array beam domain Beam domain signal, y_ibmv(k) be mentioned algorithm beamformer output signal.

8. the low complex degree minimum variance ultrasonic imaging method according to claim 1 based on power method, it is characterised in that: The value of the submatrix array element number L is L≤N/2.

Technical field

The invention belongs to ultrasonic imaging technique field, be related to a kind of low complex degree minimum variance ultrasound based on power method at Image space method.

Background technique

Ultrasonic image-forming system mainly includes that transmitting module, receiving module, Wave beam forming module and last image are shown Module, wherein beam-forming technology is the core technology of imaging system, directly determines the focusing accuracy and imaging of ultrasonic imaging Quality, the essence of Wave beam forming is a kind of array-processing techniques, i.e., by handling array signal, effectively extracts expectation Signal filters out noise interferences, realizes the Visual retrieval to desired signal region.Currently, what is be most widely used is Delay superposition (Delay and Sum, DAS) beam-forming technology, essence are the delay times fixed to the setting of each array element, To realize transmitting and received focusing, by the superposition to echo-signal so that the ultrasound echo signal of focal point reaches most strong, Ultrasonic echo other than focus area mutually weakens the clear effective imaging even offset, and then realize focus point region, to area Overseas interference signal is inhibited.But tradition DAS algorithm has that imaging resolution, contrast are insufficient, in order to improve The imaging effect of DAS algorithm, Adaptive beamformer technology are come into being, and typical Adaptive beamformer technology includes minimum Variance algorithm and generalized sidelobe cancellation algorithm.

The core concept of minimum variation algorithm is: in the case where overall gain is constant on guaranteeing desired orientation, so that output The energy of signal reaches minimum, reduces the interfering noise signal in undesired direction with this, and then derive optimal weighted value. Based on the undistorted response Beam-former of minimum power, there is scholar to propose generalized sidelobe cancellation (Generalized Sidelobe Canceller, GSC) for algorithm for noise reduction, which includes filter and Beam-former, it can be divided into Upper and lower two channels, upper channel are non-adaptive channel, and lower channel is adaptive channel, which includes blocking matrix and adaptive It answers filter for minimizing output power and then inhibiting noise residual components, is overlapped finally by Beam-former to disappear Except unwanted signal, image quality is improved.

Adaptive beam-forming algorithm is promoted obviously, still compared to traditional DAS algorithm imaging resolution and contrast Since adaptive algorithm needs that the processing of echo data real-time perfoming is calculated dynamic importance degrees, existed in operational process a large amount of Complicated matrix operation, algorithm complexity is high, has seriously affected the real-time of ultrasonic imaging；Calculating process is cumbersome to make an uproar vulnerable to environment Acoustic jamming, algorithm robustness are poor.For this problem, Asl and Mahloojifar is proposed is constructed using Toeplitz matrix form Sample covariance matrix reduces the operand of matrix inversion to reduce algorithm complexity；There is scholar to propose by with QR points Solution technology realizes the fast inversion of sample covariance matrix；The scholars such as Kim utilize principal component analysis (Principal ComponentAnalysis, PCA) principle, it chooses the corresponding feature vector of the larger characteristic value of preceding m in covariance matrix and constitutes Original high dimensional data is mapped to obtain low-dimensional data by project, while remaining desired signal by dimensionality reduction matrix；Scholar mentions Contraction AF panel least-squares algorithm seeks reduced rank processing when having gone out the adaptive space based on the optimization of parameter vector Joint iteration It is best basis set；Albulayli is proposed using the method for mixed self-adapting and non-adaptive algorithm and is reduced algorithm The threshold value of CF coefficient is arranged in computational load first, if the CF value calculated in real time is more than the given threshold, i.e., wraps in echo-signal Containing more coherent signal, then CF coefficient is combined with adaptive algorithm, otherwise uses traditional DAS algorithm；There is scholar to mention Go out the concept of subspace MV Beam-former, which has ignored the partial row of data in original sample covariance matrix, So that covariance matrix is no longer square matrix, the weight vectors obtained by the covariance matrix can be in the base for reducing algorithm complexity Guarantee the imaging effect of algorithm on plinth.

For the high problem of algorithm complexity existing for minimum variation algorithm, the invention proposes low multiple based on power method Miscellaneous degree minimum variation algorithm.Covariance matrix dimension is reduced by discrete cosine transform, using power method to the characteristic value of matrix Decomposition operation is simplified, and is extracted maximum characteristic value feature vector corresponding with its, is ignored part low energy signal data, By the inversion operation approximation abbreviation of matrix it is the multiplying of vector, and then reduces the complexity of algorithm.Calculation is proposed in order to verify The validity of method, by Field II ultrasound emulation platform to DAS algorithm, MV algorithm, ESBMV algorithm and IBMV proposed in this paper Algorithm has carried out point target and sound absorption spot target imaging, and imaging results show: the imaging time and imaging effect of mentioned algorithm are equal Better than MV algorithm, is promoted obviously compared to ESBMV algorithm imaging efficiency, there is more preferably imaging resolution and algorithm robustness, Better imaging effect is able to maintain under different center frequency, it was confirmed that the superiority of proposed algorithm.

Summary of the invention

In view of this, the purpose of the present invention is to provide a kind of low complex degree minimum variance ultrasonic imaging based on power method Method, this method can be improved the operation efficiency of original minimum variation algorithm, and the resolution ratio and contrast of imaging are also much better than most Small variance algorithm effectively overcomes the problem of traditional adaptive algorithm is limited its scope of application since runing time is longer.

In order to achieve the above objectives, the invention provides the following technical scheme:

A kind of low complex degree minimum variance ultrasonic imaging method based on power method, method includes the following steps:

S1: filtering processing and AD conversion are amplified to the received echo-signal of ultrasound element, obtain Element space number of echoes According to；

S2: carrying out discrete cosine processing to Element space echo data, determines dimensionality reduction parameter by comparing, extracts part Beam Domain echo data reduces covariance matrix dimension；

S3: being handled covariance matrix by power method, obtains maximum characteristic value and its corresponding feature vector, Obtain high-intensity echo signal；

S4: ignoring part low energy echo data, invert to covariance matrix and carry out abbreviation, by the inversion operation letter of matrix The multiplying for turning to vector reduces the complexity level of matrix inversion；

S5: by the simplification Beam Domain covariance matrix inverse matrix and Beam Domain direction vector of acquisition, Beam Domain is obtained most The expression formula of small variance algorithm weight vector；

S6: the expression formula is projected to the corresponding feature vector of maximum eigenvalue, is obtained low multiple based on power method The optimal weight vector of miscellaneous degree minimum variation algorithm；

S7: using the low complex degree minimum variation algorithm based on power method optimal weight vector to Beam Domain echo-signal into Row weighted sum obtains the output signal of the low complex degree minimum variation algorithm based on power method.

Further, in step s 2, it specifically includes:

S21: constructing the transition matrix of (1+p) × L dimension using discrete cosine transform, and concrete form is as follows:

Matrix T meets TT^H=I, wherein T^HFor the associate matrix of T, I is unit battle array, and L is the array number of subarray, ginseng Number p is dimensionality reduction parameter, and the selection of p value needs satisfaction (p+1) < L to enable dimensionality reduction by comparing with the minimum principle of amount distortion Parameter p=8；

S22: Element space echo-signal is obtained after carrying out space smoothing to echo-signal, can be obtained multiplied by transition matrix T To the echo-signal of Beam Domain, extract the sample covariance matrix that subwave beam data obtains is become from original L × L dimension matrix (p+1) × (p+1) dimension matrix is converted as the following formula by taking first of subarray as an example:

Wherein k indicates k-th of sampled point,Indicate first of received echo-signal of Element space subarray, vector dimension Number is L × 1, T_(1+p)×LIndicate the transition matrix of (1+p) × L dimension,Indicate Beam Domain subarray echo-signal；

S23: after obtaining Beam Domain subarray echo-signal by S22, sample covariance matrix accordingly becomes:

WhereinFor the Beam Domain sample covariance matrix of (p+1) × (p+1) dimension, N indicates probe array number, T_(1+p)×LIndicate the transition matrix of (1+p) × L dimension,For the Element space sample covariance matrix of L × L dimension, E [] indicates covariance operation, x_nIndicate Element space echo-signal, []^HIndicate conjugate transposition.

Further, in step s3, it specifically includes:

S31: the simplified operation of Eigenvalues Decomposition is carried out to Beam Domain covariance matrix by improving power method, seeks covariance The maximum eigenvalue of matrix feature vector corresponding with its, i.e. λ_maxAnd e_max, set e₀=[1,1 ..., 1], carries out according to the following formula Iterative solution:

Wherein, e_i+1It is the feature vector after iteration i+1 times, λ_i+1For its corresponding characteristic value, max () is greatest member Derivation after each interative computation, is normalized into next iteration obtained vector, with i-th iteration For, obtain vector e_iIt is normalized afterwards, i.e. e_i=e_i/max(e_i), iteration is executed again after normalized Operation, interative computation terminate Rule of judgment be | λ_i+1-λ_i|/|λ_i+1| < ε selectes ε value according to required precision, take ε= 0.001。

Further, in step s 4, it specifically includes:

S41: the energy of the characteristic value signal of covariance matrix in ultrasonic signal, the corresponding letter of the bigger characterization of characteristic value Number energy is stronger, and the corresponding feature vector of low energy part constitutes noise subspace and ignores portion for simplification matrix inversion operation Divide low energy signal, the corresponding characteristic value of noise subspace takes identical value in the case where covariance matrix mark is constant, guarantees super The energy constant of sound echo-signal, it may be assumed that

Wherein, q is the number of feature vector in signal subspace, and trace () asks mark operation for matrix；

S42: it enablesThen Beam Domain sample The inverse matrix of covariance matrix can be with abbreviation are as follows:

Wherein, I is unit matrix, e_iFor ith feature vector, λ_iFor its corresponding characteristic value；

S43: the solution procedure of further abbreviation S42 ignores part beam signal, q=1 is taken to turn the inversion operation of matrix It is changed to the multiplying of a vector, it may be assumed that

Above formula only gets maximum characteristic value feature vector corresponding with its, i.e. λ₁=λ_max, e₁=e_max,I is unit matrix, by matrix inversion operation It is reduced to the multiplying of vector, computational complexity is by Ο (L³) it is reduced to Ο ((p+1)²), wherein L is Element space subarray Element number of array, p be Beam Domain dimensionality reduction parameter.

Further, in step s 5, it specifically includes:

S51: the Beam Domain transition matrix obtained using S21 obtains the direction vector of Beam Domain:

a_b=Ta

Wherein, [1,1 ..., 1] a=^TIt is the Element space direction vector that N × 1 is tieed up, T is transition matrix, a_bFor Beam Domain direction Vector；

S52: the Beam Domain direction vector that the Beam Domain sample covariance matrix inverse matrix and S51 acquired using S43 is acquired, Obtain the weight vector of the low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibIt is the weight vectors of the low complex degree minimum variance ultrasonic imaging method based on power method,It is letter The Beam Domain covariance matrix inverse matrix of change, a_bFor Beam Domain direction vector.

Further, in step s 6, it specifically includes:

S61: the optimal weight vector expression formula of the low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibmvFor the optimal weighting vector of the low complex degree minimum variance ultrasonic imaging method based on power method, e_max For the corresponding feature vector of maximum characteristic value, []^HIndicate conjugate transposition operation.

Further, in the step s 7, adaptive beamformer output signal is calculated as follows:

In formula,For weight vector w_ibmvConjugate transposition, x_b(k) for after the echo-signal space smoothing of sub-array beam domain Beam Domain signal, y_ibmv(k) be mentioned algorithm beamformer output signal.

Further, the value of the submatrix array element number L is L≤N/2.

The beneficial effects of the present invention are:

Present invention employs a kind of low complex degree minimum variance ultrasonic imaging algorithm based on power method, by multiple to calculating The highest matrix inversion operation of miscellaneous degree and Eigenvalue Decomposition operation are improved, and discrete cosine transform structural transform is introduced Echo data is transformed into Beam Domain by Element space, compares selected part wave beam numeric field data by matrix, guarantees the imaging effect of algorithm Fruit reduces sample covariance matrix dimension.And improved in terms of matrix inversion, pass through the special selection to characteristic value Matrix inversion operation is converted into vector multiplication operation, reduces the complexity level of matrix inversion.Using power method to matrix Eigenvalues Decomposition improves, and seeks maximum eigenvalue feature vector corresponding with its of covariance matrix, and it is complicated to reduce operation Degree.Point target, sound absorption spot target imaging the result shows that: compare MV algorithm, IBMV point target and sound absorption the spot target imaging time It is upper to reduce 23.92% and 27.65% respectively, and there is preferable imaging resolution and contrast；Compared with ESBMV algorithm, IBMV algorithm is in point target and 71.46% and 70.31% have been respectively shortened on the sound absorption spot target imaging time, point target imaging Resolution ratio is better than ESBMV algorithm and has preferable robustness；In different center frequency, IBMV can keep it is lower at As artifact, imaging resolution is better than other algorithms, has better practicability.

Other advantages, target and feature of the invention will be illustrated in the following description to a certain extent, and And to a certain extent, based on will be apparent to those skilled in the art to investigating hereafter, Huo Zheke To be instructed from the practice of the present invention.Target of the invention and other advantages can be realized by following specification and It obtains.

Detailed description of the invention

To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention is made below in conjunction with attached drawing excellent The detailed description of choosing, in which:

Fig. 1 is the flow chart of the method for the invention；

Fig. 2 is different dimensionality reduction comparative bid parameters；

Fig. 3 is noiseless point target imaging contrast figure；

Fig. 4 is plus noise point target imaging contrast figure；

Fig. 5 is that axial distance 40mm punishes resolution comparison diagram；

Fig. 6 is noiseless sound absorption spot imaging contrast figure；

Fig. 7 is plus noise sound absorption spot imaging contrast figure；

Fig. 8 is different center frequency imaging contrast figure；

Fig. 9 is center frequency 5.0MHz resolution ratio comparison diagram.

Specific embodiment

Specific embodiments of the present invention are described in detail with reference to the accompanying drawing.

Fig. 1 is algorithm flow chart of the invention, as shown in Figure 1, the present invention provides a kind of low complexity in ultrasonic imaging Spend minimum variation algorithm, comprising the following steps:

Step S1: amplifying filtering processing and AD conversion to the received echo-signal of ultrasound element, obtains Element space and returns Wave number evidence；

Step S2: discrete cosine processing is carried out to Element space echo data, dimensionality reduction parameter is determined by comparing, is extracted Part Beam Domain echo data reduces covariance matrix dimension, specifically includes the following steps:

S21: constructing the transition matrix of (1+p) × L dimension using discrete cosine transform, and wherein L is the array number of subarray, Concrete form is as follows:

Matrix T meets TT^H=I, wherein T^HFor the associate matrix of T, I is unit battle array, and parameter p is dimensionality reduction parameter, p value Selection with the minimum principle of amount distortion, guarantee the imaging effect of Beam Domain algorithm, in order to reduce the complexity of algorithm operation, need Meet (p+1) < L and dimensionality reduction parameter p=8 is enabled by comparing, the computational complexity of algorithm not only can be effectively reduced, and And scan-line data gap compared with initial data is smaller, it is ensured that the imaging effect of algorithm, therefore algorithm proposed by the present invention Take dimensionality reduction parameter p=8.

S22: to the Element space echo-signal obtained after space smoothing multiplied by time of available Beam Domain after transition matrix T Wave signal, extracting the sample covariance matrix that subwave beam data obtains becomes (p+1) × (p+1) from original L × L dimension matrix Dimension matrix is converted as the following formula by taking first of subarray as an example:

Wherein k indicates k-th of sampled point,Indicate first of received echo-signal of Element space subarray, vector dimension Number is L × 1, T_(1+p)×LIndicate the transition matrix of (1+p) × L dimension,Indicate Beam Domain subarray echo-signal.

S23: after obtaining Beam Domain subarray echo-signal by S22, sample covariance matrix accordingly becomes:

Wherein k indicates k-th of sampled point,For L × L dimension Element space sample covariance matrix, T_(1+p)×LIndicate the transition matrix of (1+p) × L dimension,For (p+1) × (p+1) dimension Beam Domain sample covariance matrix, [·]^HIndicate conjugate transposition.

Step S3: being handled covariance matrix by improved power method, obtains maximum characteristic value and its correspondence Feature vector, obtain high-intensity echo signal, the specific steps are as follows:

S31: the simplified operation of Eigenvalues Decomposition is carried out to Beam Domain covariance matrix by improving power method, seeks covariance The maximum eigenvalue of matrix feature vector corresponding with its, i.e. λ_maxAnd e_max, set e₀=[1,1 ..., 1], carries out according to the following formula Iterative solution:

Wherein, e_i+1It is the feature vector after iteration i+1 times, λ_i+1For its corresponding characteristic value, max () is greatest member Derivation after each interative computation, is normalized into next iteration obtained vector, with i-th iteration For, obtain vector e_iIt is normalized afterwards, i.e. e_i=e_i/max(e_i), above formula is again introduced into after normalized Interative computation, interative computation terminate Rule of judgment be | λ_i+1-λ_i|/|λ_i+1| < ε can select ε value according to required precision, this Invention takes ε=0.001.

Step S4: ignoring part low energy echo data, invert to covariance matrix and carry out abbreviation, by the fortune of inverting of matrix The multiplying for being reduced to vector is calculated, the complexity level of matrix inversion is reduced, the specific steps are as follows:

S41: the energy of the characteristic value signal of covariance matrix in ultrasonic signal, the corresponding letter of the bigger characterization of characteristic value Number energy is stronger, and the corresponding feature vector of low energy part constitutes noise subspace and ignores portion for simplification matrix inversion operation Divide low energy signal, the corresponding characteristic value of noise subspace takes identical value in the case where covariance matrix mark is constant, guarantees super The energy constant of sound echo-signal, it may be assumed that

Wherein, q is the number of feature vector in signal subspace, and trace () asks mark operation for matrix；

S42: it enablesThen Beam Domain sample The inverse matrix of covariance matrix can be with abbreviation are as follows:

Wherein, I is unit matrix, e_iFor ith feature vector, λ_iFor its corresponding characteristic value.

S43: for the solution procedure of further abbreviation S42, ignore part beam signal, take inversion operation of the q=1 by matrix The multiplying of a vector is converted to, cycle-index is reduced, it may be assumed that

Above formula only gets maximum characteristic value feature vector corresponding with its, i.e. λ₁=λ_max, e₁=e_max,I is unit matrix.By matrix inversion operation It is reduced to the multiplying of vector, computational complexity is by Ο (L³) it is reduced to Ο ((p+1)²), wherein L is Element space subarray Element number of array, p be Beam Domain dimensionality reduction parameter.

Step S5: by the simplification Beam Domain covariance matrix inverse matrix and Beam Domain direction vector of acquisition, wave beam is obtained The expression formula of domain minimum variation algorithm weight vector；

S51: the Beam Domain transition matrix obtained using S21 obtains the direction vector of Beam Domain:

a_b=Ta

Wherein, [1,1 ..., 1] a=^TIt is the Element space direction vector that N × 1 is tieed up, T is transition matrix, a_bFor Beam Domain direction Vector.

S52: the Beam Domain direction vector that the Beam Domain sample covariance matrix inverse matrix and S51 acquired using S43 is acquired, Obtain the weight vector of the low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibIt is the weight vectors of the low complex degree minimum variance ultrasonic imaging method based on power method,It is letter The Beam Domain covariance matrix inverse matrix of change, a_bFor Beam Domain direction vector.

Step S6: the expression formula is projected to the corresponding feature vector of maximum eigenvalue, is obtained based on power method The optimal weight vector of low complex degree minimum variation algorithm；

S61: the optimal weight vector expression formula of the low complex degree minimum variance ultrasonic imaging method based on power method:

Wherein, w_ibmvFor the optimal weighting vector of the low complex degree minimum variance ultrasonic imaging method based on power method, e_max For the corresponding feature vector of maximum characteristic value, []^HIndicate conjugate transposition operation.

Adaptive beamformer output signal in step S7 is calculated as follows:

In formula,For weight vector w_ibmvConjugate transposition, x_b(k) for after the echo-signal space smoothing of sub-array beam domain Beam Domain signal, y_ibmv(k) be mentioned algorithm beamformer output signal.

The value of neutron array array element number L of the present invention is L≤N/2.

In order to verify effectiveness of the invention, in the present embodiment, using Field II to point common in medical imaging Scattering Targets and spot Scattering Targets be imaged and have carried out simulation imaging to the point target of different center frequency.Field II is Based on linear system roomage response principle, its simulation result and actual imaging very close to being imitative by accepting extensively in the world The standard of true ultrasonic system.In point target emulation, 15 target points are set, axial distance is distributed in 30mm~80mm, often Target point is set every 5mm, wherein 40mm and 60mm punishes 3 target points of cloth, respectively there is 1 target point in remaining position.Emulation uses Linear array emits ultrasonic signal, is emitted as fixed point and focuses, and emits focus at axial distance 50mm, reception pattern is set as dynamic State focuses, and the dynamic range of imaging is 60dB, and imaging mode is line scanning imagery.During the target area of sound absorption spot imaging is set as The heart at axial distance 25mm, radius be 3mm border circular areas, background area random distribution 100000 noise scattering points, Imaging algorithm uses space smoothing and diagonal loading technique.Be respectively adopted dynamic emission fixed point reception mode to 4 kinds of algorithms into Row imaging, and the runing time of the resolution ratio of more various imaging algorithms, contrast and algorithm.Fig. 2 is different dimensionality reduction parameters pair Than figure.

Dynamic emission pinpoints received 4 kinds of algorithm point targets in the case of noiseless and plus noise is set forth in Fig. 3 and Fig. 4 Simulation imaging result.It is 4 kinds of algorithm point target lateral resolutions at 40mm that Fig. 5, which gives axial distance,.As seen from Figure 3, Traditional DAS imaging effect is worst, and lateral pseudomorphism is serious；MV algorithm is promoted larger compared to DAS algorithm imaging effect；ESBMV algorithm Imaging resolution is further promoted on the basis of MV, and imaging artefacts significantly reduce；No matter is IBMV algorithm proposed in this paper MV algorithm is significantly better than that near field or far-field region resolution ratio；From the imaging effect of point target, ESBMV algorithm is compared Imaging resolution is close, and in far-field region, imageable target point is slightly less than ESBMV.As seen from Figure 4, echo-signal is increased and is believed After the white Gaussian noise made an uproar than 10dB, there is apparent noise hickie in background area, and image contrast is deteriorated.Fig. 5 (b) compared to (a) for, due to increasing noise signal, the total energy difference of echo-signal decreases, noise secondary lobe and noiseless When compared to more disorderly and more unsystematic, and side lobe peak compared to before not plus noise when promoted.As seen from Figure 4, addition is made an uproar After acoustical signal, background area noise hickie be increased significantly, and the discrimination of background area and target point reduces, and choose normalization herein The imaging effect that amplitude is main lobe width and the side lobe peak at the place -6dB to characterize point target, detailed numerical result such as table Shown in 1.

Main lobe width calculated result at 1 axial distance 40mm of table

It can be obtained from table 1, the main lobe width and side lobe peak of traditional DAS algorithm are significantly greater than adaptive algorithm, IBMV algorithm Main lobe width in muting situation be respectively lower than MV algorithm and ESBMV algorithm 0.47mm and 0.32mm, main lobe width drop It is low obvious；After increasing noise signal, the main lobe width of algorithms of different has promotion, and the IBMV algorithm that the present invention is mentioned is still obvious excellent In other algorithms, point target imaging resolution is promoted significant compared to MV algorithm and ESBMV algorithm.The variation of side lobe peak is similar Main lobe width in the place -6dB, the mentioned algorithm of the present invention are reduced compared to original MV algorithm in the case where there is noiseless respectively 3.64dB and 2.31dB；In conclusion can be obtained by point target imaging contrast, the imaging resolution of the mentioned IBMV algorithm of the present invention is excellent In MV algorithm；Wherein main lobe width is better than ESBMV algorithm, and side lobe peak is similar to ESBMV algorithm.

Dynamic emission pinpoints received 4 kinds of algorithms sound absorption spot in the case of noiseless and plus noise is set forth in Fig. 6 and Fig. 7 Imaging results.Adaptive algorithm image contrast deficiency is compared with DAS algorithm traditional it can be seen from Fig. 7 by Fig. 6, is inhaled in circle There are apparent background area noises for acoustic blur fringe region；Compared with ESBMV and IBMV algorithm, sound absorption spot region exists MV algorithm A small amount of noise hickie, algorithm inhibit noise signal ability slightly inadequate；In contrast, ESBMV algorithm and the IBMV of proposition are calculated Method noise inhibiting ability is stronger, and sound absorption spot region there's almost no noise jamming, and image contrast is promoted obvious.Compared to ESBMV algorithm, IBMV algorithm proposed in this paper is darker in ambient noise regional imaging effect, and characterization algorithm inhibits noise signal Ability is better than ESBMV algorithm.After the white Gaussian noise for increasing 10dB, imaging effect decline is obvious, and target area obviously occurs Noise hickie, image contrast are remarkably decreased.In order to more intuitively illustrate the contrast of algorithm imaging, table 2 and table 3 arrange respectively Go out noiseless and has the specific value result of algorithms of different image contrast under noise situations.

2 noiseless of table sound absorption spot image contrast

3 plus noise of table sound absorption spot image contrast

Wherein contrast is the absolute value of the difference of center blackening mean power and background area mean power, background area side The robustness of difference characterization algorithm, it is better to be worth smaller algorithm robustness, comparison noise ratio be contrast and background area variance it Quotient, the average power content in table 2,3 is smaller, shows that the intensity of signal is weaker, the brightness presented in image is darker.By table 3 and table 4 can obtain, adaptive algorithm under noise-free case compared to tradition DAS algorithm promoted in terms of image contrast obviously, carry on the back Scene area and target area imaging discrimination are larger, and ESBMV compares MV algorithm contrast with the IBMV algorithm of proposition and promoted respectively 6.93dB and 4.83dB.By the average power content of imaging region it is found that IBMV algorithm inhibits noise immune compared to ESBMV algorithm It is improved, the inhibiting effect of background area noise signal is stronger compared to sound absorption spot region, shows as the flat of background area Equal performance number decline becomes apparent, so image contrast is slightly inadequate compared with ESBMV algorithm.In terms of algorithm robustness, Traditional DAS algorithm is better than adaptive algorithm, therefore after increasing 10dB noise signal, and the decline of adaptive algorithm imaging effect is bright Aobvious, contrast is not as good as tradition DAS algorithm.

Table 4 lists the point target and sound absorption spot target imaging time comparing result of algorithms of different, and present invention emulation exists It is run under Matlab R2015a, the allocation of computer used is as follows: Intel i74GHz CPU, 8GB RAM.

4 imaging time contrast table of table

In simulation imaging of the invention, the investigative range of point target is in 30mm~80mm, the detection model for the spot imaging that absorbs sound Enclosing is 20mm~35mm, and the sampling number difference of two kinds of simulation imagings is obvious, therefore under identical running environment, absorb sound spot mesh It marks imaging time and is less than point target imaging time.As can be seen from Table 4, traditional DAS algorithm either point target imaging or suction Acoustic blur target imaging runing time is all much better than adaptive algorithm；IBMV is compared to MV algorithm due to reducing sample covariance square The dimension of battle array simplifies matrix inversion simultaneously, and operational efficiency is promoted；Compared to ESBMV algorithm, IBMV algorithm simplifies square Battle array is inverted and Searching Matrix Eigen Value operation, and operational efficiency is promoted obvious.In conclusion IBMV algorithm proposed by the present invention at As being superior to MV algorithm in terms of resolution ratio, contrast and operational efficiency.

During practical ultrasound detection, the difference of Chang Yinwei detection position and select different center frequency ultrasound visit Head, such as convex array probe frequency applied to abdomen is usually 2.5MHz, 3.5MHz and 5.0MHz, intracavitary probe is usually 6.5MHz, the linear probe of blood vessel detection are usually 7.5MHz, and the linear probe of high frequency can reach 10MHz and 12MHz.In order to say The practicability of bright the mentioned algorithm of the present invention has chosen 4 kinds of representative frequency probes and carries out point target simulation imaging, the velocity of sound It is set as 1540m/s, ordinate indicates that investigation depth, specific imaging results are as shown in Figure 8.

Can intuitively it be found out by Fig. 8, IBMV algorithm imaging results are better than ESBMV algorithm, there is less lateral artifact, at Picture target point is smaller, focuses more accurate；The imaging effect of MV algorithm is better than tradition DAS algorithm, calculates not as good as ESBMV and IBMV Method；With the increase of centre frequency, the imaging resolution of algorithms of different is improved, and imaging artefacts are substantially reduced, but center frequency Rate is higher, and the decaying of ultrasonic wave in the medium is more serious, and the intensity of echo-signal is also weaker, and point target brightness is shown as in image Reduction；Compared with other algorithms, IBMV algorithm can keep better imaging effect under different center frequency, be suitable for more Kind test object, has stronger stability and practicability.Selection Center frequency is the imaging data of 5.0MHz, makes imaging point Resolution comparison diagram further illustrates algorithm imaging effect, as a result as shown in Figure 9.

The imaging data that Fig. 9 chooses at axial distance 20.5mm is normalized, and sampled point is chosen laterally equidistant 40 points.As seen from Figure 9, the imaging resolution of adaptive algorithm is better than tradition DAS algorithm, under main lobe width has obviously Drop.The IBMV algorithm of proposition resolution curve compared with original MV algorithm and ESBMV algorithm is narrower, and main lobe width is substantially better than MV Algorithm, it is consistent with Fig. 8 imaging results.

Finally, it is stated that the above examples are only used to illustrate the technical scheme of the present invention and are not limiting, although referring to compared with Good embodiment describes the invention in detail, those skilled in the art should understand that, it can be to skill of the invention Art scheme is modified or replaced equivalently, and without departing from the objective and range of the technical program, should all be covered in the present invention Scope of the claims in.