CN103971127B - Forward-looking radar imaging sea-surface target key point detection and recognition method - Google Patents

Abstract

The invention discloses a forward-looking radar imaging sea-surface target key point detection and recognition method. The method includes the steps that radar echo data are quantized into a gray scale image; a region of interest is extracted from the gray scale image, partition is performed, and a target region partition image is obtained; the radar echo data and the region of interest of the image are used, and a target region peak point information matrix is obtained; information merging is conducted on the target region partition image and the target region peak point information matrix, the number K of peak points in a target region in a merging result is obtained through counting, the previous K peak points are selected and serve as target effective peak points and form a target effective peak point image in a binaryzation mode, target axial features are extracted, and a target position is determined; the center of gravity of target energy is calculated and serves as the target key point. According to forward-looking radar target features, multiple mode identification methods are used comprehensively, the inherent features of a target can be reserved, meanwhile interference factors such as artifacts and side lobes can be restrained, and the recognition accuracy and the positioning precision of radar imaging sea surface target key points are improved.

Description

A kind of forward-looking radar is imaged sea-surface target critical point detection recognition methodss

Technical field

The invention belongs to target detection, mode identification technology, and in particular to a kind of forward-looking radar is imaged sea-surface target Critical point detection recognition methodss, the method retain target inherent character while can effectively suppress artifact, secondary lobe etc. interference because Element, improves the recognition correct rate and positioning precision to radar imagery sea-surface target key point.

Background technology

The Ship Target Detection identification realized using radar imaging technology is many civilian and military field key technology, It is widely used in fields such as monitoring, military affairs.In forward-looking radar echo, strong echo-signal means that detector is searched in this place Rope is to strong scattering point.Two face reflectors and corner reflector during strong scattering point is typically by Ship Target are caused.These reflections Component distribution is presented different intensity sizes on whole Ship Target with the different of azimuth of target.However, normal radar Transmitting linear FM signal, due to the two-dimensional frequency supporting domain of imaging system it is limited so that synthetic aperture radar (Synthetic Aperture Radar, SAR) impulse response function be sinc functions to orientation in distance, cause sidelobe level very It is high.Because the process length of window of radar data is limited, and there is phase error in radar return data, form secondary lobe, lead The noise of multiplication is caused, and interference is produced with neighbouring scattering object, very big is affected on picture quality.The presence of secondary lobe causes forword-looking imaging There is artifact in the target image of middle acquisition, real target point is weaker than in intensity, this be also target critical point identification with it is high-precision The basis of degree positioning.False target can be distinguished with real goal, in addition it is also necessary to consider to target under different points of view Forword-looking imaging, the spacing of false target and real goal in image, if there is the situation for overlapping interference.Therefore, it is necessary to Before forward-looking radar target's feature-extraction, detection and identification, pretreatment is carried out to original radar echo signal data, made an uproar with reducing Sound shadow is rung, and improves the signal to noise ratio of image, and prominent target signature information recognizes target, closes so as to improve target using target characteristic The probability of key point identification, and the energy barycenter using target determines the key point of target.

In existing disclosed document, the forward-looking radar imaging sea-surface target detection recognition method mentioned is mostly by right The process of original radar echo signal is realizing.However, because by the restriction of SAR image-forming mechanisms, target slice image receives target The impact of the factors such as attitude, background characteristics value and sensor imaging attitude, shows higher changeableness, right so as to easily cause , there is erroneous judgement and misjudge in the interference of recognition result.Zhang Hong et al. exists《High Resolution SAR Images target recognition》In propose and be based on The recognition methodss of target peak feature, for selection that peak value is counted out is 20～40 taking of taking experimental data statistics to provide Value scope.It does not provide an effective criterion and method, on engineer applied for the determination that peak value is counted out Can only be empirically.Maximum entropy dividing method has the advantages that good stability in existing open source literature, but easily dry by background Disturb, it is inaccurate that segmentation result has false target, gained target information.

The content of the invention

The present invention proposes that a kind of forward-looking radar is imaged sea-surface target key point for the Objective extraction problem under complex environment Detection recognition method, specifically includes：

(1) original radar two dimension echo data is quantified as into 2-D gray image data；

(2) 2-D gray image that step (1) is obtained is carried out using the method for being based on target physical dimension and confidence level Region of interest is extracted, and obtains target area gray level image；

(3) target area gray level image is split using maximum entropy, obtains target area segmentation figure picture；

(4) using radar two dimension echo data and target area gray level image, the radar two dimension echo of target area is extracted Peak point information in data, obtains target area peak point information matrix；

(5) information fusion, statistics fusion knot are carried out to target area segmentation figure picture and target area peak point information matrix Peak value is counted out K in target area in fruit, is counted out as effective peak；

(6) peak point in the peak point information matrix of target area is sized, K peak point is used as target before choosing Effective peak point, to target area peak point information matrix two-value target effective peak value dot image is turned to；

(7) target axial direction feature in target effective peak value dot image is extracted, the interference of false-alarm point is excluded, target location is determined；

(8) using target location and target energy center of gravity, target critical point is determined.

Further, the step (1) specifically includes：

(1.1) the floating point values threshold value for being quantified is selected, if floating point values upper threshold is L_max, bottom threshold is L_min, its In：

L_max=N* (TLength*Margin)²+L_min, if L_max＜ Totalpix

L_max=Totalpix, if L_max＞ Totalpix

N is the target maximum number that may contain in background, Totalpix for original image number of pixels, TLength For target length pixel count in the picture, Margin can completely be shown for surplus with ensureing target, and minT is that target two dimension is returned The minimum of wave number evidence may floating point values；

(1.2) respectively selective value is L_maxFloating point values and value be L_minFloating point values as threshold value Level255 and Level0；To each data point in original radar two dimension echo data, if floating point values gives gray value 255 more than Level255, Gray value 0 is given less than Level0, to value in L_maxWith L_minBetween floating point values then carry out linear interpolation, determine its gray value, The formula of linear interpolation is as follows：

Wherein f (x, y) is the floating-point values of data point (x, y) place radar two dimension echo data, and g (x, y) is linear interpolation The corresponding gray value of point (x, y) afterwards.

Further, the step (2) specifically includes：

(2.1) first according to target sizes and imaging resolution, using the interval under the normal course state of target as constraint, Determine region of interest window length and width；The region of interest window centered on s points is investigated, the statistic of pixel value has mean μ in window_sWith Center of gravity G_s：

Wherein, n is the number of pixel in region of interest window, and g (x, y) is (x, y) place pixel in region of interest window Gray value, Ω is region contained by region of interest window；

(2.2) the confidence level μ of the region of interest window centered on s points is asked for_s, and center of gravity G in window_sSit with window center Mark O_sApart from d (G_s,O_s), wherein：

ρ_s=μ_s/[d(G_s,O_s)+1]

d(G_s,O_s) it is center of gravity G in window_sWith window center coordinate O_sBetween Euclidean distance, wherein x, y represents respectively the row and column of image；

(2.3) region of interest window is determined in 2-D gray image according to following principles, as target area gray level image：

If being background in the region of interest window centered on s points, because background pixel value difference is less and all than relatively low, this When apart from d (G_s,O_s) ≈ 0, confidence level ρ_s≈μ_sSuitable with background pixel value average, now region of interest window region is not Region of interest；

If a part is background for a target part in the region of interest window centered on s points, now μ in window_sIncrease, but together When window in center of gravity G_sIt is partial to the region of high pixel value, d (G_s,O_s) also increase, until region of interest window includes whole target, it is determined that For region of interest；

If including whole target in the region of interest window centered on s points, now μ_sReach maximum, maximum and target Pixel value is related, while center of gravity G in window_sWith window center O_sDistance reduces；When apart from d (G_s,O_s) ≈ 0 when confidence level ρ_sReach most Greatly, while ρ_sFor local maximum, now region of interest window region is region of interest.

Further, the step (3) specifically includes：

(3.1) determine that the segmentation threshold Th for being split to target area gray level image, segmentation threshold Th are target Entropy Entropy_OWith background entropy Entropy_BThe maximum of sum, wherein：

Background entropy

Target entropy

P (i) represents corresponding grey scale level as the probability size of i, and m represents the maximum gray scale of image, P_tAnd H_tThe back of the body is represented respectively Scape and target in the picture pixel grey scale distribution probability and；

(3.2) target area gray level image is split according to segmentation threshold Th, obtains target area segmentation figure as O (x, y), specially：

If the entropy of certain gray level is more than Th, then it is assumed that the gray level is target；If the entropy of certain gray level is less than Th, then it is assumed that The gray level is background, so as to obtain target area segmentation figure as O (x, y).

Further, the step (4) is specially：

According to radar two dimension echo data and target area gray level image, using first-order difference method, target area is extracted Radar two dimension echo data in peak point information, obtain target area peak point information matrix G (x, y), wherein：

(x, y) represents the point in two-dimentional echo-signal, and x represents distance to y represents orientation, and note radar two dimension echo is floated Point value is f (x, y).

Further, the step (5) is specially：

To target area segmentation figure as O (x, y) and target area peak point information matrix G (x, y) carry out information fusion, obtain Image R (x, y) to after fusion；

R (x, y)=G (x, y) O (x, y)/255

Image R (x, y) after fusion is 0/255 bianry image, and gray level is that 255 pixel is target area in image Peak point in regional partition objective area in image, counts these peak values and counts out K

Wherein, p (x_i,y_i) be fusion after image R (x, y) in (x_i,y_i) place gray value size.

Further, the step (6) specifically includes：

(6.1) peak point in the peak point information matrix of target area is sorted by size, chooses maximum front K peak point As the effective peak point characteristic information of target；

(6.2) binary conversion treatment is carried out to the effective peak point characteristic information of target, will this K extreme point place pixel Point assignment gray level 255, rest of pixels point is entered as 0, obtains target effective peak value dot image.

Further, the step (7) specifically includes：

(7.1) least square fitting is utilized to representing k candidate target point in target effective peak value dot image, is found The minimum straight line of residual sum of squares (RSS), the i.e. axial straight line of target are made to one, the straight line is identical with the diagonal of target；

(7.2) using the length of target, width, Diagonal Dimension as constraint, with target length, the target square of wide constraint Shape is window, allows the diagonal of window to move linearly along the axial direction of target, to find and make window on the position on axial straight line Interior contained peak value is counted out most the window's positions, and the position corresponding to window is target location；

(7.3) the peak point information flag outside window is false-alarm point, and peak point information contained in window determines target Position.

Further, the axial straight line that target is calculated in the step (7.1) is specially：

To object pixel point set V, parameter k of straight line y=kx+b, b so that residual sum of squares (RSS) are estimatedMinimum, wherein (x_i,y_i)∈V。

Further, the step (8) is specially：

Ask for the grey scale centre of gravity of target areaAs the energy barycenter of target area, the as pass of target Key point, it is as follows that the grey scale centre of gravity of wherein target area asks for formula：

F (x, y) is the floating point values at (x, y) place in radar two dimension echo data, and Τ is target area.

In general, by the contemplated above technical scheme of the present invention compared with prior art, the present invention has following Beneficial effect：

(1) while rejecting false target information using radar return data peaks point information, using OTSU image segmentations Constraint effective peak count out, so as to realize engineering applications peak value count out K self adaptation choose, so as to get result More standardization, will not change because of the different of operator；

(2) in the image quantization stage, using relevant informations such as target and background sizes, the calculating to quantization threshold is established Criterion, so that radar two dimension echo data can in the form of images show and process, it is the figure of radar two dimension echo data Pictureization process provides data and supports；

(3) according to radar image distance to the imaging characteristicses with orientation, using radar bearing to maximum point information Target area is determined, so as to realize Target Max information retrieval on the basis of performance is not affected, with respect to local maximum Extracting method reduces amount of calculation；

(4) using the axial feature of target, the positioning and false-alarm point for realizing target is excluded, and makes target location more accurate；

(5) the target region of interest extracting method of a kind of utilization target physical dimension and confidence level, the sense of extraction are proposed Region of interest is more accurate, reduces the operand of subsequent treatment；

(6) forward-looking radar imaging sponge target critical point detection recognition method proposed by the present invention is believed forward-looking radar echo Number handling process, the results show effectiveness of the flow process.

In sum, the present invention has comprehensively used various modes recognition methodss, Neng Gou according to forward-looking radar target property Suppress the interference factors such as artifact, secondary lobe while retaining target inherent character, improve the knowledge of radar imagery sea-surface target key point Other accuracy and positioning precision.Target detection identification is carried out by this method, the recognizable rate of target is high.

Description of the drawings

Fig. 1 is the overview flow chart that forward-looking radar of the present invention is imaged sea-surface target critical point detection recognition methodss；

Fig. 2 is the image in one embodiment of the invention after original radar two dimension echo data quantization；

Fig. 3 is the sharp peaks characteristic figure of original radar two dimension echo data；

Fig. 4 is that image segmentation result after original radar two dimension echo data quantization in Fig. 2 is shown；

Fig. 5 is image object effective peak hum pattern after quantifying in original radar two dimension echo data in Fig. 2；

Fig. 6 is target axial direction feature constraint result figure in image after original radar two dimension echo data quantization in Fig. 2；

Fig. 7 is target critical point location result in image after original radar two dimension echo data quantization in Fig. 2.

Specific embodiment

In order that the objects, technical solutions and advantages of the present invention become more apparent, it is right below in conjunction with drawings and Examples The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the present invention, and It is not used in the restriction present invention.As long as additionally, technical characteristic involved in invention described below each embodiment Not constituting conflict each other just can be mutually combined.

The purpose of the present invention is to suppress the interference factors such as artifact, secondary lobe while retaining target inherent character, improves radar The recognition correct rate and positioning precision of imaging sea-surface target key point.However, in forward-looking radar echo, strong echo-signal meaning Detector and search strong scattering point in this place.Strong scattering point be typically by Ship Target in two face reflectors and corner reflector Cause.These reflection parts are distributed on whole Ship Target, different intensity are presented with the different of azimuth of target big It is little.However, normal radar transmitting linear frequency modulation (Linear Frequency Modulation, LFM) signal, due to imaging system The two-dimensional frequency supporting domain of system is limited so that the impulse response function of SAR is sinc functions, secondary lobe in distance to orientation Level is very high.Because secondary lobe is likely to form the noise of multiplication, and produces interference with neighbouring scattering object, picture quality is affected very Greatly.The presence of secondary lobe causes the target image obtained in forword-looking imaging to there is artifact, and real target point is weaker than in intensity, this It is also the basis of target recognition and hi-Fix.False target can be distinguished with real goal, in addition it is also necessary to consider right Forword-looking imaging of the target under different points of view, the spacing of false target and real goal in image, if exist and overlap interference Situation.The present invention makes the assumption that estimation based on the target property and Sea background characteristic in radar image.First, it is former using radar The image of Jing over-segmentations carries out information fusion after beginning echo data and process, obtains the sharp peaks characteristic information of target area；Then, Using the axial feature knowledge exclusive PCR of target, target is recognized.Finally, target critical is positioned according to the energy barycenter of target Point.

The present invention by the feature extraction based on target peak, obtains the sharp peaks characteristic figure of target first；By original echo Data become image through quantization, are split using maximum entropy after grade resolution adjustment, obtain segmentation figure picture；By the peak of gained Value tag figure carries out fusion and obtains peak point quantity threshold with segmentation figure picture, can obtain effective target peak information, utilizes The axial feature of target and energy barycenter, recognize target, position key point.Test result indicate that, algorithm can suppress to protect in text While staying target inherent character, various interference factors are reduced, improve the recognition correct rate and positioning precision of target critical point.

The invention provides a kind of forward-looking radar is imaged sea-surface target critical point detection recognition methodss, its overall procedure is as schemed Shown in 1, the method detailed process is：

(1) original radar two dimension echo data is quantified as into 2-D gray image data

Original radar two dimension echo data is quantified first, being transformed into the gray value in the range of 0～255 can be carried out The 2-D gray image data of Digital Image Processing.Original radar two dimension echo data is mapped as into the side of 256 grades of gray level images Method is：The floating point values of original radar two dimension echo data is ranked up from small to large；Carrying out selection according to following principle is carried out The floating point values threshold value of quantization, it is assumed that floating point values upper threshold is L_max, bottom threshold is L_min, then have

L_max=N* (TLength*Margin)²+L_min, if L_max＜ Totalpix

L_max=Totalpix, if L_max＞ Totalpix

Wherein, N is the target maximum number that may contain in background, and Totalpix is the number of pixels of original image, TLength is target length pixel count in the picture, and Margin is surplus, it is ensured that target can completely show that minT is target The minimum of two-dimentional echo data may floating point values.

Respectively selective value is L_maxFloating point values and value be L_minFloating point values as threshold value Level255 and Level0；To every Individual pixel, if floating point values gives gray value 255 more than Level255, less than Level0 gray value 0 is given, to value in L_maxWith L_minBetween floating point values then carry out linear interpolation, determine its gray value.The formula of linear interpolation is as follows

Wherein f (x, y) is the floating-point values of pixel (x, y) place radar two dimension echo data, and g (x, y) is linear interpolation The corresponding gray value of point afterwards, the image after quantization is as shown in Figure 2.

(2) 2-D gray image that step (1) is obtained is carried out using the method for being based on target physical dimension and confidence level Region of interest is extracted, and obtains target area gray level image

First according to target sizes and imaging resolution, using the interval under the normal course state of target as constraint, it is determined that Suitable region of interest window length and width.The region of interest window centered on s points is investigated, the statistic of pixel value has mean μ in window_s With center of gravity G_s。

Wherein, n is the number of pixel in region of interest window, and g (x, y) is the gray value of (x, y) place pixel, and Ω is sense Region contained by region of interest window.

By μ_sIt is whether an evaluation criterion of region of interest as decision window, μ_sPixel value in higher explanation local window It is bigger, more it is likely to be target；Simultaneously as wish target lock-on at the center of region of interest, center of gravity G in window_sAnd window Mouth centre coordinate O_sApart from d (G_s,O_s) as another evaluation criterion.Region of interest confidence level of the definition centered on s points For：

ρ_s=μ_s/[d(G_s,O_s)+1]

Apart from d (G_s,O_s) it is chosen for center of gravity G in window_sWith window center coordinate O_sBetween Euclidean distance, wherein x, y difference Represent the row and column of image：

To three kinds of regions in 2-D gray image, it is respectively processed：

If being background in the region of interest window centered on s points, because background pixel value difference is less and all than relatively low, this When apart from d (G_s,O_s) ≈ 0, confidence level ρ_s≈μ_sSuitable with background pixel value average, now region of interest window region is not Region of interest；

If a part is background for a target part in the region of interest window centered on s points, now μ in window_sIncrease, but together When window in center of gravity G_sIt is partial to the region of high pixel value, d (G_s,O_s) also increase, until region of interest window includes whole target, it is determined that For region of interest；

If including whole target in the region of interest window centered on s points, now μ_sReach maximum, maximum and target Pixel value is related, while center of gravity G in window_sWith window center O_sDistance reduces；When apart from d (G_s,O_s) ≈ 0 when confidence level ρ_sReach most Greatly, while ρ_sFor local maximum, now region of interest window region is region of interest.

Here the region of interest for obtaining we be referred to as target area gray level image.

(3) target area gray level image is split using maximum entropy, obtains target area segmentation figure picture；

Target area gray level image is split using maximum entropy, target area segmentation figure picture is obtained；

If threshold value Th is divided the image into as two parts of target and background, then the entropy for defining target and background is respectively：

Background entropy

Target entropy

Wherein, P (i) represent corresponding grey scale level as i probability size, m represents the maximum gray scale of image, P_tAnd H_tRespectively Represent background and target in the picture the probability of pixel grey scale distribution and.

The maximum of target entropy and background entropy sum is calculated as threshold value

If the entropy of certain gray level is more than Th, then it is assumed that the gray level is target；If the entropy of certain gray level is less than Th, then it is assumed that The gray level is background.Target area segmentation figure can so be obtained as O (x, y), the image after segmentation is as shown in Figure 4.

(4) using radar two dimension echo data and target area gray level image, the radar two dimension echo of target area is extracted Peak point information in data, obtains target area peak point information matrix；

Using radar two dimension echo data and target area gray level image, the radar two dimension echo data of target area is extracted In peak point information, obtain target area peak point information matrix；

Scattering center can define 2 class extremal features points in radar two dimension echo data：Two-dimentional extreme vertex and one-dimensional pole Value point.Define extreme point：

Wherein U (a_i) represent with a_iCentered on local neighborhood (not including a_iPoint).Due to extreme point in orientation dynamic Scope is very big, therefore the present invention considers the one-dimensional extreme point in orientation.

Extreme point defined above is extracted using first-order difference method.To the point (x, y) in two-dimentional echo-signal, x tables Show distance to y represents orientation, and note radar two dimension echo floating point values is f (x, y), is defined:

It is target area peak point information matrix using the calculated G (x, y) of above formula, as shown in Figure 3.

(5) information fusion is carried out to target area segmentation figure picture and target area peak point information matrix, target area is counted Peak value is counted out K in the target area of regional partition image, is counted out as effective peak.

To target area segmentation figure as O (x, y) and target area peak point information matrix G (x, y) carry out information fusion, obtain Image R (x, y) to after fusion.

R (x, y)=G (x, y) O (x, y)/255

Image R (x, y) after fusion is 0/255 bianry image, and gray level is that 255 pixel is target area in image Peak point in regional partition objective area in image, counts these peak values and counts out K

Wherein, p (x_i,y_i) be fusion after image R (x, y) in (x_i,y_i) place gray value size.

(6) peak point in the peak point information matrix of target area is sized, K peak point is used as target before choosing Effective peak point, binaryzation obtains target effective peak value dot image.

Peak point in the peak point information matrix of target area is sorted by intensity, maximum front K peak point conduct is chosen The effective peak point characteristic information of target.The effective peak point characteristic information of binary conversion treatment target, will this K extreme point institute In pixel assignment gray level 255, rest of pixels point is entered as 0, obtains target effective peak value dot image, as shown in Figure 5.

(7) target axial direction feature in target effective peak value dot image is extracted, the interference of exclusive segment false-alarm point determines target position Put, as shown in Figure 6.

Noise spot or false-alarm point may be mixed in target effective peak value dot image in front k candidate target point, have been needed Carry out axis projection to screen the impact point of candidate.Axis projection make use of the axial length information of target, to waiting The spatial relation for selecting impact point enters row constraint.

Here using the data fitting method of method of least square, it finds number to this method by minimizing the quadratic sum of error According to optimal function matching.To object pixel point set V, parameter k of straight line y=kx+b, b so that residuals squares are estimated WithMinimum, wherein (x_i,y_i)∈V。

The algorithm of axial feature constraint is as follows：

1) least square fitting is utilized to representing k candidate target point in target effective peak value dot image, searches out one Bar makes the minimum straight line of residual sum of squares (RSS), the i.e. axial straight line of target, identical with the diagonal of target；

2) using the length of target, width, Diagonal Dimension as constraint, with target length, the target rectangle of wide constraint For window, allow the diagonal of window to move linearly along the axial direction of target, find and made in window on the position on axial straight line Contained peak value is counted out most the window's positions, and the position corresponding to window is target location；

3) the peak point information flag outside window is false-alarm point, and peak point information contained in window determines the position of target Put.

(8) using target location and target energy center of gravity, target critical point is determined.

By the positions and dimensions information of target, the peak point information in the position of target can be obtained, calculate these peak values Point energy barycenter as target key point, as shown in Figure 7.

According to the characteristic of radar imagery, radar echo intensity shows the information of strong scattering point, and the present invention defines target area Interior grey scale centre of gravity is positioned as the key point of target.Grey scale centre of gravity method is to regard the gray value of each pixel in region " energy " of the point, the center of gravity formula in its required region is as follows：

F (x, y) is the floating point values at (x, y) place in radar two dimension echo data, and Τ is target area,It is target The energy barycenter of the grey scale centre of gravity in region, i.e. target area.

As it will be easily appreciated by one skilled in the art that the foregoing is only presently preferred embodiments of the present invention, not to The present invention, all any modification, equivalent and improvement made within the spirit and principles in the present invention etc. are limited, all should be included Within protection scope of the present invention.

Claims (15)

1. a kind of forward-looking radar is imaged sea-surface target critical point detection recognition methodss, it is characterised in that methods described includes：

(1) original radar two dimension echo data is quantified as into 2-D gray image data；

(2) 2-D gray image obtained to step (1) is emerging using sense is carried out based on the method for target physical dimension and confidence level Interesting area extracts, and obtains target area gray level image；

(3) target area gray level image is split using maximum entropy, obtains target area segmentation figure picture；

(4) using radar two dimension echo data and target area gray level image, the radar two dimension echo data of target area is extracted In peak point information, obtain target area peak point information matrix；

(5) information fusion is carried out to target area segmentation figure picture and target area peak point information matrix, in statistics fusion results Peak value is counted out K in target area, is counted out as effective peak；

(6) peak point in the peak point information matrix of target area is sized, K peak point is used as target effective before choosing Peak point, to target area peak point information matrix two-value target effective peak value dot image is turned to；

(7) target axial direction feature in target effective peak value dot image is extracted, the interference of false-alarm point is excluded, target location is determined；

(8) using target location and target energy center of gravity, target critical point is determined.

2. the method for claim 1, it is characterised in that the step (1) specifically includes：

(1.1) the floating point values threshold value for being quantified is selected, if floating point values upper threshold is L_max, bottom threshold is L_min, wherein：

$L_{\min }=Totalpix*\frac{\min T}{Totalpix}$

L_max=N* (TLength*Margin)²+L_min, if L_max＜ Totalpix

L_max=Totalpix, if L_max＞ Totalpix

N is the target maximum number that may contain in background, and Totalpix is the number of pixels of original image, and TLength is mesh Mark length pixel count in the picture, Margin can completely show that minT is target two dimension number of echoes for surplus to ensure target According to minimum may floating point values；

(1.2) respectively selective value is L_maxFloating point values and value be L_minFloating point values as threshold value Level255 and Level0；It is right Each data point in original radar two dimension echo data, if floating point values gives gray value 255 more than Level255, less than Level0 Gray value 0 is given, to value in L_maxWith L_minBetween floating point values then carry out linear interpolation, determine its gray value, linear interpolation Formula is as follows：

$g(x,y)=\left\{\begin{array}{ccc}0,& if& \left(f\right(x,y)<Level0)\\ \frac{f(x,y)-Level0}{Level255-Level0}*255,& if& (Level0\&le;f(x,y)<Level255)\\ 255,& if& \left(f\right(x,y)>Level255)\end{array}\right.$

Wherein f (x, y) is the floating-point values of data point (x, y) place radar two dimension echo data, and g (x, y) is should after linear interpolation The corresponding gray value of point (x, y).

3. the method for claim 1, it is characterised in that the step (2) specifically includes：

(2.1) first according to target sizes and imaging resolution, using the interval under the normal course state of target as constraint, it is determined that Region of interest window length and width；The region of interest window centered on s points is investigated, the statistic of pixel value has mean μ in window_sAnd center of gravity G_s：

${\mathrm{\&mu;}}_s=\frac{1}{n}\underset{i=1}{\overset{n}{\&Sigma;}}g(x_i,y_i)$

$\stackrel{\&OverBar;}{x}=\underset{(x,y)\&Element;\mathrm{\&Omega;}}{\&Sigma;}x\&CenterDot;g(x,y)/\underset{(x,y)\&Element;\mathrm{\&Omega;}}{\&Sigma;}g(x,y)$

$\stackrel{\&OverBar;}{y}=\underset{(x,y)\&Element;\mathrm{\&Omega;}}{\&Sigma;}y\&CenterDot;g(x,y)/\underset{(x,y)\&Element;\mathrm{\&Omega;}}{\&Sigma;}g(x,y)$

$G_s=(\stackrel{\&OverBar;}{x},\stackrel{\&OverBar;}{y})$

Wherein, n is the number of pixel in region of interest window, and g (x, y) is the gray scale of (x, y) place pixel in region of interest window Value, Ω is region contained by region of interest window；

(2.2) the confidence level μ of the region of interest window centered on s points is asked for_s, and center of gravity G in window_sWith window center coordinate O_s Apart from d (G_s,O_s), wherein：

ρ_s=μ_s/[d(G_s,O_s)+1]

d(G_s,O_s) it is center of gravity G in window_sWith window center coordinate O_sBetween Europe Formula distance, wherein x, y represents respectively the row and column of image；

(2.3) region of interest window is determined in 2-D gray image according to following principles, as target area gray level image：

If being background in the region of interest window centered on s points, because background pixel value difference is less and all than relatively low, now away from From d (G_s,O_s) ≈ 0, confidence level ρ_s≈μ_sSuitable with background pixel value average, now region of interest window region is not that sense is emerging Interesting area；

If a part is background for a target part in the region of interest window centered on s points, now μ in window_sIncrease, but while window Interior center of gravity G_sIt is partial to the region of high pixel value, d (G_s,O_s) also increase, until region of interest window includes whole target, it is defined as sense Region of interest；

If including whole target in the region of interest window centered on s points, now μ_sReach the pixel of maximum, maximum and target Value is related, while center of gravity G in window_sWith window center O_sDistance reduces；When apart from d (G_s,O_s) ≈ 0 when confidence level ρ_sReach maximum, While ρ_sFor local maximum, now region of interest window region is region of interest.

4. the method for claim 1, it is characterised in that the step (3) is specially：

(3.1) determine that the segmentation threshold Th for being split to target area gray level image, segmentation threshold Th are target entropy Entropy_OWith background entropy Entropy_BThe maximum of sum, wherein：

Background entropy

Target entropy

P (i) represents corresponding grey scale level as the probability size of i, and m represents the maximum gray scale of image, P_tAnd H_tRepresent respectively background and Target in the picture pixel grey scale distribution probability and；

(3.2) target area gray level image is split according to segmentation threshold Th, obtains target area segmentation figure as O (x, y), Specially：

If the entropy of certain gray level is more than Th, then it is assumed that the gray level is target；If the entropy of certain gray level is less than Th, then it is assumed that the ash Degree level is background, so as to obtain target area segmentation figure as O (x, y).

5. the method as described in any one of Claims 1-4, it is characterised in that the step (4) is specially：

According to radar two dimension echo data and target area gray level image, using first-order difference method, the thunder of target area is extracted Up to the peak point information in two-dimentional echo data, target area peak point information matrix G (x, y) is obtained, wherein：

$G(x,y)=\left\{\begin{array}{cc}255,& if\left(f\right(x,y)-f(x,y+1)>0,f(x,y)-f(x,y-1)>0)\\ 0,& else\end{array}\right.$

(x, y) represents the point on radar two dimension echo data, and x represents distance to y represents orientation, remembers its radar two dimension echo Floating point values is f (x, y).

6. the method as described in any one of Claims 1-4, it is characterised in that the step (5) is specially：

To target area segmentation figure as O (x, y) and target area peak point information matrix G (x, y) carry out information fusion, melted Image R (x, y) after conjunction；

R (x, y)=G (x, y) O (x, y)/255

Image R (x, y) after fusion is 0/255 bianry image, and gray level is that 255 pixel is target area point in image Peak point in objective area in image is cut, these peak values is counted and is counted out K, i.e.,

$K=C_{ount}\left\{(x_i,y_i)\right|\begin{array}{c}(x_i,y_i)\&Element;R(x,y)\\ p(x_i,y_i)=255\end{array}\}$

Wherein, p (x_i,y_i) be fusion after image R (x, y) in (x_i,y_i) place gray value size.

7. method as claimed in claim 5, it is characterised in that the step (5) is specially：

To target area segmentation figure as O (x, y) and target area peak point information matrix G (x, y) carry out information fusion, melted Image R (x, y) after conjunction；

R (x, y)=G (x, y) O (x, y)/255

Image R (x, y) after fusion is 0/255 bianry image, and gray level is that 255 pixel is target area point in image Peak point in objective area in image is cut, these peak values is counted and is counted out K, i.e.,

$K=C_{ount}\left\{(x_i,y_i)\right|\begin{array}{c}(x_i,y_i)\&Element;R(x,y)\\ p(x_i,y_i)=255\end{array}\}$

Wherein, p (x_i,y_i) be fusion after image R (x, y) in (x_i,y_i) place gray value size.

8. the method as described in any one of Claims 1-4, it is characterised in that the step (6) is specially：

(6.1) peak point in the peak point information matrix of target area is sorted by size, chooses maximum front K peak point conduct The effective peak point characteristic information of target；

(6.2) binary conversion treatment is carried out to the effective peak point characteristic information of target, will this K extreme point place pixel tax Value gray level 255, rest of pixels point is entered as 0, obtains target effective peak value dot image.

9. method as claimed in claim 6, it is characterised in that the step (6) is specially：

(6.1) peak point in the peak point information matrix of target area is sorted by size, chooses maximum front K peak point conduct The effective peak point characteristic information of target；

(6.2) binary conversion treatment is carried out to the effective peak point characteristic information of target, will this K extreme point place pixel tax Value gray level 255, rest of pixels point is entered as 0, obtains target effective peak value dot image.

10. the method as described in any one of Claims 1-4, it is characterised in that the step (7) specifically includes：

(7.1) least square fitting is utilized to representing k candidate target point in target effective peak value dot image, searches out one Bar makes the minimum straight line of residual sum of squares (RSS), the i.e. axial straight line of target, and the straight line is identical with the diagonal of target；

(7.2) using the length of target, width, Diagonal Dimension as constraint, the target rectangle with target length, wide constraint is Window, allows the diagonal of window to move linearly along the axial direction of target, to find and make institute in window on the position on axial straight line Count out most the window's positions containing peak value, the position corresponding to window is target location；

(7.3) the peak point information flag outside window is false-alarm point, and peak point information contained in window determines the position of target Put.

11. methods as claimed in claim 9, it is characterised in that the step (7) specifically includes：

(7.1) least square fitting is utilized to representing k candidate target point in target effective peak value dot image, searches out one Bar makes the minimum straight line of residual sum of squares (RSS), the i.e. axial straight line of target, and the straight line is identical with the diagonal of target；

(7.2) using the length of target, width, Diagonal Dimension as constraint, the target rectangle with target length, wide constraint is Window, allows the diagonal of window to move linearly along the axial direction of target, to find and make institute in window on the position on axial straight line Count out most the window's positions containing peak value, the position corresponding to window is target location；

(7.3) the peak point information flag outside window is false-alarm point, and peak point information contained in window determines the position of target Put.

12. methods as claimed in claim 10, it is characterised in that the axial straight line tool of target is calculated in the step (7.1) Body is：

To object pixel point set V, parameter k of straight line y=kx+b, b so that residual sum of squares (RSS) are estimated Minimum, wherein (x_i,y_i)∈V。

13. methods as claimed in claim 11, it is characterised in that the axial straight line tool of target is calculated in the step (7.1) Body is：

To object pixel point set V, parameter k of straight line y=kx+b, b so that residual sum of squares (RSS) are estimated Minimum, wherein (x_i,y_i)∈V。

14. methods as described in any one of Claims 1-4, it is characterised in that the step (8) is specially：

Ask for the grey scale centre of gravity of target areaAs the energy barycenter of target area, the as key point of target, It is as follows that the grey scale centre of gravity of wherein target area asks for formula：

$\stackrel{\&OverBar;}{X}=\underset{(x,y)\&Element;T}{\&Sigma;}x\&CenterDot;f(x,y)/\underset{(x,y)\&Element;T}{\&Sigma;}f(x,y)$

$\stackrel{\&OverBar;}{Y}=\underset{(x,y)\&Element;T}{\&Sigma;}y\&CenterDot;f(x,y)/\underset{(x,y)\&Element;T}{\&Sigma;}f(x,y)$

F (x, y) is the floating point values at (x, y) place in radar two dimension echo data, and Τ is target area.

15. methods as claimed in claim 10, it is characterised in that the step (8) is specially：

Ask for the grey scale centre of gravity of target areaAs the energy barycenter of target area, the as key point of target, It is as follows that the grey scale centre of gravity of wherein target area asks for formula：

$\stackrel{\&OverBar;}{X}=\underset{(x,y)\&Element;T}{\&Sigma;}x\&CenterDot;f(x,y)/\underset{(x,y)\&Element;T}{\&Sigma;}f(x,y)$

$\stackrel{\&OverBar;}{Y}=\underset{(x,y)\&Element;T}{\&Sigma;}y\&CenterDot;f(x,y)/\underset{(x,y)\&Element;T}{\&Sigma;}f(x,y)$

F (x, y) is the floating point values at (x, y) place in radar two dimension echo data, and Τ is target area.