CN105427291A - Method for detecting vector edges of multispectral remote sensing images - Google Patents

Abstract

The invention discloses a method for detecting vector edges of multispectral remote sensing images. The method comprises the following steps: obtaining a high-spatial resolution remote sensing image; carrying out gradient vector edge detection on the high-spatial resolution remote sensing image to obtain a gradient vector characteristic image; obtaining the angle value of a maximum change or discontinuous change position point on the gradient vector characteristic image and the change rate of a corresponding position point in the direction of the angle; detecting an initial edge point set of the gradient vector characteristic image by utilizing a non-maximum suppression method according to the obtained angle value and change rate; carrying out edge contour detection to obtain an image edge contour by utilizing a double-threshold segmentation processing method according to the initial edge point set; and superposing the wave bands of the edge contour in a multidimensional color space to synthesize an edge vector and outputting the edge detection result. According to the method provided by the invention, the vector edges of multispectral remote sensing images are detected to present the normalized geographical condition of the remote sensing image edge information extraction projects, so that the truth and effectiveness of the experiment are ensured.

Description

Multi-spectrum remote sensing image vector edge detection method

Technical field

The present invention relates to a kind of multi-spectrum remote sensing image vector edge detection method.

Background technology

Rim detection is generally extracted by certain uncontinuity feature (as brightness, form and aspect, saturation degree etc.) of target in image with realizing marginal information, has a wide range of applications in remote sensing image ground object target extracts.In multispectral high spatial resolution remote sense image, due to the difference of atural object material, orientation, geometric configuration and illumination condition, reflective edges can be there is, towards edge, block edge, illumination (shade) edge, and minute surface (Gao Guang) edge; They make usually there is a large amount of pseudo-edge in edge extracting result, have a strong impact on the follow-up graphical analysis based on edge and understanding.

Famous rim detection Canny operator is well-known with its good rim detection effect, but this operator only can be applied to the edge extraction of single channel gray level image, is applied to when hyperchannel Remote Sensing Image Edge detects and then needs to carry out necessary expansion.Relative single band (single channel) remote sensing image, multiband (hyperchannel or multispectral, general wave band number is not less than three) can be used for the spectral information of rim detection in remote sensing image and feature abundanter, the feature adopted and extraction scheme need to determine neatly according to atural object classification.

The subject matter that conventional edge detection method (as Canny operator) exists when being applied to multi-spectrum remote sensing image has:

(1) restriction of wave band number and processing policy.Conventional edge detection method generally only can process single band grayscale image (or adopting the strategy processed respectively multiband image by wave band), the detection responsiveness of the corresponding ground object target of each wave band (generally can by the impact of spectral range difference) and between relationship characteristic but seldom consider in testing process, the organic connections between each wave band data of multi-spectrum remote sensing image are isolated, cause the localized loss of edge detection results, effective edge extraction cannot be carried out for all kinds of atural object.

(2) classic method is to the suitability problem of multispectral high spatial resolution remote sense image data.Multispectral high spatial resolution remote sense image data have following characteristics, data volume is large, atural object geometry is abundant with attribute detailed information, the inner spectrum heterogeneity of the even same atural object of similar atural object is higher, " the different spectrum of jljl " is general with " foreign matter is with composing " phenomenon, ground object target space structure general layout is complicated, edge transitional region is various, shade and tiny atural object serious interference etc.These problems cause usually there is a large amount of pseudo-edge in the edge detection results utilizing classic method to obtain, even sometimes, effectively cannot to extract polymorphic type ground object target edge and layering (class) is analyzed and identifies simultaneously, to have a strong impact on effectively carrying out based on follow-up works such as the multi-spectrum remote sensing image analysis of marginal information and understandings.

(3) edge detection results that conventional edge detection method is corresponding to each wave band remote sensing image lacks correlation analysis.Multiband high spatial resolution remote sense image is the differentiation record of all kinds of spectral characteristic of ground in different spectral range, and different atural object is obvious at the spectral response degree difference of different-waveband, and obvious each wave band image will produce different edge detection results.Therefore each band edge testing result lacks correlation analysis and will cause marginal information disappearance and accuracy of detection decline.

Summary of the invention

For the problems referred to above, the object of the present invention is to provide a kind of multi-spectrum remote sensing image vector edge detection method.

For achieving the above object, a kind of multi-spectrum remote sensing image vector edge detection method of the present invention, comprises the following steps:

Obtain the high spatial resolution remote sense image with n wave band;

Gradient vector rim detection is carried out to high spatial resolution remote sense image and obtains gradient vector characteristic image;

Obtain angle value and the rate of change of correspondence position point on this angle direction of maximum change on gradient vector characteristic image or discontinuous change location point;

Non-extreme value is utilized to suppress the initial edge point set of method determination gradient vector characteristic image according to the angle value obtained and rate of change;

Utilize Double Thresholding Segmentation facture to carry out edge contour according to initial edge point set and detect acquisition image edge profile;

Each wave band stacked synthesis edge vectors in multi-dimensional color space of edge profile, and export edge detection results.

Preferably, non-extreme value is utilized to suppress the step of the initial edge point set of method determination gradient vector characteristic image to comprise according to the angle value obtained and rate of change:

Tie up all pixel points in gradient image by band selection n, suppress method to carry out refinement to the non-extreme value of ridge band that the larger pixel of Grad is formed;

If the Grad on this pixel orientation angle θ is local maximum, be then left preliminary edge point;

Otherwise, this pixel is set to non-edge point.

Preferably, the step utilizing Double Thresholding Segmentation facture to carry out edge contour detection acquisition image edge profile according to initial edge point set comprises:

Selected two Grads threshold Y _hand Y _s;

Suppress to remove Grad in result by wave band in non-extreme value and be less than Y _hpixel point, and obtain strong edge point set Q;

Based on Q, marginal point is connected into initial profile;

Initial profile is searched for, at Grad between Y _hwith Y _snon-extreme value suppress in result, to find the marginal point that can be connected to current endpoint;

Utilize recurrence tracking between Y _hwith Y _sgrad in collect edge, until by Y _hin all discontinuous phases connect.

Beneficial effect of the present invention is:

The present invention by carrying out multi-spectrum remote sensing image vector edge detecting the normalization geographical environment that can present comparatively all sidedly residing for Remote Sensing Image Edge information extraction engineering, thus ensure that this true and validity of testing.

Accompanying drawing explanation

Fig. 1 is the structured flowchart of detection method described in the embodiment of the present invention;

Fig. 2 is embodiment of the present invention QuickBird high spatial resolution satellite remote-sensing image figure;

Fig. 3 carries out average drifting filtering Yunnan snub-nosed monkey result schematic diagram to Fig. 1;

Fig. 4 is the vector edge schematic diagram that the present invention extracts in rgb color space;

Fig. 5 is the vector edge schematic diagram that the present invention extracts in IHS color space;

Fig. 6 is the vector edge schematic diagram that the present invention extracts in YIQ color space;

Fig. 7 is the vector edge schematic diagram that the present invention extracts in YUV color space;

Fig. 8 is the vector edge schematic diagram that the present invention extracts in CIELUV color space;

Fig. 9 is the weight vectors edge schematic diagram that the present invention extracts in CIELUV (L)-YIQ (Y)-IHS (I) color space;

Figure 10 is the weight vectors edge schematic diagram that the present invention extracts in CIELUV.RGB (LB)-YIQ.RGB (YG)-IHS.RGB (IR) color space.

Embodiment

Below in conjunction with Figure of description, the present invention will be further described.

A kind of multi-spectrum remote sensing image vector edge detection method of the present invention, comprises the following steps.

(1) obtain the high spatial resolution remote sense image of n wave band, and carry out Yunnan snub-nosed monkey;

High spatial resolution remote sense image is due to the spatial information expressive ability of its high degree of detail, its its inner geometry detailed information while effective expression atural object semantic objects marginal information is also occurred with the form of noise (relative to this ground object target size), and " scale granular " phenomenon is very outstanding; And multispectral color information also shows obvious nonuniformity in semantic objects inside, " the different spectrum of jljl " phenomenon is also very outstanding, for rim detection work brings inconvenience.For this reason, average drifting filtering technique is adopted to carry out Yunnan snub-nosed monkey (not getting rid of other filtering techniques herein, as bilateral filtering, anisotropic diffusion filtering etc.), to reach the object removing above-mentioned noise.

(2) gradient vector rim detection is carried out to high spatial resolution remote sense image and obtain gradient vector characteristic image;

The multi-spectrum remote sensing image with n wave band (passage) can be expressed as G (x, y)=(B ₁, B ₂..., B _n) ^t.N at position (x, y) place ties up gradient vector T (x, y) and can be expressed as:

$T(x,y)=\left(\begin{array}{cc}\frac{\∂B_1}{\∂x}& \frac{\∂B_1}{\∂y}\\ \frac{\∂B_2}{\∂x}& \frac{\∂B_2}{\∂y}\\ \mathrm{...}& \mathrm{...}\\ \frac{\∂B_n}{\∂x}& \frac{\∂B_n}{\∂y}\end{array}\right)=\left(\begin{array}{c}t\left(B_1\right)\\ t\left(B_2\right)\\ \mathrm{...}\\ t\left(B_n\right)\end{array}\right)$ (formula 1)

(3) angle value and the rate of change of correspondence position point on this angle direction of maximum change on gradient vector characteristic image or discontinuous change location point is obtained;

There is in G (x, y) the eigenvector V of maximum change or uncontinuity direction (being generally present position, edge) available character pair value ^tv is expressed as:

$V^{T}V=\left(\begin{array}{cc}{u}^{T}u& u^{T}v\\ v^{T}u& v^{T}v\end{array}\right)$ (formula 2)

Wherein, n n dimensional vector n u and v can be expressed as:

$u={\left(\begin{array}{cccc}\frac{\∂B_1}{\∂x}& \frac{\∂B_2}{\∂x}& \mathrm{...}& \frac{\∂B_n}{\∂x}\end{array}\right)}^T$ (formula 3)

$v={\left(\begin{array}{cccc}\frac{\∂B_1}{\∂y}& \frac{\∂B_2}{ay}& \mathrm{...}& \frac{\∂B_n}{\∂y}\end{array}\right)}^T$ (formula 4)

In G (x, y), maximum change or uncontinuity direction Useable angles θ are expressed as:

$\mathrm{\θ}(x,y)=\frac{1}{2}arctan\frac{u^{T}v+v^{T}u}{u^{T}u-v^{T}v}$ (formula 5)

Point G (x, y) the rate of change Grad (x, y) on θ direction is expressed as:

$Grad(x,y)={\{\frac{1}{2}\[(u^{T}u+v^{T}v)+(u^{T}u-v^{T}v)\cos 2\mathrm{\θ}+(u^{T}v+v^{T}u)\sin 2\mathrm{\θ}\]\}}^{\frac{1}{2}}$ (formula 6)

(4) non-extreme value is utilized to suppress the initial edge point set of method determination gradient vector characteristic image according to the angle value obtained and rate of change;

Tie up all pixel points in gradient image by wave band traversal n, suppress method to carry out refinement to the non-extreme value of ridge band that the larger pixel of Grad is formed; If the Grad on this pixel orientation angle θ is local maximum, be then left preliminary edge point, otherwise this pixel is set to non-edge point.Because each wave band is to the difference of different atural object type light spectrum signature responsiveness, the initial edge points that general each wave band correspondence is formed also presents otherness.

(5) utilize Double Thresholding Segmentation facture to carry out edge contour according to initial edge point set and detect acquisition image edge profile;

Selected two Grads threshold Y _hand Y _s, generally have Y _s=0.4Y _h, Y _hby obtaining the statistical study of " initial edge point set ".First, suppress to remove Grad in result by wave band in non-extreme value and be less than Y _hpixel point, and obtain strong edge point set Q.Then, based on Q, marginal point is connected into initial profile, and initial profile generally has interruption.When searching profile end points, algorithm at Grad between Y _hwith Y _snon-extreme value suppress to continue in result to find the marginal point that can be connected to current endpoint; Utilize recurrence tracking between Y _hwith Y _sgrad in constantly collect edge, until by Y _hin all interruptions all couple together, and be interrupted threshold value can be given according to actual conditions.

(6) each wave band of edge profile stacked synthesis edge vectors in multi-dimensional color space, and export edge detection results.

Improved Canny operator based on output convergence strategy, the vector edges cause each band edge component in multi-dimensional color space is obtained by feature level " coupling merges " with the form being similar to stacked synthesis; The calculating of each band edge component need complete respectively in multicolour space, and obtains the edge detection results of corresponding color component implication.

Bidding amount E _ifor the weighted comprehensive edge component at remote sensing image vector edge, then by all weighted comprehensive edge component E _ithe vector edge E obtained by stacked synthesis _v(E _i) and weighted comprehensive scalar edge E _scan be expressed as:

E _v(E _i)=(E ₁e ₂... E _i) i=1,2 ..., n; (formula 7)

β ₁+ β ₂+ ... + β _i=1, β _i>=0, k≤n; (formula 8)

α ₁+ α ₂+ ... + α _i=1, α _i>=0, k≤n; (formula 9)

Wherein edge component B _ifor the edge detection results that each wave band (or different color space characteristics component) is corresponding; α _ifor corresponding to edge component B _iweight, i.e. E _iby k band edge component B _iobtain through weighted comprehensive; β _ithen for corresponding to the weight of weighted comprehensive edge component when extracting scalar edge; E _vthe edge component E comprised _ican by E _sform.

In order to show the technique effect of this method in practical engineering application, adopting QuickBird high spatial resolution remote sense image data to carry out engineering experiment and showing.

One, experimental data

Following Fig. 2, raw experimental data is 3 passages (red, green, blue) the multiband QuickBird high spatial resolution satellite remote-sensing image of 1024 × 1024 pixels, and size of data is 3M; There is the ground object target such as road, buildings, river, trees, bare area, meadow in image geographic range, present the normalization geographical environment residing for Remote Sensing Image Edge information extraction engineering comparatively all sidedly, thus ensure that this true and validity of testing.

Two, experimental result

Image semantic classification.Following Fig. 3 is the result obtained after carrying out average drifting filtering Yunnan snub-nosed monkey to Fig. 2.

Following Fig. 4-Fig. 8 is respectively this method carries out the acquisition of vector rim detection experimental result at RGB, IHS, YIQ, YUV, CIELUV color space.

Following Fig. 9-Figure 10 is respectively this method is weighted the acquisition of vector rim detection experimental result at RGB, IHS, YIQ, YUV, CIELUV color space.

In fig .9, respectively by edge detection results corresponding for the color component L in CIELUV color space, the edge detection results that the color component Y of YIQ color space is corresponding, the edge detection results that the color component I of IHS color space is corresponding, merged by other coupling of feature level and form weight vectors edge LYI, during each component L, Y, I synthesis, triangular weight is impartial each other.

In Fig. 10, respectively by edge detection results corresponding for color component L and the B in CIELUV and rgb color space, the edge detection results that color component Y and the G of YIQ and rgb color space is corresponding, the edge detection results that color component I and the R of IHS and rgb color space is corresponding, merged by other coupling of feature level and form weight vectors edge C IELUV.RGB (LB)-YIQ.RGB (YG)-IHS.RGB (IR), when each component L and B synthesizes, weight is impartial each other, during Y and G synthesis, weight is impartial each other, during I and R synthesis, weight is impartial each other, when LB and YG and IR three synthesize, weight is also impartial each other.

Above; be only preferred embodiment of the present invention, but protection scope of the present invention is not limited thereto, is anyly familiar with those skilled in the art in the technical scope that the present invention discloses; the change that can expect easily or replacement, all should be encompassed within protection scope of the present invention.Therefore, the protection domain that protection scope of the present invention should define with claim is as the criterion.

Claims (3)

1. a multi-spectrum remote sensing image vector edge detection method, is characterized in that, comprises the following steps:

Obtain the high spatial resolution remote sense image with n wave band;

Gradient vector rim detection is carried out to high spatial resolution remote sense image and obtains gradient vector characteristic image;

Obtain angle value and the rate of change of correspondence position point on this angle direction of maximum change on gradient vector characteristic image or discontinuous change location point;

Non-extreme value is utilized to suppress the initial edge point set of method determination gradient vector characteristic image according to the angle value obtained and rate of change;

Utilize Double Thresholding Segmentation facture to carry out edge contour according to initial edge point set and detect acquisition image edge profile;

Each wave band stacked synthesis edge vectors in multi-dimensional color space of edge profile, and export edge detection results.

2. detection method according to claim 1, is characterized in that, utilizes non-extreme value to suppress the step of the initial edge point set of method determination gradient vector characteristic image to comprise according to the angle value obtained and rate of change:

Tie up all pixel points in gradient image by band selection n, suppress method to carry out refinement to the non-extreme value of ridge band that the larger pixel of Grad is formed;

If the Grad on this pixel orientation angle θ is local maximum, be then left preliminary edge point;

Otherwise, this pixel is set to non-edge point.

3. detection method according to claim 1, is characterized in that, comprises according to the step that initial edge point set utilizes Double Thresholding Segmentation facture to carry out edge contour detection acquisition image edge profile:

Selected two Grads threshold Y _hand Y _s;

Suppress to remove Grad in result by wave band in non-extreme value and be less than Y _hpixel point, and obtain strong edge point set Q;

Based on Q, marginal point is connected into initial profile;

Initial profile is searched for, at Grad between Y _hwith Y _snon-extreme value suppress in result, to find the marginal point that can be connected to current endpoint;

Utilize recurrence tracking between Y _hwith Y _sgrad in collect edge, until by Y _hin all discontinuous phases connect.