CN102184536A - Method and system for extracting straight line and/or line segment end points from image - Google Patents

Abstract

The invention relates to a method and a system for extracting straight line and/or line segment end points from an image. The method comprises the following steps of: 1, performing spherical projection on the image; 2, extracting a circular arc from the large circle of a spherical surface; and 3, acquiring a straight line equation and/or the line segment end points by using the circular arc. The technical scheme provided by the invention is simple and practicable and high in noise resisting capability, and straight line parameters and/or the straight line end points can be detected at the same time.

Description

A kind of method and system of extracting image cathetus and/or line segment end points

Technical field

The present invention relates to digital image processing field, relate in particular to a kind of image cathetus and extract and extremity of segment point detecting method and system.

Background technology

Straight-line detection is widely used in computer vision and the Flame Image Process, such as target registration, target following and Target Recognition.Image packets for the interior of building outside contains a large amount of straight line information, and the information of straight line has important scientific research value and practical significance for the three-dimensional reconstruction of scene under the detection of straight lines acquisition different visual angles.Straight-line detection be used in take photo by plane or remote sensing images in the bridge, road, river, shore line, airfield runway, local horizon, buildings rim detection, and be applied in the Road Detection of vehicle-mounted autonomous navigation system.

Take photo by plane airfield runway in the image, local horizon detects in the middle of the independent landing navigation that is mainly used in aircraft, in the area image of aircraft camera place information such as local horizon and airfield runway are arranged, local horizon and position and the form of airfield runway in image can change when attitude of flight vehicle changed, determine to change corresponding relation between the two, thereby utilize local horizon and the airfield runway position form in image to find the solution the automatic navigation control that attitude of flight vehicle is realized aircraft.The image information of obtaining based on the intelligent vehicle autonomous navigation system of vision is the road information of its traffic direction, determines that by the edge of road in the detected image position of vehicle in road come the course angle of analytical calculation vehicle to realize Autonomous Control.Contain in airfield runway, local horizon, the road image, between zone such as runway, local horizon, road and the peripheral region edge feature is arranged, these edges mostly are linear feature, finish the method that detection need be adopted straight-line detection.

The method of straight-line detection mainly contains two classes: based on the straight-line detection of Hough (Hough) conversion with based on the straight-line detection of chain code.

Hough conversion basic thought is to utilize a little---the line antithesis, with the satisfied equation of constraint of point on the image space straight line, become parametric form, at the line that the some correspondence in the image space intersects in parameter space, the Hough conversion utilizes this relation the straight-line detection problem to be become the problem of asking extreme value in the parameter space.The Hough transform method is simple, even and straight line has the parameter of interruption also can detect, but exist the end points and the length information that can not obtain the image middle conductor, calculated amount is big, becomes problem such as big because the straight line parameter that the evaluated error that parameter space resolution is introduced calculates is determined the point on the straight line with the increase of straight length.Having on the basis of Hough conversion from improving improved method on its computing velocity, the method that adopts contrary Hough conversion is arranged, but these methods can not well solve the problem of fast detecting straight line parameter and line segment end points.The straight line parameter that the evaluated error that parameter space resolution is introduced calculates determines that the point on the straight line becomes big problem with the increase of straight length, does not also have effective ground method to solve.

Chain code is that the position relation of 8 of neighborhoods a bit and on every side in the image adopts coded representation.Point represents to present certain rules with the chain code sequence on the straight line, and this rule is freeman (Freeman) criterion.Straight-line detection based on chain code obtains a series of chain code sequences of image exactly, and the sequence that meets the Freeman criterion is a straight line.The Chain Code Detection method need relate to definite problem of initial point, and this method is for noise effect straight-line detection poor effect is arranged in addition.

Summary of the invention

The invention provides a kind of gray level image cathetus and extract and extremity of segment point detecting method and system, to overcome existing method calculation of complex, easy affected by noise, the problem that can not detect straight line parameter and straight line end points simultaneously.

The invention provides a kind of method of extracting image cathetus and/or line segment end points, comprising:

Step 1 is carried out spherical projection with image;

Step 2 detects the circular arc on the sphere great circle;

Step 3 utilizes described circular arc to obtain straight-line equation and/or line segment end points.

In one example, also comprise step 5 before the step 1, extract edge of image; In the step 1, the point on the edge is carried out spherical projection.。

In one example, also comprise step 6 before the step 5, image is carried out filtering.

In one example, step 1 comprises:

Step 10, setting up with the true origin is that the centre of sphere, sphere diameter are 1 ball;

Step 11 with the projection model of true origin as projection centre, and makes the geometric center of intramarginal image and sphere tangent;

Step 12 projects to the point on the image on the sphere one to one.

In one example, step 2 comprises:

Step 20, the normal vector of detection sphere great circle;

Step 21 utilizes the normal vector of sphere great circle to separate point on the sphere great circle;

Step 22 utilizes point on the sphere great circle to obtain the end points of the circular arc on the sphere great circle.

In one example, in the step 21, the inner product between the point coordinate vector on the normal vector that utilizes the sphere great circle and the sphere is separated the point on the sphere great circle; If inner product is less than preset value, then the point on the sphere is the point on the sphere great circle.

In one example, in the step 22, utilize the extreme value of inner product between the intersecting point coordinate vector of the normal vector of sphere great circle and ball and the point coordinate vector on the sphere great circle to obtain the end points of the circular arc on the sphere great circle.

In one example, establish point coordinate Q:(q on the image _xq _y), the geometric center point T:(t of image _xt _y), then the coordinate of projection model mid point is

The point on the connection layout picture and the intersection point of projection centre and sphere are the projection of point on sphere on the image, and subpoint is designated as: P:(p _xp _yp _z) satisfy:

Then

$P:\left(\begin{array}{ccc}{p}_x& p_y& p_z\end{array}\right)={(q_x^{p^2}+q_y^{p^2}+q_z^{p^2})}^{-\frac{1}{2}}\left(\begin{array}{ccc}{q}_x^p& q_y^p& q_z^p\end{array}\right);$

The normal vector of sphere great circle and the intersection point of ball at the coordinate figure on x plane more than or equal to zero.

In one example, in the step 3:

If the normal vector V=(V of sphere great circle _xV _yV _Z), sphere great circle place plane equation is V _xX+V _yY+V _ZZ=0, plane of delineation equation are z=-1, and then straight-line equation is V _xX+V _yY=V _Z

If sphere orthodrome end points coordinate P:(p _xp _yp _z), then extremity of segment point coordinate is:

The invention provides a kind of system that extracts image cathetus and/or line segment end points, comprising:

Projection module is used for image is carried out spherical projection;

Module arc extraction module is used to extract the circular arc on the sphere great circle;

Computing module is used to utilize described circular arc to obtain straight-line equation and/or line segment end points.

In one example, also comprise the edge extracting module, be used to extract edge of image.

In one example, also comprise filtration module, be used for image is carried out filtering.

Technical scheme provided by the invention is simple, and antinoise power can be strong, and can detect straight line parameter and/or straight line end points simultaneously.

Description of drawings

Come the present invention is described in further detail below in conjunction with accompanying drawing, wherein:

Fig. 1 is that straight line provided by the invention extracts and extremity of segment point detecting method process flow diagram

Fig. 2 is the spherical projection illustraton of model;

Fig. 3 is a spherical arc end-point detection schematic diagram.

Embodiment

Method provided by the invention comprises as shown in Figure 1:

Step 101 is carried out pre-service to image.Original image is designated as f (x y), and the method for medium filtering is adopted in filtering, is expressed as:

S is a window, and the window of selecting for use is 3 * 3.View data can exist to disturb produce noise in obtaining transmission course, the noise of pre-service in can the filtering image.

Step 102 is carried out rim detection to image.Adopt the method for Tuscany (Canny) rim detection to detect edge of image, the image behind the edge extracting is carried out binary conversion treatment, the pixel at edge is recorded as 1, and non-marginal point is 0.Straight line is included among the edge, and data volume reduces behind the extraction edge, thereby can improve the computing velocity of algorithm.

Step 103 is carried out spherical projection to the marginal point in the binary image after handling.Choose in the space a bit as projection centre, the spot projection on the image on the same straight line to the sphere after, in the same plane, the intersection of plane and sphere is circle.Straight line on the original image is a finite length, and projecting on the sphere is one section circular arc.Circular arc and straight line are corresponding one by one, detect circular arc and just detect straight line.The different circular arcs that obtain of position relation on the projection model and the plane of delineation and projecting plane are also different.Choose the position relation on suitable subpoint, projecting plane and image and projecting plane, make that the straight line on the image is one section circular arc on the sphere great circle in the projection on the sphere.

In order to realize spherical projection, at first to set up the projection sphere.Setting up with the true origin among the present invention is the centre of sphere, and sphere diameter is 1 ball.Sphere is designated as S, and the coordinate of putting on the sphere is: N:(n _xn _yn _z), have

Secondly, set up projection model, as shown in Figure 1, wherein C is a projection centre, and O is the centre of sphere, and C overlaps with O, and S is that the centre of sphere is the sphere of O, Q ^PBe the point on the image, L is the line of picture point and projection centre C, and P is that the intersection point of L and sphere is Q ^PSubpoint on sphere.Among the present invention, projection centre is got the centre of sphere (true origin), and the plane of delineation and sphere are tangent, and real image is a limited area, makes the geometric center of image-region and point of contact overlap.

At last, utilize projection model with man-to-man the projecting on the sphere of point on the plane of delineation.Pixel coordinate in the image-region, Q:(q _xq _y), geometric center point T:(t _xt _y), the coordinate of picture point is in the projection model: The intersection point of point and projection centre and sphere is the projection of point on sphere on the image on the connection layout picture, and subpoint is designated as: P:(p _xp _yp _z) satisfy:

$\frac{q_x^p}{p_x}=\frac{q_y^p}{p_y}=\frac{q_z^p}{p_z},$ Then $P:\left(\begin{array}{ccc}{p}_x& p_y& p_z\end{array}\right)={(q_x^{p^2}+q_y^{p^2}+q_z^{p^2})}^{-\frac{1}{2}}\left(\begin{array}{ccc}{q}_x^p& q_y^p& q_z^p\end{array}\right).$

Step 104 detects sphere great circle parameter.The normal vector on plane, great circle place is as the parameter of great circle.The inner product of the normal vector of each point is zero on the planar process vector sum great circle.The coordinate of planar process vector is a parameter to be asked, and the vector of the point on the great circle is a known quantity.Utilizing inner product is zero can set up equation of constraint.Can obtain the normal vector on the plane at great circle place by the method for sphere Hough transformation.

Particularly, set up the normal equation of constraint of sphere great circle, utilize Hough change detection great circle parameter, the parameter of great circle is the normal vector on plane, great circle place.The inner product of vectors of putting on the normal vector on plane, great circle place and the great circle is zero, chooses unit normal vector as parameter, and the point on the great circle is a variable.The plane at note sphere great circle place is π, and unit normal vector is: V=(v _xv _yv _z), the coordinate vector of putting on the great circle: P:(p _xp _yp _z), equation of constraint:

(v _x?v _y?v _z)·(p _x?p _y?p _z)＝0。

Utilize the random Hough transformation on the sphere in parameter space, to ask extreme value to determine parameter.

A) two points of stochastic sampling on sphere are remembered 2 coordinate vector: P _iP _j, the cross product of compute vector, P _Ij=P _j* P _j, repeated sampling obtains the cross product vector space N time, and the cross product vector space is the great circle parameter space, if there be m the not orthodrome of coplanar on the sphere, then naming a person for a particular job with this m great circle parameter at parameter space is that the center becomes clustering distribution.

B) utilize the method for K rank average to calculate the cluster centre of big space of circles.The value at center is the great circle normal vector.Note: V ¹V ²V ^m

Step 106 detects end points.By conformal spherical projection, point on the image on the straight line is being on the sphere on the same great circle, the parameter that detects the sphere great circle just detects the straight line parameter, sphere great circle parameter utilizes inner product of vectors can faster isolate the point that belongs on the sphere great circle after determining, straight line on the image is the line segment of finite length, is one section arc on the sphere great circle after these line segments project on the sphere.Set reference vector, ask for every direction vector and the angle of reference vector on the arc, by the rule of angle monotone variation, two extreme values of angle are the end points of arc, and the end points of this arc is corresponding to the end points of straight line.This method is obtained the detection of straight line end points by the method for the computing of matrix, its quick and precisely property improve greatly.

In order to detect end points, need to separate the point on the sphere great circle.After obtaining the normal vector on plane, sphere great circle place, separating the some employing inner product that belongs to this great circle on the sphere is zero equation.Because the influence of sum of errors image resolution ratio, inner product equation of constraint are zero inapplicable, need be adjusted into one near zero threshold value, be point on the great circle less than the point of threshold value.

Particularly, calculate the inner product of umbilical point and great circle parameter, the coordinate vector P that puts on the sphere ₁P ₂P _n, note The C of great circle ^jThe normal vector parameter is V ^j, as shown in Figure 3: If g _i≤ τ (threshold tau is less than 0.015) is P then _i∈ C ^j

After isolating the point on the sphere great circle, detect sphere orthodrome end points.Particularly, great circle C ^jBe designated as with the intersection point vector of sphere: (intersection point has two, selects for use

), the point on the isolated great circle is designated as: P ¹P ²P ^k,

If: $P^f{\left(P^0\right)}^T=\underset{i=\mathrm{1,2},...k}{\max }\left(P^i{\left(P^0\right)}^T\right),$ $P^e{\left(P^0\right)}^T=\underset{i=\mathrm{1,2},...k}{\min }\left(P^i{\left(P^0\right)}^T\right)$

P then ^f, P ^eEnd points for line segment.

Step 107, carry out straight line parameter and end points and calculate.The plane at great circle place and the intersection of the plane of delineation are the straight line in the image, set up projection model and coordinate system after, plane of delineation equation is fixed, plane, great circle place can obtain by a some French equation on the sphere.The equation on two planes goes the Z value just can get the equation of straight line.End points on the great circle is directly tried to achieve by the projective transformation transforming relationship.

Particularly, utilize the transformation relation of umbilical point and straight line to restore straight line in the original image.

Sphere great circle place planar process vector: V=(V _xV _yV _Z),

Sphere great circle place plane equation is: V _xX+V _yY+V _ZZ=0,

Plane of delineation equation is: z=-1,

Then the equation of straight line is: V _xX+V _yY=V _Z

Sphere orthodrome end points coordinate is: P:(p _xp _yp _z), then

Point coordinate on the orthodrome end points correspondence image is:

The above only is a preferred implementation of the present invention, but protection domain of the present invention is not limited thereto.Any those skilled in the art all can carry out suitable change or variation to it in technical scope disclosed by the invention, and this change or variation all should be encompassed within protection scope of the present invention.

Claims (12)

1. a method of extracting image cathetus and/or line segment end points is characterized in that, comprising:

Step 1 is carried out spherical projection with image;

Step 2 is extracted the circular arc on the sphere great circle;

Step 3 utilizes described circular arc to obtain straight-line equation and/or line segment end points.

2. the method for claim 1 is characterized in that, also comprises step 5 before the step 1, extracts edge of image; In the step 1, the point on the edge is carried out spherical projection.

3. method as claimed in claim 2 is characterized in that, also comprises step 6 before the step 5, and image is carried out filtering.

4. method as claimed in claim 2 is characterized in that step 1 comprises:

Step 10, setting up with the true origin is that the centre of sphere, sphere diameter are 1 ball;

Step 11 with the projection model of true origin as projection centre, and makes the geometric center of intramarginal image and sphere tangent;

Step 12 projects to the point on the image on the sphere one to one.

5. method as claimed in claim 4 is characterized in that step 2 comprises:

Step 20, the normal vector of detection sphere great circle;

Step 21 utilizes the normal vector of sphere great circle to separate point on the sphere great circle;

Step 22 utilizes point on the sphere great circle to obtain the end points of the circular arc on the sphere great circle.

6. method as claimed in claim 5 is characterized in that, in the step 21, the inner product between the point coordinate vector on the normal vector that utilizes the sphere great circle and the sphere is separated the point on the sphere great circle; If inner product is less than preset value, then the point on the sphere is the point on the sphere great circle.

7. method as claimed in claim 5 is characterized in that, in the step 22, utilizes the extreme value of inner product between the intersecting point coordinate vector of the normal vector of sphere great circle and ball and the point coordinate vector on the sphere great circle to obtain the end points of the circular arc on the sphere great circle.

8. method as claimed in claim 7 is characterized in that, establishes the point coordinate Q:(q on the image _xq _y), the geometric center point T:(t of image _xt _y), then the coordinate of projection model mid point is

The point on the connection layout picture and the intersection point of projection centre and sphere are the projection of point on sphere on the image, and subpoint is designated as: P:(p _xp _yp _z) satisfy:

Then

$P:\left(\begin{array}{ccc}{p}_x& p_y& p_z\end{array}\right)={(q_x^{p^2}+q_y^{p^2}+q_z^{p^2})}^{-\frac{1}{2}}\left(\begin{array}{ccc}{q}_x^p& q_y^p& q_z^p\end{array}\right);$

The normal vector of sphere great circle and the intersection point of ball at the coordinate figure on x plane more than or equal to zero.

9. method as claimed in claim 8 is characterized in that, in the step 3:

If the normal vector V=(V of sphere great circle _xV _yV _Z), sphere great circle place plane equation is V _xX+V _yY+V _ZZ=0, plane of delineation equation are z=-1, and then straight-line equation is V _xX+V _yY=V _Z

If sphere orthodrome end points coordinate P:(p _xp _yp _z), then extremity of segment point coordinate is:

10. a system that extracts image cathetus and/or line segment end points is characterized in that, comprising:

Projection module is used for image is carried out spherical projection;

Module arc extraction module is used to extract the circular arc on the sphere great circle;

Computing module is used to utilize described circular arc to obtain straight-line equation and/or line segment end points.

11. system as claimed in claim 10 is characterized in that, also comprises the edge extracting module, is used to extract edge of image.

12., it is characterized in that as claim 10 or 11 described systems, also comprise filtration module, be used for image is carried out filtering.

Images