CN106934832A - A kind of simple straight line automatic positioning method towards vision line walking - Google Patents

Abstract

The invention discloses a kind of simple straight line automatic positioning method towards vision line walking, belong to the machine vision applications fields such as robot, dolly line walking and unmanned plane cruise, when solving to be described video using linear feature in existing line walking technology, extract all straight line information in image, ignore contacting between frame and frame, so as to increased the time of straight line positioning, and the result for obtaining is caused to have the problem of larger error.The present invention based on centre of form formula and point-slope form straight-line equation, by straight line segmentation into two parts, seeks position of form center, and then obtain slope and the position of straight line respectively in scene bianry image.It is different from traditional cut-off rule and region immobilizes pattern, the present invention takes the positioning result with reference to former frame straight line mark, automatically updates cut-off rule equation and region, so as to reach the effect of automatic vision positioning.

Description

A kind of simple straight line automatic positioning method towards vision line walking

Technical field

A kind of simple straight line automatic positioning method towards vision line walking, it is fast automatic for being carried out to the straight line in video Positioning, and equivalent straight line equation is extracted, belong to the machine vision applications fields such as robot, dolly line walking and unmanned plane cruise.

Background technology

Linear feature as highly important middle level descriptor in video, in characteristic matching, aviation and navigation image analysing computer, The fields such as machine learning suffer from important researching value.

In currently for the technology in the machine vision applications fields such as robot, dolly line walking and unmanned plane cruise, use more Directly extract the straight line of a certain frame in video, conventional Straight Line Extraction such as Hough transformation (Hough Transform), phase Marshalling algorithm etc., the position and slope guided robot, dolly and unmanned plane for reusing extracted straight line carry out line walking or cruise.

Algorithm and phase coding sequences algorithm etc. are before the use for Straight Line Extraction such as Hough transformation (Hough Transform) Often also need to carry out Canny rim detections, Canny edge detection operators are that John F.Canny were developed in 1986 One multistage edge detection algorithm, effect is preferable.

The specific method of Hough transformation (Hough Transform) algorithm line walking is：Pretreatment image is used first Canny operators carry out rim detection, obtain the edge image of binaryzation；Then will be in a space with coordinate space transformations Curve or straight line with same shape are mapped on a point of another coordinate space and form peak value；Then by threshold value Reason, the linear order that acquisition Hough transformation (Hough Transform) responds larger several peak points representative is defeated as a result Go out；The respective slope of linear order and position are finally obtained, guided robot, dolly and unmanned plane carry out line walking or cruise.

The specific method of phase coding sequences algorithm line walking is：Edge inspection is carried out using Canny operators to pretreatment image first Survey, obtain the edge image of binaryzation；Then the gradient direction of edge image is calculated, phase coding sequences are carried out；Then edge is connected Position is adjacent on image and phase identical pixel, obtains a series of linear orders；Finally obtain linear order respective tiltedly Rate and position, guided robot, dolly and unmanned plane carry out line walking or cruise.

Both the above Straight Line Extraction ignores contacting between frame and frame, and is searched based on rim detection and full figure Rope, extracts all possible straight line on a certain two field picture, wherein not only include target line, and including unrelated straight line.And it is right In line walking technical field, handled object is typically video, and contact is certainly existed between each frame of video, and ignores frame Contacting between frame, each two field picture of isolated treatment, such as both the above algorithm, what is brought is Time ＆ Space Complexity Rise the increase with error.

The content of the invention

The present invention provides a kind of simple straight line automatic positioning method towards vision line walking, solution for above-mentioned weak point When being described to video using linear feature in certainly existing line walking technology, all straight line information in image are extracted, ignored Contacting between frame and frame, so as to increased the time of straight line positioning, and causes the result for obtaining to have the problem of larger error.

The technical solution adopted by the present invention is as follows：

A kind of simple straight line automatic positioning method towards vision line walking, it is characterised in that comprise the following steps：

Step 1：Video is obtained, the initial position and gradient identified according to the straight line to be positioned in video, initialization The center point P of cut-off rule equation_ex, slope k_exWith parameter param_ex, obtain cut-off rule equation；

Step 2：The first two field picture f is read from video_t(x,y)；

Step 3：To the first two field picture f_t(x, y) carries out threshold process, obtains bianry image b (x, y)；

Step 4：In bianry image b (x, y), obtain with point P_exIt is the center of circle, R_minIt is the cut zone C (P of radius_ex, R_min)；

Step 5：According to cut-off rule equation, cut-off rule is obtained, using cut-off rule by cut zone C (P_ex,R_min) it is divided into C₁ And C₂Two regions, judge that straight line is identified whether in C₁And C₂In two regions, if it is, performing step 6, step is otherwise performed Rapid 2, the next frame in treatment video；

Step 6：Respectively in C₁And C₂Inside obtain centre of form P₁And P₂, according to centre of form P₁And P₂Calculate the center of equivalent straight line equation Point P_tmp, slope K^DWith parameter paramD, equivalent straight line equation is obtained, wherein, when parameter paramD is 0, represent that equivalent straight line hangs down It is straight in x-axis, and paramD represents that equivalent straight line is not orthogonal to x-axis when being 1；

Step 7：Judge whether video terminates, if terminated, equivalent straight line is obtained according to equivalent straight line equation, will be equivalent straight Time and again positioning result is stored in text line, otherwise, calculates the midpoint P of equivalent straight line and bianry image b (x, y) intersection point_c, Update the center point P of cut-off rule_exIt is P_c, update the slope k of cut-off rule_exWith parameter param_ex, the cut-off rule side after being updated Journey, performs step 2, the next frame in treatment video.

Further, in the step 1, the formula of cut-off rule equation is as follows：

L_ex(P_ex(x_ex,y_ex),k_ex,param_ex)=param_ex(y-y_ex)-k_ex(x-x_ex)=0；

Wherein, L_exRepresent cut-off rule, P_exRepresent the central point of cut-off rule, x_ex,y_exP is represented respectively_exHorizontal stroke, ordinate, X, y represent the horizontal stroke at any point, ordinate, k on cut-off rule respectively_exRepresent the slope of cut-off rule, param_exRepresenting cut-off rule is It is no perpendicular to x-axis.

Further, in the step 4, cut zone C (P_ex,R_min) radius R_minComputing formula is as follows：

R_min=min (Wide/2-x_ex,Wide/2+x_ex,Height/2-y_ex,Height/2+y_ex)；

Wherein, Wide is the width of bianry image b (x, y), and Height is the height of bianry image b (x, y).

Further, in the step 6, centre of form P₁And P₂Computing formula it is as follows：

Wherein, X and Y are respectively centre of form P₁And P₂Horizontal stroke, ordinate, x_iAnd y_jRespectively straight line mark is in C₁And C₂Interior pixel Horizontal stroke, the ordinate of point, N represent region C₁Or C₂Interior pixel sum；

The center point P of equivalent straight line equation_tmpComputing formula it is as follows：

Wherein, x_tmp、y_tmpCentered on point P_tmpHorizontal stroke, ordinate, x₁、x₂Respectively centre of form P₁And P₂Abscissa, y₁And y₂ It is centre of form P₁And P₂Ordinate；

Work as x₁=x₂, order：

ParamD=0, k^D=1；

Work as x₁≠x₂, order：

Obtaining expression formula of the equivalent straight line in rectangular coordinate system xoy is：

L^D(P_tmp(x_tmp,y_tmp),k^D, paramD) and=paramD (y-y_tmp)-k^D(x-x_tmp)=0,

Wherein, L^DRepresent equivalent straight line, k^DEquivalent straight line slope is represented, whether paramD represents equivalent straight line perpendicular to x Axle.

Further, in the step 7, the slope k of cut-off rule_exWith parameter param_exRenewal it is specific as follows:

Work as paramD=0, make：

param_ex=1, k_ex=0；

Work as k^D=0, order：

param_ex=0, k_ex=1；

As paramD ≠ 0 and k^D≠ 0, order：

In sum, by adopting the above-described technical solution, the beneficial effects of the invention are as follows：

First, relative to line walking algorithm conventional at present, each frame of the treatment video different from isolating more emphasizes frame to the present invention Contacting between frame, more than 0.99999, Euclidean distance generally in 2 pixels or so, is ensureing the COS distance of straight line positioning On the premise of algorithm steps are refined, there is a degree of accuracy higher；

2nd, the characteristics of present invention is split with followability, followability is mainly reflected in following two：

(1) position for following straight line to identify, automatically updates cut-off rule equation, relative to the changeless side of cut-off rule equation Formula, to the vision positioning of the straight line mark at optional position and inclination angle, all with the degree of accuracy higher；

(2) position for following straight line to identify, automatically updates cut zone, during reducing background to straight line marker extraction Influence, improve the degree of accuracy；

3rd, the present invention is extremely low to hardware platform requirements, and processing speed is fast, real-time.

Brief description of the drawings

Fig. 1 is the overall flow in the present invention；

Fig. 2 is the geometrical analysis figure of cathetus positioning of the present invention；

Fig. 3 is the original image of the first two field picture in the present invention；

Fig. 4 is the binary image of the first two field picture in the present invention；

Fig. 5 is the design sketch for carrying out straight line positioning of the first two field picture in the present invention；

Fig. 6 is the time and again datagram of the COS distance of straight line positioning in the present invention；

Fig. 7 is the time and again datagram of the Euclidean distance of straight line positioning in the present invention.

Specific embodiment

In order to make the purpose , technical scheme and advantage of the present invention be clearer, 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 used to explain the present invention, not For limiting the present invention.

A kind of simple straight line automatic positioning method towards vision line walking, in scene bianry image, with centre of form formula and Based on point-slope form straight-line equation, by straight line segmentation into two parts, position of form center is sought respectively, and then obtain the slope of straight line And position.It is different from traditional cut-off rule and region immobilizes pattern, the present invention is taken with reference to determining that former frame straight line is identified Position result, automatically updates cut-off rule equation and cut zone, so as to reach the effect of automatic vision positioning, to hardware platform requirements Extremely low, processing speed is very fast.Idiographic flow as shown in figure 1, a kind of simple straight line automatic positioning method towards vision line walking, such as Lower step：

Step 1：Video is obtained, the initial position and gradient identified according to the straight line to be positioned in video, initialization The center point P of cut-off rule equation_ex, slope k_exWith parameter param_ex, cut-off rule equation is obtained, as shown in Fig. 2 being video image Rectangular coordinate system XOY is set up, polygon EFGH represents the geometric profile of straight line mark, and ab is cut-off rule；Wherein cut-off rule equation Formula it is as follows：

L_ex(P_ex(x_ex,y_ex),k_ex,param_ex)=param_ex(y-y_ex)-k_ex(x-x_ex)=0；

Wherein, L_exRepresent cut-off rule, P_exRepresent central point, (x_ex,y_ex) P is represented respectively_exTransverse and longitudinal coordinate, k_exRepresent and divide Secant slope, param_exWhether cut-off rule is represented perpendicular to x-axis, x, y represent that the horizontal, vertical of any point sits on cut-off rule respectively Mark,.

The initial position P of the straight line to be positioned in setting video mark₀(x₀,y₀) and gradient k₀, P₀(x₀,y₀) represent straight The position of form center of line mark, k₀The tangent value of straight line mark and video window angle is represented, the initialization of cut-off rule equation parameter is public Formula is as follows：

P_ex=P₀；

Work as k₀=0, order：

param_ex=0, k_ex=1；

Work as k₀=∞, order：

param_ex=1, k_ex=0.

Step 2：The first two field picture f is read from video_t(x, y), as shown in Figure 3；

Step 3：To the first two field picture f_t(x, y) carries out threshold process, obtains bianry image b (x, y), as shown in Figure 4；

Step 4：In bianry image b (x, y), obtain with point P_exIt is the center of circle, R_minIt is the cut zone C (P of radius_ex, R_min)；

Cut zone C (P_ex,R_min) radius R_minComputing formula is as follows：

R_min=min (Wide/2-x_ex,Wide/2+x_ex,Height/2-y_ex,Height/2+y_ex)；

Wherein, Wide is the width of bianry image b (x, y), and Height is the height of bianry image b (x, y).

Step 5：According to cut-off rule equation, cut-off rule is obtained, using cut-off rule by cut zone C (P_ex,R_min) it is divided into C₁ And C₂Two regions, judge that straight line is identified whether in C₁And C₂In two regions, if it is, performing step 6, step is otherwise performed Rapid 2, the next frame in treatment video；

Step 6：Respectively in C₁And C₂Inside obtain centre of form P₁And P₂, according to centre of form P₁And P₂Calculate the center of equivalent straight line equation Point P_tmp, slope K^DWith parameter paramD, equivalent straight line equation is obtained, wherein, when parameter paramD is 0, represent that equivalent straight line hangs down It is straight in x-axis, and paramD represents that equivalent straight line is not orthogonal to x-axis when being 1；As shown in Fig. 2 being obtained by equivalent straight line equation When equivalent straight line mn, parameter paramD are 0, equivalent straight line is represented perpendicular to x-axis, and paramD represents equivalent straight line not when being 1 Perpendicular to x-axis.The design sketch of the first frame straight line positioning is as shown in Figure 5 in video；

Wherein, centre of form P₁And P₂Computing formula it is as follows：

Wherein, X and Y are respectively centre of form P₁And P₂Horizontal stroke, ordinate, x_iAnd y_jRespectively straight line mark is in C₁And C₂Interior pixel Horizontal stroke, the ordinate of point, N represent region C₁Or C₂Interior pixel sum；

The center point P of equivalent straight line equation_tmpComputing formula it is as follows：

Wherein, x_tmp、y_tmpCentered on point P_tmpHorizontal stroke, ordinate, x₁、x₂Respectively centre of form P₁And P₂Abscissa, y₁And y₂ It is centre of form P₁And P₂Ordinate；

Work as x₁=x₂, order：

ParamD=0, k^D=1；

Work as x₁≠x₂, order：

Obtaining expression formula of the equivalent straight line in rectangular coordinate system XOY is：

L^D(P_tmp(x_tmp,y_tmp),k^D, paramD) and=paramD (y-y_tmp)-k^D(x-x_tmp)=0.

Wherein, L^DRepresent equivalent straight line, k^DEquivalent straight line slope is represented, whether paramD represents equivalent straight line perpendicular to x Axle.

Step 7：Judge whether video terminates, if terminated, equivalent straight line is obtained according to equivalent straight line equation, will be equivalent straight Time and again positioning result is stored in text line, otherwise, calculates equivalent straight line L^DWith the midpoint of bianry image b (x, y) intersection point P_c, update the center point P of cut-off rule_exIt is P_c, update the slope k of cut-off rule_exWith parameter param_ex, the cut-off rule after being updated Equation, performs step 2, the next frame in treatment video.

Wherein, the slope slope k of cut-off rule_exWith parameter param_exRenewal it is specific as follows:

Work as paramD=0, make：

param_nex=1, k_nex=0；

Work as k^D=0, order：

param_nex=0, k_nex=1；

As paramD ≠ 0 and k^D≠ 0, order：

The straight line mark of this case represents tracker wire, and equivalent straight line represents the center line of tracker wire two edges.Equivalent straight line is time and again Locating effect is represented that Fig. 6 represents time and again the COS distance of straight line positioning by Fig. 6 and Fig. 7, and Fig. 7 represents time and again the Europe of straight line positioning Formula distance.COS distance is the cosine value of straight line positioning tilt angle error, from shown in Fig. 6, it can be seen that straight line positioning of the present invention COS distance reached more than 0.99999；Euclidean distance be straight line positioning central point to the distance between the centre of form, can from Fig. 2 With find out when straight line identify EFGH by image b (x, y) window turning when, centre of form q has relative to the central point p of equivalent straight line Larger skew, and the Euclidean distance of Fig. 7 cathetus positioning is that straight line mark have passed through image window more than the frame number of 2 pixels B (x, y) turning, the Euclidean distance of remaining frame number is below 2 pixels, and precision is higher.

Presently preferred embodiments of the present invention is the foregoing is only, is not intended to limit the invention, it is all in essence of the invention Any modification, equivalent and improvement made within god and principle etc., should be included within the scope of the present invention.

Claims (5)

1. a kind of simple straight line automatic positioning method towards vision line walking, it is characterised in that comprise the following steps：

Step 1：Video is obtained, the initial position and gradient identified according to the straight line to be positioned in video, initialization segmentation The center point P of line equation_ex, slope k_exWith parameter param_ex, obtain cut-off rule equation；

Step 2：The first two field picture f is read from video_t(x,y)；

Step 3：To the first two field picture f_t(x, y) carries out threshold process, obtains bianry image b (x, y)；

Step 4：In bianry image b (x, y), obtain with point P_exIt is the center of circle, R_minIt is the cut zone C (P of radius_ex,R_min)；

Step 5：According to cut-off rule equation, cut-off rule is obtained, using cut-off rule by cut zone C (P_ex,R_min) it is divided into C₁And C₂ Two regions, judge that straight line is identified whether in C₁And C₂In two regions, if it is, performing step 6, step 2 is otherwise performed, Next frame in treatment video；

Step 6：Respectively in C₁And C₂Inside obtain centre of form P₁And P₂, according to centre of form P₁And P₂Calculate the central point of equivalent straight line equation P_tmp, slope K^DWith parameter paramD, equivalent straight line equation is obtained, wherein, when parameter paramD is 0, represent that equivalent straight line is vertical In x-axis, and paramD represents that equivalent straight line is not orthogonal to x-axis when being 1；

Step 7：Judge whether video terminates, if terminated, equivalent straight line is obtained according to equivalent straight line equation, equivalent straight line is tired out Secondary positioning result is stored in text, otherwise, calculates the midpoint P of equivalent straight line and bianry image b (x, y) intersection point_c, update The center point P of cut-off rule_exIt is P_c, update the slope k of cut-off rule_exWith parameter param_ex, the cut-off rule equation after being updated, Perform step 2, the next frame in treatment video.

2. a kind of simple straight line automatic positioning method towards vision line walking according to claim 1, it is characterised in that：Institute State in step 1, the formula of cut-off rule equation is as follows：

L_ex(P_ex(x_ex,y_ex),k_ex,param_ex)=param_ex(y-y_ex)-k_ex(x-x_ex)=0；

Wherein, L_exRepresent cut-off rule, P_exRepresent the central point of cut-off rule, x_ex,y_exP is represented respectively_exHorizontal stroke, ordinate, x, y point Not Biao Shi on cut-off rule any point horizontal stroke, ordinate, k_exRepresent the slope of cut-off rule, param_exRepresent whether cut-off rule hangs down Directly in x-axis.

3. a kind of simple straight line automatic positioning method towards vision line walking according to claim 2, it is characterised in that：Institute In stating step 4, cut zone C (P_ex,R_min) radius R_minComputing formula is as follows：

R_min=min (Wide/2-x_ex,Wide/2+x_ex,Height/2-y_ex,Height/2+y_ex)；

Wherein, Wide is the width of bianry image b (x, y), and Height is the height of bianry image b (x, y).

4. a kind of simple straight line automatic positioning method towards vision line walking according to claim 1, it is characterised in that institute In stating step 6, centre of form P₁And P₂Computing formula it is as follows：

$\left\{\begin{array}{c}X=\frac{\underset{i=1}{\overset{N}{\Σ}}{x}_i}{N};\\ Y=\frac{\underset{j=1}{\overset{N}{\Σ}}{y}_j}{N};\end{array}\right.$

Wherein, X and Y are respectively centre of form P₁And P₂Horizontal stroke, ordinate, x_iAnd y_jRespectively straight line mark is in C₁And C₂Interior pixel Horizontal, ordinate, N represents region C₁Or C₂Interior pixel sum；

The center point P of equivalent straight line equation_tmpComputing formula it is as follows：

$\left\{\begin{array}{c}{x}_{tmp}=\frac{x_1+x_2}{2};\\ y_{tmp}=\frac{y_1+y_2}{2};\end{array}\right.$

Wherein, x_tmp、y_tmpCentered on point P_tmpHorizontal stroke, ordinate, x₁、x₂Respectively centre of form P₁And P₂Abscissa, y₁And y₂It is shape Heart P₁And P₂Ordinate；

Work as x₁=x₂, order：

ParamD=0, k^D=1；

Work as x₁≠x₂, order：

$\begin{array}{cc}paramD=1,& k^D=\frac{y_1-y_2}{x_1-x_2};\end{array}$

Obtaining expression formula of the equivalent straight line in rectangular coordinate system xoy is：

L^D(P_tmp(x_tmp,y_tmp),k^D, paramD) and=paramD (y-y_tmp)-k^D(x-x_tmp)=0,

Wherein, L^DRepresent equivalent straight line, k^DEquivalent straight line slope is represented, whether paramD represents equivalent straight line perpendicular to x-axis.

5. a kind of simple straight line automatic positioning method towards vision line walking according to claim 4, it is characterised in that institute In stating step 7, the slope k of cut-off rule_exWith parameter param_exRenewal it is specific as follows:

Work as paramD=0, make：

param_ex=1, k_ex=0；

Work as k^D=0, order：

param_ex=0, k_ex=1；

As paramD ≠ 0 and k^D≠ 0, order：

$\begin{array}{cc}{\mathrm{param}}_{ex}=1,& k_{ex}=\frac{-1}{k^D}.\end{array}$