CN105211034A - A kind of vehicular forest three-dimensional colour imaging target spraying method - Google Patents

Abstract

The invention provides a kind of vehicular forest three-dimensional colour imaging target spraying method that forest discrimination is high, speed is fast, cost is low, its step: when vehicular platform moves, two-dimensional laser radar scans up and down the object under test of vehicular platform sidepiece in the plane of scanning motion, obtain the two-dimensional position data of object under test in the plane of scanning motion, and two-dimensional position data is passed to calculator; Calculator obtains the mobile position data of vehicular platform according to the rate signal that vehicle speed sensor is measured, and mobile position data and two-dimensional position data are combined the three-dimensional location data being converted into object under test; The colour picture signal of the object under test obtained from colored CCD and three-dimensional location data map by geometrical relationship and merge by calculator again, form the three-dimensional colour cloud point data of object under test, again by calculator to the data analysis of three-dimensional colour cloud point, identify the forest being applicable to target spraying operation, control sprayer unit and carry out target spraying operation.

Description

A kind of vehicular forest three-dimensional colour imaging target spraying method

Technical field

The present invention relates to a kind of vehicular forest three-dimensional colour imaging target spraying method.

Background technology

Vehicular target spraying is control technology, the concentrated expression of spray technique in agricultural, forestry applications, is the typical apply of closely forestry remote sensing.Vehicular target spraying machine is one of pesticide dispersal equipment being widely used in the trees such as orchard, street tree, ornamental plantation.The targets such as the sensing mode such as general employing ultrasonic, microwave, optical imagery, infrared, laser ranging detection trees, realize using for the pesticide spray of particular target by control system, thus realize target spraying, this mode is also referred to as precision spraying.When reaching identical control efficiency, exterior spraying decrease agricultural chemicals usage amount, reduce environmental pollution and people and animals harm, obtain in recent years and promote widely.But due to the scrambling between the polytropy of forest profile, forest, particularly cities and towns street tree, its road vehicle on both sides of the road, building etc. all have impact on the identification of target spraying system to forest, therefore existing forest exterior spraying adopts rough target recognition methods to identify forest mostly, to coordinate the feature of spraying operation relative poor.Rough target recognition methods has mainly carried out the identification with or without target by simple feature, and its False Rate is higher.Therefore, need a kind of target recognition system being suitable for forest target spraying more accurately, require that its cost is low, speed fast, forest discrimination is high.

Summary of the invention

The object of this invention is to provide a kind of vehicular forest three-dimensional colour imaging target spraying method that forest discrimination is high, speed is fast, cost is low.

Vehicular forest three-dimensional colour imaging target spraying method of the present invention, vehicular platform 7 is provided with two-dimensional laser radar 2, has the camera of colored CCD 3, vehicle speed sensor 4, calculator 5 and sprayer unit 6; Vehicle speed sensor 4 measures the translational speed of vehicular platform 7 in travel direction 8; This target spraying method comprises the steps:

When a, vehicular platform 7 move, two-dimensional laser radar 2 scans the object under test about 1 of vehicular platform sidepiece in the plane of scanning motion, obtains the two-dimensional position data P of object under test in the plane of scanning motion _{i, j}(x _i, y _i), and two-dimensional position data is passed to calculator 5;

B, calculator 5 obtain the mobile position data of vehicular platform according to the rate signal that vehicle speed sensor 4 is measured, and mobile position data and two-dimensional position data are combined the three-dimensional location data P being converted into object under test 1 _i,j(x _i, y _i, z _i);

The colour picture signal P of object under test that c, calculator 5 will obtain from colored CCD 3 again _{xi, Yi}(R _i, G _i, B _i), maps with three-dimensional location data merge by geometrical relationship, the three-dimensional colour cloud point data P of formation object under test 1 _i,j(x _i, y _i, z _i, R _i, G _i, B _i),

D, again by the data analysis of calculator 5 pairs of three-dimensional colour cloud points, identify and be applicable to the forest of target spraying operation, control sprayer unit and carry out target spraying operation.

Beneficial effect of the present invention: install colour imagery shot and two-dimensional laser radar on the vehicular platform of spraying vehicle, scanning plane and the vehicle heading of this laser radar are perpendicular, calculator is simultaneously by the travel speed of vehicle speed sensor measuring vehicle, form three-dimensional coordinate (position) data of external environment, and map with the colour picture signal that colour imagery shot is taken the three-dimensional colour data merging and obtain environment to external world by geometrical relationship, from three-dimensional colour data, forest target to be sprayed is found out again by calculator, calculate corresponding spray amount, final control spraying, to reach the object of target spraying.

Two-dimensional laser radar can record the range data of object point, colored CCD records image coordinate and the color data of object point, these two groups of data have and describe the difference of identical object point, and namely the range information that records of laser radar and CCD chrominance information are coordinate data, the description of chroma data under different coordinates of same object point.Vehicle translational speed, two-dimensional coordinate data, chroma data three combine by the present invention, achieve the three-dimensional colour imaging of environment to external world, are applicable to target spraying job requirements.

In the process, calculator reads high-speed line scanning laser radar and obtains two-dimentional ranging data, in conjunction with GES, forms three-dimensional laser radar ranging data stream; Combine the operation area picture signal of the corresponding moment colour imagery shot acquisition read again, according to the mapping relations of measurement target laser cloud point and colored CCD measurement data, the image information that the coordinate datas such as the distance that laser radar is recorded, orientation and colored CCD obtain carries out data fusion, achieves the three-dimensional colour imaging data stream meeting target spraying job requirements.

The object of the present invention is achieved like this: this vehicular forest three-dimensional colour imaging target spraying method, vehicular platform 7 is provided with two-dimensional laser radar 2, has the camera of colored CCD 3, vehicle speed sensor 4, calculator 5 and sprayer unit 6; Vehicle speed sensor 4 measures the translational speed of vehicular platform 7 in travel direction 8; When vehicular platform 7 moves, two-dimensional laser radar 2 scans the object under test about 1 of vehicular platform sidepiece in the plane of scanning motion, obtains the two-dimensional position data of object under test in the plane of scanning motion, and two-dimensional position data is passed to calculator 5; Calculator 5 obtains the mobile position data of vehicular platform according to the rate signal that vehicle speed sensor 4 is measured, and mobile position data and two-dimensional position data are combined the three-dimensional location data being converted into object under test 1; Calculator 5 is again by the colour picture signal of the colored CCD 3 of acquisition, map with three-dimensional location data merge by geometrical relationship, form the three-dimensional colour cloud point data of object under test 1, again by the data analysis of calculator 5 pairs of three-dimensional colour cloud points, identify the forest being applicable to target spraying operation, control sprayer unit and carry out target spraying operation.This vehicular forest three-dimensional colour imaging target spraying method is high to forest discrimination, and automaticity is high, and operating speed is fast, cost is low.

Above-mentioned target spraying method, the plane of scanning motion of two-dimensional laser radar 2 is vertical with the travel direction of vehicular platform.

Above-mentioned target spraying method, in steps d, calculator 5 pairs of three-dimensional colour cloud point data are analyzed as follows, and the method identifying the forest of target spraying operation is:

As three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) meet following three conditions simultaneously: 45 < R _i+ G _i+ B _i< 720; $\frac{2G_i}{R_i+B_i+1}\&GreaterEqual;1;R_{min}<\sqrt{{X_i}^2+{Y_i}^2+{Z_i}^2}<R_{max};$ Then object under test 1 is the forest of target spraying operation.Due to the polytropy of Tree image, the features the such as particularly comparison of light and shade of trees gap and shade dark place is outstanding, the present invention spy adopts light and shade Color Picking to adopt weighting algorithm, can adapt to bright with excessively dark picture point preferably, too bright or too dark non-trees object is got rid of in spread of spray.R _min, R _maxrepresent that two-dimensional laser radar is to beeline, the maximum distance of tested fixed point, can be set by the user. ensure that target is in suitable distance range. ensure that from color background, extract green, or approximate green part.

Above-mentioned target spraying method, after step a, before step b in steps a1: the two-dimensional position data P of calculator 5 measuring targets in the plane of scanning motion _{i, j}(x _i, y _i), screen, if X _min< x _i< X _max, Y _min< y _i< Y _maxthen retain this two-dimensional position data P _{i, j}(x _i, y _i), otherwise then abandon this two-dimensional position data P _{i, j}(x _i, y _i).X _min, X _maxthe minimum distance on the direction (as x-axis direction) of object under test in the plane of scanning motion of two-dimensional laser radar 2 and maximum distance, Y _min, Y _maxthe minimum distance on object under test another direction (as y-axis direction) in the plane of scanning motion of two-dimensional laser radar 2 and maximum distance.X _min, X _max, Y _min, Y _maxbe artificial setting, ensure that target is in suitable distance range.

Above-mentioned target spraying method, the detailed process of step c: the colour picture signal P of the object under test that calculator 5 obtains _{xi, Yi}(R _i, G _i, B _i) after, as follows green tint extraction process is carried out to the RGB three primary color components of these data, the picture signal P after process _{xi, Yi}(R _i, G _i, B _i) only have P _{xi, Yi}(0,0,0) or P _{xi, Yi}(0,255,0) two kinds:

$(R_i,G_i,B_i)=\left\{\begin{array}{cc}(0,0,0)& R_i+G_i+B_i\&GreaterEqual;720\\ (0,0,0)& \frac{2G_i}{R_i+B_i+1}<1\\ (0,0,0)& R_i+G_i+B_i\&le;45\\ (0,255,0)& 45<R_i+G_i+B_i<720\&cap;\frac{2G_i}{R_i+B_i+1}\&GreaterEqual;1\end{array}\right.$ Formula (1)

Then calculator is again the picture signal P after process _{xi, Yi}(R _i, G _i, B _i) maps with three-dimensional location data merge by geometrical relationship, the three-dimensional colour cloud point data P of formation object under test (1) _i,j(x _i, y _i, z _i, R _i, G _i, B _i);

The detailed process of steps d: calculator is to three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) be processed into target cloud point Q by the following method _i,j:

$Q_{i,j}=\left\{\begin{array}{cc}1& R_{min}<\sqrt{{X_i}^2+{Y_i}^2+{Z_i}^2}<R_{\max }\&cap;(R_i,G_i,B_i)=(0,255,0)\\ 0& else\end{array}\right.$ Formula (2)

Then calculator reads target cloud point Q _i,jcalculate as follows: $S=\left\{\begin{array}{cc}1& \underset{i=0,j=0}{\overset{i=I,j=J}{\mathrm{\&Sigma;}}}{Q}_{i,j}\&GreaterEqual;M\\ 0& \end{array}\right.;$ If S=1, control sprayer unit and carry out spraying operation, if S=0, control sprayer unit and do not carry out spraying operation; I, J are respectively the size between spraying area, and M is amount threshold constant, scope 0 ~ I*J.

In step c, calculator obtains the color image data P of colored CCD simultaneously _{xi, Yi}(R _i, G _i, B _i), green tint extraction is carried out to the RGB three primary color components of these data.Due to the polytropy of Tree image, the features the such as particularly comparison of light and shade of trees gap and shade dark place is outstanding, light and shade Color Picking adopts weighting algorithm, sees formula (1), can adapt to bright with excessively dark picture point preferably.Image picture point after this formula (1) process only has black (0,0,0) and green (0,255,0) two kinds, namely facilitates and shows on screen, be also convenient to subsequent treatment.

In steps d, to three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) be processed into target cloud point Q according to formula (2) _i,j, namely ensure that target is in suitable distance range, in turn ensure that the green character of target.

Accompanying drawing explanation

Fig. 1 is overall construction drawing of the present invention.

In figure: 1, object under test 2, two-dimensional laser radar 3, colored CCD 4, vehicle speed sensor 5, calculator 6, sprayer unit 7, vehicular platform 8, travel direction

Embodiment

Below in conjunction with accompanying drawing, the present invention is further described.

In Fig. 1, object under test 1 is the unknowns such as the target of target spraying, its size, position, complexity.Vehicular platform 7 is provided with high-speed line scanning two-dimensional laser radar 2, there is the camera of colored CCD 3, calculator 5 and sprayer unit 6.The plane of scanning motion and the vehicle heading of this laser radar are perpendicular, two-dimensional laser radar 2 is arranged on the front end of vehicular platform 7 with the camera with colored CCD 3, object obtains data message in advance, carries out decision-making by calculator 5, finally controls the sprayer unit 6 being arranged in vehicular platform 7 rear portion.Vehicle speed sensor 4 measures the translational speed of vehicular platform 7 in travel direction 8.Vehicular platform 7 is when travel direction 8 shown in Fig. 1 moves, and the plane of scanning motion of two-dimensional laser radar 2 scans object under test 1 successively from right to left, and will scan 2-D data stream P _{i, j}(x _i, y _i) pass to calculator 5; At the same time, calculator 5, in conjunction with the vehicular platform 7 translational speed signal of the measurement of vehicle speed sensor 4, calculates corresponding displacement, is three-dimensional location data P by data transformations _i,j(x _i, y _i, z _i).Calculator 5 is by the colour picture signal P of the colored CCD 3 of acquisition _{xi, Yi}(R _i, G _i, B _i), with three-dimensional location data P _i,j(x _i, y _i, z _i) map fusion by geometrical relationship, define three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i).Again by calculator 5 by color recognition in conjunction with cloud point data analysis, finally realize the identification to forest target.

Two-dimensional laser radar can record the range data of object point, color CCD image records image coordinate and the color data of object point, these two groups of data have and describe the difference of identical object point, and namely the range information that records of laser radar and CCD chrominance information are coordinate data, the description of chroma data under different coordinates of same object point.Car vehicle translational speed, two-dimensional coordinate data, chroma data three combine by the present invention, achieve the three-dimensional colour imaging of environment to external world, are applicable to target spraying job requirements.

In the process, calculator reads high-speed line scanning laser radar and obtains two-dimentional ranging data, in conjunction with GES, forms three-dimensional laser radar ranging data stream; Combine the operation area picture signal of the corresponding moment colour imagery shot acquisition read again, according to the mapping relations of measurement target laser cloud point and colored CCD measurement data, the image information that the coordinate datas such as the distance that laser radar is recorded, orientation and colored CCD obtain carries out data fusion, achieves the three-dimensional colour imaging data stream meeting target spraying job requirements.

Described vehicular forest three-dimensional colour imaging target spraying method, detailed process is as follows.

1, calculator obtains the data of two-dimensional laser radar

Calculator obtains the data cloud point P of two-dimensional laser radar _{i, j}(x _i, y _i) after, first carry out threshold filtering, namely collection result is multiplied with rectangular window function, to obtain the data formulating range intervals, removes hypotelorism and data excessively far away.Again in conjunction with the translational speed of vehicle and the frame per second of two-dimensional laser radar, the data the most at last after process save as three-dimensional location data P _i,j(x _i, y _i, z _i).

2, obtain the view data of colored CCD

Calculator obtains the view data P of colored CCD simultaneously _{xi, Yi}(R _i, G _i, B _i), green tint extraction is carried out to the RGB three primary color components of these data.Due to the polytropy of Tree image, the features the such as particularly comparison of light and shade of trees gap and shade dark place is outstanding, the special light and shade Color Picking that adopts adopts weighting algorithm, sees formula (1), can adapt to cross bright with excessively dark picture point preferably.

$(R_i,G_i,B_i)=\left\{\begin{array}{cc}(0,0,0)& R_i+G_i+B_i\&GreaterEqual;720\\ (0,0,0)& \frac{2G_i}{R_i+B_i+1}<1\\ (0,0,0)& R_i+G_i+B_i\&le;45\\ (0,255,0)& 45<R_i+G_i+B_i<720\&cap;\frac{2G_i}{R_i+B_i+1}\&GreaterEqual;1\end{array}\right.$ Formula (1)

According to company (1), R _i, G _i, B _i45 < R must be met simultaneously _i+ G _i+ B _i< 720, during this two conditions, (R _i, G _i, B _i)=(0,255,0).Under other situation, (R _i, G _i, B _i)=(0,0,0).

Image picture point after this formula manipulation only has black (0,0,0) and green (0,255,0) two kinds, namely facilitates and shows on screen, be also convenient to subsequent treatment.

3, define three-dimensional colour cloud point data

Colour picture signal P _{xi, Yi}(R _i, G _i, B _i), with three-dimensional location data P _i,j(x _i, y _i, z _i) map fusion by geometrical relationship, define three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i), i.e. any one some P _i,jcontain the spatial positional information (x of data _i, y _i, z _i) also contains the color information (R of this point _i, G _i, B _i).To data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) carry out threshold process according to formula (2) and become target cloud point Q _i,j, namely this process ensure that target is in suitable distance range in turn ensure that the green character ((R of target _i, G _i, B _i)=(0,255,0)), that is, target must meet in suitable distance range simultaneously, this two conditions of green character time, Q _i,j=1.If target is not in suitable distance range, or target is not green, Q _i,j=0.R _min, R _maxrepresent that two-dimensional laser radar is to beeline, the maximum distance of tested fixed point, can be set by the user.

$Q_{i,j}=\left\{\begin{array}{cc}1& R_{min}<\sqrt{{X_i}^2+{Y_i}^2+{Z_i}^2}<R_{\max }\&cap;(R_i,G_i,B_i)=(0,255,0)\\ 0& else\end{array}\right.$ Formula (2)

4, control the spray solenoid valve in sprayer unit

As the vehicle moves, calculator reads target cloud point Q _i,jenter amount threshold to detect by formula (3), to control the unlatching of spray solenoid valve with closed, finally realize target spraying.

$S=\left\{\begin{array}{cc}1& \underset{i=0,j=0}{\overset{i=I,j=J}{\mathrm{\&Sigma;}}}{Q}_{i,j}\&GreaterEqual;M\\ 0& \end{array}\right.$ Formula (3)

In formula (3), if meet then S=1, represents that spray solenoid valve is opened and carries out target spraying; If do not meet then S=0, represents that spray solenoid valve is closed.

I, J are respectively the size between spraying area, and M is amount threshold constant, and scope 0 ~ I*J can be set by the user.

Claims (5)

1. a vehicular forest three-dimensional colour imaging target spraying method, is characterized in that: vehicular platform (7) is provided with two-dimensional laser radar (2), the camera with colored CCD (3), vehicle speed sensor (4), calculator (5) and sprayer unit (6); Vehicle speed sensor (4) measures vehicular platform (7) translational speed in travel direction (8); This target spraying method comprises the steps:

When a, vehicular platform (7) are mobile, two-dimensional laser radar (2) scans up and down the object under test (1) of vehicular platform sidepiece in the plane of scanning motion, obtains the two-dimensional position data P of object under test in the plane of scanning motion _{i, j}(x _i, y _i), and two-dimensional position data is passed to calculator (5);

B, calculator (5) obtain the mobile position data of vehicular platform according to the rate signal that vehicle speed sensor (4) is measured, and mobile position data and two-dimensional position data are combined the three-dimensional location data P being converted into object under test (1) _i,j(x _i, y _i, z _i);

The colour picture signal P of the object under test that c, calculator (5) will obtain from colored CCD (3) again _{xi, Yi}(R _i, G _i, B _i), maps with three-dimensional location data merge by geometrical relationship, the three-dimensional colour cloud point data P of formation object under test (1) _i,j(x _i, y _i, z _i, R _i, G _i, B _i),

D, again by calculator (5) to the data analysis of three-dimensional colour cloud point, identify and be applicable to the forest of target spraying operation, control sprayer unit and carry out target spraying operation.

2. vehicular forest three-dimensional colour imaging target spraying method according to claim 1, is characterized in that: the plane of scanning motion of two-dimensional laser radar (2) is vertical with the travel direction of vehicular platform.

3. vehicular forest three-dimensional colour imaging target spraying method according to claim 1, it is characterized in that: in steps d, calculator (5) is analyzed as follows to three-dimensional colour cloud point data, and the method identifying the forest of target spraying operation is:

As three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) meet following three conditions simultaneously: 15*3 < R _i+ G _i+ B _i< 240*3; then object under test (1) forest that is target spraying operation.

4. the vehicular forest three-dimensional colour imaging target spraying method according to claim 1,2 or 3, is characterized in that: after step a, before step b in steps a1: the two-dimensional position data P of calculator (5) measuring targets in the plane of scanning motion _{i, j}(x _i, y _i), screen, if X _min< x _i< X _max, Y _min< y _i< Y _maxthen retain this two-dimensional position data P _{i, j}(x _i, y _i), otherwise then abandon this two-dimensional position data P _{i, j}(x _i, y _i).

5. vehicular forest three-dimensional colour imaging target spraying method according to claim 1, is characterized in that:

The detailed process of step c: the colour picture signal P of the object under test that calculator (5) obtains _{xi, Yi}(R _i, G _i, B _i) after, as follows green tint extraction process is carried out to the RGB three primary color components of these data, the picture signal P after process _{xi, Yi}(R _i, G _i, B _i) only have P _{xi, Yi}(0,0,0) or P _{xi, Yi}(0,255,0) two kinds:

$(R_i,G_i,B_i)=\left\{\begin{array}{cc}(0,0,0)& R_i+G_i+B_i\&GreaterEqual;240*3\\ (0,0,0)& \frac{2G_i}{R_i+B_i+1}<1\\ (0,0,0)& R_i+G_i+B_i\&le;15*3\\ (0,255,0)& 15*3<R_i+G_i+B_i<240*3\&cap;\frac{2G_i}{R_i+B_i+1}\&GreaterEqual;1\end{array}\right.$

Then calculator is again the picture signal P after process _{xi, Yi}(R _i, G _i, B _i) maps with three-dimensional location data merge by geometrical relationship, the three-dimensional colour cloud point data P of formation object under test (1) _i,j(x _i, y _i, z _i, R _i, G _i, B _i);

The detailed process of steps d: calculator is to three-dimensional colour cloud point data P _i,j(x _i, y _i, z _i, R _i, G _i, B _i) be processed into target cloud point Q by the following method _i,j: $Q_{i,j}=\{\begin{array}{cc}1& R_{min}<\sqrt{{X_i}^2+{Y_i}^2+{Z_i}^2}<R_{\max }\&cap;(R_i,G_i,B_i)=(0,255,0)\\ 0& else\end{array};$ Then calculator reads target cloud point Q _i,jcalculate as follows: $S=\left\{\begin{array}{cc}1& \underset{i=0,j=0}{\overset{i=I,j=J}{\mathrm{\&Sigma;}}}{Q}_{i,j}\&GreaterEqual;M\\ 0& \end{array}\right.;$ If S=1, control sprayer unit and carry out spraying operation, if S=0, control sprayer unit and do not carry out spraying operation; I, J are respectively the size between spraying area, and M is amount threshold constant, scope 0 ~ I*J.