CN100513996C - Method for detecting field water obstacle detection based on polarizing information - Google Patents

Abstract

This invention relates to a method of detecting outdoors water substance impediment base on polarization information, the process is: fix polarizationlum filter at camera front; pick three pieces different polarization physic images in identity region; count the polarization degree chart, polarization phase map and phase similarity measure chart of this region, and at the phase similarity measure chart verdict whether water substance impediment existing; if existing, then through self-adapting threshold partition algorithm to line out possible range of water substance impediment, at last combine polarization degree chart to definite true water substance impediment region. This device by detecting polarization characteristic of water-reflected light, could detect rounded water surface region, especially possess favorable identify effect to state of water surface existing reflection .This invention adopt relatively simple image segmentation algorithm, small calculated amount, quickly, high accuracy rating, and wide detecting celerity, wide applicability.

Description

Detect the method for field water barrier based on polarization information

Technical field

The present invention relates to a kind of method, be applicable to when unmanned autonomous car moves, the detection of water body types of obstructions based on polarization information detection field water barrier.

Background technology

When vehicle moves in complicated field environment, to discern various types of barriers.Various types of water bodys, big because of the probability of occurrence height to the car body injury, be the important object of unmanned autonomous car detection of obstacles in the field environment.

At present, the defective of himself is arranged all: as discerning water body by brightness, saturation analysis in the coloured image, but can't discern the water surface that contains inverted image to the detection method of various water body types of obstructions; Discern water body by having or not of rreturn value in the radar image, but the accuracy rate that detects is not high; Discern water body by contrast water body part with the day and night temperature of surrounding environment in the hot sensed image, but detection time is long, can't be used for real-time application.

Summary of the invention

The purpose of this invention is to provide a kind of detection fast, accuracy rate is high, what applicability was wide detects the method for field water barrier based on polarization information.

Method based on polarization information detection field water barrier of the present invention may further comprise the steps:

1) with the camera that the linear polarization filter is installed polarization image is taken in same pre-detection zone, the angle that obtains polarization filter light transmission shaft and surface level is respectively 0 °, 45 °, 90 ° three width of cloth images, remembers that the gray-scale value of the picture element of corresponding same position in 0 °, 45 °, 90 ° these three width of cloth images is respectively I ₀, I ₄₅And I ₉₀, the polarization phase θ and the degree of polarization P of all picture elements calculated in (1), (2) pointwise by formula respectively, obtains the polarization phase figure and the degree of polarization figure in whole pre-detection zone;

(1)

$P=\frac{I_{90}-I_0}{(I_{90}+I_0)\cos 2\mathrm{\θ}}$ (2)

2) in whole polarization phase figure, in 7 * 7～15 * 15 pixel window ranges, choose the pixel window of any size, by formula (3) calculate the correlativity that each should select picture element polarization phase in the pixel window successively, obtain the similarity measure figure of whole pre-detection zone polarization phase

$\mathrm{\Δ}=\underset{i,j}{\mathrm{\Σ}}|\mathrm{\θ}(i,j)-\stackrel{\‾}{\mathrm{\θ}}|---\left(3\right)$

Δ is for characterizing the similarity measure of polarization phase figure in the formula, and θ (i is respectively i for horizontal ordinate among the polarization phase figure and ordinate j), the polarization phase of the picture element of j, and θ is for there being the arithmetic mean of a polarization phase in the pixel window of selected size;

3) judge in the pre-detection zone whether have water barrier with the size of Δ, in polarization phase similarity measure figure, have any one Δ≤1, then judge in this zone to have water barrier, otherwise judge and do not have water barrier in this zone;

4) may there be water barrier if judge before, estimate at phase portrait and to carry out self-adapting threshold segmentation on the figure, judge that Δ is a water barrier less than the part of final threshold value, and the result who obtains carried out 4 connections or 8 connected component labelings that are communicated with, ask the specific algorithm of threshold value to be:

4.1) obtain the minimum value Δ of similarity measure figure ₁With the maximal value Δ _m, make the threshold value initial value:

T ₀＝(Δ ₁+Δ _m)/2 (4)

4.2) note T _kBe the k+1 time iteration threshold value (k=0,1,2...), according to threshold value T _kFigure is divided into greater than T with similarity measure _kWith less than T _kTwo parts, by formula (5) obtain two-part average measure value Δ _OAnd Δ _R,

${\mathrm{\&Delta;}}_O=\frac{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}N(i,j)}$ ${\mathrm{\&Delta;}}_B=\frac{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}N(i,j)}$ (5)

Δ in the following formula (i is that horizontal ordinate and ordinate are respectively i among the similarity measure figure j), the similarity measure value of the picture element of j, N (i is that horizontal ordinate and ordinate are respectively i among the similarity measure figure j), the weight of the picture element of j, and General N gets 1;

4.3) by formula (6) obtain new threshold value:

T _k+1＝(Δ _O+Δ _B)/2 (6)

4.4) if T _K+1With T _kError in 1%, promptly | T _K+1-T _k|≤T _k* 1%, then iterative process finishes, otherwise makes T _k=T _K+1, change step 4.2 over to) and the continuation iteration;

5) process of similar step 4) is that object carries out self-adapting threshold segmentation on degree of polarization figure with degree of polarization P, and degree of polarization P is judged to water barrier greater than the part of final threshold value, and the result who obtains is carried out 4 is communicated with or 8 connected component labelings that are communicated with;

6) connected region of the different labels that step 4) is obtained compares with all connected regions that step 5) obtains successively, if all connected regions that a certain zone and step 5) obtain all do not have intersection, then the connected region of this label is removed as erroneous judgement; If any one connected region that a certain zone and step 5) obtain has intersection, then keep the segmentation result of the connected region of this label as water barrier.

Above-mentioned steps 4) said 4 connected component labelings that are communicated with are meant if the final threshold value that the Δ of any one pixel of pixel upper and lower, left and right 4 directions adjacent with it all obtains less than step 4) then divides into a slice connected region; 8 connected component labelings that are communicated with are meant if the final threshold value that the Δ of any one pixel of pixel upper and lower, left and right adjacent with it, upper left, upper right, lower-left, 8 directions in bottom right all obtains less than step 4) then divides into a slice connected region;

Above-mentioned steps 5) said 4 connected component labelings that are communicated with are meant if the final threshold value that the degree of polarization P of any one pixel of pixel upper and lower, left and right 4 directions adjacent with it all obtains greater than step 5) then divides into a slice connected region; 8 connected component labelings that are communicated with are meant if the final threshold value that the degree of polarization P of any one pixel of pixel upper and lower, left and right adjacent with it, upper left, upper right, lower-left, 8 directions in bottom right all obtains less than step 5) then divides into a slice connected region;

For the error that reduces to cause by noise and other various factorss in the shooting process, can adopt the result that step 6) is obtained to open the morphologic filtering computing of afterwards closing earlier with 5 * 5 disk template, so that remove most erroneous judgements, and keep partitioning boundary information as much as possible.

Among the present invention, take polarization image and can adopt a camera, a linear polarization filter is installed before camera, the angle of taking polarization filter light transmission shaft and surface level respectively is 0 °, 45 °, 90 ° three width of cloth images.Perhaps also can adopt three cameras, a linear polarization filter is installed before every camera, the angle of taking polarization filter light transmission shaft and surface level simultaneously is 0 °, 45 °, 90 ° three width of cloth images.

Beneficial effect of the present invention is:

The present invention can detect complete water surface zone by detecting water-reflected polarisation of light characteristic, particularly exists the situation of inverted image to have good identification effect to the water surface; The present invention adopts better simply image segmentation algorithm, and calculated amount is little, and is quick to the detection of water barrier, the accuracy rate height, and applicability is wide.

Description of drawings

Fig. 1 is a kind of water barrier preliminary examination area schematic;

Fig. 2 is another kind of water barrier preliminary examination area schematic;

Fig. 3 is the operational flowchart that the inventive method detects water barrier.

Embodiment

Further specify the inventive method below in conjunction with embodiment.

Embodiment 1

In preliminary examination zone A as shown in Figure 1, there is water barrier B (shown in the dash area).Below in conjunction with Fig. 3 flowchart text the inventive method, detect water barrier by analyzing polarization information, and depict the scope of water barrier.

This embodiment uses a camera to gather polarization image.A linear polarization filter is installed before camera, is aimed at the same area and keep camera position motionless, the angle of adjusting polarization filter light transmission shaft and surface level is respectively 0 °, 45 °, 90 °, takes the polarization image of three width of cloth the same areas.Three width of cloth image sizes are identical, and the visual angle is identical, and the position of the object in the zone in each image is identical.

The gray-scale value of remembering the picture element of corresponding same position in 0 °, 45 °, 90 ° these three width of cloth images is respectively I ₀, I ₄₅And I ₉₀, be that the picture element of (1,1) is an example with upper left corner coordinate, the gray-scale value of this point is respectively I in three width of cloth images ₀=31, I ₄₅=28, I ₉₀=43, by formula (1), (2)

　　　　　　　(1)

$P=\frac{I_{90}-I_0}{(I_{90}+I_0)\cos 2\mathrm{\&theta;}}$ (2)

The polarization phase θ that calculates this point in this example is $\frac{1}{2}\arctan \left(\frac{31+43-2*28}{43-31}\right)=0.4914,$ Degree of polarization P is $\frac{43-31}{(43+31)\cos (2\&times;0.4914)}=0.2941.$

The polarization phase and the degree of polarization of all picture elements calculated in pointwise, obtains the polarization phase figure and the degree of polarization figure in whole pre-detection zone.This two width of cloth figure is the image identical with the pre-detection area size, the polarization phase and the degree of polarization of each some expression pre-detection zone corresponding point of the inside.

Then, the similarity measure with polarization phase is that criterion judges whether water barrier exists.This method is based on following physical criteria: the relative surrounding environment of polarisation of light phase place by water-reflected has bigger similarity.Definition characterizes the similarity of polarization phase figure as the Δ of formula (3).By formula (3)

$\mathrm{\&Delta;}=\underset{i,j}{\mathrm{\&Sigma;}}|{\mathrm{\&theta;}}_{i,j}-\stackrel{\&OverBar;}{\mathrm{\&theta;}}|$ (3)

Calculate the correlativity of each point polarization phase in 11 * 11 pixel windows of the upper left corner of polarization phase figure, obtain the upper left corner in this example $\mathrm{\&Delta;}=\underset{j=1..11}{\underset{i=1..11}{\mathrm{\&Sigma;}}}|{\mathrm{\&theta;}}_{i,j}-\stackrel{\&OverBar;}{\mathrm{\&theta;}}|=50.3199.$

Calculate the correlativity of each point polarization phase in each 11 * 11 pixel window successively, obtain the similarity measure figure of whole pre-detection zone polarization phase.This figure is the width of cloth image identical with the pre-detection area size, the similarity degree of the polarization phase of 11 * 11 pixel windows around each some expression pre-detection zone corresponding point of the inside.Δ is more for a short time to show that the phase portrait of this Δ institute corresponding region is high more, and vice versa.Similar because of the polarization phase of water body part, then the size of the Δ of each picture element is all less in the water body scope.

Estimate among the figure to judge with the size of Δ whether water barrier exists at phase portrait,, then judge in this zone to have water barrier when there being any one Δ≤1.In the present embodiment, calculate, phase portrait is estimated the point for (100,80) of coordinate among the figure, and its phase portrait is estimated Δ=0.3269.So judge in this preliminary examination zone and may have water barrier.

Then, estimate at phase portrait and carry out self-adapting threshold segmentation on the figure.The phase portrait that obtains in the present embodiment is estimated the figure minimum value and maximal value is respectively Δ ₁=0.1427 and Δ _m=113.2764, make threshold value initial value T ₀=(Δ ₁+ Δ _m)/2=56.7096,

By formula calculate (5), (6)

${\mathrm{\&Delta;}}_O=\frac{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}N(i,j)}$ ? ${\mathrm{\&Delta;}}_B=\frac{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}N(i,j)}$ (5)

T _k+1＝(Δ _O+Δ _B)/2 (6)

Obtain $T_1=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}N(i,j)})/2=46.9064.$

At this moment | T ₁-T ₀| T ₀* 1%, thus need substitution formula (5), (6) to continue to calculate,

Obtain $T_2=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}N(i,j)})/2=44.6468.$

At this moment | T ₂-T ₁| T ₁* 1%, so still need bring formula (5), (6) continuation calculating into.

Through 5 iteration, obtain T in the present embodiment ₄=43.1143,

$T_5=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_4}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_4}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_4}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_4}{\mathrm{\&Sigma;}}N(i,j)})/2=42.7481.$

At this moment | T ₅-T ₄|＜T ₄* 1%, iteration finishes, and drawing the final gray threshold of cutting apart Δ in this example is 42.7481, judges that then Δ is a water barrier less than 42.7481 part.

The result of in the present embodiment Δ being cut apart has adopted 8 connected component labelings that are communicated with, and has only a connected region, and its Δ is less than final gray threshold 42.7481, i.e. B zone as shown in Figure 1.

According to similar adaptive threshold dividing method, degree of polarization figure is carried out Threshold Segmentation, the C zone (getting ready among Fig. 1 shown in the part) that obtains as shown in Figure 1 is the part of degree of polarization greater than final threshold value 0.3561.

Segmentation result to degree of polarization P in the present embodiment has adopted 8 connected component labelings that are communicated with, and has only a connected region, and its degree of polarization P is greater than final gray threshold 0.3561, i.e. C zone as shown in Figure 1.

Result of calculation shows, the result that degree of polarization is cut apart is contained among the result that polarization phase similarity measure figure cuts apart, promptly intersection is arranged, so keep the segmentation result that whole phase portraits are estimated figure based on polarization phase similarity measure figure result of cutting apart and the result of cutting apart based on degree of polarization.The result shows that Δ coincide with the B zone less than 42.7481 part, judges that the B zone is a water barrier.

For the error that reduces to cause by noise and other various factorss in the shooting process, segmentation result has been carried out morphologic filtering.Disk template with 5 * 5 has carried out opening the morphology operations that afterwards closes earlier to segmentation result, has removed most erroneous judgements, and has kept partitioning boundary information as much as possible.

So far, in pre-detection zone as shown in Figure 1, detect water barrier, and obtained complete water barrier scope.

Embodiment 2

In preliminary examination zone A as shown in Figure 2, there is water barrier B (shown in the dash area), the non-water barrier in D zone among the figure.

This embodiment uses a camera to gather polarization image.A linear polarization filter is installed before camera, is aimed at the same area and keep camera position motionless, the angle of adjusting polarization filter light transmission shaft and surface level is respectively 0 °, 45 °, 90 °, takes the polarization image of three width of cloth the same areas.Three width of cloth image sizes are identical, and the visual angle is identical, and the position of the object in the zone in each image is identical.

The gray-scale value of remembering the picture element of corresponding same position in 0 °, 45 °, 90 ° these three width of cloth images is respectively I ₀, I ₄₅And I ₉₀, be that the picture element of (1,1) is an example with upper left corner coordinate, the gray-scale value of this point is respectively I in three width of cloth images ₀=73, I ₄₅=60, I ₉₀=96, by formula (1), (2)

(1)

$P=\frac{I_{90}-I_0}{(I_{90}+I_0)\cos 2\mathrm{\&theta;}}$ (2)

The polarization phase θ that calculates this point in this example is $\frac{1}{2}\arctan \left(\frac{73+96-2\&times;60}{96-73}\right)=0.5660,$ Degree of polarization P is $\frac{96-73}{(96+73)\cos (2\&times;0.5660)}=0.3203.$

The polarization phase and the degree of polarization of all picture elements calculated in pointwise, obtains the polarization phase figure and the degree of polarization figure in whole pre-detection zone.This two width of cloth figure is the image identical with the pre-detection area size, the polarization phase and the degree of polarization of each some expression pre-detection zone corresponding point of the inside.

Then, the similarity measure with polarization phase is that criterion judges whether water barrier exists.This method is based on following physical criteria: the relative surrounding environment of polarisation of light phase place by water-reflected has bigger similarity.Definition characterizes the similarity of polarization phase figure as the Δ of formula (3).By formula (3)

$\mathrm{\&Delta;}=\underset{i,j}{\mathrm{\&Sigma;}}|{\mathrm{\&theta;}}_{i,j}-\stackrel{\&OverBar;}{\mathrm{\&theta;}}|$ (3)

Calculate the correlativity of each point polarization phase in 11 * 11 pixel windows of the upper left corner of polarization phase figure, obtain the upper left corner in this example $\mathrm{\&Delta;}=\underset{j=1..11}{\underset{i=1..11}{\mathrm{\&Sigma;}}}|{\mathrm{\&theta;}}_{i,j}-\stackrel{\&OverBar;}{\mathrm{\&theta;}}|=31.0804.$

Calculate the correlativity of each point polarization phase in each 11 * 11 pixel window successively, obtain the similarity measure figure of whole pre-detection zone polarization phase.This figure is the width of cloth image identical with the pre-detection area size, the similarity degree of the polarization phase of 11 * 11 pixel windows around each some expression pre-detection zone corresponding point of the inside.Δ is more for a short time to show that the phase portrait of this Δ institute corresponding region is high more, and vice versa.Similar because of the polarization phase of water body part, then the size of the Δ of each picture element is all less in the water body scope.

Estimate among the figure to judge with the size of Δ whether water barrier exists at phase portrait,, then judge in this zone to have water barrier when there being any one Δ≤1.In the present embodiment, calculate, phase portrait is estimated the point for (160,120) of coordinate among the figure, and its phase portrait is estimated Δ=0.7642.So judge in this preliminary examination zone and may have water barrier.

Then, estimate at phase portrait and carry out self-adapting threshold segmentation on the figure.The phase portrait that obtains in the present embodiment is estimated the figure minimum value and maximal value is respectively Δ ₁=0.2220 and Δ _m=91.6732, make threshold value initial value T ₀=(Δ ₁+ Δ _m)/2=45.9476,

By formula calculate (5), (6)

${\mathrm{\&Delta;}}_O=\frac{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}N(i,j)}$ ${\mathrm{\&Delta;}}_B=\frac{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}N(i,j)}$ (5)

T _k+1＝(Δ _O+Δ _B)/2 (6)

Obtain $T_1=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_0}{\mathrm{\&Sigma;}}N(i,j)})/2=33.9782.$

At this moment | T ₁-T ₀| T ₀* 1%, thus need substitution formula (5), (6) to continue to calculate,

Obtain $T_2=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_1}{\mathrm{\&Sigma;}}N(i,j)})/2=31.4126.$

At this moment | T ₂-T ₁| T ₁* 1%, so still need bring formula (5), (6) continuation calculating into.

Through 6 iteration, obtain T in the present embodiment ₅=29.3377,

$T_6=(\frac{\underset{\mathrm{\&Delta;}(i,j)<T_5}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_5}{\mathrm{\&Sigma;}}N(i,j)}+\frac{\underset{\mathrm{\&Delta;}(i,j)<T_5}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_5}{\mathrm{\&Sigma;}}N(i,j)})/2=29.0576.$

At this moment | T ₆-T ₅|＜T ₅* 1%, iteration finishes, and drawing the final gray threshold of cutting apart Δ in this example is 29.0576, judges that then Δ is a water barrier less than 29.0576 part.

The result of in the present embodiment Δ being cut apart has adopted 8 connected component labelings that are communicated with, and two connected regions are arranged, and its Δ is less than final gray threshold 29.0576, i.e. B zone and D zone as shown in Figure 2.

According to similar adaptive threshold dividing method, degree of polarization figure is carried out Threshold Segmentation, the C zone (getting ready among Fig. 2 shown in the part) that obtains as shown in Figure 2 is the part of degree of polarization greater than final threshold value 0.4708.

Segmentation result to degree of polarization P in the present embodiment has adopted 8 connected component labelings that are communicated with, and has only a connected region, and its degree of polarization P is greater than final gray threshold 0.4708, i.e. C zone as shown in Figure 2.

This moment, there was intersection in the B zone with the C zone, was kept; And the D zone does not have intersection with the C zone, and the D zone is removed as erroneous judgement.The final decision Δ is a water barrier less than 29.0576 B zone, coincide with the fact.

Claims (5)

1. detect the method for field water barrier based on polarization information, may further comprise the steps:

1) with the camera that the linear polarization filter is installed polarization image is taken in same pre-detection zone, the angle that obtains polarization filter light transmission shaft and surface level is respectively 0 °, 45 °, 90 ° three width of cloth images, remembers that the gray-scale value of the picture element of corresponding same position in 0 °, 45 °, 90 ° these three width of cloth images is respectively I ₀, I ₄₅And I ₉₀, respectively by formula (1),

(2) the polarization phase θ and the degree of polarization P of all picture elements calculated in pointwise, obtains the polarization phase figure and the degree of polarization figure in whole pre-detection zone;

$P=\frac{I_{90}-I_0}{(I_{90}+I_0)\cos 2\mathrm{\&theta;}}---\left(2\right)$

2) in whole polarization phase figure, in 7 * 7～15 * 15 pixel window ranges, choose the pixel window of any size, by formula (3) calculate the correlativity that each should select picture element polarization phase in the pixel window successively, obtain the similarity measure figure of whole pre-detection zone polarization phase

$\mathrm{\&Delta;}=\underset{i,j}{\mathrm{\&Sigma;}}|\mathrm{\&theta;}(i,j)-\stackrel{\&OverBar;}{\mathrm{\&theta;}}|---\left(3\right)$

Δ is for characterizing the similarity measure of polarization phase figure in the formula, and θ (i is respectively i for horizontal ordinate among the polarization phase figure and ordinate j), the polarization phase of the picture element of j, and θ is for there being the arithmetic mean of a polarization phase in the pixel window of selected size;

3) judge in the pre-detection zone whether have water barrier with the size of Δ, in polarization phase similarity measure figure, have any one Δ≤1, then judge in this zone to have water barrier, otherwise judge and do not have water barrier in this zone;

4) may there be water barrier if judge before, estimate at phase portrait and to carry out self-adapting threshold segmentation on the figure, judge that Δ is a water barrier less than the part of final threshold value, and the result who obtains carried out 4 connections or 8 connected component labelings that are communicated with, ask the specific algorithm of threshold value to be:

4.1) obtain the minimum value Δ of similarity measure figure ₁With the maximal value Δ _m, make the threshold value initial value:

T ₀＝(Δ ₁+Δ _m)/2 (4)

4.2) note T _kBe the threshold value of the k+1 time iteration, k=0,1,2... is according to threshold value T _kFigure is divided into greater than T with similarity measure _kWith less than T _kTwo parts, by formula (5) obtain two-part average measure value Δ _OAnd Δ _B,

${\mathrm{\&Delta;}}_O=\frac{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)<T_k}{\mathrm{\&Sigma;}}N(i,j)}$ ${\mathrm{\&Delta;}}_B=\frac{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}\mathrm{\&Delta;}(i,j)\&times;N(i,j)}{\underset{\mathrm{\&Delta;}(i,j)>T_k}{\mathrm{\&Sigma;}}N(i,j)}---\left(5\right)$

Δ in the following formula (i is that horizontal ordinate and ordinate are respectively i among the similarity measure figure j), the similarity measure value of the picture element of j, N (i is that horizontal ordinate and ordinate are respectively i among the similarity measure figure j), the weight of the picture element of j, and N gets 1;

4.3) by formula (6) obtain new threshold value:

T _k+1＝(Δ _O+Δ _B)/2 (6)

4.4) if T _K+1With T _kError in 1%, promptly | T _K+1-T _k|≤T _k* 1%, then iterative process finishes, otherwise makes T _k=T _K+1, change step 4.2 over to) and the continuation iteration;

5) process of similar step 4) is that object carries out self-adapting threshold segmentation on degree of polarization figure with degree of polarization P, and degree of polarization P is judged to water barrier greater than the part of final threshold value, and the result who obtains is carried out 4 is communicated with or 8 connected component labelings that are communicated with;

6) connected region of the different labels that step 4) is obtained, all connected regions that obtain with step 5) compare successively, if a certain zone of the connected region that step 4) obtains and all connected regions that step 5) obtains all do not have intersection, then the connected region of this label is removed as erroneous judgement; If a certain zone of the connected region that step 4) obtains and any one connected region that step 5) obtains have intersection, then keep the segmentation result of the connected region of this label as water barrier.

2. according to the said method of claim 1, it is characterized in that the result that step 6) is obtained opens the morphologic filtering computing of afterwards closing earlier with 5 * 5 disk template based on polarization information detection field water barrier.

3. according to the said method that detects field water barrier based on polarization information of claim 1, it is characterized in that adopting a camera, a linear polarization filter is installed before camera, and the angle of taking polarization filter light transmission shaft and surface level respectively is 0 °, 45 °, 90 ° three width of cloth images.

4. according to the said method that detects field water barrier based on polarization information of claim 1, it is characterized in that adopting three cameras, a linear polarization filter is installed before every camera, and the angle of taking polarization filter light transmission shaft and surface level simultaneously is 0 °, 45 °, 90 ° three width of cloth images.

Images