CN102930244B - A kind of natural landmark based on pixel may differentiate and touchdown area defining method - Google Patents

Abstract

The present invention relates to a kind of natural landmark based on pixel may differentiate and touchdown area defining method, belong to field of deep space exploration.The method is lower based on the may differentiate of the pixel in the pixel in flat site in image and its neighborhood, and meteorite crater, the feature that the may differentiate of the pixel in the pixel in the regions such as rock and its neighborhood is higher, by the may differentiate of pixel each in computed image, using may differentiate lower than the pixel of given threshold value as the candidate point of touchdown area, using may differentiate higher than the pixel of given threshold value as natural landmark candidate point, using the set of the natural landmark candidate point in given radius as natural landmark; By carrying out morphological dilations to the pixel in natural landmark candidate point and image boundary, to be finally inflated the touchdown area candidate point of operation covering as the center of circle, with this center of circle to the distance of the natural landmark candidate point nearest apart from it for radius, obtain the maximum circle that lands.The present invention the determination of fast and reliable can manage to make do region, land.

Description

A kind of natural landmark based on pixel may differentiate and touchdown area defining method

Technical field

The invention belongs to field of deep space exploration, relate to the correlation technique of image procossing.

Background technology

Natural landmark detects and touchdown area determines it is the major issue in deep-space detection field.In survey of deep space process, because the lander distance earth is very remote, signal transmission has larger time delay, therefore needs the attitude of the determination self that lander can be autonomous and position and autonomous finding can touchdown area.

Natural landmark refers in environment existing, and unartificial setting, can in order to identify references object or the pattern of varying environment.Natural landmark detects attitude and the position of lander being determined to self, and carrying out independent navigation has important effect.(as meteorite crater, rock etc., image appearance features as shown in Figure 1) carries out natural landmark detection to the various landform mostly existing natural landmark detection method is for planetary surface.But these class methods are subject to the impact of illumination and concrete landform, cannot obtain good result.

And touchdown area determines for detector it is also extremely important automatically.In lander landing period, because the distance earth is very remote, rely on ground communications can not meet the demand of actual time safety landing, therefore need detector to have the stronger ability independently determining touchdown area.Existing independent landing area determination method is mainly by detecting the meteorite crater of planetary surface, and the barriers such as rock, determine touchdown area by avoiding obstacles.Also the textural characteristics of with good grounds planetary surface carries out the defining method of touchdown area.The problem of carrying out the method that touchdown area is determined by detecting barrier is that the detection of barrier is difficult to accomplish accurately and robust, simultaneously larger by lighting conditions.Existing texture analysis method is undertaken by features such as the contrasts in statistical regions, is also subject to the impact of the conditions such as illumination.

We make discovery from observation, the barrier of planetary surface, no matter be meteorite crater (Fig. 1 (a)), rock (Fig. 1 (b)), or crackle (Fig. 1 (c)), all there is following characteristics: the point in these obstacles borders is (as 1 in Fig. 1,2,3,4), very high with degree distinguished of other points in its neighborhood, be both easy to these points to make a distinction with other point of surrounding.And the point (as 5 in Fig. 1,6,7) in flat site, more similar with the point around it, substantially cannot with around make a distinction.We are based on this observation, carry out under natural landmark being detected the may differentiate framework determining to be unified in pixel with touchdown area.

Summary of the invention

The object of the invention is to propose a kind of natural landmark based on pixel may differentiate for the deficiency that above-mentioned natural landmark detects and touchdown area defining method exists to detect and touchdown area defining method.

Main contents of the present invention are: employing video camera obtains the image information below lander in lander decline process, in the image that each frame gets, calculate the may differentiate of each pixel, the degree distinguished of the pixel both in this pixel and its neighborhood.Setting threshold value, may differentiate higher than the pixel of given threshold value as natural landmark candidate point, using the set of the natural landmark candidate point in certain radius as natural landmark; May differentiate lower than the pixel of given threshold value as can the candidate point of safe landing locations, by carrying out morphological dilation to the pixel in natural landmark candidate point and image boundary, to be finally inflated the touchdown area candidate point of operation covering as the center of circle, with this center of circle to the distance of the natural landmark candidate point nearest apart from it for radius, obtain the maximum circle that lands.

The object of the invention is to be achieved through the following technical solutions.

Natural landmark based on pixel may differentiate detects and a touchdown area defining method, and specific implementation step is as follows:

Step one: Image Acquisition

Video camera is arranged on the below of lander, can obtain the image on ground by continuous print in lander decline process.

Step 2: the may differentiate calculating pixel

If s is (p, p ') for pixel p and pixel p ' similarity degree, if p (s (p, p ')) represents the probability of pixel p and pixel p ' similar, the may differentiate of definition pixel is the entropy that in this pixel and its neighborhood, pixel likelihood probability distributes, both

${\mathrm{distinguishbility}}_p=-{\mathrm{\Σ}}_{p^{\′}\∈N_p}p\left(s(p,p^{\′})\right)\log p\left(s(p,p^{\′})\right)---\left(1\right)$

Wherein N _pfor the neighborhood of pixel p.Entropy is larger, then may differentiate is lower.

Step 3: the may differentiate based on pixel is determined can touchdown area

The region that may differentiate is high and that may differentiate is medium pixel covers all regards barrier region as, and the region that the pixel that may differentiate is low covers is regarded as can touchdown area.Also regard the point in image boundary as pixel that may differentiate is high, both also regard barrier zone as.Morphological dilations is carried out to the pixel of barrier region at every turn, pixel around barrier is also set to barrier region, along with the carrying out of expanding, can touchdown area more and more less, be finally inflated that operation covers can touchdown area be exactly maximum can the center of circle of touchdown area.Maximum land radius of a circle by calculate with the nearest expansion of this point before high may differentiate or the distance of pixel of medium discrimination obtain

Step 4: the natural landmark based on high discrimination pixel detects

Have high may differentiate pixel can with its around pixel region separate, can as natural landmark.We using the set with the pixel of high may differentiate in certain radius as natural landmark.

Advantage of the present invention

The present invention, compared with other natural landmark and touchdown area defining method, has the advantage of the following aspects:

(1) by meteorite crater, the detection of the various natural landmark such as rock and can the detection of touchdown area be unified in the framework of may differentiate under process, natural landmark and touchdown area can be obtained simultaneously.

(2) obtain the maximum round center of circle and the radius of landing by carrying out morphologic expansive working to non-natural road sign candidate point, the determination of fast and reliable can manage to make do region, land.

Accompanying drawing explanation

The various landform of Fig. 1 planetary surface;

Fig. 2 is based on the natural landmark of pixel may differentiate and touchdown area defining method process flow diagram;

The likelihood probability distribution plan of pixel in Fig. 3 pixel and its neighborhood;

The maximum round defining method schematic diagram that lands of Fig. 4.

Embodiment

Below in conjunction with drawings and Examples, the present invention is elaborated.

As shown in Figure 2, concrete implementation step is as follows for the detection of the natural landmark based on pixel may differentiate that the present invention proposes and the process flow diagram of touchdown area defining method:

Step one: Image Acquisition

Video camera is arranged on the below of lander, can obtain the image on ground by continuous print in lander decline process.

Step 2: the may differentiate calculating pixel

If p represents a pixel in image, its coordinate is (px, py), and its pixel value is I (px, py), then the may differentiate of pixel p is the degree distinguished of other pixels in p and its neighborhood.If the neighborhood of pixel p is N _p, N _pfor the rectangle of centered by pixel p, its length and width can set according to actual conditions, if its length and width are set as width and the height of image, are then the may differentiate carrying out this pixel in whole sub-picture, arrange neighborhood N _plength and width be respectively picture traverse and height 1/3rd.Definition s (p, p ') for pixel p and pixel p ' similarity degree, herein, we adopt SSD to calculate the similarity degree of two pixels, SSD value is less, show that two pixels are more similar, the Size of Neighborhood adopted when SSD is calculated in design is (2n+1) × (2m+1), then

$s(p,p^{\′})=\underset{i=-n}{\overset{n}{\mathrm{\Σ}}}\underset{j=-m}{\overset{m}{\mathrm{\Σ}}}{[I(\mathrm{px}+i,\mathrm{py}+j)-I(p^{\′}x+i,p^{\′}y+j)]}^2---\left(2\right)$

Wherein (px, py) is pixel p coordinate in the picture, the coordinate that (p ' x, p ' y) is pixel p ' in the picture.If p (s (p, p ')) represents the probability of pixel p and pixel p ' similar, then

$p\left(s(p,p^{\′})\right)=\frac{\exp \{-s(p,p^{\′})\}}{{\mathrm{\Σ}}_{p^{\′}\∈N_p}\exp \{-s(p,p^{\′})\}}---\left(3\right)$

The may differentiate of definition pixel is the entropy that in this pixel and its neighborhood, pixel likelihood probability distributes, both

${\mathrm{distinguishbility}}_p=-{\mathrm{\Σ}}_{p^{\′}\∈N_p}p\left(s(p,p^{\′})\right)\log p\left(s(p,p^{\′})\right)---\left(4\right)$

The pixel marked in Fig. 1 is calculated and distributes with the likelihood probability of pixel in its neighborhood, as shown in Figure 3, can find out, be distributed with three kinds of situations:

1. spot distribution, as shown in Fig. 3 (1-2), these pixels are generally the angle points in image, and the entropy of its likelihood probability distribution is less, and may differentiate is high.

2. wire distribution, as shown in Fig. 3 (3-4), these pixels are generally the edges in image, and the entropy of its likelihood probability distribution is placed in the middle, and may differentiate is medium.

3. planar distribution, as shown in Fig. 3 (5-7), these pixels generally belong to flat site, and the entropy of its likelihood probability distribution is comparatively large, and may differentiate is low.

Step 3: the may differentiate based on pixel is determined can touchdown area

The region that the pixel that we are high by may differentiate and may differentiate is medium covers all regards barrier region as, and the region that the pixel that may differentiate is low covers is regarded as can touchdown area.Can should meet following two requirements by touchdown area, one is that area is enough large, so that lander lands; Two is that distance barrier is enough far away, to guarantee the safety of landing mission.Because the region outside image boundary is zone of ignorance, therefore desirable touchdown area should range image border also enough far away.Therefore the may differentiate of the point in image boundary is also set to height by us, and both borderline point also regarded the pixel that may differentiate is high as.As shown in Figure 4, some barrier region red in figure, borderline point also thinks barrier region.Green portion is the pixel that may differentiate is low, can be used as touchdown area, and our target finds a maximum circle, makes this circle only comprise green area as can touchdown area.At every turn specific practice carries out morphological dilations to the pixel of barrier region, pixel around barrier is also set to barrier region, as Fig. 4 (b), (c), shown in (d), can see, along with the carrying out of expanding, green area is more and more less, the green area finally reddened be exactly maximum can the center of circle of touchdown area.Idiographic flow is:

1) successively morphological dilations is carried out to pixel (pixel of high may differentiate and the medium discrimination) point in barrier region, each pixel that expands; Both the region of a pixel around barrier region was also set to barrier region.

2) judge the number of the pixel that may differentiate is low, if number is greater than 1, then turn the first step; If number equals 1, then turn the 3rd step, if number is 0, then turn the 4th step.

3) if the number of the low pixel of may differentiate equals 1, then this pixel is the maximum round center of circle of landing, maximum land radius of a circle by calculate with the nearest expansion of this point before high may differentiate or the distance of pixel of medium discrimination obtain, as shown in Fig. 4 (d).

4) if the number of the low pixel of may differentiate equals 0, then choose before this time expanding, the low pixel of not capped may differentiate is as can the center of circle of touchdown area, if any multiple such pixel, then get closest to one of picture centre as can the center of circle of touchdown area, radius obtains by the distance of the pixel calculating high may differentiate before the expansion nearest with this point or medium discrimination.

Step 4: the natural landmark based on high discrimination pixel detects

Natural landmark refers in environment existing, and unartificial setting, can in order to identify references object or the pattern of varying environment.Can be found by the definition of natural landmark, the pixel with high may differentiate meets the definition of natural landmark, both had high may differentiate pixel can with its around pixel region separate, can as natural landmark.We using the set with the pixel of high may differentiate in certain radius as natural landmark.Because the may differentiate of these points is very high, therefore these pixels more guaranteedly can be traced in follow-up image.

Claims (1)

1. based on natural landmark and the touchdown area defining method of pixel may differentiate, it is characterized in that: a kind of natural landmark based on pixel may differentiate detects and touchdown area defining method, and specific implementation step is as follows:

Step one: Image Acquisition

Video camera is arranged on the below of lander, and in lander decline process, continuous print obtains the image on ground;

Step 2: the may differentiate of the pixel in the image that calculation procedure one obtains

If p represents a pixel in image, its coordinate is (px, py), and its pixel value is I (px, py), then the may differentiate of pixel p is the degree distinguished of other pixels in pixel p and its neighborhood; If the neighborhood of pixel p is N _p, N _pfor the rectangle of centered by pixel p, its length and width can set according to actual conditions, if its length and width are set as width and the height of image, then in whole sub-picture, carry out may differentiate differentiation to this pixel, arrange neighborhood N _plength and width be respectively picture traverse and height 1/3rd, definition s (p, p') be the similarity degree of pixel p and pixel p', SSD is adopted to calculate the similarity degree of two pixels, SSD value is less, show that two pixels are more similar, the Size of Neighborhood adopted when SSD is calculated in design is (2n+1) × (2m+1), then

$s(p,p^{\′})=\underset{i=-n}{\overset{n}{\Σ}}\underset{j=-m}{\overset{m}{\Σ}}{\[I(px+i,py+j)-I(p^{\′}x+i,p^{\′}y+j)\]}^2---\left(2\right)$

Wherein (px, py) is pixel p coordinate in the picture, and (p'x, p'y) is pixel p' coordinate in the picture, if p (s (p, p')) represents the probability that pixel p is similar to pixel p', then

$p\left(s\right(p,p^{\′}\left)\right)=\frac{\exp \{-s(p,p^{\′})\}}{{\mathrm{\Σ}}_{p^{\′}\∈N_p}\exp \{-s(p,p^{\′})\}}---\left(3\right)$

The may differentiate of definition pixel is the entropy that in this pixel and its neighborhood, pixel likelihood probability distributes, both

${\mathrm{distinguishbility}}_p=-{\mathrm{\Σ}}_{p^{\′}\∈N_p}p\left(s\right(p,p^{\′}\left)\right)\log p\left(s\right(p,p^{\′}\left)\right)---\left(4\right)$

Wherein, entropy is larger, and the may differentiate of pixel is lower, therefore draws following three kinds of distributions about pixel may differentiate:

(1) spot distribution, these pixels are the angle points in image, and the entropy of its likelihood probability distribution is less, is may differentiate height pixel;

(2) wire distribution, these pixels are the edges in image, and the entropy of its likelihood probability distribution is placed in the middle, is the medium pixel of may differentiate;

(3) planar distribution, these pixels are flat sites in image, and the entropy of its likelihood probability distribution is comparatively large, is the low pixel of may differentiate;

Step 3: the may differentiate based on the pixel of step 2 calculating is determined can touchdown area

The region that may differentiate is high and that may differentiate is medium pixel covers all regards barrier region as, and the region that the pixel that may differentiate is low covers is regarded as can touchdown area; Also regard the point in image boundary as pixel that may differentiate is high, both also regard barrier region as; Morphological dilations is carried out to the pixel of barrier region at every turn, pixel around barrier is also set to barrier region, along with the carrying out of expanding, can touchdown area more and more less, be finally inflated that operation covers can touchdown area be exactly maximum can the center of circle of touchdown area; The maximum radius of a circle that lands is obtained by the distance calculating the pixel that may differentiate is high or may differentiate is medium before the expansion nearest with the center of circle, and the determining step of its maximum the land round center of circle and radius is:

1) successively morphological dilations is carried out to the pixel in barrier region and may differentiate is high and may differentiate is medium pixel, each pixel that expands; Both the region of a pixel around barrier region was also set to barrier region;

2) judge the number of the pixel that may differentiate is low, if number is greater than 1, then turn the first step; If number equals 1, then turn the 3rd step, if number is 0, then turn the 4th step;

3) if the number of the low pixel of may differentiate equals 1, then this pixel is the maximum round center of circle of landing, and the maximum radius of a circle that lands is obtained by the distance calculating the pixel that may differentiate is high or may differentiate is medium before the expansion nearest with this pixel;

4) if the number of the low pixel of may differentiate equals 0, then choose before this time expanding, the low pixel of not capped may differentiate is as can the center of circle of touchdown area, if any multiple such pixel, then get closest to one of picture centre as can the center of circle of touchdown area, radius obtains by the distance calculating high may differentiate before the expansion nearest with this pixel or the medium pixel of may differentiate;

Step 4: the high discrimination pixel obtained based on step 2 carries out natural landmark detection

Have high may differentiate pixel can with its around pixel make a distinction, meet the requirement as road sign, as natural landmark; Using the pixel set with high may differentiate in certain radius as natural landmark.