A kind of method extracting lichen coverage based on Lab color space algorithm
Technical field
The present invention relates to a kind of method obtaining lichen coverage, particularly relate to a kind of based on Lab color space algorithm
The method extracting lichen coverage, relates to Computer Image Processing and field of ecology.
Background technology
At present, in the research of vegetation coverage information retrieval, the method utilizing computer digital image to process is to compare
One of universal method, and simple.Traditional lichen coverage extracting method is grid eye estimating method, draws with will studying sample
It is divided into the sample prescription of some area equation, is the most directly differentiated by naked eyes, then estimate the vegetation coverage of sample prescription, finally
Average as the vegetation coverage studying sample ground.The sampling frame typically utilizing 0.5m*0.5m selectes lichen region, has in frame
26*26 cross point, estimates the vegetation coverage determining survey region, though the method is simple by the number in statistics cross point
Easy, but subjective random big, and visual estimation precision is closely related with the experience measuring people, and research is thought, what single visual estimation obtained covers
Cover degree data maximum absolute error up to 40.4% [Zhang Wenbo, Fu Suhua, Liu Baoyuan. eye estimating method measures the essence of vegetation coverage
Degree analyzes [J]. Beijing Normal University's journal (natural science edition), 2001,03:402-408.].Supervised classification needs to select training
Sample, as classification reference, therefore can introduce a lot of anthropic factor, it is impossible to accomplish automation mechanized operation in categorizing process, time-consuming takes
Power;Unsupervised classification only relies on inhomogeneity terrestrial object information on image and carries out feature extraction, though need not more priori, but
Need its result is carried out big component analysis and post processing;Although and visual interpretation precision is higher, however it is necessary that and manually choose image
In all vegetation informations, therefore also bigger than the workload of supervised classification (Lu Ping Jun, Liu Honggu, symbol element China etc. photographic process combines
Digital image technology calculates vegetation coverage Research on Accuracy [J]. water and soil conservation).
Utilizing photographic process to extract vegetation coverage in measurement near the ground is currently with one of more method, and its principle is
Vertically is taken a picture in surveyed region, then extract the information on photo and calculate vegetation percentage ratio.It is usually and uses at image
Photo is interpreted by reason software, calculates vegetation pixel number and accounts for the percentage ratio of total pixel number.Divided by spectrum texture such as Zhou etc.
Class device automatically extracts grassland vegetation coverage (Zhou Q, Robson M.Automated rangeland vegetation
cover and density estimation using ground digital images and a spectral-
contextual classifier[J].Remote Sensing,2001,22(17):3457-3470).Anatoly etc. use
Wheat coverage (Anatoly A.Gitelson, the Yoram of United States Nebraska estimated by digital camera
J.Kaufman,Robert Stark,Don Rundquist.Novel algorithms for remote estimation
of vegetation fraction[J].Remote Sensing of Environment,2002,801:.).Friedl etc.
Utilizing digital camera to extract U.S.'s arid ecosystems system vegetation coverage, research shows to utilize digital camera to extract vegetation coverage
Be effectively and accurately (M.A.Friedl, J.Michaelsen, F.W.Davis, H.Walker,
D.S.Schimel.Estimating grassland biomass and leaf area index using ground and
satellite data[J].International Journal of Remote Sensing,1994,157:.).Photographic process
In conjunction with computer disposal, the most quickly, extraction accuracy is high, it is seen that the method utilizing digital photo technologies extraction vegetation coverage is non-
Often there are prospect and researching value.
The method extracting vegetation coverage based on photographic process has a lot, is roughly divided into three classes.The first kind is to directly utilize number
The feature of word photo color, extracts vegetation coverage, as Zhang Yun's rosy clouds etc. (Zhang Yun's rosy clouds, Li Xiaobing, Zhang Yunfei. based on numeral phase
Machine, ASTER and MODIS image integration measure vegetation cover degree [J]. Acta Phytoecologica Sinica, 2007,05:842-849.) propose
RGB threshold method, meadow, Inner Mongolia area carried out research, and thinks visual with by image processing software of RGB threshold method
Interpretative method precision is suitable.Zhang Chaobin etc. (Zhang Chaobin, Li Jianlong, Zhang Ying, Zhou Wei, Qian Yurong, Yang Feng. one based on RGB pattern
Meadow cover degree quantitative rapid assay methods research [J]. Acta Prataculture, 2013,04:220-226.) utilize RGB decision tree method to extract
The grassland region vegetation coverage in Fukang, Xinjiang city, extraction accuracy is more than 95%.
Equations of The Second Kind is to build vegetation index to extract vegetation coverage, only comprises the letter of tri-wave bands of RGB due to digital photos
Breath, being usually used in extracting the normalized differential vegetation index of vegetation, ratio vegetation index etc. needs near infrared band to participate in the vegetation calculated
Index cannot extract the vegetation coverage of digital photos, so some scholar utilizes the vegetation only comprising RGB band class information to refer to
Number extracts vegetation coverage.As Hu Jianbo etc. (Hu Jianbo, Zhang Lu, Huang Wei etc. grassland vegetation coverage based on digital photograph
Rapid extracting method [J]. Practaculture Science, 2011,09:1661-1665.) choose super green feature as vegetation index, by setting
Threshold value extracts Grass cover degree, extracts result for pure green plants preferable, but plant unconspicuous to green characteristic, exist low
The phenomenon estimated.Ren Jie etc. (appoint outstanding person, Bai Yanchen, Wang Jin ground. the technique study of rapid extraction vegetation coverage from digital photograph
[J]. remote sensing technology and application, 2010,05:719-724.) utilize normalization differential index (di) (Normalized Difference
Index, NDI) carry out processing extraction vegetation coverage information to digital photos, result shows to utilize NDI classification can reach prison
Superintend and direct the precision of classification, and rapidly and efficiently.
3rd class is that digital photos is converted to other color spaces, by the feature extraction vegetation of different color spaces
Coverage.The most frequently used color space is HLS color space, as Li Cunjun etc. (Li Cunjun, Wang Jihua, Liu Liangyun etc. based on number
The wheat coverage of word photo eigen automatically extracts research [J]. journal of Zhejiang university (agricultural and life sciences version), and 2004,06:
64-70.) setting tone (H) value and extract wheat coverage, eliminate brightness and effect of saturation degree in image, major part soil is all
Can be non-vegetation by successful classification, but red soil there is also certain interference to Semen Tritici aestivi classification.Xing Bailing etc. (Xing Bailing,
Occupying as the people, Zhu Gaolong etc. the actual measurement of Hulunbuir Pasture Land vegetation coverage ground studies [J] with remote sensing appraising. Agriculture in Jiangxi journal,
2012,05:142-147+163.) utilizing HLS color transformed and RGB combined techniques extracts meadow and is coated to cover degree, extraction accuracy is higher,
But the partially yellow plant of small part can be classified as non-vegetation.HSV and Lab color space is also utilized to extract vegetation coverage, as
Richardson etc. utilize the tone (H) in hsv color space to be worth setting the circular turfgrass extracting golf course, Liu Yaokai
Deng (Yaokai Liu, Xihan Mu, Haoxing et al.Anovel method for extracting green
Fractional vegetation cover from digital images [J] .J Veg Sci, 2012,233:.) utilize
Digital photos is classified by the feature of Lab color space, and extraction effect is better than supervised classification.
In the above sorting technique, the sorting technique of RGB and HLS color space is utilized to have luminance factor interference, and
Utilize Lab color space can also separate brightness, reduce the brightness impact on classification;Lab and the tone (H) utilizing HLS and HSV
Component is compared with the sorting technique of RGB, and its tone colour gamut is broad, also compensate for the deficiency of other patterns such as RGB.Existing based on number
The vegetation coverage research of word photo is extracted, for the extraction of bryophyte mainly for the crops such as Oryza sativa L., Semen Maydis and meadow
The most not relevant research.
Summary of the invention
It is an object of the invention to provide a kind of method extracting lichen coverage based on Lab color space algorithm, should
Method, in addition to can extracting for single vegetation color, is particularly suited for the covering with the bryophyte that multiple color mixes
Degree extracts, and method is simple, and has degree of precision.
The present invention based on Lab color space algorithm extraction lichen coverage method, described extracting method include with
Lower step: the lichen digital photos of acquisition is converted to Lab color space by RGB color by (1);
(2) bryophyte in digital photos is carried out color analysis, utilize the red green component extraction of Lab color space to go out
Lichen region;
(3) obtain the pixel count in lichen region, and calculate the ratio of itself and photo total pixel number, it is thus achieved that lichen coverage.
In technique scheme, the method for described step (1) is: first transfer RGB color to color tristimulus values
X, Y, Z, the computing formula further according to CIE 1976 homogeneous color space Lab tries to achieve Lab color space chromacity value.
Described step (2), particularly as follows: the red green section a component only chosen in Lab color space, arranges the ash to be rejected
Angle value threshold value [T1, T2], by [T in a component1, T2] pixel be set to 0, in a component, rest of pixels point is set to 1, it is thus achieved that
Two-value black white image, extracts lichen region.
In described two-value black white image, white is lichen region, and black is non-lichen region, calculates lichen area pixel
Number and the ratio of photo total pixel number, it is thus achieved that lichen coverage.
Described gray value threshold value [T1, T2] when arranging often according to the COLOR COMPOSITION THROUGH DISTRIBUTION of real figure photo and incorporate experience into,
Generally could be arranged to [-1,1].
It is bigger that Lab color space compares RGB color, and comprises all perceived colors of the almost mankind, passes through
Analyzing the feature of bryophyte of digital photos, the color of bryophyte is the most single, sorting technique based on Lab color space with
Other digitized video sorting techniques are compared, and utilize Lab color space can be good at removing outside plant intrinsic colour other
The result of classification is disturbed and impact by color, i.e. merely with a component, can remove brightness (L) component and blue yellow section (b component)
Interference.
The invention has the beneficial effects as follows:
The present invention utilizes the color component feature of Lab color space, i.e. a component in Lab only comprises red green two colors
Component, is applied to lichen coverage and extracts, and by chance can make up lichen due to trophophase difference, green and red lichen meeting
Growth mixed in together causes the defect that vegetation color is the most single, and avoids the impact on classification of other colors simultaneously, it is possible to very
Good extraction lichen coverage, is effectively improved bryophyte extraction accuracy.The inventive method is by carrying out RGB-to digital photos
Lab color notation conversion space processes, and in conjunction with respective algorithms rapid extraction lichen coverage from digital photos, is effectively reduced artificial
Operate thus reduce workload, improve efficiency;And the precision using the method to extract coverage information is also generally higher than other
Method, the method extracting lichen coverage of a kind of accurate objective of can yet be regarded as.
Accompanying drawing explanation
Fig. 1 is the technology path schematic diagram of the inventive method;
Fig. 2 be the present invention based on sorting technique contrast schematic diagram, photo lichen kind with other sorting technique classification results
Class is little plumage moss (Bryohaplocladium angustifolium), is collected in Areas In Tianshan Region In Xinjiang in by the end of March, 2014, all
Photo all uses Nikon D90 digital camera to shoot under the cloudless state of fine day.The digital photos of shooting stores with JPGE form.
Fig. 3 is that the present invention contrasts schematic diagram with other sorting technique niceties of grading, utilizes the error moments tactical deployment of troops to sorting technique
It is evaluated.
Detailed description of the invention
Say that the invention will be further described below in conjunction with the accompanying drawings
Fig. 1 is sorting technique technology path schematic diagram of the present invention, by analyzing bryophyte color feature, utilizes Lab face
The red green component (a*) of the colour space extracts the lichen coverage in digital photos.First by the RGB of the lichen digital photos of acquisition
Color space conversion is Lab color space;Secondly, according to lichen color feature on digital photos, Lab color space is utilized
Red green component (a*), sets gray value threshold value and extracts lichen coverage;Finally, by the lichen part extracted and non-lichen
(background) is partially converted to two-value and adds up all kinds of pixel count, calculates lichen coverage.
Digital photos has 3 different colors, spectrum sensitive wave bands different in each color correspondence visible ray, these three
Wave band is respectively R (red) G (green) B (blue), and scope is 0~255, describes and generates the specific colorant required for color in a device
Quantity.
Lab color space is that International Commission on Illumination is at the 1976 a kind of color modes formulated.Any one of nature
Color can be expressed in Lab color space, and its color space is bigger than rgb space.It does not relies on equipment, and color
Field width is wealthy, compensate for the shortcoming that rgb color model COLOR COMPOSITION THROUGH DISTRIBUTION is uneven.Lab color space is an outstanding brightness and colour
Separator, it is very useful in terms of compression of images.What Lab color space described is the display mode of color, wherein, L
(Lightness) representing brightness, codomain is [0,100], and a, b are color component, and a component color is green to redness, positive number generation
Table is red, and negative terminal represents green;B component color is blue to yellow, and positive number represents yellow, and negative terminal represents blueness.Two colors
Component codomain is [-120 ,+120].Owing to having separated the monochrome information of digital photos, the number of shooting under different brightness can be reduced
The impact that lichen vegetation coverage is extracted by word photo, has so utilizing Lab color space that lichen digital photos is carried out process
The biggest advantage.And digital picture is transformed into Lab color space from RGB color, need first by RGB patten transformation to XYZ
Tristimulus values, then it is transformed into Lab color space by XYZ.Concrete transformational relation is following (to be specifically referred to: Wang Ke, Lu Changde, pleasure
Wan De. the research of Color Similarity Measurement and application [J]. computer utility is studied, 2005,10:168-170.):
Colour brightness equation:
Y=0.2963R+0.6192G+0.0845B (1)
In formula, R, G, B are respectively the content that red, green, blue is trichroism, and span is 0~100, according to CIE spectrum tristimulus
Formula:
The transition matrix of RGB with XYZ is:
Turning Lab color space is:
In its Chinese style (2) (3) (4), X0、Y0、Z0For the tristimulus values of standard light source D65, its value is X0=95.045, Y0=
100、Z0=108.255.
After the color space conversion of lichen digital photos, red green component (a) is only utilized to extract lichen coverage, i.e. to red
Green component (a) sets threshold value [T1, T2], can be by the lichen extracting section in digital photos out red according to Lab color space
Green component (a) feature, sets T1、T2Two threshold values should avoid intermediate value, according to the analysis to lichen image, avoids [-1,1]
Gray areas, to remove the interference of greyish white blackness.Then meet condition then for lichen region when pixel, be otherwise non-lichen (background) district
Territory, i.e.
Utilize above-mentioned formula, extract the lichen coverage in digital photos and non-lichen region.
By adding up each lichen and the number of pixels in non-lichen region, thus draw lichen coverage, i.e. lichen pixel
Number accounts for the ratio of the total number of pixels of photo.
Fig. 2 lists sorting technique and other classification results being usually used in remote sensing and ecological approach herein contrast, from dividing
From the point of view of class result effect, method of maximum likelihood and this research sorting technique effect are preferable, automated taxonomy based on HLS color space
Relatively big to the non-green lichen fractional error extracted in digital photos, and ISODATA sorting technique is disturbed relatively by other atural objects
Greatly, classification results is the most bad.
Fig. 3 lists the present invention and contrasts schematic diagram with other sorting technique niceties of grading, compare from each method nicety of grading
From the point of view of, ISODATA method precision is minimum, and the effect of classification is the most bad, and dash area is by mistake point more;Method of maximum likelihood and classifying herein
Method result is close, and precision is higher;The precision of automated taxonomy is slightly lower, although automated taxonomy extracts vegetation coverage in theoretical essence
Up to more than 90% on Du, but owing to the method is to classify pure green plants, it is contemplated that to extracting lichen not only
Only extract green portion, so precision is the highest when extracting lichen, there is the phenomenon underestimated.So, classification results comes the most herein
Seeing, it is poor that method of maximum likelihood and automated taxonomy extract result, and precision is relatively low, and method of maximum likelihood and Lab classification method precision are higher.
Grid eye estimating method only compares from lichen coverage data, and grid eye estimating method can ignore the lichen that area is less, and due to people's
Experience is different, thus there is bigger error.