CN103632148A - Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant - Google Patents
Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant Download PDFInfo
- Publication number
- CN103632148A CN103632148A CN201310683599.9A CN201310683599A CN103632148A CN 103632148 A CN103632148 A CN 103632148A CN 201310683599 A CN201310683599 A CN 201310683599A CN 103632148 A CN103632148 A CN 103632148A
- Authority
- CN
- China
- Prior art keywords
- skeleton
- image
- blade face
- leaf
- blade
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 7
- 238000000034 method Methods 0.000 claims abstract description 65
- 239000011159 matrix material Substances 0.000 claims abstract description 16
- 241000196324 Embryophyta Species 0.000 claims description 10
- 238000013507 mapping Methods 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 8
- 238000002372 labelling Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 241000510091 Quadrula quadrula Species 0.000 claims description 2
- 238000002474 experimental method Methods 0.000 claims description 2
- 238000005194 fractionation Methods 0.000 claims description 2
- 239000000463 material Substances 0.000 claims description 2
- 238000010845 search algorithm Methods 0.000 claims description 2
- 238000005259 measurement Methods 0.000 abstract description 11
- 238000012545 processing Methods 0.000 abstract description 6
- 230000011218 segmentation Effects 0.000 abstract description 4
- 238000004364 calculation method Methods 0.000 abstract description 3
- 238000004458 analytical method Methods 0.000 abstract description 2
- 238000003384 imaging method Methods 0.000 abstract 1
- 230000000877 morphologic effect Effects 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 210000000056 organ Anatomy 0.000 description 2
- 230000029553 photosynthesis Effects 0.000 description 2
- 238000010672 photosynthesis Methods 0.000 description 2
- 235000010469 Glycine max Nutrition 0.000 description 1
- 244000068988 Glycine max Species 0.000 description 1
- 241000238631 Hexapoda Species 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000000691 measurement method Methods 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000005068 transpiration Effects 0.000 description 1
- LENZDBCJOHFCAS-UHFFFAOYSA-N tris Chemical compound OCC(N)(CO)CO LENZDBCJOHFCAS-UHFFFAOYSA-N 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses an image-based method for automatically detecting a middle axis and the length of a leaf of a long-petiole axisymmetric plant. The method comprises the following steps: acquiring a leaf image containing a petiole by a scanner; carrying out threshold segmentation and morphological processing to obtain a leaf region; detecting a petiole region according to thin and long characteristics of the shape of the petiole and removing the petiole region to obtain a leaf surface region; extracting a leaf skeleton and reconstructing a leaf surface region by using a leaf surface skeleton by taking a skeleton adjacent to the petiole as a middle axis starting point; expanding and marking the skeleton and solving an adjacent matrix of a skeleton unit; exhausting all skeleton paths from the middle axis starting point to a skeleton end point by adopting the adjacent matrix; determining the middle axis by an equal area criterion and a maximum curvature criterion; fitting the middle axis by using a broken line; and determining the length of the leaf surface by using the length of the fitted broken line and a scanning resolution. According to the method disclosed by the invention, the detection of the middle axis of the leaf of the long-petiole axisymmetric plant and the calculation of the length of the leaf can be finished automatically; the measurement accuracy can be improved and the labor intensity is relieved; the automatic management analysis of related plant information can be realized conveniently.
Description
Technical field
The present invention relates to the fields such as agricultural production and management, biology, ecology and ecomanagement, specifically a kind of axis of the symmetrical plant leaf blade of arbor that comes into leaves based on image detects and blade face length measurement method.
Background technology
Blade is the major organs that plant carries out photosynthesis and transpiration, also be the major organs that photosynthesis carries out dry-matter accumulation, it is subject to the impact of the envirment factors such as moisture, temperature, illumination remarkable, therefore, the form parameter of blade as area, shape, thickness, leaf is long, leaf is wide and leaf index etc., and the important indicator of evaluating plant environment factorial effect is provided.Wherein length of blade, can be among the estimation of leaf area and the tasks such as automatic identification of plant leaf blade as important form parameter.
Current visible length of blade image measuring method, is substantially all the long edge lengths of calculating the minimum boundary rectangle of leaf area, or usining petiole/blade face point of interface sets out between other blade face marginal points the maximum in distance as the measurement result of length of blade.When not excessive blade exists certain distortion, such result and the length of blade of direct feel will disagree, and more suitably do rule, be to utilize the length of the axis of symmetry of blade or the axle that is right to be used as the description of length of blade.In addition, by blade axis itself, also can derive more blade shape feature, proportion as shared in the area of the width distribution feature along axis, blade different piece etc.
Therefore, be necessary to find a kind of method of processing based on image, can automatically detect the axis of rotational symmetry blade, calculate length of blade, and can improve the accuracy of measurement of length of blade and other associated vanes form parameters, alleviate worker's labour intensity, be convenient to realize automatic management and the analysis of relevant plant information.
Summary of the invention
Technical matters to be solved by this invention is on the basis of existing axis detection method (skeletonization method), for the plant leaf blade with come into leaves handle and axisymmetric shape, provides a kind of method of full automatic blade axis detection and blade face linear measure longimetry.
For solving the problems of the technologies described above, the solution that the present invention proposes is: the leaf image that utilizes scanner collection to comprise petiole; Through Threshold segmentation and morphology processing, obtain the bianry image of leaf area; The feature according to petiole with elongated shape, detects petiole region, obtains the bianry image in region, blade face after then removing petiole region, and blade is carried out obtaining the blade face skeleton point adjacent with petiole part as the starting point of axis after skeletal extraction; Utilize blade face skeleton to carry out blade face regional restructuring, and skeleton is expanded and mark, ask for the adjacency matrix between skeleton unit; Utilize this adjacency matrix, exhaustive go out all by axis starting point the skeleton path to skeleton end points; Utilize homalographic criterion and maximum curvature criterion to determine axis; Finally use segmented fitting axis, and utilize the length of matching broken line and the physical resolution in when scanning to determine blade face length.Specifically comprise the following steps:
I. utilize the image I of the arbor symmetrical blading that comes into leaves that scanner collection wins;
Ii. leaf image I is carried out to binaryzation, obtain bianry image B; Mode by experiment, for used concrete scanning background, determines a gray threshold t, and the pixel that then gray level in gray level image is not less than to t is set to 1,
In formula, 1≤x≤N and 1≤y≤M are respectively row subscript and the row subscript of each pixel in image, and N and M are respectively width and the height of image;
Iii. utilize a radius to be
Disc-shaped structure element S
preb is carried out to mathematical morphology open operator, obtain bianry image B
open,
B
open=BoS
pre?（3）
In formula, o represents opening operation in the mathematical morphology of bianry image;
Iv. to B
openutilize the hole fill method (realization of the bwfill function providing in can the image handling implement bag with reference to MATLAB 7.0) of existing standard to fill the hole in leaf area, then utilize the connected component labeling algorithm (realization of the bwlabel function providing in can the image handling implement bag with reference to MATLAB 7.0) of existing standard to find out each 8-connected region in the bianry image of having filled hole, retain wherein area the maximum and, as leaf area, obtain thus the bianry image L of blade;
V. utilize the skeletonization method realization of skel ' subfunction (in the bwmorph function providing in can the image handling implement bag with reference to MATLAB 7.0 ') of existing standard to extract the skeleton K of blade L
full;
Vi. in leaf image L, extract petiole region P and region, blade face F;
Vii. utilize region, blade face F and intact leaf skeleton K
fullobtain the skeleton K on blade face with operation,
K(x,y)=F(x,y)AND?K
full(x,y),1≤x≤N,1≤y≤M?（4）
In K, find a skeleton point (being white point) p
root=(x
root, y
root), K (x
root, y
root)=1, makes at p
root8 neighborhood positions at least there is a white point in P,
In formula, N
8location sets in 8 neighborhoods of (x, y) expression point (x, y), is defined as follows
N
8(x,y)={(x+1,y),(x+1,y-1),(x,y-1),(x-1,y-1),?（6）
(x-1,y),(x-1,y+1),(x,y+1),(x+1,y+1)}
If there is the skeleton point in a plurality of such K, get first run into point as p
root;
Viii. the m-neighbours of calculating K count mapping graph N
m, m-neighbours count mapping graph N
mbe defined as
M-abutment points wherein defines as follows: if by white point p=(x, y) 8 neighborhood position (x+1, y), (x+1, y-1), (x, y-1), (x-1, y-1), (x-1, y), (x-1, y+1), (x, y+1) and (x+1, y+1) number consecutively is position 0～7; If certain white point q has dropped on position c=0,2,4 or 6, q is the m-abutment points of p; If certain white point q has dropped on position c=1,3,5 or 7, and all there is no white point on c-1He position, position (c+1) mod 8, q is also the m-abutment points of p, otherwise q is not just the m-abutment points of p;
Ix. according to N
mto blade face, skeleton K carries out mark, obtains skeleton signature M
kand the adjacency matrix A of skeleton unit
k;
X. according to blade face skeleton K, carry out blade face regional restructuring, obtain region, reconstruct blade face bianry image F
rc, and according to N
mcarry out the expansion of blade face skeleton, so that the end points of skeleton all drops on the edge in region, blade face of reconstruct, the blade face skeleton signature of resulting expansion is designated as M
ex;
Xi. from M
exin read p
rootthe label l of point
root, then according to adjacency matrix A
k, from l
rootcorresponding row starts, and according to the depth-first tree search algorithm of existing standard, searches for, exhaustive all in blade face expansion skeleton l
rootcorresponding skeleton unit sets out, till the skeleton path of a certain other skeleton end points; The set in the exhaustive skeleton path obtaining is designated as C;
Xii. according to homalographic criterion and maximum curvature criterion, from C, select a paths a
med, this path is detects the blade face axis obtaining;
Xiii. given error of fitting is limit ε
fit, utilize existing standard fractionation segmented fitting method (can with reference to " R.C.Gonzalez, R.E.Woods, work. Ruan Qiuqi, Ruan Yuzhi, waits and translates. Digital Image Processing (second edition). Beijing: Electronic Industry Press, 2004, pp.524-525 "), obtain a
meda matching broken line; The value of error of fitting limit is generally corresponding to the physical size of 0.2～1.0mm;
Xiv. calculate the matching broken line length l in pixel, then, according to the physical resolution s of the image of image acquisition phase record, by following formula, obtaining actual blade face AL is blade face length l
w:
Described leaf image collection (i step) specifically comprises the following steps:
(i.1) blade is carried out certain smooth and clean, then utilize scanner scanning leaf image;
(i.2) when scanning adopt black not reflectorized material (as shaggy black paper) as scanning background, during scanning, should use the object with constant weight to push down blade to keep it smooth;
(i.3) if blade exists obvious branch (as maple leaf etc.), in scanning process, the blade-section of different branches must not overlap, in order to avoid people is the hole causing in leaf area, changes the shape of blade;
(i.4) Image Saving that scanning obtains is the gray level image of 256 gray levels;
(i.5) record the resolution s that when scanning adopts, the point/inch (dpi) of take is unit, as 300dpi etc.
Described petiole region and blade face method for extracting region (vi step) core is to utilize the series of steps that is equivalent to mathematical morphology open operator to realize the detection in petiole region, specifically comprises the following steps:
(vi.1) utilize existing standard bianry image contour extraction method (realization of the bwperim function that can provide with reference to the image handling implement bag of MATLAB 7.0) to extract the inverse image L of leaf image L
rev8-be communicated with profile C
l, rev; The inverse image L of image L
revbe defined as follows:
(vi.2) the digital disc-shaped structure element B that to utilize radius be r
rpto C
l, revcarry out the expansive working of mathematical morphology, obtain C
l, erode,
C
L,erode=C
L,rev⊕B
rp?（10）
In formula, ⊕ represents mathematical morphology expansive working; The value of r should be a bit larger tham 1/2 of petiole width, can determine as follows: obtain C
l, revin distance between 2 of all different white point centerings, then find one maximum in these distances, be made as d; Then get
r=βd?（11）
In formula, β is a default constant, and general value can be between 0.01～0.05;
(vi.3) obtain C
l, erodeand L
revexclusive disjunction image L
erode, rev,
L
erode,rev(x,y)=C
L,erode(x,y)OR?L
rev(x,y)?（12）
Then get L
erode, revinverse image L
erode, and extract L with existing standard bianry image contour extraction method
erode8-be communicated with profile C
erode;
(vi.4) again utilize B
rpto C
erodecarry out the expansive working of mathematical morphology, then ask for result images and L
erodeexclusive disjunction image L
open;
(vi.5) by L
opencarry out XOR with L, obtain R
1,
R
1(x,y)=L
open(x,y)XOR?L(x,y)?（13）
Then utilize the connected component labeling algorithm of existing standard to find out R
1in each connected region, retain wherein area the maximum as petiole region P;
(vi.6) P and L are carried out to XOR, obtain R
2, then utilize the connected component labeling algorithm of existing standard to find out R
2in each connected region, retain wherein area the maximum as region, blade face F.
The essence of described skeleton mark and the acquisition methods of skeleton unit adjacency matrix (ix step) is to utilize m-adjacency to define the adjacency between the connective and different skeleton connected unit between different skeleton points in skeleton K.According to m-neighbours' mapping graph N
mskeleton point can be divided into dissimilar connected unit or skeleton unit: by N
melement value equals the m-that 1 or 2 skeleton point forms and is communicated with the branch that skeleton is partly called skeleton, by N
melement value is not less than the m-that 3 skeleton point forms and is communicated with the fork block that skeleton is partly called skeleton.For each skeleton unit in K, utilize the method for graph search to mark it with the continuous positive integer label by 1 beginning.M-adjacency between skeleton unit is defined as follows: if for two skeleton unit A and B, at least have some p in an A and the some q in a B, making p and q is m-adjacency, and these two skeleton units of A and B are m-adjacency so.According to the adjacency between skeleton unit, can construct the adjacency matrix A of each skeleton unit of K
k.Concrete steps are as follows:
(ix.1) one of initialization is counted mapping graph N with the m-neighbours of page skeleton
mthe integer matrix M that size is identical
kas skeleton signature, wherein the value of all elements is set to 0; Storehouse A of initialization
stk; Initialization skeleton unit label i=0;
(ix.2) first investigate again N from top to bottom from left to right
min each element value, until run into a N
mposition (the x that element value equals 1
0, y
0), by (x
0, y
0) be pressed into storehouse A
stk;
(ix.3) if A
stkbe not empty, to (ix.4); Otherwise to (ix.11);
(ix.4) put i=i+1; Eject A
stkstack top element (x
c, y
c);
(ix.5) put M
k(x
c, y
c)=i;
(ix.6) if N
m(x
c, y
c) <3, to (ix.7); Otherwise to (ix.8);
(ix.7) search N
min with (x
c, y
c) be that m-adjacency and element value are not equal to 0 point (x
n, y
n); If N
m(x
n, y
n) <3 and M
k(x
n, y
n)=0, puts x
c=x
n, y
c=y
n, to (ix.5); If N
m(x
n, y
n)>=3 and M
k(x
n, y
n)=0, by (x
n, y
n) be pressed into A
stk, to (ix.3); If fail, find such abutment points, to (ix.3);
(ix.8) search N
min all and (x
c, y
c) be that m-adjacency and element value are not equal to 0 point (x
j, y
j) (1≤j≤n
c, n
cfor (x
c, y
c) the number of above-mentioned abutment points);
(ix.9) find (x
j, y
j) (1≤j≤n
c) in meet N
m(x
j, y
j) <3 and M
k(x
j, y
jthe point of)=0, and be pressed into A
stk;
(ix.10) find (x
j, y
j) (1≤j≤n
c) in meet N
m(x
j, y
j)>=3 and M
k(x
j, y
jfirst point (x of)=0
n, y
n), put x
c=x
n, y
c=y
n, to (ix.5); If fail to find such point, to (ix.3);
(ix.11) put n
u=i; Size of initialization is n
u* n
ulogical type matrix A
kas skeleton unit adjacency matrix, element wherein all sets to 0;
(ix.12) make i=1;
(ix.13) find M
kthe point that middle all elements value is i, finds the m-abutment points (x of all these points
k, y
k) (1≤k≤n
t, n
tnumber for all these abutment points); Investigate each (x
k, y
k), if there is M
k(x
k, y
k)=j ≠ i, puts A
k(i, j)=A
k(j, i)=1;
(ix.14) put i=i+1; If i<n
u, to (ix.13); Otherwise finish.
The basic object of described blade face regional restructuring and skeleton extended method (x step) is each end points (N that guarantees skeleton
mthe point that value equals 1) can drop on the edge in region, blade face.Basic ideas are: the skeleton end points that utilizes range conversion to find out not drop on blade face edges of regions, need to expand; The skeleton end points that need to expand each, finds its corresponding blade face edges of regions circular arc, if there are many such circular arcs, gets wherein elder; The angular bisector at angle and the intersection point of circular arc that find circular arc and the skeleton end points that is positioned at circle centre position to form, with straight-line segment, skeleton end points and this intersection point are coupled together, completed skeleton expansion, the skeleton part of expansion belongs to a skeleton unit together with corresponding skeleton end points.Concrete steps are as follows:
(x.1) utilize existing gauged distance mapping algorithm to ask for the inverse image L of leaf area L
reveuclidean Distance Transform matrix D;
(x.2) the reconstruct blade face bianry image F that size of initialization is identical with region, blade face bianry image F
rc, element wherein all sets to 0;
(x.3), to each the skeleton point (x, y) in the skeleton image K of blade face, find all distances to this skeleton point to be no more than the position (x of D (x, y)
i, y
i) (1≤i≤n
d, n
dnumber for above-mentioned position), then put F
rc(x
i, y
i)=1;
(x.4), after all skeleton points in K have all been carried out to the processing described in (x.3), utilize existing standard two-value region hole fill method to fill F
rcin hole, obtain the region, blade face of reconstruct;
(x.5) utilize existing standard two-value region contour extracting method, extract F
rcprofile C
rc;
(x.6) find N
mposition (the x that middle all elements value equals 1
j, y
j) (1≤j≤n
f, n
fnumber for above-mentioned position);
(x.7) to each position (x
j, y
j), if
do not expand; Otherwise at C
rcin find and retain and allly meet
White point (x, y), obtain thus the point bianry image C remaining
arc;
(x.8) utilize existing standard connected domain extracting method, find out C
arcin all 8-connected regions, and retain wherein area the maximum, the bianry image of result gained is C
major_arc;
(x.9) at C
major_arcin find two white points, and in their 8-neighborhoods separately, only have other a white point; Using these 2 end points as circular arc, do them to (x respectively
j, y
j) straight line, obtain thus an angle; Then find angular bisector and the C at this angle
major_arcintersection point (the x of middle white portion
m, y
m); Make (x
j, y
j) to (x
m, y
m) digitizing straight-line segment, and in K, the element value on this straight-line segment is made as to 1, at M
kmiddle element value on this straight-line segment is made as to M
k(x
j, y
j);
(x.10) to all (x
j, y
j) all carry out the operation described in (x.7)～(x.9) after, make M
ex=M
k, the skeleton signature M that has got final product to such an extent that pass through expansion
ex.
The key of described axis detection method (xii step) is to have introduced homalographic criterion and maximum curvature criterion, with them, determines that a paths in the exhaustive skeleton path obtaining is as blade axis.To every skeleton path a, utilize a that blade face is split into at least two area parts, find area maximum in these regions two, then to obtain compared with small size and larger area ratio, this ratio is not less than the given homalographic threshold value t of certain user
a, corresponding path a retains the candidate as axis, otherwise directly gets rid of, and the criterion of this judgement axis is called homalographic criterion; For candidate's axis path a, find its terminal that drops on blade face edges of regions, and calculate the curvature (can with reference to " D.G.Lowe.Organization of smooth image curves at multiple scales.International Journal of Computer Vision; 1989; vol.1; pp.119-130 ") of the blade face edges of regions of this destination county, in all candidate's axis path, finding this curvature is the maximum, is resulting blade face axis a
med; The criterion of this judgement axis is called maximum curvature criterion.Concrete steps are as follows:
(xii.1) empty skeleton set of paths C of initialization
a;
(xii.2) to the every paths a in skeleton set of paths C, by F
rcin all elements value that drops on a set to 0, obtain the blade face bianry image F of division
split;
(xii.3) utilize existing standard connected domain extracting method, find F
splitin all 4-connected regions, and find two of area maximum in these regions, the area of establishing them is respectively A
1and A
2, and A
1>=A
2;
(xii.4) calculate homalographic criterion function value f
a(a) as follows:
f
a(a)=A
2/A
1?（15）
If to the given homalographic threshold value t of user
a≤ 1, there is f
a(a)>=t
a, a is added to C
a; t
agenerally can be in [0.8,0.95] interval value;
(xii.5) if also have path not investigated in C, to (xii.2); Otherwise to (xii.6);
(xii.6) to C
ain every paths a, utilize existing curvature computing method, ask for the reconstruction blade F of this paths destination county
rcedge curvature κ (a); Then according to maximal margin curvature criterion, determine axis a
med,
In sum, method provided by the present invention can be automatically determined blade axis, and is calculated AL as length of blade in the tree-like skeleton of blade, thus realize blade axis and length accurately, automatically detect.But because skeletonization method is more responsive to the variation at edges of regions place, therefore method provided by the present invention be applicable to edge comparatively complete, without insect, gnaw and bite or gather the symmetrical plant leaf blade of the arbor that comes into leaves improper and breach that cause.
Accompanying drawing explanation
Fig. 1 is the overall procedure block diagram of institute's extracting method in the present invention;
Fig. 2 is the embodiment leaf image that utilizes the step I of institute's extracting method in the present invention to collect;
Fig. 3 utilizes the step I i～iv of institute's extracting method in the present invention to cut apart the embodiment leaf area bianry image obtaining;
Fig. 4 is the embodiment leaf area skeleton image of utilizing institute's extracting method step v gained in the present invention;
Fig. 5 is the image that the leaf area stack described in the leaf area skeleton described in Fig. 4 and Fig. 3 obtains;
Fig. 6 is the FB(flow block) of petiole and blade face method for extracting region in institute's extracting method step vi in the present invention;
Fig. 7 is the petiole region bianry image that utilizes the embodiment blade that in the present invention, institute's extracting method step vi extraction obtains;
Fig. 8 is region, the blade face bianry image that utilizes the embodiment blade that in the present invention, institute's extracting method step vi extraction obtains;
Fig. 9 is for utilizing institute's extracting method step vii in the present invention to extract the image that described in region, the blade face skeleton of the embodiment blade obtaining and Fig. 8, the stack of region, blade face obtains;
Figure 10 is region, the reconstruct blade face bianry image that utilizes institute's extracting method step viii～x gained in the present invention;
The image of Figure 11 for utilizing the stack of region, reconstruct blade face described in the expansion blade face skeleton of institute's extracting method step viii～x gained in the present invention and Figure 10 to obtain;
Figure 12 is for utilizing the blade axis of institute's extracting method step xi～xii gained in the present invention and the image that the stack of region, reconstruct blade face obtains.
Embodiment
Below with reference to the drawings and specific embodiments, the present invention is described in further details.
The overall procedure block diagram of institute of the present invention extracting method as shown in Figure 1.
The embodiment leaf image that the step I of institute's extracting method according to the present invention collects as shown in Figure 2.The physical resolution adopting during scanning is s=300dpi.Embodiment image is read in to MATLAB 7.0 in computing machine, and adopt gray-scale value 96 as segmentation threshold, and actionradius r
prethe disc-shaped structure element of=3 pixels, carries out the leaf area image that obtains after Threshold segmentation and aftertreatment as shown in Figure 3 by the step I i～iv of institute of the present invention extracting method.The bwmorph function that utilizes MATLAB 7.0 image handling implement bags to provide carries out the leaf area skeletonizing in the step v of institute of the present invention extracting method, and as shown in Figure 4, skeleton is superimposed upon effect on leaf area as shown in Figure 5 to the leaf area skeleton obtaining.
The FB(flow block) of petiole and blade face method for extracting region in the step vi of institute of the present invention extracting method as shown in Figure 6.Utilize petiole region bianry image and region, the blade face bianry image that this step obtains on embodiment image to distinguish as shown in Figure 7 and Figure 8, the radius of the disc-shaped structure element that wherein used is 25 pixels.
To utilize region, blade face skeleton that the step vii of institute of the present invention extracting method obtains and the superimposed image in region, blade face as shown in Figure 9.
Be region, the reconstruct blade face bianry image that the m-neighbours that utilize the step viii～x of institute of the present invention extracting method to carry out blade face skeleton count gained after mapping calculation, blade face skeleton mark, blade face regional restructuring and the expansion of blade face skeleton as shown in figure 10, after expansion, the skeleton of gained and the superimposed image in region, reconstruct blade face are as shown in figure 11.
That the step xi～xii that utilizes institute of the present invention extracting method detects the axis of embodiment blade and the superimposed image in region, reconstruct blade face, wherein the threshold value t of homalographic functional value obtaining as shown in figure 12
a=0.85.
According to the present invention, the step xiii～xiv of institute's extracting method, utilizes ε
fitthe limits of error of=6 pixels (corresponding to the physical size of 0.51mm) is carried out after segmented fitting extracted axis, the length of blade that the broken line length in pixels that utilization calculates and formula (8) conversion obtain is 111.0mm, and the length of blade manually being recorded by expert is 117.4mm.Take expert's measurement result as conventional true value, and the relative measurement error of institute of the present invention extracting method is 5.43%.
As a comparison, also utilize " Tang Xiaodong, Liu Manhua; Zhao Hui, etc. the soybean leaves identification under complex background. electronic surveying and instrument journal, 2010; vol.24, no.4, pp.385-390 " in method (being designated as MER-D method) He“ garden Wei outstanding; Hu Di. utilize square to realize the measurement of plant leaf blade length and width. computer engineering and application; 2013, vol.49, no.16; pp.188-191,231 " in method (being designated as MER-M method) embodiment blade has been carried out to linear measure longimetry.The measurement result of MER-D method is 97.9mm, and relative measurement error is 16.61%; The measurement result of MER-M method is 97.1mm, and relative measurement error is 17.26%.
In the present embodiment visible, institute of the present invention extracting method has been obtained measurement result more accurately, and has realized the fully automatic operation of image processing and computation process.
Claims (6)
- Based on image, for the axis with the rotational symmetry plant leaf blade of the handle that comes into leaves, detect and the method for blade face linear measure longimetry, comprise the following steps:I. utilize the image I of the arbor symmetrical blading that comes into leaves that scanner collection wins; Physical resolution s during writing scan, the point/inch (dpi) of take is unit;Ii. to leaf image I, utilize fixed threshold to carry out binaryzation, obtain bianry image B; Threshold value by experiment mode obtains;Iii. utilize a radius to be
Disc-shaped structure element B is carried out to mathematical morphology open operator, obtain bianry image B open;Iv. to B openutilize the hole fill method of existing standard to fill the hole in leaf area, then utilize the connected component labeling algorithm of existing standard to find out each 8-connected region in the bianry image of having filled hole, retain wherein area the maximum and, as leaf area, obtain thus the bianry image L of blade;V. utilize the skeleton K of the skeletonization method extraction blade L of existing standard full;Vi. in leaf image L, extract petiole region P and region, blade face F;Vii. utilize region, blade face F and intact leaf skeleton K fullobtain the skeleton K on blade face with operation; In K, find a skeleton point (being white point) p root=(x root, y root), K (x root, y root)=1, makes at p root8 neighborhood positions at least there is a white point in P; If there is the skeleton point in a plurality of such K, get first run into point as p root;Viii. the m-neighbours of calculating K count mapping graph N m, m-neighbours count mapping graph N mbe defined as
Ix. according to N mto blade face, skeleton K carries out mark, obtains skeleton signature M kand the adjacency matrix A of skeleton unit k;X. according to blade face skeleton K, carry out blade face regional restructuring, obtain region, reconstruct blade face bianry image F rc, and according to N mcarry out the expansion of blade face skeleton, so that the end points of skeleton all drops on the edge in region, blade face of reconstruct, the blade face skeleton signature of resulting expansion is designated as M ex;Xi. from M exin read p rootthe label l of point root, then according to adjacency matrix A k, from l rootcorresponding row starts, and according to the depth-first tree search algorithm of existing standard, searches for, exhaustive all in blade face expansion skeleton l rootcorresponding skeleton unit sets out, till the skeleton path of a certain other skeleton end points; The set in the exhaustive skeleton path obtaining is designated as C;Xii. according to homalographic criterion and maximum curvature criterion, from C, select a paths a med, this path is detects the blade face axis obtaining;Xiii. given error of fitting is limit ε fit, utilize the fractionation segmented fitting method of existing standard, obtain a meda matching broken line;Xiv. calculate the matching broken line length l in pixel, then, according to the physical resolution s of the image of image acquisition phase record, by following formula, obtaining actual blade face AL is page length l w:。
- 2. the leaf image acquisition method in i according to claim 1 step, is characterized in that following steps:A. blade is carried out certain smoothly and clean, then utilize scanner scanning leaf image;While b. scanning, adopt black not reflectorized material (as shaggy black paper) as scanning background, during scanning, should use the object with constant weight to push down blade to keep it smooth;If c. blade exists obvious branch (as maple leaf etc.), in scanning process, the blade-section of different branches must not overlap, in order to avoid people is the hole causing in leaf area, changes the shape of blade;D. the Image Saving that scanning obtains is the gray level image of 256 gray levels;E. record the resolution s that when scanning adopts, the point/inch (dpi) of take is unit, as 300dpi etc.
- 3. petiole region and the blade face method for extracting region in vi according to claim 1 step, is characterized in that following steps:A. utilize existing standard bianry image contour extraction method to extract the inverse image L of leaf image L rev8-be communicated with profile C l, rev;B. the digital disc-shaped structure element B that to utilize radius be r rpto C l, revcarry out the expansive working of mathematical morphology, obtain C l, erode; R can determine as follows: obtain C l, revin distance between 2 of all different white point centerings, then find one maximum in these distances, be made as d; Then getr=βdIn formula, β is a default constant, and general value can be between 0.01～0.05;C. obtain C l, erodeand L revexclusive disjunction image L erode, rev, then get L erode, revinverse image L erode, and extract L with existing standard bianry image contour extraction method erode8-be communicated with profile C erode;D. again utilize B rpto C erodecarry out the expansive working of mathematical morphology, then ask for result images and L erodeexclusive disjunction image L open;E. by L opencarry out XOR with L, obtain R 1, then utilize the connected component labeling algorithm of existing standard to find out R 1in each connected region, retain wherein area the maximum as petiole region P;F. P and L are carried out to XOR, obtain R 2, then utilize the connected component labeling algorithm of existing standard to find out R 2in each connected region, retain wherein area the maximum as region, blade face F.
- 4. the skeleton mark in ix according to claim 1 step and the acquisition methods of skeleton unit adjacency matrix, is characterized in that:A. utilize m-adjacency to define the connectedness between different skeleton points in skeleton K, then according to m-neighbours' mapping graph N mskeleton point is divided into dissimilar connected unit or skeleton unit: by N melement value equals the m-that 1 or 2 skeleton point forms and is communicated with the branch that skeleton is partly called skeleton, by N melement value is not less than the m-that 3 skeleton point forms and is communicated with the fork block that skeleton is partly called skeleton; Utilize the method for graph search that each skeleton unit in K is carried out to label, label is the continuous positive integer by 1 beginning;B. investigate m-adjacency between each skeleton unit and carry out the adjacency matrix A of Tectonic Framework unit k; M-adjacency between skeleton unit is defined as follows: if for two skeleton unit A and B, at least have some p in an A and the some q in a B, making p and q is m-adjacency, and these two skeleton units of A and B are m-adjacency so.
- 5. blade face regional restructuring and the skeleton extended method in x according to claim 1 step, is characterized in that:A. the skeleton end points that utilizes range conversion to find out not drop on blade face edges of regions, need to expand;B. the skeleton end points that need to expand each, finds its corresponding blade face edges of regions circular arc; If there are many such circular arcs, get wherein elder;C. find circular arc and the angular bisector at angle and the intersection point of circular arc that the skeleton end points that is positioned at circle centre position forms, with straight-line segment, skeleton end points and this intersection point are coupled together, completed skeleton expansion; The skeleton part of expansion belongs to a skeleton unit together with corresponding skeleton end points.
- 6. the axis detection method in xii according to claim 1 step, is characterized in that:A. to every skeleton path a, utilize a that blade face is split into at least two area parts; Find area maximum in these regions two, then obtain compared with small size and larger area ratio; This ratio is not less than the given homalographic threshold value t of certain user a, corresponding path a retains the candidate as axis, otherwise directly gets rid of; The criterion of this judgement axis is called homalographic criterion;B. for candidate's axis path a, find its terminal that drops on blade face edges of regions, and calculate the curvature of the blade face edges of regions of this destination county; In all candidate's axis path, finding this curvature is the maximum, is resulting blade face axis a med; The criterion of this judgement axis is called maximum curvature criterion.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310683599.9A CN103632148B (en) | 2013-12-13 | 2013-12-13 | Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310683599.9A CN103632148B (en) | 2013-12-13 | 2013-12-13 | Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103632148A true CN103632148A (en) | 2014-03-12 |
| CN103632148B CN103632148B (en) | 2017-02-08 |
Family
ID=50213177
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310683599.9A Expired - Fee Related CN103632148B (en) | 2013-12-13 | 2013-12-13 | Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103632148B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363784A (en) * | 2019-06-28 | 2019-10-22 | 青岛理工大学 | Identification method of overlapped fruits |
| CN114985283A (en) * | 2022-04-27 | 2022-09-02 | 韩山师范学院 | Tea leaf identification method based on multi-feature image identification technology |
-
2013
- 2013-12-13 CN CN201310683599.9A patent/CN103632148B/en not_active Expired - Fee Related
Non-Patent Citations (5)
| Title |
|---|
| LEE S ETC: "Curved glide-reflection symmetry detection", 《IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE》 * |
| 孙杰等: "目标识别中轴对称物体特征面积的定位及提取", 《光电子 激光》 * |
| 张金瑞: "基于图像处理的树叶面积特征提取的研究", 《中国优秀硕士学位论文全文数据库》 * |
| 郭斯羽: "一种用于植物叶片图像骨架提取的去毛刺方法", 《电子测量与仪器学报》 * |
| 魏敏: "序列图像轴对称物体中轴线提取方法", 《半导体光电》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110363784A (en) * | 2019-06-28 | 2019-10-22 | 青岛理工大学 | Identification method of overlapped fruits |
| CN114985283A (en) * | 2022-04-27 | 2022-09-02 | 韩山师范学院 | Tea leaf identification method based on multi-feature image identification technology |
| CN114985283B (en) * | 2022-04-27 | 2024-03-29 | 韩山师范学院 | Tea leaf identification method based on multi-feature image identification technology |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103632148B (en) | 2017-02-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wäldchen et al. | Plant species identification using computer vision techniques: a systematic literature review | |
| CN106846344B (en) | A kind of image segmentation optimal identification method based on the complete degree in edge | |
| CN105046252B (en) | A kind of RMB prefix code recognition methods | |
| CN103400151B (en) | The optical remote sensing image of integration and GIS autoregistration and Clean water withdraw method | |
| CN103077529B (en) | Based on the plant leaf blade characteristic analysis system of image scanning | |
| CN106600600A (en) | Wafer defect detection method based on characteristic matching | |
| CN104680161A (en) | Digit recognition method for identification cards | |
| CN105095822B (en) | A kind of Chinese letter co pattern image detection method and system | |
| CN101826204B (en) | Quick particle image segmentation method based on improved waterline algorithm | |
| CN101334896B (en) | Processing method for measuring sub-pixel rim of digital picture | |
| CN103279945A (en) | Quality map guide method and branch cutting method-based interferogram unwrapping method | |
| CN102750531B (en) | Method for detecting handwriting mark symbols for bill document positioning grids | |
| CN103995860B (en) | A kind of image indexing system and its method based on color feature extracted | |
| CN111007013B (en) | Crop rotation fallow remote sensing monitoring method and device for northeast cold region | |
| CN106813576A (en) | A kind of leaf area, girth, leaf measuring method long and leaf width | |
| CN106228136A (en) | Panorama streetscape method for secret protection based on converging channels feature | |
| CN105574265B (en) | Entire assembly model quantitative description towards model index | |
| CN116823811B (en) | Functional jacket surface quality detection method | |
| CN110211128A (en) | Loess plateau terraced fields extracting method based on remote sensing image and DEM | |
| CN116229265A (en) | Method for automatically and nondestructively extracting phenotype of soybean plants | |
| CN105956544A (en) | Remote sensing image road intersection extraction method based on structural index characteristic | |
| CN103632148A (en) | Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant | |
| CN107330886A (en) | A kind of high-precision quantization method of surface microlesion | |
| CN106324708B (en) | Digitizing solution, the device of rainfall record drawing | |
| Tamrakar et al. | No grouping left behind: From edges to curve fragments |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170208 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |