CN109684938A - It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies - Google Patents
It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies Download PDFInfo
- Publication number
- CN109684938A CN109684938A CN201811487764.2A CN201811487764A CN109684938A CN 109684938 A CN109684938 A CN 109684938A CN 201811487764 A CN201811487764 A CN 201811487764A CN 109684938 A CN109684938 A CN 109684938A
- Authority
- CN
- China
- Prior art keywords
- curve
- sugarcane
- endpoint
- white
- strain
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/188—Vegetation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/68—Food, e.g. fruit or vegetables
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
It is taken photo by plane the sugarcane strain number automatic identifying method of top view the present invention provides a kind of based on crop canopies, its field sugarcane canopy image for overlooking shooting under the conditions of natural light using unmanned plane carries out the removal of field complex background and the preliminary extraction of sugarcane Ye Baijing using digital image processing techniques as research object to it；The endpoint distribution map of white warp is obtained with the Bai Jinghou that root the is broken endpoint for extracting white warp by connecting, the last distribution character intensive according to strain heart district domain endpoint identifies the sugarcane strain heart using DBSCAN clustering algorithm from the endpoint figure of white warp；The average value approximation in all kinds of strain heart districts domain is calculated as strain heart coordinate and to count coordinate number be strain number, is achieved in field sugarcane strain number automatic identification.Field sugarcane is counted using this method, flexibility is high, and precision is good, and implementation cost is low, can effectively reduce and count mistakes and omissions caused by artificial subjective factor, and improve the automatic detection degree of sugarcane early stage plant quantity.
Description
Technical field
It is specifically a kind of to be taken photo by plane top view based on crop canopies the present invention relates to a kind of sugarcane strain number automatic identifying method
Sugarcane strain number automatic identifying method, belongs to Digital Image Processing, image segmentation and technical field of machine vision.
Background technique
Sugarcane is the important sugar crop in the world.Cultivation of sugar cane thinks that reasonable specification of planting can promote liquid manure photo-thermal
Etc. ecological factors reasonable distribution, to obtain higher yield and more preferably quality.Therefore, the sugarcane planting strain number in field is obtained
Measure the planting density of sugar cane breeds different for optimization, forecast production is of great significance.The major way that sugarcane counts at present
It is still based on and manually counts on the spot, i.e., carried out in such a way that artificial field estimates and counts or sample before harvesting sugarcane
Quantity estimation, this method of counting is time-consuming and laborious, inefficiency, and artificial field counts the method for counting pair of this intrusive mood
There is destructiveness in field conditions.Divide the number of evil different additionally, due to every plant of sugarcane and distance be compact between sugarcane strain and strain,
This method of counting is easy to produce visual fatigue, error-prone.
Currently, carrying out the quantity statistics of object based on image processing techniques, domestic and foreign scholars are to research in this respect
Have very much, but is essentially all the counting research for regular targets object, such as seed, apple, citrus, dragon fruit etc., it is such
Object has the characteristics that the shape and less overlapping of simple structure, rule.And it is this for sugarcane with different
Posture, the research for the method for counting for being overlapped and being difficult to the irregular object divided are then relatively fewer, and are mostly based on remote sensing
Image prediction sugarcane yield, implementation cost are high.Therefore need to develop a kind of convenience for sugarcane plant, lossless, efficient meter
Counting method provides technical support with the prediction for later period sugarcane quality and yield.
Summary of the invention
It is taken photo by plane the sugarcane strain number automatic identifying method of top view the object of the present invention is to provide a kind of based on crop canopies,
The canopy image that shooting sugarcane is just overlooked using unmanned plane high-altitude of taking photo by plane, is had using sugarcane influences of plant crown sugarcane leaf by strain heart Xiang Si
It is white through obvious feature among the trend and blade that week dissipates, it is extracted among sugarcane top clearly by the processing to image
Bai Jing, and the strain heart of the diverging central point approximate substitution sugarcane of white warp is found, it is achieved in the automatic identification of sugarcane strain number, from
And a kind of better choice is provided for the counting mode of field sugarcane.
The specific technical solution of the present invention is as follows:
It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies, comprising the following steps:
Step A: the canopy that field sugarcane crop is acquired in the form of unmanned plane overlooks color image, and to image
The positive vertical view part of middle sugarcane leaf carries out shear treatment, obtains target original image；
Step B: handling target original image, the binary map after obtaining sugarcane Ye Baijing primary segmentation；
Step C: ambient noise and impurity treatment are carried out to the binary map that step B is obtained；
Step D: deburring processing is carried out to the binary map handled well through step C；
Step E: it is white through being broken through reconnection operation to being broken in the binary map handled well through step D with root, and reject
Shorter white warp obtains more clean white extracted figure；
Step F: the endpoint of sugarcane Ye Baijing is extracted from the white extracted figure that step E is obtained, obtains endpoint extraction figure；
Step G: on the basis of endpoint extracts figure, according to strain heart district domain endpoint concentration, DBSCAN cluster is carried out simultaneously
The average value approximation in all kinds of strain heart districts domain is calculated as strain heart coordinate, counting coordinate number is strain number.
Further, the step B is specifically included:
Step B1: gray processing processing and morphological operation are successively carried out to target original image, respectively obtain initial gray image
With Background (no Bai Jing)；
Step B2: image subtraction operation is carried out to initial gray figure and Background (no Bai Jing), obtains grayscale image；
Step B3: the optimal segmenting threshold of grayscale image is found, after progress binary conversion treatment obtains sugarcane Ye Baijing primary segmentation
Binary map.
Further, the step C is specifically included:
Step C1: skeleton micronization processes, denoising, hole rejecting processing are successively carried out to binary map；
Step C2: Contour tracing processing is carried out to connected domain each in binary map, obtains the minimum circumscribed rectangle of connected domain；
Step C3: judging whether each connected domain is the white warp of sugarcane leaf, if it is not, then deleting the connected domain；If so, retaining
The connected domain.
Further, the step D is specifically included:
Step D1: to the binary map classification marker handled well through step C go out boundary point in image and endpoint, bifurcation point and
Crosspoint；
Step D2: checking each endpoint and boundary point, nearest in connection branch where counting endpoint or boundary point to its
The length in bifurcation or crosspoint, the as endpoint or the corresponding burr length of boundary point；
Step D3: the burr that burr length threshold is less than setting value len1 is deleted.
Further, the step E is specifically included:
Step E1: whether each connected domain, which is (white often one section of the region Bai Jing, is judged to the binary map handled well through step D
Smooth curve), if the region Bai Jing, then endpoint and region intermediate point of the dialogue through region are marked, then carry out primary song
Line process of fitting treatment；If it is not, not dealing with then；
Step E2: the endpoint for the white warp that need to be connected is screened；
Step E3: determination can junction curve uniqueness: every reference curve at most only have a curve can connect, to
The smallest endpoint of Euclidean distance between corresponding reference curve right endpoint is selected in the curve left end point of connection；
Step E4: broken in binary map through reconnection to the endpoint of the white warp of fracture extracted；
Step E5: the white warp that length threshold is less than setting value len2 is rejected, more clean white extracted figure is obtained.
Wherein, the specific steps of the step E1 are as follows:
Step E11: judge whether each connected domain is the region Bai Jing, if so, storing all coordinates of the connected domain, and is marked
Remember the left end point (X of white warp0L,Y0L) and right endpoint (X0R,Y0R)；If it is not, not dealing with then；
Step E12: the coordinate of the connected domain of each white warp is subjected to descending arrangement according to the coordinate in main trend direction, and is stored
The coordinate of the upper intermediate point of Bai Jing；Wherein, if | X0L-X0R|≥|Y0L-Y0R|, then the main trend direction of the white warp is line direction；It is no
It then, is column direction；
Step E13: a curve matching is carried out to the connected domain of each white warp.
Wherein, the specific steps of the step E2 are as follows:
Step E21: curve location judgement: curve to be detected is located at the lower right or upper right of reference curve, i.e.,
Curve to be detected is located on the extending direction of reference curve；
Step E22: two slope of a curve jack per lines to be compared are determined and fluctuation range is in given threshold k_vary:
kref, ktestA respectively curve matching slope of reference curve and curve to be detected meets:
Step E23: determine both reference curve and curve to be detected length and (pixel and) setting range p_min~
It is white through length violation with reality to prevent the curve of connection too short or too long in p_max；
Step E24: keeping the slope of the midpoint line of two curves approximate with the fit slope of two curves, if reference curve midpoint
The slope of line segment is k after connecting with mid point of curve to be detectedmid, then meet:
Step E25: make the slope of new line segment obtained by the line of reference curve right endpoint and curve left end point to be detected with
The fit slope jack per line of reference curve；
Step E26: if the Euclidean distance of reference curve right endpoint and curve left end point to be detected is less than given threshold d_
Min then stores the left end point of the curve to be detected to corresponding reference curve right endpoint to connection end point centering.
The present invention overlooks the field sugarcane canopy image of shooting using unmanned plane as research object under the conditions of natural light, uses
Digital image processing techniques carry out the removal of field complex background and the preliminary extraction of sugarcane Ye Baijing to it；It is same by connecting
The endpoint that the Bai Jinghou of root fracture extracts white warp obtains the endpoint distribution map of white warp, finally intensive according to strain heart district domain endpoint
Distribution character identifies the sugarcane strain heart using DBSCAN clustering algorithm from the endpoint figure of white warp；Calculate the flat of all kinds of strain heart districts domain
Mean approximation is as strain heart coordinate and to count coordinate number be strain number, is achieved in field sugarcane strain number automatic identification.This
It is convenient, lossless, efficient that method has the characteristics that, is counted using this method to field sugarcane, and flexibility is high, and precision is good, real
It applies at low cost, can effectively reduce and count mistakes and omissions caused by artificial subjective factor, and improve the automation of sugarcane early stage plant quantity
Detection level.
Detailed description of the invention
Fig. 1 is the flow chart of sugarcane strain number automatic identifying method of the present invention.
Fig. 2 is the schematic diagram in the non-white minimum circumscribed rectangle region through region.
Fig. 3 is the disconnected schematic diagram through reconnection.
Fig. 4 is the schematic diagram for being broken white warp.
Fig. 5 is the effect picture being broken after Bai Jingjing algorithm reconnection operation.
Fig. 6 is the target original image of sugarcane canopy after cropped.
Fig. 7 is the sugarcane strain strain heart position detected through this method and the white extracted superimposed effect picture of figure.
Fig. 8 is the sugarcane strain strain heart position detected through this method and the superimposed effect picture of target original image.
Specific embodiment
Present invention is further described in detail with reference to the accompanying drawings and examples.
As shown in Figure 1, sugarcane strain number automatic identifying method of the invention, comprising the following steps:
Step A: the canopy that field sugarcane crop is acquired in the form of unmanned plane overlooks color image, and to image
The positive vertical view part of middle sugarcane leaf carries out shear treatment, obtains target original image.
The camera that the embodiment of the present invention is carried using big boundary spirit phantom 2+vision unmanned plane is in 10 meters of a high-altitude left side
The right side is acquired the positive overhead view image of field sugarcane canopy, and original image is the color image of 4384*3288 pixel, due to
The limitation of the height and camera itself shooting angle of tillering stage sugarcane itself leads to there was only middle part in original image
Sugarcane is positive overhead view image, is carried out the positive vertical view part in original image at shearing using the clipping function of Photoshop
Reason obtains the target original image (color image) of sugarcane canopy, and image size is 371*451 pixel, as shown in Figure 6.
Step B: handling target original image, the binary map after obtaining sugarcane Ye Baijing primary segmentation.This step is specifically wrapped
It includes:
Step B1: gray processing processing and morphological operation are successively carried out to target original image, respectively obtain initial gray image
With Background (no Bai Jing).
Gray processing is carried out using weighted average to R, G, the B component of target original image (color image) to handle to obtain initial gray
Image, wherein the weight component of R, G, B are respectively 0.2989,0.587,0.114.According to the shape size of sugarcane leaf, half is chosen
The planar circular structure that diameter is 2 carries out morphology to initial gray image and opens operation, and obtained image is as Background (without white
Through).
Step B2: image subtraction behaviour is carried out to initial gray figure and Background (no Bai Jing) using imsubtract function
Make, obtains grayscale image；The purpose of image subtraction operation is to weaken the influence of complex background as far as possible.
Step B3: the optimal segmenting threshold of grayscale image is found, after progress binary conversion treatment obtains sugarcane Ye Baijing primary segmentation
Binary map.
The optimal segmenting threshold t for finding grayscale image in the present embodiment using maximum variance between clusters, according to formula:
Wherein t is the segmentation threshold of foreground and background, and it is w that prospect points, which account for image scaled,0, average gray u0；Background dot
It is w that number, which accounts for image scaled,1, average gray u1.The average gray of image is u, and objective function g indicates foreground and background image
Variance.When g maximum, gray scale t at this time is optimal segmenting threshold.
Step C: ambient noise and impurity treatment are carried out to the binary map that step B is obtained.This step specifically includes:
Step C1: skeleton micronization processes, denoising, hole rejecting processing are successively carried out to binary map.
Binary map is refined using bwmorph function, the pixel number of each connected domain in the binary map after statistics refinement
And Euler's numbers, connected domain of the pixel number less than 15 are defined as noise, deleting these noises (i.e. will be in corresponding connected domain
0) pixel value is set；The figure number (constant 1) that the Euler's numbers of each connected domain are equal to the connected domain subtracts the hole in the connected domain
Number, connected domain of the Euler's numbers less than 1 are defined as background impurities (region comprising hole), delete these background impurities and (i.e. will
0) pixel value in corresponding connected domain is set.
Step C2: profile is carried out to connected domain each in binary map using the boundingbox attribute of regionprops function
Tracking process obtains the minimum circumscribed rectangle of connected domain.
Fig. 2 is the schematic diagram in the non-white minimum circumscribed rectangle region through region, the minimum circumscribed rectangle area row
Width on direction is denoted as length, and the length on column direction is denoted as width.
Step C3: judging whether each connected domain is the white warp of sugarcane leaf, if it is not, then deleting the connected domain；If so, retaining
The connected domain.
Empirically judge, the non-white actual pixels through in region and the length and width much larger than its minimum circumscribed rectangle,
It is non-white to meet Rule of judgment through region such as Fig. 2:
Wherein, max (length, width) indicates the maximum value of both length and width, PsumIndicate non-white through region
Actual pixels and.
Step D: deburring processing is carried out to the binary map handled well through step C.This step specifically includes:
Step D1: to the binary map classification marker handled well through step C go out boundary point in image and endpoint, bifurcation point and
Crosspoint.
The 0-1 pattern count for calculating each pixel in image to the binary map that step C is handled well (is pressed in 8 neighborhoods of certain pixel
According to prescribed direction (for example being set as clock-wise fashion) run one week its adjacent pixel pixel value from 0 become 1 number), distinguish
Storage 0-1 pattern count is the coordinate of 1 (boundary point, endpoint), 3 (bifurcation points) and 4 (crosspoints).
Step D2: checking each endpoint and boundary point, nearest in connection branch where counting endpoint or boundary point to its
The length in bifurcation or crosspoint, the as endpoint or the corresponding burr length of boundary point.
A label matrix J (identical as original image matrix) is created, the pixel value in bifurcation and crosspoint is all marked as 3
(being considered as same type of point).The point (boundary point and endpoint) that pixel value is 1 in traversal label matrix J, when central point meets 8
Pixel value is not equal to 3 and when at least one pixel value is 1 in neighborhood, then marks the pixel value of the point to be in label matrix J
2 (preventing from being accessed again) record the coordinate, and length counter counter adds 1, then checks pixel value with same method
For 8 neighborhoods of 1 point, until 8 neighborhoods of central point are unsatisfactory for condition, counter, which is counted, to be stopped, at this point, counter
Numerical value be the endpoint or the corresponding burr length of boundary point.
Step D3: the burr that burr length threshold is less than setting value len1 is deleted.
Through testing, len1=15 in the present embodiment.Length counter counter is reset, step D2 is executed, until marking square
All endpoint and boundary point are traversed in battle array J, and algorithm terminates.In the present embodiment, deburring processing is repeated 2 times, is
In order to be eliminated as much as burr completely.
Step E: it is white through being broken through reconnection operation to being broken in the binary map handled well through step D with root, and reject
Shorter white warp obtains more clean white extracted figure.This step specifically includes:
Step E1: whether each connected domain, which is (white often one section of the region Bai Jing, is judged to the binary map handled well through step D
Smooth curve), if the region Bai Jing, then endpoint and region intermediate point of the dialogue through region are marked, then carry out primary song
Line process of fitting treatment；If it is not, not dealing with then.
The specific steps of step E1 are as follows:
Step E11: judge whether each connected domain is the region Bai Jing, if so, storing all coordinates of the connected domain, and is marked
Remember the left end point (X of white warp0L,Y0L) and right endpoint (X0R,Y0R)；If it is not, not dealing with then.
Judge the number of endpoint of connected domain, if number of endpoint is 2, which is white warp, stores all seats of the connected domain
Mark, and mark the left end point (X of white warp0L,Y0L) and right endpoint (X0R,Y0R), wherein X0R>X0L>0；Otherwise, it does not deal with.
Step E12: the coordinate of each connected domain of white warp is subjected to descending arrangement according to the coordinate in main trend direction, and is stored
The coordinate of the upper intermediate point of Bai Jing；Wherein, if | X0L-X0R|≥|Y0L-Y0R|, then the main trend direction of the white warp is line direction；It is no
It then, is column direction.
Step E13: a curve matching is carried out to the connected domain of each white warp.
Extract it is white be one section of gentle curve, its approximation is regarded as straight line section, therefore uses least square method
A curve matching is carried out to it.In the present embodiment, the purpose of a curve matching is exactly to seek and give one group of point
{(xi,yi) square distance and a smallest curve y=k of (i=1,2 ..., m)px+bp。
Step E2: the endpoint for the white warp that need to be connected is screened.
Fig. 3 is the disconnected schematic diagram through reconnection, it is assumed that Bai Jing EA is reference curve lEA, fit slope kref, left end point E
(X0L,Y0L), right endpoint A (X0R,Y0R), C (X0M,Y0M) it is the white intermediate point through EA；White through BF is curve l to be detectedBF, fitting
Slope is ktest, left end point B (XiL,YiL), right endpoint F (XiR,YiR), D (XiM,YiM) it is the white intermediate point through BF.
The specific steps of step E2 are as follows:
Step E21: curve location judgement: curve to be detected is located at the lower right or upper right of reference curve, i.e.,
Curve to be detected is located on the extending direction of reference curve.
Such as the reference curve l of Fig. 3EAWith curve l to be detectedBFExtreme coordinates relationship meet judgement formula:
XiL≥X0RAnd YiL≥max(Y0R,Y0L) or XiL≥X0RAnd YiL≤min(Y0R,Y0L)
Curve l i.e. to be detectedBFPositioned at reference curve lEAExtending direction on.
Step E22: two slope of a curve jack per lines to be compared are determined and fluctuation range is in given threshold k_vary:
The value of k_vary is 0.42, k in the present embodimentref, ktestRespectively reference curve lEAWith curve l to be detectedBF's
Curve matching slope, then meet:
Step E23: determine both reference curve and curve to be detected length and (pixel and) setting range p_min~
It is white through length violation with reality to prevent the curve of connection too short or too long in p_max.
In the present embodiment, p_min=38, p_max=162.
Step E24: keeping the slope of the midpoint line of two curves approximate with the fit slope of two curves, if reference curve midpoint
The slope of line segment is k after connecting with mid point of curve to be detectedmid, then meet:
Such as Fig. 3, kCDThe slope of line segment, k after being connect for reference curve midpoint with mid point of curve to be detectedmid=kCD, through trying
It tests, λmin=0.9, λmax=1.1, then meet:
Step E25: make the slope of new line segment obtained by the line of reference curve right endpoint and curve left end point to be detected with
The fit slope jack per line of reference curve.
Such as Fig. 3, if new line segment AB, slope k that point A is connected with point BAB, then have kAB×kref>0。
Step E26: if the Euclidean distance of reference curve right endpoint and curve left end point to be detected is less than given threshold d_
Min then stores the left end point of the curve to be detected to corresponding reference curve right endpoint to connection end point centering.
In the present embodiment, through testing, the Euclidean distance threshold value d_min=50 of new line segment AB.
Step E3: determination can junction curve uniqueness: every reference curve at most only have a curve can connect, to
The smallest endpoint of Euclidean distance between corresponding reference curve right endpoint is selected in the curve left end point of connection.
In the present embodiment, the right endpoint of certain reference curve is corresponding sometimes has several curves to be detected to be connected (i.e.
Have the left end point of several curves to be detected), terminal B when not meeting actual conditions, therefore line segment AB Euclidean distance minimum need to be selected
Uniquely coupled endpoint as the reference curve.
Step E4: broken in binary map through reconnection to the endpoint of the white warp of fracture extracted.
In the present embodiment, through above-mentioned steps, the endpoint pair through reconnection that need to break is obtained, using an interpolation method in binary map
On to these endpoints to being attached.Fig. 4 is the schematic diagram of the white warp of fracture, and Fig. 5 is the fracture Bai Jingjing algorithm reconnection
Effect picture after operation.
Step E5: the white warp that length threshold is less than setting value len2 is rejected, more clean white extracted figure is obtained.
In the present embodiment, through testing, len2=33 is deleted in pixel and Bai Jingcong binary map no more than 33.
Step F: the endpoint of sugarcane Ye Baijing is extracted from the white extracted figure that step E is obtained, obtains endpoint extraction figure.
It to the white extracted binary map that step E is obtained, traverses all pixels value and is 1 pixel, and calculate the picture of its 8 neighborhood
Element and, only when pixel and 1 indicate current point be endpoint and mark, retain binary map in all endpoints, delete rest of pixels value
Figure is extracted for 1 pixel to get to endpoint.
Step G: on the basis of endpoint extracts figure, according to strain heart district domain endpoint concentration, DBSCAN cluster is carried out simultaneously
The average value approximation in all kinds of strain heart districts domain is calculated as strain heart coordinate, counting coordinate number is strain number.
The algorithmic procedure of DBSCAN cluster is as follows:
Step1: 3 input parameters of DBSCAN cluster are determined: given radius (Eps), minimum point of destination (Minpts)
And data set Data.
In the present embodiment, Eps=30, Minpts=3, Eps indicate the maximum distance in classification between endpoint, and Minpts's takes
Value depends on the item number of every plant of white stem of sugarcane leaf, the sugarcane in tillering stage according to sugarcane leaf number generally take 3 or 4.
Step2: traversing whether all input points are kernel object, if so, finding out the institute in the Eps neighborhood of kernel object
There is direct density accessible point.
Step3: traversing the Eps neighborhood of all kernel objects, reachable according to all direct density in kernel object E neighborhood
Point merges density achievable pair as finding the connected object set of maximal density.
DBSCAN cluster the advantages of be can filter density regions by dense region division be cluster, it can find arbitrary shape
The cluster of shape and the effectively interference of shielding noise spot.The foundation of cluster is big outside the density ratio classification put in classification, the pass realized
Key point is whether the point for judging that every bit includes in the range of given radius (Eps) in classification is not less than minimum point of destination
(Minpts).Fig. 7 is the sugarcane strain strain heart position that detects through this method and the white extracted superimposed effect picture of figure, Fig. 8 be through
The sugarcane strain strain heart position and the superimposed effect picture of target original image that this method detects.
In order to verify the effect of the method for the present invention, We conducted verification tests, in an experiment, have chosen sugarcane strain number not
Deng image 10 open, the statistics of canopy sugarcane strain number is carried out using the method for the present invention, statistical result is as shown in table 1.It is sent out in test
Existing, the method for the present invention is best for the strain heart recognition effect for just overlooking the sugarcane canopy of shooting, in such photo the characteristics of sugarcane strain
It is that the white stem of sugarcane leaf is clear, quantity is enough and sugarcane leaf is dispersed around around the strain heart, counting accuracy can reach 91.5%.
The experimental result that 1 sugarcane canopy strain number of table counts
Above-mentioned legend is only exemplary embodiments of the invention, is not intended to restrict the invention, it is all in spirit of the invention and
Made any modification or equivalent replacement and improvement etc., should all be included in the protection scope of the present invention within principle.
Claims (7)
1. a kind of taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies, which is characterized in that including following step
It is rapid:
Step A: the canopy that field sugarcane crop is acquired in the form of unmanned plane overlooks color image, and to sugarcane in image
The positive vertical view part of leaf carries out shear treatment, obtains target original image；
Step B: handling target original image, the binary map after obtaining sugarcane Ye Baijing primary segmentation；
Step C: ambient noise and impurity treatment are carried out to the binary map that step B is obtained；
Step D: deburring processing is carried out to the binary map handled well through step C；
Step E: broken with the white of root fracture through reconnection operation in the binary map handled well through step D, and rejected shorter
White warp, obtain more clean white extracted figure；
Step F: the endpoint of sugarcane Ye Baijing is extracted from the white extracted figure that step E is obtained, obtains endpoint extraction figure；
Step G: it on the basis of endpoint extracts figure, according to strain heart district domain endpoint concentration, carries out DBSCAN cluster and calculates
The average value approximation in all kinds of strain heart districts domain is as strain heart coordinate and to count coordinate number be strain number.
2. sugarcane strain number automatic identifying method according to claim 1, which is characterized in that the step B is specifically included:
Step B1: gray processing processing and morphological operation are successively carried out to target original image, respectively obtain initial gray image and back
Scape figure (no Bai Jing)；
Step B2: image subtraction operation is carried out to initial gray figure and Background (no Bai Jing), obtains grayscale image；
Step B3: finding the optimal segmenting threshold of grayscale image, carries out binary conversion treatment and obtains two after sugarcane Ye Baijing primary segmentation
Value figure.
3. sugarcane strain number automatic identifying method according to claim 1, which is characterized in that the step C is specifically included:
Step C1: skeleton micronization processes, denoising, hole rejecting processing are successively carried out to binary map；
Step C2: Contour tracing processing is carried out to connected domain each in binary map, obtains the minimum circumscribed rectangle of connected domain；
Step C3: judging whether each connected domain is the white warp of sugarcane leaf, if it is not, then deleting the connected domain；If so, retaining the company
Logical domain.
4. sugarcane strain number automatic identifying method according to claim 1, which is characterized in that the step D is specifically included:
Step D1: go out boundary point and endpoint, the bifurcation point and intersection in image to the binary map classification marker handled well through step C
Point；
Step D2: checking each endpoint and boundary point, bifurcated nearest in connection branch where counting endpoint or boundary point to its
The length in point or crosspoint, the as corresponding burr length of the endpoint or boundary point；
Step D3: the burr that burr length threshold is less than setting value len1 is deleted.
5. sugarcane strain number automatic identifying method according to claim 1, which is characterized in that the step E is specifically included:
Step E1: whether each connected domain, which is that (white often one section smooth in the region Bai Jing, is judged to the binary map handled well through step D
Curve), if the region Bai Jing, then endpoint and region intermediate point of the dialogue through region are marked, then to carry out a curve quasi-
Conjunction processing；If it is not, not dealing with then；
Step E2: the endpoint for the white warp that need to be connected is screened；
Step E3: determination can junction curve uniqueness: every reference curve at most only have a curve can connect, from wait connect
Curve left end point in select the smallest endpoint of Euclidean distance between corresponding reference curve right endpoint；
Step E4: broken in binary map through reconnection to the endpoint of the white warp of fracture extracted；
Step E5: the white warp that length threshold is less than setting value len2 is rejected, more clean white extracted figure is obtained.
6. sugarcane strain number automatic identifying method according to claim 5, which is characterized in that the specific steps of the step E1
Are as follows:
Step E11: judge whether each connected domain is the region Bai Jing, if so, storing all coordinates of the connected domain, and is marked white
Left end point (the X of warp0L,Y0L) and right endpoint (X0R,Y0R)；If it is not, not dealing with then；
Step E12: the coordinate of the connected domain of each white warp is subjected to descending arrangement according to the coordinate in main trend direction, and stores white warp
The coordinate of upper intermediate point；Wherein, if | X0L-X0R|≥|Y0L-Y0R|, then the main trend direction of the white warp is line direction；Otherwise, it is
Column direction；
Step E13: a curve matching is carried out to the connected domain of each white warp.
7. sugarcane strain number automatic identifying method according to claim 5, which is characterized in that the specific steps of the step E2
Are as follows:
Step E21: curve location judgement: curve to be detected is located at the lower right or upper right of reference curve, i.e., to be checked
Curve is surveyed to be located on the extending direction of reference curve；
Step E22: two slope of a curve jack per lines to be compared are determined and fluctuation range is in given threshold k_vary:
kref, ktestA respectively curve matching slope of reference curve and curve to be detected meets:
kref×ktest> 0 and
Step E23: the length and (pixel and) for determining both reference curve and curve to be detected are in setting range p_min~p_
It is white through length violation with reality to prevent the curve of connection too short or too long in max；
Step E24: keeping the slope of the midpoint line of two curves approximate with the fit slope of two curves, if reference curve midpoint with to
The slope of line segment is k after the connection of detection curve midpointmid, then meet:
Step E25: make the slope of new line segment and reference obtained by the line of reference curve right endpoint and curve left end point to be detected
The fit slope jack per line of curve；
Step E26: if the Euclidean distance of reference curve right endpoint and curve left end point to be detected is less than given threshold d_min,
By the left end point storage of the curve to be detected to corresponding reference curve right endpoint to connection end point centering.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811487764.2A CN109684938A (en) | 2018-12-06 | 2018-12-06 | It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811487764.2A CN109684938A (en) | 2018-12-06 | 2018-12-06 | It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109684938A true CN109684938A (en) | 2019-04-26 |
Family
ID=66187312
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811487764.2A Pending CN109684938A (en) | 2018-12-06 | 2018-12-06 | It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109684938A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298366A (en) * | 2019-07-05 | 2019-10-01 | 北华航天工业学院 | Crops are distributed extracting method and device |
| CN110610438A (en) * | 2019-09-16 | 2019-12-24 | 黑龙江八一农垦大学 | Crop canopy petiole included angle calculation method and system |
| CN110660075A (en) * | 2019-09-20 | 2020-01-07 | 中国计量大学 | Method for row segmentation of soybean crops adhered to aerial images of unmanned aerial vehicle |
| CN110827393A (en) * | 2019-07-17 | 2020-02-21 | 当家移动绿色互联网技术集团有限公司 | Image processing method, image processing device, storage medium and electronic equipment |
| CN111724354A (en) * | 2020-06-02 | 2020-09-29 | 浙江托普云农科技股份有限公司 | Image processing-based method for measuring spike length and small spike number of multiple wheat |
| CN112270708A (en) * | 2020-10-26 | 2021-01-26 | 中国计量大学 | Vegetable and fruit plant side branch point identification method based on different edge type intersection points |
| CN113421301A (en) * | 2021-07-08 | 2021-09-21 | 浙江大学 | Method and system for positioning central area of field crop |
| CN116147508A (en) * | 2022-10-18 | 2023-05-23 | 广西科技大学 | Visual measurement method for sugarcane height of sugarcane harvester |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101750051A (en) * | 2010-01-04 | 2010-06-23 | 中国农业大学 | Visual navigation based multi-crop row detection method |
| CN104899553A (en) * | 2015-04-30 | 2015-09-09 | 浙江理工大学 | Field crop row extraction method capable of resisting dense weed interference |
| CN105957115A (en) * | 2016-05-18 | 2016-09-21 | 河北工程大学 | Crop planting structure remote sensing extraction method under generalized DEM idea |
| CN106910197A (en) * | 2017-01-13 | 2017-06-30 | 广州中医药大学 | A kind of dividing method of the complex background leaf image in single goal region |
| CN108369635A (en) * | 2015-11-08 | 2018-08-03 | 阿格洛英公司 | The method with analysis is obtained for aerial image |
| CN108596165A (en) * | 2018-08-21 | 2018-09-28 | 湖南鲲鹏智汇无人机技术有限公司 | Road traffic marking detection method based on unmanned plane low latitude Aerial Images and system |
-
2018
- 2018-12-06 CN CN201811487764.2A patent/CN109684938A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101750051A (en) * | 2010-01-04 | 2010-06-23 | 中国农业大学 | Visual navigation based multi-crop row detection method |
| CN104899553A (en) * | 2015-04-30 | 2015-09-09 | 浙江理工大学 | Field crop row extraction method capable of resisting dense weed interference |
| CN108369635A (en) * | 2015-11-08 | 2018-08-03 | 阿格洛英公司 | The method with analysis is obtained for aerial image |
| CN105957115A (en) * | 2016-05-18 | 2016-09-21 | 河北工程大学 | Crop planting structure remote sensing extraction method under generalized DEM idea |
| CN106910197A (en) * | 2017-01-13 | 2017-06-30 | 广州中医药大学 | A kind of dividing method of the complex background leaf image in single goal region |
| CN108596165A (en) * | 2018-08-21 | 2018-09-28 | 湖南鲲鹏智汇无人机技术有限公司 | Road traffic marking detection method based on unmanned plane low latitude Aerial Images and system |
Non-Patent Citations (3)
| Title |
|---|
| LIJIA WANG 等: "Canopy Center and Plant Quantity Detection of Sugarcane Based on Aerial Visible Images", 《AMERICAN SOCIETY OF AGRICULTURAL AND BIOLOGICAL ENGINEERS ANNUAL INTERNATIONAL MEETING 2018》 * |
| 贾洪雷 等: "基于机器视觉的玉米植株数量获取方法与试验", 《农业工程学报》 * |
| 郭显久 等: "基于图像处理技术的海洋微藻数量统计方法", 《大连海洋大学学报》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110298366A (en) * | 2019-07-05 | 2019-10-01 | 北华航天工业学院 | Crops are distributed extracting method and device |
| CN110827393A (en) * | 2019-07-17 | 2020-02-21 | 当家移动绿色互联网技术集团有限公司 | Image processing method, image processing device, storage medium and electronic equipment |
| CN110610438A (en) * | 2019-09-16 | 2019-12-24 | 黑龙江八一农垦大学 | Crop canopy petiole included angle calculation method and system |
| CN110610438B (en) * | 2019-09-16 | 2022-02-15 | 黑龙江八一农垦大学 | Crop canopy petiole included angle calculation method and system |
| CN110660075A (en) * | 2019-09-20 | 2020-01-07 | 中国计量大学 | Method for row segmentation of soybean crops adhered to aerial images of unmanned aerial vehicle |
| CN110660075B (en) * | 2019-09-20 | 2023-03-24 | 中国计量大学 | Method for row segmentation of soybean crops adhered to aerial images of unmanned aerial vehicle |
| CN111724354B (en) * | 2020-06-02 | 2023-05-30 | 浙江托普云农科技股份有限公司 | Image processing-based method for measuring wheat ear length and wheat ear number of multiple wheat plants |
| CN111724354A (en) * | 2020-06-02 | 2020-09-29 | 浙江托普云农科技股份有限公司 | Image processing-based method for measuring spike length and small spike number of multiple wheat |
| CN112270708A (en) * | 2020-10-26 | 2021-01-26 | 中国计量大学 | Vegetable and fruit plant side branch point identification method based on different edge type intersection points |
| CN112270708B (en) * | 2020-10-26 | 2024-02-02 | 中国计量大学 | Vegetable and fruit plant lateral branch point identification method based on intersection points of different edge types |
| CN113421301A (en) * | 2021-07-08 | 2021-09-21 | 浙江大学 | Method and system for positioning central area of field crop |
| CN116147508B (en) * | 2022-10-18 | 2023-09-19 | 广西科技大学 | Visual measurement method for sugarcane height of sugarcane harvester |
| CN116147508A (en) * | 2022-10-18 | 2023-05-23 | 广西科技大学 | Visual measurement method for sugarcane height of sugarcane harvester |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109684938A (en) | It is a kind of to be taken photo by plane the sugarcane strain number automatic identifying method of top view based on crop canopies | |
| CN107316289B (en) | Method for dividing rice ears in field based on deep learning and superpixel division | |
| CN109978822B (en) | Banana maturity judging modeling method and judging method based on machine vision | |
| CN112101159B (en) | Multi-temporal forest remote sensing image change monitoring method | |
| CN109447945B (en) | Quick counting method for basic wheat seedlings based on machine vision and graphic processing | |
| CN110120042B (en) | Crop image pest and disease damage area extraction method based on SLIC super-pixel and automatic threshold segmentation | |
| CN109409205A (en) | Video road driveway line detecting method of taking photo by plane based on line pitch characteristics point cluster | |
| CN109886277B (en) | Contour analysis-based fresh tea leaf identification method | |
| CN103077529A (en) | System for analyzing plant leaf features based on image scanning | |
| CN104657706B (en) | The fracture of high ferro circuit line bar and draw bail body method for detecting abnormality based on image | |
| CN108052886A (en) | A kind of puccinia striiformis uredospore programming count method of counting | |
| CN105868717B (en) | A kind of high-resolution remote sensing image list wood crown information extraction method based on classification | |
| CN108230307B (en) | Corn broken grain detection method based on contour centroid distance and neural network | |
| CN114067114B (en) | Area nuclear segmentation counting method based on algae cell morphology | |
| AU2020103260A4 (en) | Rice blast grading system and method | |
| CN106023199B (en) | A kind of flue gas blackness intelligent detecting method based on image analysis technology | |
| CN111523511B (en) | Video image Chinese wolfberry branch detection method for Chinese wolfberry harvesting and clamping device | |
| CN110059713A (en) | Precipitation type identification method based on precipitation particle multi-feature parameters | |
| CN103177266A (en) | Intelligent stock pest identification system | |
| CN107680098A (en) | A kind of recognition methods of sugarcane sugarcane section feature | |
| CN109871900A (en) | The recognition positioning method of apple under a kind of complex background based on image procossing | |
| CN111582237A (en) | High-resolution image airplane type identification method based on ATSM model | |
| CN112347894A (en) | Single-plant vegetation extraction method based on transfer learning and Gaussian mixture model separation | |
| CN105260559B (en) | A kind of paper pulp fiber morphological parameters calculation method based on contour area and contour thinning | |
| CN110866547B (en) | Automatic classification system and method for traditional Chinese medicine decoction pieces based on multiple features and random forests |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190426 |
|
| RJ01 | Rejection of invention patent application after publication |