CN107067430A - A kind of wheatland crop row detection method of distinguished point based cluster - Google Patents
A kind of wheatland crop row detection method of distinguished point based cluster Download PDFInfo
- Publication number
- CN107067430A CN107067430A CN201710238165.6A CN201710238165A CN107067430A CN 107067430 A CN107067430 A CN 107067430A CN 201710238165 A CN201710238165 A CN 201710238165A CN 107067430 A CN107067430 A CN 107067430A
- Authority
- CN
- China
- Prior art keywords
- point
- crop
- crop row
- characteristic point
- straight line
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/60—Analysis of geometric attributes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30172—Centreline of tubular or elongated structure
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Geometry (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The present invention carries out feature point extraction using 2G R B chromatic imagies gray processing, Otsu image binaryzations, left and right edges center line detecting method to image, can obtain crop row information, the feature dot image of relatively fewer data point is had again；And new clustering method is proposed, characteristic point is clustered using the distance feature of characteristic point, can accurately obtain representing the characteristic point of each crop row；Fitting a straight line is carried out with least square method to the characteristic point of each class.This method can be carried out effectively crop row and extracted, disclosure satisfy that the demand of farm working machinery real-time navigation on the image under the complicated farm environment such as crop missing and weeds influence.
Description
Technical field
The present invention relates to a kind of wheatland crop row detection method, specially a kind of wheatland crop row of distinguished point based cluster
Detection method, belongs to data handling utility field.
Background technology
As an important component of precision agriculture, the vision guided navigation technology of farm working machinery increasingly enjoys pass
Note, and be widely used in terms of proportion of crop planting, fertilising, tillage and weeding.Detection crop row center line is to carry out vision to lead
The basis of boat.Conventional crop row detection algorithm has Hough transform method and least square method at present.Hough transform algorithm is made an uproar
Sound shadow rings small, strong robustness, has the disadvantage that the peakvalue's checking of algorithm is difficult, time complexity and space complexity are larger.In order to subtract
Lack amount of calculation, Xu etc. proposes randomized hough transform, reduces amount of calculation using more to one mapping method, is dropped with dynamic link table
Low internal memory.Though the innovatory algorithm can reduce internal memory to a certain extent, the speed of service is improved, for the crop map with weeds
Picture, crop row accuracy of detection is not still high；Least square method when for extracting leading line, can quick detection go out crop row,
But the shortcoming of this method is susceptible to picture noise interference, when image includes multiple crop rows, it is impossible to directly using minimum
Square law carries out straight-line detection.
The content of the invention
In order to solve the problem above of prior art presence, the present invention proposes a kind of wheatland crop of distinguished point based cluster
Row detection method, is mainly included the following steps that：
(1) wheat crop image is shot with color camera；
(2) wheat crop bianry image is obtained；
(3) obtain representing the candidate feature point of wheat crop row with left and right edges center line detecting method；
(4) constrained procedure for being respectively less than a certain threshold value to the distance of the straight line with the characteristic point being distributed in around straight line enters
Row feature points clustering, obtains validity feature point；
(5) fitting a straight line is carried out to the validity feature point in each class with least square method, so as to detect each crop
OK.
Specifically, in step (4),
1) wheat crop bianry image from top to bottom, is from left to right scanned, the characteristic point (x that all pixels value is 1 is foundi,
yi), there are n characteristic point, and i=1,2,3 ..., n in data space V；
2) distance threshold dbound, wheatland crop line number num are initialized.
3) for any combination (km,bl), line correspondence y=kmx+bl, calculate all characteristic point (x in Vi,yi) arrive straight line y
=kmx+blApart from d,Wherein 1≤m≤max_m, kmStep-length be Sk, 1≤l≤max_l, blStep
A length of Sb, and as m=1, blFrom b1Traverse bmax_l, and bmax_l=bmax_l-1+Sb, kmax_m=kmax_m-1+Sk, until kmFrom k1
Traverse kmax_m；
If being less than threshold value dbound apart from d, this feature point is considered as belonging to straight line l validity feature point.If one tires out
Plus device records combination (k all the timem,bl) corresponding validity feature point number maximum, while recording the corresponding k of the maximummWith
blValue；
4) step 3) it is finished, the corresponding k of the maximum in accumulatormAnd blValue, find straight line y=kmx
+blDistance is less than dbound all validity feature points, and stores it in one-dimension array, the validity feature of the first crop row
Point cluster is finished；
5) characteristic point of the participation cluster in deleting 4), updates the data collection V, and circulation performs step 3) -4), until the n-th um
Capable feature points clustering is finished.
More specifically, in above-mentioned steps, dbound is the half of wheatland width.
Because this method according to distance restraint selects a several most straight lines of covering point, as long as therefore choosing suitable distance
Threshold value d, the cluster results of these characteristic points is based on can represent the trend of crop row well.After the present invention, amount of calculation subtracts
It is few, overcome the problem of peakvalue's checking is difficult, and calculating speed faster, more accurate, more to the noise spot image tool of fitting effect
There is very strong robustness.
Brief description of the drawings
Fig. 1 is growth image when wheat is in the jointing stage.
Fig. 2 is Fig. 1 pre-processed results.
Fig. 3 is characterized a cluster result.
Embodiment
1. image is obtained.During picture collection, shot using Samsung S750 color cameras, camera distance
Ground level is 1.1 meters, and camera optical axis and horizontal line angle are 30 degree or 80 degree, image be size be 640pixel ×
480pixel or 480pixel × 640pixel coloured image.Experiment allocation of computer used is CPU frequency 2.60GHz,
Inside save as 1.88GB.Software used in image procossing is Matlab R2009a.
2. image preprocessing
2.1 coloured image gray processings
The present invention is that 2G-R-B characterization factors are split using super green method.To Fig. 1 gray processing result such as Fig. 2 a.
2.2 image binaryzation
In order to which Crop Information is separated from soil, background, binary conversion treatment is carried out to Fig. 2 a with Otsu methods,
Binaryzation result such as Fig. 2 b.
2.3 feature point extraction
In order to reduce image procossing later stage work amount, left and right edges center is used from bianry image figure (2b) crop row
Line detecting method extracts Partial Feature point and represents crop row.It is suitable using from top to bottom, from left to right during feature point extraction
Ordered pair bianry image is scanned.Often in row pixel, crop is regarded as and is made up of the different white line segment of length, each line segment is taken
Midpoint be used as the characteristic point for representing crop row.Specific method is as follows:Current line is from left to right scanned, when pixel value from
0 (background) jumps to 1 (crop), then pixel value is that 1 pixel is considered as the starting point leftx of line segment, and image continues to sweep to the right
Retouch, when pixel value from 1 (crop) jumps to 0 (background), then the pixel value is considered as the terminal of line segment for 1 pixel
rightx.One length threshold is set, and length is less than into all pixels on the line segment of threshold value is considered as pseudomorphism vegetarian refreshments, conversely, taking this
The midpoint of line segmentFor the characteristic point to be looked for.Then swept from the second pixel point after rightx
Retouch, until the end of scan.Feature point extraction image such as Fig. 2 c.
3rd, feature points clustering.After target image is handled through above step, the feature dot image for representing crop row is obtained.According to
The characteristics of characteristic point of each crop row should be substantially distributed near the straight line determined by characteristic point, proposes following cluster side
Method：Belong to of a sort characteristic point and represent the data space that characteristic point is constituted very close to, it is assumed that V apart from d values to target line,
Y=kx+b represents the discrete straight line in image such as Fig. 2 c, calculates all characteristic points in V and, to the distance of these straight lines, finds special
Several straight lines for levying a cover-most are target crop row center line.
Comprise the following steps that：
(1) from top to bottom, from left to right scanning feature dot image (Fig. 2 c), finds the characteristic point that all pixels value is 1, deposits
Store up its positional information.
(2) span for setting two parameters k, b, k is k1To km, step-length is w1, b span is b1To bl, step
A length of w2.It is to have the discrete straight line of many bars in y=kx+b, this feature dot image that linear equation is represented with the slope-intercept form of straight line.
(3) there is n characteristic point (x in data space V1,y1),(x2,y2)…,(xn,yn).Initialize 0 matrix
NUM, size is M × N, wherein
To any combination (k, b), from top to bottom, from left to right scan image, calculates characteristic point (x in Vn,yn) arrive straight line y
=kx+b apart from d,
A distance threshold dbound is set, according to wheatland feature, dbound is the half of wheatland width.If d is less than
Dbound, then NUM [i, j]=NUM [i, j]+1.
(4) position corresponding k, the b in NUM corresponding to maximum are found out, the current a beeline y=kx+b is found apart from small
In dbound institute a little, and by its location index record in ntemp [], ntemp [] be 1 × n one-dimension array, n's takes
Value is determined by the number for meeting the characteristic point of distance constraints.One class cluster is finished.
(5) the corresponding all characteristic points of location index of the deletion record in ntemp [], circulation performs (1)-(4), and this is followed
Ring number of times is determined by the crop line number to be detected.
Characteristic point is divided into six classes by Fig. 3, rejects the point for being unsatisfactory for distance constraints, leaving can represent in target crop row
The characteristic point of heart line.
4th, the crop row center line detection based on least square method.
Because the discrete straight line in feature dot image is limited, easily disturbed by noise spot.Therefore, the present invention passes through elder generation
Cluster, then least square fitting is carried out, so as to improve the adaptability of least square method.First use and be distributed in around certain straight line
Point to the straight line distance be respectively less than a certain threshold value distance restraint to feature points clustering, point set is obtained, then on point set
Use least square fitting straight line.Specific method is as follows：
(1) the corresponding k in position in NUM corresponding to maximum is found out, b value finds and a little arrives the current a beeline y=kx+b
Distance is less than dbound institute a little, and by its location index record in ntemp [].
(2) the corresponding all characteristic points of location index value that will be recorded in ntemp [] carry out straight line with least square method
Fitting.
(3) characteristic point in above-mentioned steps (2) is deleted, circulation performs (1)-(4) in the 3rd step, until finding out all
Target line.
Because this method according to distance restraint selects a several most straight lines of covering point, as long as therefore choosing suitable distance
Threshold value d, the cluster results of these characteristic points is based on can represent the trend of crop row well.After the present invention, amount of calculation subtracts
It is few, overcome the problem of peakvalue's checking is difficult, and calculating speed faster, more accurate, more to the noise spot image tool of fitting effect
There is very strong robustness.
The row detection of the present invention also suitable for program request crop such as soybean, corn etc..
Claims (3)
1. a kind of wheatland crop row detection method of distinguished point based cluster, it is characterised in that：This method step is as follows：
(1) wheat crop image is shot with color camera；
(2) wheat crop bianry image is obtained；
(3) obtain representing the candidate feature point of wheat crop row with left and right edges center line detecting method；
(4) constrained procedure for being respectively less than a certain threshold value with the characteristic point being distributed in around straight line to the distance of the straight line carries out special
A cluster is levied, validity feature point is obtained；
(5) fitting a straight line is carried out to the validity feature point in each class with least square method, so as to detect each crop row.
2. a kind of wheatland crop row detection method of distinguished point based cluster according to claim 1, it is characterised in that step
Suddenly (4) are specially：
1) wheat crop bianry image from top to bottom, is from left to right scanned, the characteristic point (x that all pixels value is 1 is foundi,yi),
There are n characteristic point, and i=1,2,3 ..., n in data space V；
2) distance threshold dbound, wheatland crop line number num are initialized.
3) for any combination (km,bl), line correspondence y=kmx+bl, calculate all characteristic point (x in Vi,yi) arrive straight line y=kmx
+blApart from d,Wherein 1≤m≤max_m, kmStep-length be Sk, 1≤l≤max_l, blStep-length be
Sb, and as m=1, blFrom b1Traverse bmax_l, andkmax_m=kmax_m-1+Sk, until kmFrom k1
Traverse kmax_m；
If being less than threshold value dbound apart from d, this feature point is considered as belonging to straight line l validity feature point.If an accumulator
All the time record combination (km,bl) corresponding validity feature point number maximum, while recording the corresponding k of the maximummAnd bl's
Value；
4) step 3) it is finished, the corresponding k of the maximum in accumulatormAnd blValue, find straight line y=kmx+blAway from
From all validity feature points less than dbound, and store it in one-dimension array, the validity feature point of the first crop row gathers
Class is finished；
5) characteristic point of the participation cluster in deleting 4), updates the data collection V, and circulation performs step 3) -4), until the n-th um rows
Feature points clustering is finished.
3. a kind of wheatland crop row detection method of distinguished point based cluster according to claim 1, it is characterised in that：
Dbound is the half of wheatland width.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710238165.6A CN107067430B (en) | 2017-04-13 | 2017-04-13 | Wheat field crop row detection method based on feature point clustering |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710238165.6A CN107067430B (en) | 2017-04-13 | 2017-04-13 | Wheat field crop row detection method based on feature point clustering |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107067430A true CN107067430A (en) | 2017-08-18 |
| CN107067430B CN107067430B (en) | 2020-04-21 |
Family
ID=59599418
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710238165.6A Expired - Fee Related CN107067430B (en) | 2017-04-13 | 2017-04-13 | Wheat field crop row detection method based on feature point clustering |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107067430B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109344843A (en) * | 2018-09-07 | 2019-02-15 | 华南农业大学 | Rice seedling line extracting method, device, computer equipment and storage medium |
| CN111080649A (en) * | 2019-12-10 | 2020-04-28 | 桂林电子科技大学 | Image segmentation processing method and system based on Riemann manifold space |
| CN113111892A (en) * | 2021-05-12 | 2021-07-13 | 中国科学院地理科学与资源研究所 | Crop planting row extraction method based on unmanned aerial vehicle image |
| CN113128576A (en) * | 2021-04-02 | 2021-07-16 | 中国农业大学 | Crop row detection method and device based on deep learning image segmentation |
| CN115170949A (en) * | 2022-06-07 | 2022-10-11 | 浙江大学 | Early rice seedling row detection method based on root positioning |
Citations (4)
| 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 |
| CN101807252A (en) * | 2010-03-24 | 2010-08-18 | 中国农业大学 | Crop row center line extraction method and system |
| CN105046229A (en) * | 2015-07-27 | 2015-11-11 | 浙江理工大学 | Crop row identification method and apparatus |
| CN108133471A (en) * | 2016-11-30 | 2018-06-08 | 天津职业技术师范大学 | Agriculture Mobile Robot guidance path extracting method and device based on artificial bee colony algorithm under the conditions of a kind of natural lighting |
-
2017
- 2017-04-13 CN CN201710238165.6A patent/CN107067430B/en not_active Expired - Fee Related
Patent Citations (4)
| 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 |
| CN101807252A (en) * | 2010-03-24 | 2010-08-18 | 中国农业大学 | Crop row center line extraction method and system |
| CN105046229A (en) * | 2015-07-27 | 2015-11-11 | 浙江理工大学 | Crop row identification method and apparatus |
| CN108133471A (en) * | 2016-11-30 | 2018-06-08 | 天津职业技术师范大学 | Agriculture Mobile Robot guidance path extracting method and device based on artificial bee colony algorithm under the conditions of a kind of natural lighting |
Non-Patent Citations (3)
| Title |
|---|
| JIANG G等: "Automatic detection of crop rows based on multi-ROIs", 《EXPERT SYSTEMS WITH APPLICATIONS》 * |
| 司永胜等: "基于最小二乘法的早期作物行中心线检测方法", 《农业机械学报》 * |
| 姜国权等: "基于已知点的作物行检测方法", 《应用基础与工程科学学报》 * |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109344843A (en) * | 2018-09-07 | 2019-02-15 | 华南农业大学 | Rice seedling line extracting method, device, computer equipment and storage medium |
| CN109344843B (en) * | 2018-09-07 | 2020-09-25 | 华南农业大学 | Method and device for extracting rice seedling row line, computer equipment and storage medium |
| CN111080649A (en) * | 2019-12-10 | 2020-04-28 | 桂林电子科技大学 | Image segmentation processing method and system based on Riemann manifold space |
| CN111080649B (en) * | 2019-12-10 | 2023-05-30 | 桂林电子科技大学 | Image segmentation processing method and system based on Riemann manifold space |
| CN113128576A (en) * | 2021-04-02 | 2021-07-16 | 中国农业大学 | Crop row detection method and device based on deep learning image segmentation |
| CN113111892A (en) * | 2021-05-12 | 2021-07-13 | 中国科学院地理科学与资源研究所 | Crop planting row extraction method based on unmanned aerial vehicle image |
| CN115170949A (en) * | 2022-06-07 | 2022-10-11 | 浙江大学 | Early rice seedling row detection method based on root positioning |
| CN115170949B (en) * | 2022-06-07 | 2024-09-17 | 浙江大学 | Early rice seedling row detection method based on root positioning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107067430B (en) | 2020-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107067430A (en) | A kind of wheatland crop row detection method of distinguished point based cluster | |
| CN109215071B (en) | Intelligent rice and wheat harvester swath measuring method based on vision | |
| CN111462225B (en) | Centroid identification and positioning method of infrared light spot image | |
| CN109447945B (en) | Quick counting method for basic wheat seedlings based on machine vision and graphic processing | |
| CN110853081B (en) | Ground and airborne LiDAR point cloud registration method based on single-tree segmentation | |
| CN105117701B (en) | Corn crop row framework extraction method based on largest square principle | |
| CN111753577A (en) | Apple identification and positioning method in automatic picking robot | |
| CN104182974B (en) | A speeded up method of executing image matching based on feature points | |
| CN107833206B (en) | Accurate power line extraction method under complex background | |
| CN109522929A (en) | It is a kind of based on multi-source image fusion natural environment under wheat weeds recognition methods | |
| CN108133471B (en) | Robot navigation path extraction method and device based on artificial bee colony algorithm | |
| CN109190452B (en) | Crop row identification method and device | |
| CN114627367B (en) | Sea bottom line detection method for side-scan sonar image | |
| CN105279745A (en) | Normalization method for multi-feature point constraint histogram of remote sensing image color normalization | |
| CN110335282A (en) | A kind of profile line segment feature extraction algorithm based on grid | |
| CN108182700A (en) | A kind of method for registering images based on the detection of feature twice | |
| CN105046229B (en) | A kind of recognition methods of crops row and device | |
| CN110689586B (en) | Tongue image identification method in traditional Chinese medicine intelligent tongue diagnosis and portable correction color card used for same | |
| CN113421301B (en) | Method and system for positioning central area of field crop | |
| CN112039439B (en) | Automatic detection system and detection method for defects of solar cell | |
| CN113963255A (en) | Ridge operation machinery leading line identification method based on ridge color characteristics | |
| CN112446912B (en) | Financial bill width calculation method | |
| Germain et al. | Non destructive counting of wheatear with picture analysis | |
| Du et al. | Computation method of phenotypic parameters based on distribution map of kernels for corn ears | |
| CN115131387A (en) | Gasoline engine spray wall collision parameter automatic extraction method and system based on image processing |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20200421 Termination date: 20210413 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |