CN110363784A - A kind of recognition methods being overlapped fruit - Google Patents

A kind of recognition methods being overlapped fruit Download PDF

Info

Publication number
CN110363784A
CN110363784A CN201910571409.1A CN201910571409A CN110363784A CN 110363784 A CN110363784 A CN 110363784A CN 201910571409 A CN201910571409 A CN 201910571409A CN 110363784 A CN110363784 A CN 110363784A
Authority
CN
China
Prior art keywords
image
fruit
pixel
recognition methods
overlapped
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
Application number
CN201910571409.1A
Other languages
Chinese (zh)
Other versions
CN110363784B (en
Inventor
张栋
周涛
郗厚印
陈涛
谢凯胜
莫言
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Qingdao University of Technology
Original Assignee
Qingdao University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Qingdao University of Technology filed Critical Qingdao University of Technology
Priority to CN201910571409.1A priority Critical patent/CN110363784B/en
Publication of CN110363784A publication Critical patent/CN110363784A/en
Application granted granted Critical
Publication of CN110363784B publication Critical patent/CN110363784B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • 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
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • 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
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)

Abstract

The invention discloses a kind of recognition methods for being overlapped fruit, comprising the following steps: (1) Image Acquisition；(2) image of acquisition is filtered and removes dryness processing；(3) overlay target fruit extracts；(4) overlay target fruit separates；(5) actual profile extracts；(6) edge fitting and centroid detection.Recognition methods disclosed in this invention can effectively identify that large area is overlapped fruit, and accuracy of identification is high, helps to improve picking efficiency, has a good application prospect.

Description

A kind of recognition methods being overlapped fruit
Technical field
The present invention relates to image identification technical field, in particular to a kind of recognition methods for being overlapped fruit.
Background technique
With the development of agricultural automation, more and more Agriculture pick robots are applied in picking operation.Wherein, Fruit identification technology is extremely important for Agriculture pick robot, and the height of recognition accuracy directly affects picking effect Rate.
The case where overlapping inevitably will appear due to the fruit grown in natural environment, this difficulty for identifying fruit It greatly increases, the success rate for identifying fruit reduces, to reduce the efficiency of agricultural picking.
Summary of the invention
In order to solve the above technical problems, the present invention provides a kind of recognition methods for being overlapped fruit, to reach effective identification Large area is overlapped fruit, improves the purpose of picking efficiency.
In order to achieve the above objectives, technical scheme is as follows:
A kind of recognition methods being overlapped fruit, comprising the following steps:
(1) Image Acquisition；
(2) image of acquisition is filtered and removes dryness processing；
(3) overlay target fruit extracts, and carries out operation to each pixel in image first in RGB color, Background is split using maximum variance between clusters later；Then Morphological scale-space is carried out to the region after segmentation, finally will Fruit region is reacted in original image；
(4) overlay target fruit separates, and edge detection is carried out to fruit region first, then by the picture of Morphological scale-space XOR operation is carried out with the picture of edge detection, the picture after XOR operation is subjected to corrosion treatment, so that two objective fruits Separation it is more obvious；Connected domain analysis is carried out to the image after corrosion later, the binary map of two fruits is respectively displayed on In two images, if the case where wrong segmentation does not remove, by the small connected domain of connected domain analysis area of detection, carries out it Removal；
(5) actual profile extracts, after carrying out convex closure processing to the two images in previous step and flood filling processing, into The removal of row false contouring；
(6) edge fitting and centroid detection.
In above scheme, in the step (3), image is carried out using c=k*r-g-b operator in RGB color Operation, wherein c is the pixel value after operation, and k is chromatic aberration coefficient, and r, g, b respectively indicate red, green, blue in RGB color Value.
In above scheme, in the step (3), the method for Morphological scale-space is as follows: first carrying out closed operation to it and carries out again Operation is opened, the hole that fruit region occurs is filled and accidentally the local of segmentation removes by fraction, if there is The excessive situation of hole utilizes the method for flooding filling；If accidentally the region of segmentation is larger, will be smaller using connected domain analysis Connected domain removal.
In above scheme, in the step (4), canny operator edge detection is carried out to fruit region, by lap Edge effectively identifies.
In above scheme, in the step (4), XOR operation is that the pixel for the same position for being directed to two images is transported It calculates, when two pixels are all white or are all black, then output black；If two pixels are one black one white, export white Color；Here, this two images is the bianry image img1 of Morphological scale-space and the image img2 of edge detection, exclusive or fortune respectively The formula of calculation are as follows:
Wherein, img indicates the image after XOR operation, and i, j respectively indicate the transverse and longitudinal position of pixel.
In above scheme, in the step (4), connected domain analysis is carried out using se ed filling algorithm, and the specific method is as follows:
1) pixel in scan image, until pixel img [i, j]=1；
2) it regard current pixel img [i, j] as seed, and is labeled as a label, by seed adjoining and pixel It is worth in identical pixel indentation stack；
3) stack top pixel is popped up, and marks identical label, then will be adjacent and the identical pixel indentation stack of pixel value In；
4) repeat step 3), until stack be sky, just obtained one of the connected domain in image at this time, and by a label mark Note；
5) whole connected domains in image can be obtained until the end of scan to step 4) by repeating step 1).
In above scheme, in the step (5), specific step is as follows for convex closure processing:
1) the smallest pixel of ordinate in image is found, m is labeled as0；
2) residual pixel point and m are calculated0Line and horizontal axis between angle cosine value, from big to small according to cosine value Sequence, by these point be respectively labeled as m1, m2, m3…；
3) by m0And m1It is first pressed into stack, then from m2Start, calculates the vector of the two o'clock of stack top relative to the point and stack top Whether the vector of point is to rotate counterclockwise, and if it is counterclockwise relationship, is then pressed this in pushing on；Otherwise, stack top element is popped up；
4) all pixels point in last stack is the vertex of convex closure.
In above scheme, in the step (5), flooding filling, the specific method is as follows: being first filled, will fill out to background The pixel filled is recorded and is reflected on new picture, then the pixel value of this picture is negated, after convex closure processing can be obtained The picture that hole fills up.
In above scheme, in the step (5), the removal of false contouring is carried out by Hough straight-line detection.
In above scheme, in the step (6), the edge of fruit is fitted with least square method ellipses detection, To rebuild the profile for the fruit that is blocked；The mass center of fruit is obtained using elliptical geometric formula.
Through the above technical solutions, a kind of recognition methods for being overlapped fruit provided by the invention is simple to operation, versatility It is good, there is very high discrimination, can effectively improve the picking efficiency of fruit, has broad application prospects.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below There is attached drawing needed in technical description to be briefly described.
Fig. 1 is a kind of recognition methods flow diagram of overlapping fruit disclosed in the embodiment of the present invention；
Fig. 2 is the image of camera acquisition；
Fig. 3 is the image after median filter process；
Fig. 4 is the bianry image after the segmentation of new operator；
Fig. 5 is the image after Morphological scale-space；
Fig. 6 is the result of Objective extraction；
Fig. 7 is the edge-detected image in tomato region；
Fig. 8 is the image after XOR operation；
Fig. 9 is the image after corrosion；
Figure 10 is the fruit image that is not blocked after connected domain analysis；
Figure 11 is the fruit image that is blocked after connected domain analysis；
Figure 12 is the fruit image not being blocked after actual profile extracts；
Figure 13 is the fruit image being blocked after actual profile extracts；
Figure 14 be least square method elliptical edge be fitted and centroid detection after the fruit image that is not blocked；
Figure 15 be least square method elliptical edge be fitted and centroid detection after the fruit image that is blocked；
Figure 16 is final detection result image.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete Site preparation description.
The present invention provides a kind of recognition methods for being overlapped fruit, process as shown in Figure 1, specific embodiment is as follows:
The present invention carries out image recognition for using tamato fruit.
1, it is overlapped the Image Acquisition of tomato
It is acquired using Tomato Image of the colored binocular camera to overlapping, as shown in Fig. 2, extracting the two dimension of target Information.
2, image filtering denoising
During camera obtains image, some unnecessary interference informations can be inevitably generated, they are largely On affect the quality of image, and to the processing result of image there may be large effect, the present embodiment gets camera Original image carry out median filter process, treated, and image is shown in that Fig. 3, median filtering can not only effectively remove noise, also protect Edge details are protected.
3, Objective extraction:
1) the background segment method of the OTSU based on kr-g-b: since mature tomato takes on a red color, so in RGB color Operation is carried out to image using c=k*r-g-b operator, the red color component value of pixel in picture can be improved in this formula, reduces Green and blue component value, to enhance the tomato part in picture and the contrast of non-tomato part, later using between maximum kind Variance method (OTSU) is split background, has good segmentation effect when k takes 1.8 by testing discovery, after segmentation Image is as shown in Figure 4.
2) Morphological scale-space: the image after segmentation can inevitably have some holes and the accidentally region of segmentation.Utilize form It learns the bianry image for obtaining previous step to handle, closed operation is first carried out to it and carries out out operation again, can incite somebody to action in this way The hole occurred in tomato region is filled and accidentally the local of segmentation removes by fraction.It is most of to divide faulty feelings Condition can almost be removed using morphology.If there is the excessive situation in cavity, the method that can use flooding filling；If accidentally The region of segmentation is larger, can use connected domain analysis, is that can reach effect by the removal of lesser connected domain, after Morphological scale-space Image as shown in Figure 5.
3) Objective extraction result: tomato region is reacted in original image, as shown in Figure 6.
4, overlay target separates:
1) edge detection: carrying out canny operator edge detection to tomato region, can be effective by the edge of lap It identifies, the image after detection is shown in Fig. 7.
2) picture of Morphological scale-space and the picture of edge detection XOR operation: are subjected to XOR operation.XOR operation is Operation is carried out for the pixel of the same position of two images, when two pixels are all white or be all black, is then exported Black；If two pixels are one black one white, white is exported.Here, this two images is the image of Morphological scale-space respectively Edge-detected image (Fig. 7) img2 of (Fig. 5) img1 and overlapping tomato, the formula of XOR operation are as follows:
Wherein, img indicates the image after XOR operation, and i, j respectively indicate the transverse and longitudinal position of pixel.After XOR operation Image it is as shown in Figure 8.
3) the connected domain separation based on erosion operation: the picture after XOR operation is subjected to corrosion treatment, so that two mesh The separation for marking tomato is more obvious.Connected domain analysis is carried out to the image (Fig. 9) after corrosion, the binary map of two fruits is distinguished In two images (as shown in Figure 10 and Figure 11), if the case where wrong segmentation does not remove, can be examined by connected domain analysis The small connected domain of area is surveyed, it is removed, to can reach the effect that accidentally partitioning portion is removed.
Connected domain analysis is carried out using se ed filling algorithm, basic ideas are to choose a certain pixel as seed, then The pixel being connected to the seed is attributed to the same set of pixels by the adjacent and equal pixel value principle according to the position of connected region In conjunction, this pixel set is a connected region.Its analysis method is as follows:
(1) pixel in scan image, until pixel img [i, j]=1；
(2) it regard current pixel img [i, j] as seed, and is labeled as a label (label), by seed neighbour It connects and in the identical pixel indentation stack of pixel value；
(3) stack top pixel is popped up, and marks identical label, then will be adjacent and the identical pixel indentation stack of pixel value In；
(4) repeat step 3, until stack be sky, just obtained one of the connected domain in image at this time, and by a label mark Note；
(5) whole connected domains in image can be obtained until the end of scan to step 4 by repeating step 1.
5, the extraction of actual profile: if directly extracting edge, the fruit being blocked to the two images in previous step A part of false contouring can be obtained, it is carried out to carry out edge extracting after convex closure and flooding filling processing, false contouring at this time Part is straight line, is removed by straight-line detection to it.
1) convex closure is handled: specific step is as follows for algorithm of convex hull:
(1) the smallest pixel of ordinate in image is found, m is labeled as0；
(2) residual pixel point and m are calculated0Line and horizontal axis between angle cosine value, according to cosine value from greatly to These points are respectively labeled as m by small sequence1, m2, m3…
(3) by m0And m1It is first pressed into stack, then from m2Start, calculates the vector of the two o'clock of stack top relative to the point and stack Vector (such as the vector on vertexRelative to vector) it whether is to rotate counterclockwise.If it is counterclockwise relationship, then will In this presses and pushes on；Otherwise, stack top element is popped up.
(4) all pixels point in last stack is the vertex of convex closure.
2) flood filling: after convex closure processing, it may appear that some holes first carry out background using the method for the filling that floods Filling, the pixel of filling is recorded and is reflected on new picture, then the pixel value of this picture is negated, can be obtained convex The picture that hole fills up after packet processing.
3) removal of false contouring: false contouring part at this time is straight line, is gone by Hough straight-line detection to it It removes.Straight line is indicated with (r, θ), wherein r is the shortest distance of this straight line to origin, and θ is the vertical line and x of this straight line The angle of axis.Assuming that two o'clock is respectively i, j in image, then the straight line for crossing this two o'clock is (ri,θi) and (rj,θj), since two o'clock is true Straight line is determined, so there must be ri=rjAnd θi=θjThe case where.To whether point-blank judge multiple points, then only need Judge whether these r and θ have equal situation.When more points has above situation, then illustrate point point-blank More, the quantity that can be point-blank put by setting is threshold value to determine whether to detect straight line.Actual profile Image after extraction is as shown in Figure 12 and Figure 13.
6, edge fitting and centroid detection:
1) due to tamato fruit approximate ellipse body, the approximate ellipse in two dimensional image, so being examined with least square method ellipse It surveys to be fitted to the edge of fruit, so as to rebuild the profile for the fruit that is blocked.
Write the ellipse in plane as non-standard form first
x2+Axy+By2+ Cx+Dy+E=0
In formula, there are five unknown numbers of A, B, C, D, E respectively.To find out this five unknown numbers, at least require have five groups and The above sampled point, if Pi(xi,yi) (i=1,2 ..., N) be ellipse on point, wherein N >=5.According to principle of least square method, The objective function for needing to be fitted are as follows:
In order to arrive the minimum value of objective function F, enable
After calculating as, its result is write to the form of matrix:
By the value for solving equation available five unknown numbers.
2) mass center of the available fruit of elliptical geometric knowledge is utilized:
Edge fitting and the image of centroid detection are as shown in Figure 14 and Figure 15.Final testing result is shown in Figure 16.
The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, as defined herein General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, of the invention It is not intended to be limited to the embodiments shown herein, and is to fit to and the principles and novel features disclosed herein phase one The widest scope of cause.

Claims (10)

1. a kind of recognition methods for being overlapped fruit, which comprises the following steps:
(1) Image Acquisition；
(2) image of acquisition is filtered and removes dryness processing；
(3) overlay target fruit extracts, and carries out operation to each pixel in image first in RGB color, later Background is split using maximum variance between clusters；Then Morphological scale-space is carried out to the region after segmentation, finally by fruit Region is reacted in original image；
(4) overlay target fruit separates, and edge detection is carried out to fruit region first, then by the picture of Morphological scale-space and side The picture of edge detection carries out XOR operation, and the picture after XOR operation is carried out corrosion treatment, so that point of two objective fruits From more obvious；Connected domain analysis is carried out to the image after corrosion later, the binary map of two fruits is respectively displayed on two width On image, if the case where wrong segmentation does not remove, by the small connected domain of connected domain analysis area of detection, goes it It removes；
(5) actual profile extracts, and carries out convex closure processing with after flooding filling processing to the two images in previous step, carries out pseudo- The removal of profile；
(6) edge fitting and centroid detection.
2. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (3), Operation is carried out to image using c=k*r-g-b operator in RGB color, wherein c is the pixel value after operation, and k is color difference Coefficient, r, g, b respectively indicate the value of red, green, blue in RGB color.
3. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (3), form The method for learning processing is as follows: first carrying out closed operation to it and carries out out operation again, the hole that fruit region occurs is filled out It fills and accidentally the local of segmentation removes by fraction, if there is the excessive situation of hole, the method for utilizing flooding filling；If Accidentally the region of segmentation is larger, and using connected domain analysis, lesser connected domain is removed.
4. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (4), to fruit Real region carries out canny operator edge detection, and the edge of lap is effectively identified.
5. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (4), exclusive or Operation is to carry out operation for the pixel of the same position of two images, when two pixels are all white or are all black, Then output black；If two pixels are one black one white, white is exported；Here, this two images is Morphological scale-space respectively Bianry image img1 and edge detection image img2, the formula of XOR operation are as follows:
Wherein, img indicates the image after XOR operation, and i, j respectively indicate the transverse and longitudinal position of pixel.
6. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (4), connection Domain analysis is carried out using se ed filling algorithm, and the specific method is as follows:
1) pixel in scan image, until pixel img [i, j]=1；
2) it regard current pixel img [i, j] as seed, and is labeled as a label, by seed adjoining and pixel value phase In same pixel indentation stack；
3) stack top pixel is popped up, and marks identical label, then will be adjacent and in the identical pixel indentation stack of pixel value；
4) step 3) is repeated, until stack is sky, has just obtained one of the connected domain in image at this time, and marked by a label；
5) whole connected domains in image can be obtained until the end of scan to step 4) by repeating step 1).
7. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (5), convex closure Specific step is as follows for processing:
1) the smallest pixel of ordinate in image is found, m is labeled as0；
2) residual pixel point and m are calculated0Line and horizontal axis between angle cosine value, according to cosine value from big to small suitable These points are respectively labeled as m by sequence1, m2, m3…；
3) by m0And m1It is first pressed into stack, then from m2Start, calculates the vector of the two o'clock of stack top relative to the point and stack top point Whether vector is to rotate counterclockwise, and if it is counterclockwise relationship, is then pressed this in pushing on；Otherwise, stack top element is popped up；
4) all pixels point in last stack is the vertex of convex closure.
8. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (5), flooding The specific method is as follows for filling: being first filled to background, the pixel of filling is recorded and is reflected on new picture, then will The pixel value of this picture negates, and the picture that hole fills up after convex closure is handled can be obtained.
9. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (5), false ring Wide removal is carried out by Hough straight-line detection.
10. a kind of recognition methods for being overlapped fruit according to claim 1, which is characterized in that in the step (6), use Least square method ellipses detection is fitted the edge of fruit, to rebuild the profile for the fruit that is blocked；Using elliptical Geometric formula obtains the mass center of fruit.
CN201910571409.1A 2019-06-28 2019-06-28 Identification method of overlapped fruits Active CN110363784B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910571409.1A CN110363784B (en) 2019-06-28 2019-06-28 Identification method of overlapped fruits

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910571409.1A CN110363784B (en) 2019-06-28 2019-06-28 Identification method of overlapped fruits

Publications (2)

Publication Number Publication Date
CN110363784A true CN110363784A (en) 2019-10-22
CN110363784B CN110363784B (en) 2022-12-27

Family

ID=68215963

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910571409.1A Active CN110363784B (en) 2019-06-28 2019-06-28 Identification method of overlapped fruits

Country Status (1)

Country Link
CN (1) CN110363784B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112183230A (en) * 2020-09-09 2021-01-05 上海大学 Identification and central point positioning method for pears in natural pear orchard environment

Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102043957A (en) * 2011-01-11 2011-05-04 北京邮电大学 Vehicle segmentation method based on concave spots of image
CN103632148A (en) * 2013-12-13 2014-03-12 湖南大学 Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant
US20140313206A1 (en) * 2013-04-23 2014-10-23 Adobe Systems Incorporated Fast high-fidelity flood-filling on vector artwork
CN104318546A (en) * 2014-09-29 2015-01-28 中国农业大学 Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system
CN105005761A (en) * 2015-06-16 2015-10-28 北京师范大学 Panchromatic high-resolution remote sensing image road detection method in combination with significance analysis
CN105518743A (en) * 2013-03-13 2016-04-20 微软技术许可有限责任公司 Depth image processing
CN107038446A (en) * 2017-03-23 2017-08-11 中国计量大学 A kind of night double fruits overlapping tomato recognition methods detected under active illumination based on overlapping edge
CN107766784A (en) * 2016-08-20 2018-03-06 宋坤骏 A kind of novel video people counting algorithm
CN108827181A (en) * 2018-03-14 2018-11-16 浙江大学山东工业技术研究院 A kind of plate surface detection method of view-based access control model
CN109711284A (en) * 2018-12-11 2019-05-03 江苏博墨教育科技有限公司 A kind of test answer sheet system intelligent recognition analysis method

Patent Citations (10)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102043957A (en) * 2011-01-11 2011-05-04 北京邮电大学 Vehicle segmentation method based on concave spots of image
CN105518743A (en) * 2013-03-13 2016-04-20 微软技术许可有限责任公司 Depth image processing
US20140313206A1 (en) * 2013-04-23 2014-10-23 Adobe Systems Incorporated Fast high-fidelity flood-filling on vector artwork
CN103632148A (en) * 2013-12-13 2014-03-12 湖南大学 Image detection method for middle axis and length of leaf of long-petiole axisymmetric plant
CN104318546A (en) * 2014-09-29 2015-01-28 中国农业大学 Multi-scale analysis-based greenhouse field plant leaf margin extraction method and system
CN105005761A (en) * 2015-06-16 2015-10-28 北京师范大学 Panchromatic high-resolution remote sensing image road detection method in combination with significance analysis
CN107766784A (en) * 2016-08-20 2018-03-06 宋坤骏 A kind of novel video people counting algorithm
CN107038446A (en) * 2017-03-23 2017-08-11 中国计量大学 A kind of night double fruits overlapping tomato recognition methods detected under active illumination based on overlapping edge
CN108827181A (en) * 2018-03-14 2018-11-16 浙江大学山东工业技术研究院 A kind of plate surface detection method of view-based access control model
CN109711284A (en) * 2018-12-11 2019-05-03 江苏博墨教育科技有限公司 A kind of test answer sheet system intelligent recognition analysis method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
夏雪 等: "自然环境下苹果作业机器人双果重叠目标侦测方法", 《中国农业科技导报》 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112183230A (en) * 2020-09-09 2021-01-05 上海大学 Identification and central point positioning method for pears in natural pear orchard environment

Also Published As

Publication number Publication date
CN110363784B (en) 2022-12-27

Similar Documents

Publication Publication Date Title
CN102132323B (en) System and method for automatic image straightening
CN105894503B (en) A kind of restorative procedure of pair of Kinect plant colour and depth detection image
CN107038446B (en) Night double-fruit overlapping tomato identification method based on overlapping edge detection under active illumination
CN111753577B (en) Apple identification and positioning method in automatic picking robot
CN108319973A (en) Citrusfruit detection method on a kind of tree
CN105260693A (en) Laser two-dimensional code positioning method
CN107492094A (en) A kind of unmanned plane visible detection method of high voltage line insulator
CN110866932A (en) Multi-channel tongue edge detection device and method and storage medium
CN112395984B (en) Method for detecting seedling guide line of unmanned agricultural machine
CN114331986A (en) Dam crack identification and measurement method based on unmanned aerial vehicle vision
CN104766319A (en) Method for improving registration precision of images photographed at night
Feng et al. A separating method of adjacent apples based on machine vision and chain code information
CN112146646A (en) Method for detecting field leading line after crop ridge sealing
CN111476723B (en) Remote sensing image lost pixel recovery method for failure of Landsat-7 scanning line corrector
CN112085037A (en) Infrared thermal fault feature extraction and digital expression method for power transformation equipment
Yang et al. Vehicle color recognition using monocular camera
CN110363784A (en) A kind of recognition methods being overlapped fruit
CN114241436A (en) Lane line detection method and system for improving color space and search window
CN103593840A (en) Method for detecting phenotype of Arabidopsis
CN112541383B (en) Method and device for identifying weed area
CN109685773A (en) A kind of determination method of synoptic sites point in immunocyte image
CN117058018A (en) Method for repairing suspended impurity vision shielding area facing underwater structure detection
CN114612333B (en) Microscopic image-based micro-capsule algae cell statistical method
CN107437254B (en) Orchard adjacent overlapping shape fruit distinguishing method
Weng et al. MSER and connected domain analysis based algorithm for container code locating process

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