CN109658379A - A method of orange yield is quickly calculated by picture - Google Patents
A method of orange yield is quickly calculated by picture Download PDFInfo
- Publication number
- CN109658379A CN109658379A CN201811328800.0A CN201811328800A CN109658379A CN 109658379 A CN109658379 A CN 109658379A CN 201811328800 A CN201811328800 A CN 201811328800A CN 109658379 A CN109658379 A CN 109658379A
- Authority
- CN
- China
- Prior art keywords
- picture
- tree
- fruit
- coefficient
- citrus
- 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
- 238000000034 method Methods 0.000 title claims abstract description 20
- 235000013399 edible fruits Nutrition 0.000 claims abstract description 59
- 241000207199 Citrus Species 0.000 claims abstract description 43
- 235000020971 citrus fruits Nutrition 0.000 claims abstract description 40
- 241000196324 Embryophyta Species 0.000 claims abstract description 34
- 238000012360 testing method Methods 0.000 claims description 39
- 238000013528 artificial neural network Methods 0.000 claims description 18
- 238000012549 training Methods 0.000 claims description 18
- 230000002708 enhancing effect Effects 0.000 claims description 9
- 238000007781 pre-processing Methods 0.000 claims description 9
- 241001672694 Citrus reticulata Species 0.000 claims description 5
- 230000005855 radiation Effects 0.000 claims description 5
- 238000012937 correction Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims 1
- 238000010606 normalization Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000007493 shaping process Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
-
- 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
- G06T2207/30242—Counting objects in image
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of methods that orange yield is quickly calculated by picture, comprising the following steps: S1: intelligent terminal obtains single plant citrus picture；S2: the S1 single plant citrus picture obtained is identified by network model, is returned to single plant and is identified fruit quantity n, k (m) age of tree coefficient, the tree-like coefficient of k (w)；S3: citrus quantity: P=n* { e is obtained by following formula[k(n)+k(m)+k(w)]- 1 } * [ln (x+1)] * y*z；Wherein, P is orange yield, and n is that single plant identifies fruit quantity, and k (n) is fruit coefficient, and k (m) is age of tree coefficient, and k (w) is tree-like coefficient, and x is average single weight, and y is average strain number per acre, and z is mu number.This method realizes the photo that the single plant citrus in fruiting stage is shot by intelligent terminal, intelligent recognition goes out the quantity of citrus fruit, according to fruit coefficient, the age of tree, tree-like, single fruit weight characteristics, the fruit of single plant citrus fruit tree scientific can be estimated, it is accurate to calculate growing area orange yield.
Description
Technical field
The invention belongs to agricultural product image recognitions and intelligence computation field, in particular to a kind of quickly to calculate mandarin orange by picture
The method of tangerine yield.
Technical background
Before citrus receives fruit, generally by way of several or estimation, obtains single plant citrus and produce fruit quantity, then conversion produces
Amount.In the existing method for calculating yield by estimation, generally with single plant estimation fruit quantity, average single weight, average every
Mu strain number and the product of mu number obtain.By the single plant citrus fruit grain number amount that counts, counts, yield is then estimated, low efficiency, accurately
Rate is low.
The existing intelligent identification Method to fruit quantity in shooting image, is only capable of the fruit grain number of statistics shooting face appearance
Amount, it is difficult to which the fruit quantity of single plant fruit tree is obtained according to picture.
Summary of the invention
According to above-mentioned technical problem, the present invention shoots the photo of the single plant citrus in fruiting stage, root by intelligent terminal
According to the fruit coefficient, age of tree, tree-like, can science estimate the fruit of single plant citrus fruit tree, and by single weight of citrus
Amendment, it is accurate to calculate growing area orange yield.Specific technical solution is as follows:
A method of orange yield is quickly calculated by picture, comprising the following steps:
S1: intelligent terminal obtains single plant citrus picture；
S2: the S1 single plant citrus picture obtained is identified by network model, returns to single plant identification fruit quantity n, k (m) tree
The tree-like coefficient of age coefficient, k (w)；
S3: citrus quantity is obtained by following formula:
P=n* { e[k(n)+k(m)+k(w)]-1}*[ln(x+1)]*y*z
Wherein, P is orange yield, and n is that single plant identifies fruit quantity, and k (n) is fruit coefficient, and k (m) is age of tree coefficient, and k (w) is
Tree-like coefficient, x are average single weight, and y is average strain number per acre, and z is mu number.
Since the picture of single plant citrus only takes the fruiting quantity of single plant citrus fruit tree unilateral side outer layer, therefore, it is necessary to right
Feature, the citrus age of tree and tree-like fruiting feature, the weight characteristics of the batch fruit of fruit are modified, and are reasonably estimated
The quantity of single plant citrus.
Further, further include following steps:
S0: establish n single plant identification fruit quantity, k (m) age of tree coefficient, the tree-like coefficient of k (w) network model.
Further, fruit coefficient k (n) is obtained by following calculation method:
Wherein, n is that single plant identifies fruit quantity, nsFor true single plant fruit quantity, fsFor correction value.
Further, age of tree coefficient k (m), wherein m >=1.5,0.85≤k (m)≤1.
Further, tree-like coefficient k (w), wherein 0.98≤k (w)≤1.12.
Further, average single citrus weight x is obtained by following formula:
Wherein, X is citrus sample size, and x is that single plant identifies fruit quantity, xiFor the weight of single citrus in sample.
Further, specific step is as follows for the network model foundation of single plant identification fruit quantity n:
S0-1-1: different several photos containing citrus fruit are shot；
S0-1-2: the citrus fruit in picture obtained using labelimg annotation tool to S0-1-1 is labeled, and picture is outlined
In fruit；
S0-1-3: pre-processing the picture that S0-1-2 has been marked, including data enhancing and normalized；
S0-1-4: data are divided into training set and test set according to the ratio of 5:1 by the pretreated picture of S0-1-3；
S0-1-5: the data in training set extract the characteristic information of picture, feature letter by faster rcnn deep neural network
Breath includes texture information, color information, shape information, then in predicted pictures target location information, it is raw after successive ignition
At network parameter；
S0-1-6: the network parameter that data in test set generate S0-1-5 by faster rcnn deep neural network into
Row test, the location information for obtaining test picture is compared with test picture in the location information that S0-1-2 is marked, final logical
It crosses mean accuracy map and recall rate recall is assessed；
S0-1-7: repeating step S0-1-5 and S0-1-6, until map and recall reaches demand；
S0-1-8: final trained network parameter is loaded into faster rcnn network model.
Further, specific step is as follows for the network model foundation of age of tree coefficient k (m):
S0-2-1: several photos of the single plant citrus of the different specific age of tree sections of shooting, specific age of tree section includes: 1.5≤m < 2.5,
2.5≤m < 3.5,3.5≤m < 4.5,4.5≤y, and picture is divided into 4 classes according to age of tree section；
S0-2-2: pre-processing the step S0-2-1 picture classified, including data enhancing and normalized；
S0-2-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-2-2；
S0-2-4: the data in training set extract the characteristic information of picture, characteristic information by GoogleNet deep neural network
Including texture information, color information, shape information, the then affiliated classification of predicted pictures generates network ginseng after successive ignition
Number；
S0-2-5: the data in test set carry out the network parameter that S0-2-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-2-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-2-6: repeating step S0-2-4 and S0-2-5, until recall and accuracy of identification reach demand；
S0-2-7: final trained network parameter is loaded into GoogleNet network model.
Further, specific step is as follows for the network model foundation of tree-like coefficient k (w):
S0-3-1: several photos of the single plant mandarin tree of the different specific tree-like w of shooting, specific tree-like w include: more major branch radiation
Shape, natural open centre shape, cylinder, trunk shape, natural globular model, and picture is divided into 5 classes according to tree-like；
S0-3-2: pre-processing the step S0-3-1 picture classified, including data enhancing and normalized；
S0-3-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-3-2；
S0-3-4: the data in training set extract the characteristic information of picture, characteristic information by GoogleNet deep neural network
Including texture information, color information, shape information, the then affiliated classification of predicted pictures generates network ginseng after successive ignition
Number；
S0-3-5: the data in test set carry out the network parameter that S0-3-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-3-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-3-6: repeating step S0-3-4 and S0-3-5, until recall and accuracy of identification reach demand；
S0-3-7: final trained network parameter is loaded into GoogleNet network model.
The invention has the benefit that
This method realizes the photo that the single plant citrus in fruiting stage is shot by intelligent terminal, and intelligent recognition goes out the number of citrus fruit
Amount scientific can estimate the fruit of single plant citrus fruit tree according to fruit coefficient, the age of tree, tree-like, single fruit weight characteristics,
It is accurate to calculate growing area orange yield.
Detailed description of the invention
Attached drawing 1 is a kind of implementation flow chart of method that orange yield is quickly calculated by picture of the invention.
Specific embodiment
Specifically the solution of the present invention is illustrated below by way of attached drawing.
A method of orange yield is quickly calculated by picture, comprising the following steps:
S0: establish n single plant identification fruit quantity, k (m) age of tree coefficient, the tree-like coefficient of k (w) network model；
S0-1: the network model of single plant identification fruit quantity n is established；
S0-1-1: the different several photos for containing citrus fruit are shot as citrus fruit sample database, picture number is no less than 1000
?；
S0-1-2: the citrus fruit in picture obtained using labelimg annotation tool to S0-1-1 is labeled, and picture is outlined
In fruit；
S0-1-3: pre-processing the picture that S0-1-2 has been marked, and is revolved using the RotateImage of OpenCV to picture
Turn, picture luminance is handled using LightImage, realizes data enhancing；Using caffe Scale layer and
BatchNorm layer realizes data normalization processing；Increase sample size to enable the network to more to improve algorithm robustness
Good optimizes；
S0-1-4: data are divided into training set and test set according to the ratio of 5:1 by the pretreated picture of S0-1-3；
S0-1-5: the data in training set extract the characteristic information of picture, feature letter by faster rcnn deep neural network
Breath include texture information, color information, shape information, then in predicted pictures target location information, after 500,000 iteration
Generate network parameter；
S0-1-6: the network parameter that data in test set generate S0-1-5 by faster rcnn deep neural network into
Row test, the location information for obtaining test picture is compared with test picture in the location information that S0-1-2 is marked, final logical
It crosses mean accuracy map and recall rate recall is assessed；
S0-1-7: repeating step S0-1-5 and S0-1-6, until map and recall reaches demand；
S0-1-8: final trained network parameter is loaded into faster rcnn network model.
S0-2: the network model of k (m) age of tree coefficient is established；
S0-2-1: it is the specific age of tree sample database of citrus that several photos of the single plant citrus of the different specific age of tree sections of shooting, which are used as, often
Kind certain tree age picture number is no less than 1000；
Specific age of tree section includes: 1.5≤m < 2.5,2.5≤m < 3.5,3.5≤m < 4.5,4.5≤y, and by picture according to tree
Age section is divided into 4 classes；
According to well known citrus from growth of seedling by 4.5 years, tree-like and branch just tends towards stability, tree-like size and luxuriant influence meter
It calculates.It is sampled and is corrected according to mass data, age of tree coefficient and age of tree relationship are as follows:
| Age of tree m | Age of tree coefficient k (m) |
| 1.5-2.5 | 0.85 |
| 2.5-3.5 | 0.93 |
| 3.5-4.5 | 0.97 |
| 4.5 and on | 1 |
S0-2-2: pre-processing the step S0-2-1 picture classified, using OpenCV RotateImage to picture into
Row rotation is handled picture luminance using LightImage, realizes data enhancing；Using caffe Scale layer and
BatchNorm layer realizes data normalization processing；Increase sample size to enable the network to more to improve algorithm robustness
Good optimizes；
S0-2-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-2-2；
S0-2-4: the data in training set extract the characteristic information of picture, characteristic information by GoogleNet deep neural network
Including texture information, color information, shape information, then the affiliated classification of predicted pictures, generates network after 500,000 iteration
Parameter；
S0-2-5: the data in test set carry out the network parameter that S0-2-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-2-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-2-6: repeating step S0-2-4 and S0-2-5, until recall and accuracy of identification reach demand；
S0-2-7: final trained network parameter is loaded into GoogleNet network model.
S0-3: the network model of the tree-like coefficient of k (w) is established；
S0-3-1: several photos of the single plant mandarin tree of the different specific tree-like w of shooting are as the specific tree-like sample database of citrus, and every kind
Specific tree-like picture number is no less than 1000；
Specific tree-like w includes: more major branch radiations, natural open centre shape, cylinder, trunk shape, natural globular model, and picture is pressed
It is divided into 5 classes according to tree-like；
It is tree-like there are many shape according to well known citrus, it is trimmed according to difference to Citrus Cultivars.Common classification has:
1) more major branch radiations: without class central backbone, have major branch source several；
2) natural globular model: closest to the citrus shaping grown naturally, branch is more and close；
3) natural open centre shape: skeleton branch subordinate is obvious, and branch is evenly distributed, and illumination is good；
4) trunk shape: naturally, tree-like higher, Shoot number is more for branch；
5) cylindrical: upper and lower major branch difference in length is smaller；
Different tree form structure influences to calculate.It is sampled and is corrected according to mass data, tree-like coefficient and tree-like relationship are as follows:
| Tree-like w | Tree-like coefficient k (w) |
| More major branch radiations | 0.98 |
| Natural open centre shape | 1 |
| It is cylindrical | 1.02 |
| Trunk shape | 1.05 |
| Natural globular model | 1.12 |
S0-3-2: pre-processing the step S0-3-1 picture classified, using OpenCV RotateImage to picture into
Row rotation is handled picture luminance using LightImage, realizes data enhancing；Using caffe Scale layer and
BatchNorm layer realizes data normalization processing；Increase sample size to enable the network to more to improve algorithm robustness
Good optimizes；
S0-3-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-3-2；
S0-3-4: the data in training set extract the characteristic information of picture, characteristic information by GoogleNet deep neural network
Including texture information, color information, shape information, then the affiliated classification of predicted pictures, generates network after 500,000 iteration
Parameter；
S0-3-5: the data in test set carry out the network parameter that S0-3-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-3-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-3-6: repeating step S0-3-4 and S0-3-5, until recall and accuracy of identification reach demand；
S0-3-7: final trained network parameter is loaded into GoogleNet network model.
S1: intelligent terminal obtains single plant citrus picture, when taking pictures, keeps single plant citrus main body clear, position is located at picture
Middle.
S2: picture is identified by faster rcnn network model, is returned to the citrus target detected, is obtained citrus
Single plant identifies fruit quantity n；Picture is identified by GoogleNet network model, returns k (m) age of tree coefficient, the k that obtain citrus
(w) classification results and confidence level of tree-like coefficient.
S3: citrus quantity is obtained by following formula:
P=n* { e[k(n)+k(m)+k(w)]-1}*[ln(x+1)]*y*z
Wherein, P is orange yield, and n is that single plant identifies fruit quantity, and k (n) is fruit coefficient, and k (m) is age of tree coefficient, and k (w) is
Tree-like coefficient, x are average single weight, and y is average strain number per acre, and z is mu number.
Average single citrus weight x is obtained by following formula:
Wherein, X is citrus sample size, and x is that single plant identifies fruit quantity, xiFor the weight of single citrus in sample.
Fruit coefficient k (n) is obtained by following calculation method:
Wherein, n is that single plant identifies fruit quantity, nsFor true single plant fruit quantity, fsFor correction value；
Fruit coefficient, that is, single plant identifies fruit quantity, is the quantity that citrus plant unilateral side outer layer detects, needs multiplied by fruit system
Column, approaching to reality single plant fruit quantity ns.It is grown in fruit tree outer layer according to most of well known citrus fruit, and is had substantially point
Cloth, internal layer fruit are less.Fruit coefficient can be according to true single plant fruit quantity nsFruit quantity linear relationship is identified with single plant
And correction value is added and subtracted, it can obtain fruit coefficient.
Claims (9)
1. a kind of method for quickly calculating orange yield by picture, which comprises the following steps:
S1: intelligent terminal obtains single plant citrus picture；
S2: the S1 single plant citrus picture obtained is identified by network model, returns to single plant identification fruit quantity n, k (m) tree
The tree-like coefficient of age coefficient, k (w)；
S3: citrus quantity is obtained by following formula:
P=n* { e[k(n)+k(m)+k(w)]-1}*[ln(x+1)]*y*z
Wherein, P is orange yield, and n is that single plant identifies fruit quantity, and k (n) is fruit coefficient, and k (m) is age of tree coefficient, and k (w) is
Tree-like coefficient, x are average single weight, and y is average strain number per acre, and z is mu number.
2. a kind of method for quickly calculating orange yield by picture according to claim 2, which is characterized in that further include
Following steps:
S0: establish n single plant identification fruit quantity, k (m) age of tree coefficient, the tree-like coefficient of k (w) network model.
3. a kind of method for quickly calculating orange yield by picture according to claim 1, which is characterized in that
Fruit coefficient k (n) is obtained by following calculation method:
Wherein, n is that single plant identifies fruit quantity, nsFor true single plant fruit quantity, fsFor correction value.
4. a kind of method for quickly calculating orange yield by picture according to claim 1, which is characterized in that
Age of tree coefficient k (m), wherein m >=1.5,0.85≤k (m)≤1.
5. a kind of method for quickly calculating orange yield by picture according to claim 1, which is characterized in that
Tree-like coefficient k (w), wherein 0.98≤k (w)≤1.12.
6. a kind of method for quickly calculating orange yield by picture according to claim 1, which is characterized in that
Average single citrus weight x is obtained by following formula:
Wherein, X is citrus sample size, and x is that single plant identifies fruit quantity, xiFor the weight of single citrus in sample.
7. a kind of method for quickly calculating orange yield by picture according to claim 2, which is characterized in that single plant is known
Specific step is as follows for the network model foundation of other fruit quantity n:
S0-1-1: different several photos containing citrus fruit are shot；
S0-1-2: the citrus fruit in picture obtained using labelimg annotation tool to S0-1-1 is labeled, and picture is outlined
In fruit；
S0-1-3: pre-processing the picture that S0-1-2 has been marked, including data enhancing and normalized；
S0-1-4: data are divided into training set and test set according to the ratio of 5:1 by the pretreated picture of S0-1-3；
S0-1-5: the data in training set extract the characteristic information of picture by faster rcnn deep neural network, then in advance
The location information of target, generates network parameter after successive ignition in mapping piece；
S0-1-6: the network parameter that data in test set generate S0-1-5 by faster rcnn deep neural network into
Row test, the location information for obtaining test picture is compared with test picture in the location information that S0-1-2 is marked, final logical
It crosses mean accuracy map and recall rate recall is assessed；
S0-1-7: repeating step S0-1-5 and S0-1-6, until map and recall reaches demand；
S0-1-8: final trained network parameter is loaded into faster rcnn network model.
8. a kind of method for quickly calculating orange yield by picture according to claim 2, which is characterized in that age of tree system
Specific step is as follows for the network model foundation of number k (m):
S0-2-1: several photos of the single plant citrus of the different specific age of tree sections of shooting, specific age of tree section includes: 1.5≤m < 2.5,
2.5≤m < 3.5,3.5≤m < 4.5,4.5≤y, and picture is divided into 4 classes according to age of tree section；
S0-2-2: pre-processing the step S0-2-1 picture classified, including data enhancing and normalized；
S0-2-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-2-2；
S0-2-4: the data in training set extract the characteristic information of picture by GoogleNet deep neural network, then predict
The affiliated classification of picture, generates network parameter after successive ignition；
S0-2-5: the data in test set carry out the network parameter that S0-2-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-2-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-2-6: repeating step S0-2-4 and S0-2-5, until recall and accuracy of identification reach demand；
S0-2-7: final trained network parameter is loaded into GoogleNet network model.
9. a kind of method for quickly calculating orange yield by picture according to claim 2, which is characterized in that tree-like system
Specific step is as follows for the network model foundation of number k (w):
S0-3-1: several photos of the single plant mandarin tree of the different specific tree-like w of shooting, specific tree-like w include: more major branch radiation
Shape, natural open centre shape, cylinder, trunk shape, natural globular model, and picture is divided into 5 classes according to tree-like；
S0-3-2: pre-processing the step S0-3-1 picture classified, including data enhancing and normalized；
S0-3-3: data are divided into training set and test set according to the ratio of 5:1 by the picture of different classifications in S0-3-2；
S0-3-4: the data in training set extract the characteristic information of picture by GoogleNet deep neural network, then predict
The affiliated classification of picture, generates network parameter after successive ignition；
S0-3-5: the data in test set carry out the network parameter that S0-3-4 is generated by GoogleNet deep neural network
Test, the affiliated classification for obtaining test picture are compared with picture affiliated classification in S0-3-1 is tested, final recall rate
Recall and accuracy of identification are assessed；
S0-3-6: repeating step S0-3-4 and S0-3-5, until recall and accuracy of identification reach demand；
S0-3-7: final trained network parameter is loaded into GoogleNet network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811328800.0A CN109658379A (en) | 2018-11-09 | 2018-11-09 | A method of orange yield is quickly calculated by picture |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811328800.0A CN109658379A (en) | 2018-11-09 | 2018-11-09 | A method of orange yield is quickly calculated by picture |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109658379A true CN109658379A (en) | 2019-04-19 |
Family
ID=66110730
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811328800.0A Pending CN109658379A (en) | 2018-11-09 | 2018-11-09 | A method of orange yield is quickly calculated by picture |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109658379A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428114A (en) * | 2019-08-12 | 2019-11-08 | 深圳前海微众银行股份有限公司 | Output of the fruit tree prediction technique, device, equipment and computer readable storage medium |
| CN110570408A (en) * | 2019-09-04 | 2019-12-13 | 南京大学 | System and method for counting fine targets on outer surface of cylinder |
| CN112364739A (en) * | 2020-10-31 | 2021-02-12 | 成都新潮传媒集团有限公司 | People counting method and device and computer readable storage medium |
| CN113920474A (en) * | 2021-10-28 | 2022-01-11 | 成都信息工程大学 | Internet of things system and method for intelligently monitoring citrus planting situation |
| CN114746866A (en) * | 2020-09-16 | 2022-07-12 | Ｆｓｘ株式会社 | Portable terminal and wet towel management system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104200193A (en) * | 2014-08-05 | 2014-12-10 | 北京农业信息技术研究中心 | Fruit tree yield estimation method and device |
| CN104616004A (en) * | 2015-03-08 | 2015-05-13 | 无锡桑尼安科技有限公司 | Citrus yield estimation method based on multi-estimation parameters |
| CN104881626A (en) * | 2015-01-19 | 2015-09-02 | 新疆农业大学 | Recognition method for fruit of fruit tree |
-
2018
- 2018-11-09 CN CN201811328800.0A patent/CN109658379A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104200193A (en) * | 2014-08-05 | 2014-12-10 | 北京农业信息技术研究中心 | Fruit tree yield estimation method and device |
| CN104881626A (en) * | 2015-01-19 | 2015-09-02 | 新疆农业大学 | Recognition method for fruit of fruit tree |
| CN104616004A (en) * | 2015-03-08 | 2015-05-13 | 无锡桑尼安科技有限公司 | Citrus yield estimation method based on multi-estimation parameters |
Non-Patent Citations (1)
| Title |
|---|
| 邱小兰: "《龙海市杨梅资源资产评估》", 《中国优秀硕士学位论文全文数据库农业科技辑》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110428114A (en) * | 2019-08-12 | 2019-11-08 | 深圳前海微众银行股份有限公司 | Output of the fruit tree prediction technique, device, equipment and computer readable storage medium |
| CN110570408A (en) * | 2019-09-04 | 2019-12-13 | 南京大学 | System and method for counting fine targets on outer surface of cylinder |
| CN110570408B (en) * | 2019-09-04 | 2022-04-22 | 南京大学 | System and method for counting fine targets on outer surface of cylinder |
| CN114746866A (en) * | 2020-09-16 | 2022-07-12 | Ｆｓｘ株式会社 | Portable terminal and wet towel management system |
| CN114746866B (en) * | 2020-09-16 | 2023-05-23 | Ｆｓｘ株式会社 | Portable terminal and wet towel management system |
| CN112364739A (en) * | 2020-10-31 | 2021-02-12 | 成都新潮传媒集团有限公司 | People counting method and device and computer readable storage medium |
| CN112364739B (en) * | 2020-10-31 | 2023-08-08 | 成都新潮传媒集团有限公司 | People counting method and device and computer readable storage medium |
| CN113920474A (en) * | 2021-10-28 | 2022-01-11 | 成都信息工程大学 | Internet of things system and method for intelligently monitoring citrus planting situation |
| CN113920474B (en) * | 2021-10-28 | 2024-04-30 | 成都信息工程大学 | Internet of things system and method for intelligently supervising citrus planting situation |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109658379A (en) | A method of orange yield is quickly calculated by picture | |
| WO2022160771A1 (en) | Method for classifying hyperspectral images on basis of adaptive multi-scale feature extraction model | |
| CN110222215B (en) | Crop pest detection method based on F-SSD-IV3 | |
| CN110009043B (en) | Disease and insect pest detection method based on deep convolutional neural network | |
| CN111369540B (en) | Plant leaf disease identification method based on mask convolutional neural network | |
| Yao et al. | Application of convolutional neural network in classification of high resolution agricultural remote sensing images | |
| CN111696101A (en) | Light-weight solanaceae disease identification method based on SE-Inception | |
| CN108898577A (en) | Based on the good malign lung nodules identification device and method for improving capsule network | |
| CN111984817B (en) | Fine-grained image retrieval method based on self-attention mechanism weighting | |
| CN113627282A (en) | Tea disease identification method based on deep migration learning | |
| CN115311316A (en) | Small watermelon identification and positioning method in three-dimensional cultivation mode based on deep learning | |
| CN116703932A (en) | CBAM-HRNet model wheat spike grain segmentation and counting method based on convolution attention mechanism | |
| CN116524279A (en) | Artificial intelligent image recognition crop growth condition analysis method for digital agriculture | |
| CN115861844A (en) | Rice early-stage remote sensing identification method based on planting probability | |
| CN110132865B (en) | Method for establishing Vis-NIR spectral depth characteristic model based on SAE-LSSVR crop cadmium content | |
| CN114708492A (en) | Fruit tree pest and disease damage image identification method | |
| CN110705698B (en) | Target counting depth network design method for scale self-adaptive perception | |
| CN117975278A (en) | Pest detection and identification method based on improved YOLOv model | |
| CN113627240A (en) | Unmanned aerial vehicle tree species identification method based on improved SSD learning model | |
| CN117392535A (en) | Fruit tree flower bud target detection and white point rate estimation method oriented to complex environment | |
| CN110555379B (en) | Human face pleasure degree estimation method capable of dynamically adjusting features according to gender | |
| CN116189175A (en) | Crop disease and pest classification and identification method based on InheretofectNet algorithm | |
| CN115019162B (en) | Silkworm detection method based on deep learning | |
| CN114049346B (en) | Citrus psyllid detection and identification method based on cutting YOLOv3-SPP3 | |
| CN113723482B (en) | Hyperspectral target detection method based on multi-example twin network |
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: 20190419 |
|
| RJ01 | Rejection of invention patent application after publication |