CN103793686B - A kind of method of output of the fruit tree early prediction - Google Patents
A kind of method of output of the fruit tree early prediction Download PDFInfo
- Publication number
- CN103793686B CN103793686B CN201410020528.5A CN201410020528A CN103793686B CN 103793686 B CN103793686 B CN 103793686B CN 201410020528 A CN201410020528 A CN 201410020528A CN 103793686 B CN103793686 B CN 103793686B
- Authority
- CN
- China
- Prior art keywords
- fruit
- region
- image
- area
- tree
- 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.)
- Expired - Fee Related
Links
Abstract
The invention provides a kind of method of output of the fruit tree early prediction, including step：S1：According to orchard management it needs to be determined that image acquisition time, in certain IMAQ condition, using portable image capture equipment it is determined that time collection treat the image of production forecast fruit tree；S2：Fruit region recognition is carried out according to feature of image；S3：Leaf region recognition is carried out according to feature of image；S4：The feature in fruit region and leaf region is extracted as top fruit sprayer feature；S5：Top fruit sprayer feature input artificial neural network is surveyed into production model to be predicted the yield of fruit tree.This method compensate for the simple deficiency being predicted using image procossing and identification technology, the yield of apple in orchard be predicted in early stage so as to accurate using effective combination of image procossing and identification technology and artificial intelligence technology.
Description
Technical field
The present invention relates to agricultura1 calculation technical field, specifically, is related to a kind of method of output of the fruit tree early prediction.
Background technology
Apple is one of edible widest fruit in the world, and world's apple annual production is about 32,000,000 tons.China is generation
More than the 40% of maximum apple production state and country of consumption in boundary, cultivated area of the apple and yield Jun Zhan worlds total amount, in the world
Occupy critical role in Apple Industry.It is required during arranging various harvests in order to estimate apple production before apple harvests
Various human and material resources resources, currently used method be it is artificial extract part fruit tree, count apple quantity, Ran Hou great by
Cause estimation yield.Such a manual method is time-consuming, laborious and precision is not high.In research to apple production early prediction, computer
The information processing technology turns into one of focus of current research as a kind of means.Due to the color, shape, Texture eigenvalue of apple
Different from the branches and leaves on tree, current research process mainly identifies apple using these feature combination image processing techniques,
Estimate yield.But to realize and carry out forecast production to set the accurate identification of upper apple solve overlapping apple and leaf to apple
The problem of blocking, it is a very big challenge to only rely on image processing techniques.
The content of the invention
In order to solve problems of the prior art, it is an object of the invention to provide output of the fruit tree early stage in a kind of orchard
The method of prediction, realize and apple production is more accurately predicted.
In order to realize the object of the invention, the present invention provides a kind of method of output of the fruit tree early prediction, including step：
S1：According to orchard management it needs to be determined that image acquisition time, in certain IMAQ condition, using portable image
Collecting device it is determined that time collection it is to be measured production fruit tree image；
S2：Fruit region recognition is carried out according to feature of image；
S3：Leaf region recognition is carried out according to feature of image；
S4：The feature in fruit region and leaf region is extracted as top fruit sprayer feature；
S5：Top fruit sprayer feature input artificial neural network is surveyed into production model to be predicted the yield of fruit tree.
Further, the step S1 shoots fruit tree one-sided image in specific fruit growth period, and the whole tree crown of fruit tree should
It is contained in taken the photograph image, places white curtain as background after tree during image taking.
Further, the step S2 includes step：
S21：Analyze color characteristic, shape facility and the textural characteristics of different objects in image.
S22：According to the analysis of step 1, the qualifications that fruit region can be identified from image are drawn；
S23：Qualifications in step 2 obtain fruit area image.
Further, the step S3 includes step：
S31：Fruit area image is subtracted from fruit tree image and obtains leaf region initial pictures；
S32：The background in the initial pictures of leaf region is removed, obtains leaf region without background image；
S33：Leaf region is removed without the limb in background image.
Further, the step S31 is specially：By the R values of pixel in fruit area image more than 0, in apple
Corresponding pixel R, G, B value is set to 0 in tree Image, obtains the initial pictures in leaf region；The step S32 is specially：Profit
With the G in RGB color model, B component, the G-B values of each pixel in image are calculated, the pixel that difference is less than 10 belongs to background
Region, background is removed with this；The step S33 is specially：RGB color is transformed into HSI spaces, the H values and branch of leaf
Dry H values have bigger difference, using OTSU automatic threshold segmentation algorithms, are partitioned into leaf region.
Further, the step S4 includes step：
S41：Calculate total image area；
S42：Count fruit areal；
S43：Calculate fruit region proportion；
S44：Calculate small area fruit region proportion；
S45：Calculate leaf region proportion；
S46：Set fruit growth period parameter value.
Wherein, in the step S41：Total pixel count of statistical picture, as its area.
Wherein, in the step S42：In fruit area image, several fruit region (one of fruit areas are included
The corresponding fruit in domain or a fruit cluster), the fruit areal in image is counted, comprised the following steps that：
Step 1：By fruit area image binaryzation；
Step 2：Using neighborhood method mark fruit region；
Step 3：Statistics obtains the total number in fruit region.
Wherein, in the step S43：In the bianry image of fruit region, non-zero number of pixels is counted, as fruit region
The gross area；The fruit region proportion=fruit region gross area/total image area.
Wherein, in the step S44：The small area fruit region proportion=small area fruit region gross area/total face of image
Product；The small area fruit region is the 0.2% fruit region that fruit region area is less than total image area.
Wherein, the step S45 is specially：By leaf area image binaryzation；In bianry image, non-zero pixel is counted
Number, as leaf region area；Leaf region proportion=leaf region area/total image area.
Wherein, fruit tree growth period parameters value surveys growth period in production model according to artificial neural network in the step S46
Arranges value is set.
Further, the step S5 using the tree crown feature extracted from fruit tree image as estimation output of the fruit tree according to
According to；With the number in fruit region, fruit region proportion, small area fruit region proportion, leaf region proportion and fruit growth
Period is the input of Artificial Neural Network Prediction Model, and the forecast production (kilogram/tree) of fruit tree is output.
The beneficial effects of the present invention are：
The invention provides apple early yield Forecasting Methodology in a kind of orchard, this method utilizes image procossing and identification skill
Effective combination of art and artificial intelligence technology, the simple deficiency being predicted using image procossing and identification technology is compensate for, from
And accurate the yield of apple in orchard can be predicted in early stage.This method, compare existing in current orchard
Forecasting Methodology, it is more objective, more efficiently.
Brief description of the drawings
Fig. 1 is that output of the fruit tree described in the embodiment of the present invention predicts flow chart；
Fig. 2 is leaf region recognition flow chart described in the embodiment of the present invention；
Fig. 3 is tree crown feature extraction flow chart described in the embodiment of the present invention；
Fig. 4 is apple early yield forecast model schematic diagram described in the embodiment of the present invention.
Embodiment
Following examples are used to illustrate the present invention, but are not limited to the scope of the present invention.
Embodiment 1
By taking loud, high-pitched sound apple tree in orchard as an example, to illustrate that process is produced in the survey of the present invention.Orchard is positioned at Germany in the present embodiment
Bonn University Klein Altendorf test orchard, and involved fruit tree is the apple tree of label 170 in embodiment, apple in orchard
Fruit tree north and south is embarked on journey, and fruit tree west is relatively sufficient by sunlight in the east.No. 170 apple trees are completed about in its physiological fallen fruit phase
After 1.5 months, survey production is carried out, process needs following steps：
S1：Obtain production fruit tree image to be measured
Step 1：Determine image acquisition time, IMAQ was in August in 2010 19 days in the present embodiment.
Step 2：When image obtains, strong illumination is avoided, places 2 meters of length after tree, high 1.5 off-white painting cloth curtain is as background.
Shooting angle is 1.2 meters of the height off the ground apart from 1.7 meters of fruit tree, and perpendicular to planting fruit trees direction.
Step 3：Using Cannon PowerShot SX110ISS digital cameras, fruit tree is shot in the case of auto-focusing
Image, picture size are 2112 × 2816.
S2：Fruit region recognition is handled in image, is comprised the following steps that：
Step 1：Original image is read in, reduces its size to 512 × 683.
Step 2：In RGB color, using cuing open different objects in the image mentioned in line-plot method analytical procedure two
The wave characteristic of each color component.
Step 3：According to the analysis of step 2, using color component difference analytic approach, the fruit region defined in image is drawn
Condition.Meet reddish blue difference R-B>40 and red green difference G-R<The pixel of 20 two conditions belongs to fruit region, non-fruit
R, G, B value of real area pixel are set to 0.
Step 4：Image is switched into gray level image, isolated spot noise is removed using 3 × 3 median filter.
Step 5：Opening operation processing is carried out to image, structural element uses disc structure, and its radius is 2, in smoothed image
The profile in each region, disconnect narrow connection.
Step 6：The apple region obtained after step 5 is handled is corresponded in original image, obtains the RGB figures in apple region
Picture.
S3：Leaf region recognition in image, its concrete scheme are as shown in Figure 2.
S31：By the R values of pixel in fruit area image more than 0, by corresponding pixel R, G, B in apple tree Image
Value is set to 0, obtains the initial pictures in leaf region.
S32：White background and the fluctuation of each color component of tree in the initial pictures of RGB color analysis leaf region
Situation, if the turquoise value of chromatism G-B of pixel<10, judge that pixel belongs to white background area, R, G, B value of pixel are set to 0 respectively,
So as to obtain leaf region without background image.
S33：Conversion formula using RGB color to HIS color spaces, leaf region is obtained without background image
HIS schemes.
S34：Analyze the fluctuation and histogram of the H components of image, I component and S components.
S35：According to the analysis of step 4, using OTSU automatic threshold methods, the H components of segmentation figure picture, can divide the image into
Two parts：Trunk and branch region and leaf region, by trunk and branch in picture of the leaf region without the part in background image
R, G, B value of plain value put O.
S4：The feature of apple tree tree crown is extracted, its concrete scheme is as shown in Figure 3.
S41：Total number of pixels in statistical picture, as the gross area of image, its size is：512 × 683=
349696。
S42：In fruit area image, comprising several apple regions, the corresponding apple in one of apple region
Or an apple cluster, the fruit areal in image is counted, comprised the following steps that：
Step 1：By fruit area image binaryzation, binary-state threshold is arranged to 0；
Step 2：The object in bianry image is marked using 8 neighborhood methods；
Step 3：The number for counting tagged object is 65, its total number as fruit region.
S43：In the bianry image of fruit region, it is 9408 to count non-zero number of pixels, i.e. the fruit region gross area is
9408.The fruit region proportion=total image area 349696=0.0269 of the fruit region gross area 9408/.
S44：The summation of the area in the apple region that an apple is only corresponded in image is calculated, is comprised the following steps that：
Step 1：The ≈ 700 of total image area 349696 × 0.2% is as small less than 700 region in apple area image
Area apple region；
Step 2：The area that each object in image is marked in S42 is calculated, area of the area less than 700 is added up,
Obtain the gross area 7785 in small area apple region.
Step 3：The ≈ of 7785/ total image area of the gross area 349696 in fruit region proportion=small area fruit region
0.0223。
S45：The gross area in leaf region is calculated, is comprised the following steps that：
Step 1：By leaf area image binaryzation, binary-state threshold is arranged to 0.
Step 2：In bianry image, non-zero number of pixels 216673 is counted, i.e. leaf region area is 216673.
Step 3：The leaf region proportion=≈ 0.6196 of leaf region 216673/ total image area of the gross area 349696.
S46：Fruit growth period parameter value is set to 0.5.
S5：Production model is surveyed using BP neural network apple early stage and carries out survey production, as shown in Figure 4.
By fruit areal 65, fruit region proportion 0.0269, small area fruit region proportion 0.0223, leaf region
Proportion 0.6196 and the input model of fruit growth period 0.5, it is 22.09, the i.e. forecast production to this apple tree to obtain output
For 22.09 kilograms.Wherein BP neural network apple early yield forecast model can predict the yield in three periods of fruit tree：Fruit tree
At the end of physiological fallen fruit, after fruit physiology shedding terminates 1.5 months, 2 weeks before fruit harvest.Join corresponding to three fruit tree growth phases
Numerical value is respectively 0,0.5,1；Image of the model according to 150 fruit trees in orchard in 2009, three periods herein, with every fruit
Set fruit areal in each period image, fruit region proportion, small area fruit region proportion, leaf region proportion with
And fruit growth period is input, it is trained and is obtained with the fruit tree actual production of every (kilogram/tree) target.The model is
Three-layer neural network structure (input layer, hidden layer, output layer), 15 neurons of input layer, output layer neuron, intermediate layer
Neuron number 11.Intermediate layer neural transferring function uses tangent sigmoid function, and output layer neural transferring function is using linear
Function.
Although above the present invention is described in detail with a general description of the specific embodiments,
On the basis of the present invention, it can be made some modifications or improvements, this will be apparent to those skilled in the art.Cause
This, these modifications or improvements, belong to the scope of protection of present invention without departing from theon the basis of the spirit of the present invention.
Claims (8)
- A kind of 1. method of output of the fruit tree early prediction, it is characterised in that including step：S1：According to orchard management it needs to be determined that image acquisition time, in certain IMAQ condition, using portable image capture Equipment it is determined that time collection it is to be measured production fruit tree image；The acquisition condition should be contained in by the whole tree crown of fruit tree and take the photograph figure As in, white curtain is placed as background after tree during image taking；S2：Fruit region recognition is carried out according to feature of image；S3：Leaf region recognition is carried out according to feature of image；S4：The feature in fruit region and leaf region is extracted as top fruit sprayer feature：S41：Calculate total image area；S42：Count fruit areal；S43：Calculate fruit region proportion；S44：Calculate small area fruit region proportion；The small area fruit region is that fruit region area is less than total image area 0.2% fruit region；S45：Calculate leaf region proportion；S46：Set fruit growth period parameter value；S5：Top fruit sprayer feature input artificial neural network is surveyed into production model to be predicted the yield of fruit tree.
- 2. according to the method for claim 1, it is characterised in that the step S3 includes step：S31：Fruit area image is subtracted from fruit tree image and obtains leaf region initial pictures；S32：The background in the initial pictures of leaf region is removed, obtains leaf region without background image；S33：Leaf region is removed without the limb in background image.
- 3. according to the method for claim 2, it is characterised in that the step S31 is specially：By picture in fruit area image The R values of element are more than 0, and corresponding pixel R, G, B value is set into 0 in the image of fruit tree, obtains the initial pictures in leaf region； The step S32 is specially：Using the G in RGB color model, B component, the G-B values of each pixel, difference in calculating image Pixel less than 10 belongs to background area, and background is removed with this；The step S33 is specially：RGB color is transformed into HSI spaces, for the H values and the H values of limb of leaf, using OTSU automatic threshold segmentation algorithms, it is partitioned into leaf region.
- 4. according to the method for claim 1, it is characterised in that in the step S43：Fruit region proportion=fruit region The gross area/total image area.
- 5. according to the method for claim 1, it is characterised in that in the step S44：Small area fruit region proportion=small The area fruit region gross area/total image area；The small area fruit region is that fruit region area is less than total image area 0.2% fruit region.
- 6. according to the method for claim 1, it is characterised in that the step S45 is specially：By leaf area image two-value Change；In bianry image, non-zero number of pixels is counted, as leaf region area；Leaf region proportion=leaf region area/ Total image area.
- 7. according to the method for claim 1, it is characterised in that fruit tree growth period parameters value is according to people in the step S46 The arranges value that artificial neural networks survey growth period in production model is set.
- 8. according to the method for claim 1, it is characterised in that the step S5 is special by the tree crown extracted from fruit tree image Levy the foundation as estimation output of the fruit tree；With the number in fruit region, fruit region proportion, small area fruit region proportion, tree Leaf region proportion and the input that fruit growth period is Artificial Neural Network Prediction Model, the forecast production of fruit tree is output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410020528.5A CN103793686B (en) | 2014-01-16 | 2014-01-16 | A kind of method of output of the fruit tree early prediction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410020528.5A CN103793686B (en) | 2014-01-16 | 2014-01-16 | A kind of method of output of the fruit tree early prediction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103793686A CN103793686A (en) | 2014-05-14 |
| CN103793686B true CN103793686B (en) | 2017-12-15 |
Family
ID=50669331
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410020528.5A Expired - Fee Related CN103793686B (en) | 2014-01-16 | 2014-01-16 | A kind of method of output of the fruit tree early prediction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103793686B (en) |
Families Citing this family (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104881626B (en) * | 2015-01-19 | 2017-12-29 | 新疆农业大学 | The recognition methods of Tree Fruit |
| EP3529771A1 (en) * | 2016-10-19 | 2019-08-28 | BASF Agro Trademarks GmbH | Determining the grain weight of an ear |
| CN106570770B (en) * | 2016-11-11 | 2020-12-18 | 山东农业大学 | Orchard water and fertilizer integrated topdressing amount accurate estimation method |
| CN107832655A (en) * | 2017-02-21 | 2018-03-23 | 石河子大学 | A kind of take photo by plane system and output of cotton estimating and measuring method based on unmanned plane imaging near the ground |
| CN107092891A (en) * | 2017-04-25 | 2017-08-25 | 无锡中科智能农业发展有限责任公司 | A kind of paddy rice yield estimation system and method based on machine vision technique |
| WO2019097892A1 (en) | 2017-11-15 | 2019-05-23 | Ｎｅｃソリューションイノベータ株式会社 | Device for collecting breeding data in farm field, device for analyzing feature in breeding, method for collecting breeding data in farm field, program, and recording medium |
| CN108053078A (en) * | 2017-12-28 | 2018-05-18 | 深圳春沐源控股有限公司 | A kind of production prediction method, server and computer readable storage medium |
| JP7144779B2 (en) | 2018-03-23 | 2022-09-30 | Ｎｅｃソリューションイノベータ株式会社 | Tree canopy identification device, identification method, program, and recording medium |
| CN109937733B (en) * | 2019-03-28 | 2021-04-06 | 北京农业智能装备技术研究中心 | Automatic measuring device of orchard output |
| CN111311573B (en) * | 2020-02-12 | 2024-01-30 | 贵州理工学院 | Branch determination method and device and electronic equipment |
| CN113673279A (en) * | 2020-05-14 | 2021-11-19 | 明谷农业生技股份有限公司 | Plant growth identification method and system |
| CN114155526B (en) * | 2021-11-09 | 2024-04-16 | 中国农业大学 | Tomato fruit growth prediction method, device, equipment and product |
| CN115238964A (en) * | 2022-06-28 | 2022-10-25 | 安徽未来种业有限公司 | Rice harvest prediction evaluation method and system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102954816A (en) * | 2012-01-13 | 2013-03-06 | 北京盈胜泰科技术有限公司 | Crop growth monitoring method |
| CN103310218A (en) * | 2013-05-21 | 2013-09-18 | 常州大学 | Precise recognition method of overlapped shielded fruits |
-
2014
- 2014-01-16 CN CN201410020528.5A patent/CN103793686B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102954816A (en) * | 2012-01-13 | 2013-03-06 | 北京盈胜泰科技术有限公司 | Crop growth monitoring method |
| CN103310218A (en) * | 2013-05-21 | 2013-09-18 | 常州大学 | Precise recognition method of overlapped shielded fruits |
Non-Patent Citations (6)
| Title |
|---|
| Application of Neural Networks and Image Visualization for Early Forecast of Apple Yield;Rozman Črtomir等;《Erwerbs-Obstbau》;20120614;第54卷(第2期);第70页右栏第4段，第71页左栏第1段、右栏第2段，第72页右栏第2段、第3段 * |
| Estimation of mango crop yield using image analysis – Segmentation;A.B. Payne等;《Computers and Electronics in Agriculture》;20130201;第91卷(第2期);全文 * |
| Modelling Apple Fruit Yield Using Image Analysis for Fruit Colour,;D. Stajnko等;《European Journal of Horticultural Science 》;20091231;第74卷(第6期);全文 * |
| Using colour features of cv. ‘Gala’ apple fruits in an orchard in image processing to predict yield;Rong Zhou等;《Precision Agriculture》;20120609;第13卷(第5期);全文 * |
| 基于matlab的苹果树枝图像分割方法研究;李莹莹等;《科学技术与工程》;20120508;第12卷(第13期);第2节第1段，第3.2节，第3.3节第1-2段，图4 * |
| 自然场景下成熟水果的计算机视觉识别;蔡健荣等;《农业机械学报》;20050225;第36卷(第2期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103793686A (en) | 2014-05-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103793686B (en) | A kind of method of output of the fruit tree early prediction | |
| Wu et al. | Using channel pruning-based YOLO v4 deep learning algorithm for the real-time and accurate detection of apple flowers in natural environments | |
| Stajnko et al. | Estimation of number and diameter of apple fruits in an orchard during the growing season by thermal imaging | |
| Payne et al. | Estimation of mango crop yield using image analysis–segmentation method | |
| Aquino et al. | Automated early yield prediction in vineyards from on-the-go image acquisition | |
| Bai et al. | Crop segmentation from images by morphology modeling in the CIE L* a* b* color space | |
| RU2735151C2 (en) | Weeds identification in natural environment | |
| Hernández-Hernández et al. | Optimal color space selection method for plant/soil segmentation in agriculture | |
| Nuske et al. | Automated visual yield estimation in vineyards | |
| Diago et al. | Assessment of flower number per inflorescence in grapevine by image analysis under field conditions | |
| Liu et al. | A method of segmenting apples at night based on color and position information | |
| Zhou et al. | Strawberry maturity classification from UAV and near-ground imaging using deep learning | |
| Zhou et al. | Using colour features of cv.‘Gala’apple fruits in an orchard in image processing to predict yield | |
| Skovsen et al. | The GrassClover image dataset for semantic and hierarchical species understanding in agriculture | |
| CN105718945B (en) | Apple picking robot night image recognition method based on watershed and neural network | |
| Sarkate et al. | Application of computer vision and color image segmentation for yield prediction precision | |
| Zhou et al. | Early detection and continuous quantization of plant disease using template matching and support vector machine algorithms | |
| CN110427798B (en) | Intelligent green vegetable identification and leaf environment change prediction method | |
| Lv et al. | A visual identification method for the apple growth forms in the orchard | |
| CN111199192A (en) | Method for detecting integral maturity of field red globe grapes by adopting parallel line sampling | |
| Revathi et al. | Homogenous segmentation based edge detection techniques for proficient identification of the cotton leaf spot diseases | |
| Shin et al. | Postharvest citrus mass and size estimation using a logistic classification model and a watershed algorithm | |
| Chen et al. | A surface defect detection system for golden diamond pineapple based on CycleGAN and YOLOv4 | |
| Gao et al. | Recognition and Detection of Greenhouse Tomatoes in Complex Environment. | |
| Victorino et al. | Grapevine yield prediction using image analysis-improving the estimation of non-visible bunches |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20171215 Termination date: 20190116 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |