CN111325767B - Real scene-based citrus fruit tree image set synthesis method - Google Patents
Real scene-based citrus fruit tree image set synthesis method Download PDFInfo
- Publication number
- CN111325767B CN111325767B CN202010097186.2A CN202010097186A CN111325767B CN 111325767 B CN111325767 B CN 111325767B CN 202010097186 A CN202010097186 A CN 202010097186A CN 111325767 B CN111325767 B CN 111325767B
- Authority
- CN
- China
- Prior art keywords
- image
- fruit tree
- fruit
- image set
- steps
- 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.)
- Active
Links
Images
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
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
-
- 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
-
- 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
- G06T2207/20112—Image segmentation details
- G06T2207/20152—Watershed segmentation
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention belongs to the technical field of agricultural images, and particularly relates to a method for synthesizing an orange fruit tree image set based on a real scene, which comprises the steps of carrying out threshold segmentation processing on each image F in a fruit image set F by adopting a threshold segmentation method; performing threshold segmentation processing on each image T in the fruit tree image set T by adopting a mark control watershed method; making an image mask set containing known fruit numbers; and (3) synthesizing the fruit images and the fruit tree images based on the mask set. The invention synthesizes the fruit tree image set containing the fixed number of the fruits by using the processed simple images of the fruits and the fruit trees, saves the process of analyzing and labeling data, has high universality and higher application value.
Description
Technical Field
The invention belongs to the technical field of agricultural images, and particularly relates to a method for synthesizing an image set of citrus fruit trees based on a real scene.
Background
The citrus fruit tree estimation has gradually changed from the traditional modes of manual field sampling, visual inspection and counting of the number of single fruits and the like to the automatic identification combined with the image processing technology, and has developed from manual rough agriculture to intensive fine agriculture. When combining with the automatic identification of the image processing technology, the image of the single fruit tree needs to be acquired and an accurate test image set is established, which is a key step for establishing the linear relation between the image of the fruit tree and the number of fruits. The accurate image collection needs to be carried out by combining a specific orchard, so that the universality is poor.
Disclosure of Invention
The invention aims to solve the problems that in the prior art, more manpower and material resources are needed to be input in the process of collecting and processing an accurate image set, and the image set is required to be combined with a specific orchard for carrying out and has poor universality, and provides a real-scene-based citrus fruit tree image set synthesizing method.
In order to achieve the technical purpose, the invention adopts the following technical scheme:
the synthetic method of the citrus fruit tree image set based on the real scene is characterized by comprising the following steps of:
s1, carrying out threshold segmentation processing on each image F in a fruit image set F by adopting a threshold segmentation method;
s2, carrying out threshold segmentation processing on each image T in the fruit tree image set T by adopting a mark control watershed method;
s3, manufacturing an image mask set containing known fruit numbers;
and S4, synthesizing fruit tree graphics based on a mask set for the fruit image and the fruit tree image.
Further, the fruit image set F is a complete citrus image in the mature period, each image only contains one citrus, and the Number of images number_F is more than 50.
Further, the fruit tree image set T is a single citrus fruit tree image with complete growth period, each image contains a complete fruit tree crown, and the Number of images number_T is more than 50.
Further, the number of known fruits in the set of image masks M ranges from 1 to 40.
Further, in step S1, the threshold segmentation process for each image F in the set of fruit images F includes the steps of,
s11: converting the RGB color space into Lab color space;
s12: OSTU automatic threshold segmentation is carried out on the basis of Lab color space to obtain a fruit region ROI_F;
s13: the center position center_roi_f of the fruit region roi_f is calculated and saved.
Further, in step S2, the threshold segmentation processing of each image T in the fruit tree image set T comprises the following steps,
s21: marking the crown area with M_C;
s22: marking the background area with M_B;
s23: making a local minimum region VI of the image t based on the m_c and m_b marks;
s24: and obtaining a crown region ROI_T by using watershed transformation on VI.
Further, in step S3, the image mask fabrication for a given fruit number includes the steps of,
s31: setting the number N of fruits, randomly selecting N images from a fruit image set F to form an F-N image set for standby, and extracting the center position center_ROI_F of a fruit region ROI_F of the N images from the F-N image set;
s32: randomly selecting any image I in the fruit tree image set T, and acquiring a crown region ROI_T in the image I;
s33: n positions are randomly selected from the crown region ROI_T as marking positions center_mask_F of the fruit region ROI_F.
Further, in the step S4, the synthesis of the fruit tree image comprises the following steps,
s41: sequentially overlapping the center positions center_ROI_F of the fruit areas of N images in the F-N image set with the mark positions center_mask_F in the crown area ROI_T;
s42: and sequentially overlapping the ith image of the N images in the F-N image set with the image I according to the position of the center_mask_F to obtain a composite image.
Further, the synthesis of the fruit tree images is repeated for a plurality of times to obtain an image set, and the number of fruits in each image in the image set is known.
Compared with the prior art, the invention has the beneficial technical effects that:
the invention synthesizes the fruit tree image set containing the known fruit number by using the processed simple images of the fruits and the fruit trees, saves the process of analyzing and labeling data, has high universality and higher application value.
Drawings
FIG. 1 is a synthetic flow chart of a real scene based citrus fruit tree image set;
FIG. 2 is a flow chart of a process for an image of a citrus fruit;
FIG. 3 is a flow chart of the processing of an image of a citrus fruit tree;
fig. 4 is a flow chart of image synthesis of fruit tree based on an image mask.
Detailed Description
The invention will be further described with reference to specific examples, but the scope of the invention is not limited thereto.
As shown in fig. 1-4, the method for synthesizing the citrus fruit tree image set based on the real scene in the embodiment comprises the following steps:
s1, carrying out threshold segmentation processing on each image F in a fruit image set F by adopting a threshold segmentation method;
s2, carrying out threshold segmentation processing on each image T in the fruit tree image set T by adopting a mark control watershed method;
s3, manufacturing an image mask set containing known fruits.
And S4, synthesizing the fruit images and the fruit tree images based on the mask set so as to complete fruit tree estimation.
The fruit image set F is a complete orange image in the mature period, each image only contains one orange, and the Number of images number_F is more than 50. The fruit tree image set T is a complete single citrus fruit tree image in a growing period, each image must contain a complete fruit tree crown, and the Number of images number_T is more than 50. The larger the number of the fruit image sets and the fruit tree image sets is, the more the synthesized image sets can be ensured to simulate the real orchard scene. In order for the synthetic fruit image set to cover the case of fruits in a real orchard, the number of known fruits in the image mask set M ranges from 1 to 40.
Mature citrus fruits are bright in color and large in distinction from background images, and extraction of fruit areas is achieved by using automatic threshold segmentation. In step S1, a threshold segmentation process is performed for each image F in the set F of fruit images, including the steps of,
s11: converting the RGB color space into Lab color space;
s12: OSTU automatic threshold segmentation is carried out on the basis of Lab color space to obtain a fruit region ROI_F, wherein ROI_F= { (x_f, y_f) |a1 is not less than x and not more than b1, c1 is not less than y and not more than d1}, the Number of pixel points in the region is number_f, and a1, b1, c1 and d1 represent the boundary range of a fruit image;
s13: the center position center_roi_f of the fruit region roi_f is calculated and saved, wherein center_roi_f (x, y) =sum ({ (x_f, y_f) |a1.ltoreq.x.ltoreq.b1, c1.ltoreq.y.ltoreq.d1)/number_f.
The image background of the citrus fruit tree is complex, and the image background comprises areas such as the ground, fruit tree branch parts and the like, so that the difficulty in extracting the fruit tree crown parts is increased, and the method of marking and controlling the watershed is needed to realize segmentation. In the step S2, the threshold segmentation processing is carried out on each image T in the fruit tree image set T by adopting the following steps,
s21: marking the crown area with M_C;
s22: marking the background area with M_B;
s23: making a local minimum region VI of the image t based on the m_c and m_b marks;
s24: and transforming VI by using a watershed to obtain a crown region ROI_T, wherein the ROI_T= { (x, y) |a2 is not less than x and not more than b2, c2 is not less than y and not more than d2, and a2, b2, c2 and d2 represent the boundary range of the fruit tree image.
In step S3, the image mask production for a given number of fruits is performed by the following steps,
s31: setting the number N of fruits, randomly selecting N images from a fruit image set F to form an F-N image set for standby, extracting the center position center_ROI_F of a fruit region ROI_F of the N images from the F-N image set, wherein the center position center_ROI_F of the fruit region ROI_F is expressed as { x_f i ,y_f i I is equal to or more than 1 and is equal to or less than N;
s32: and randomly selecting any image I in the fruit tree image set T, and acquiring a crown region ROI_T in the image I, wherein ROI_T= { (x, y) |a2 is not less than x and not more than b2, and c2 is not less than y and not more than d 2.
S33: randomly selecting N positions in a crown region ROI_T as marking positions center_mask_F of a fruit region ROI_F, wherein center_mask_F= { x_M i ,y_M i },1≤i≤N,a≤x_M i ≤b,c≤y_M i ≤d。
In step S4, the composition of the fruit tree image based on the mask set includes the following steps,
s41: sequentially overlapping the center positions center_ROI_F of the fruit areas of N images in the F-N image set and the mark positions center_mask_F in the crown area ROI_T to make the image center positions { x_f in the F-N image set i ,y_f i Location { x_M } and center_mask_F set i ,y_M i Overlapping;
s42: sequentially setting the ith image of N images in the F-N image set according to the position { x_M of center_mask_F i ,y_M i And overlapping the image I in sequence to obtain a composite image, wherein I is more than or equal to 1 and less than or equal to N.
The synthesis of the fruit tree images is repeated for a plurality of times to obtain an image set, and the number of fruits in each image in the image set is known for fruit tree estimation.
While the embodiments of the present invention have been described in detail, those skilled in the art will appreciate that many modifications are possible in the specific embodiments, and that such modifications are intended to be within the scope of the present invention.
Claims (8)
1. The synthetic method of the citrus fruit tree image set based on the real scene is characterized by comprising the following steps of:
s1, carrying out threshold segmentation processing on each image F in a fruit image set F by adopting a threshold segmentation method;
s2, carrying out threshold segmentation processing on each image T in the fruit tree image set T by adopting a mark control watershed method;
s3, manufacturing an image mask set containing known fruit numbers;
s4, synthesizing fruit tree graphics based on a mask set on the fruit image and the fruit tree image;
step S3 comprises the steps of,
s31: setting the number N of fruits, randomly selecting N images from a fruit image set F to form an F-N image set for standby, and extracting the center position center_ROI_F of a fruit region ROI_F of the N images from the F-N image set;
s32: randomly selecting any image I in the fruit tree image set T, and acquiring a crown region ROI_T in the image I;
s33: n positions are randomly selected from the crown region ROI_T as marking positions center_mask_F of the fruit region ROI_F.
2. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: the fruit image set F is a complete orange image in the mature period, each image only contains one orange, and the Number number_F of the images is more than 50.
3. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: the fruit tree image set T is a complete single citrus fruit tree image in a growing period, each image comprises a complete fruit tree crown, and the Number of images number_T is more than 50.
4. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: the number of known fruits in the set of image masks ranges from 1 to 40.
5. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: in step S1, the threshold segmentation processing for each image F in the fruit image set F includes the steps of:
s11: converting the RGB color space into Lab color space;
s12: OSTU automatic threshold segmentation is carried out on the basis of Lab color space to obtain a fruit region ROI_F;
s13: the center position center_roi_f of the fruit region roi_f is calculated and saved.
6. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: in step S2, the threshold segmentation processing for each image T in the fruit tree image set T includes the following steps:
s21: marking the crown area with M_C;
s22: marking the background area with M_B;
s23: making a local minimum region VI of the image t based on the m_c and m_b marks;
s24: and obtaining a crown region ROI_T by using watershed transformation on VI.
7. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: in step S4, the synthesizing of the fruit tree image includes the following steps:
s41: sequentially overlapping the center positions center_ROI_F of the fruit areas of N images in the F-N image set with the mark positions center_mask_F in the crown area ROI_T;
s42: and sequentially overlapping the ith image of the N images in the F-N image set with the image I according to the position of the center_mask_F to obtain a composite image.
8. The method for synthesizing the real scene-based citrus fruit tree image set according to claim 1, wherein the method comprises the following steps of: the synthesis of the fruit tree images is repeated for a plurality of times to obtain an image set, and the number of fruits in each image in the image set is known.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010097186.2A CN111325767B (en) | 2020-02-17 | 2020-02-17 | Real scene-based citrus fruit tree image set synthesis method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010097186.2A CN111325767B (en) | 2020-02-17 | 2020-02-17 | Real scene-based citrus fruit tree image set synthesis method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111325767A CN111325767A (en) | 2020-06-23 |
CN111325767B true CN111325767B (en) | 2023-06-02 |
Family
ID=71168809
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010097186.2A Active CN111325767B (en) | 2020-02-17 | 2020-02-17 | Real scene-based citrus fruit tree image set synthesis method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111325767B (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011041924A1 (en) * | 2009-10-09 | 2011-04-14 | 江苏大学 | Device and method for identifying ripe oranges in nature scene by filter spectral image technology |
WO2019062846A1 (en) * | 2017-09-28 | 2019-04-04 | 北京西格码列顿信息技术有限公司 | Medical image aided diagnosis method and system combining image recognition and report editing |
CN109711317A (en) * | 2018-12-21 | 2019-05-03 | 重庆理工大学 | The segmentation recognition method of mature orange fruit and branches and leaves based on provincial characteristics |
CN109948444A (en) * | 2019-02-19 | 2019-06-28 | 重庆理工大学 | Method for synchronously recognizing, system and the robot of fruit and barrier based on CNN |
Family Cites Families (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8903167B2 (en) * | 2011-05-12 | 2014-12-02 | Microsoft Corporation | Synthesizing training samples for object recognition |
JP5762600B1 (en) * | 2014-06-06 | 2015-08-12 | キヤノン株式会社 | Information processing apparatus and information processing method |
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 |
US10242449B2 (en) * | 2017-01-04 | 2019-03-26 | Cisco Technology, Inc. | Automated generation of pre-labeled training data |
CN108664782B (en) * | 2017-03-28 | 2023-09-12 | 三星电子株式会社 | Face verification method and device |
CN108319973B (en) * | 2018-01-18 | 2022-01-28 | 仲恺农业工程学院 | Detection method for citrus fruits on tree |
JP2019212106A (en) * | 2018-06-06 | 2019-12-12 | 日本電信電話株式会社 | Area extraction model learning device, area extraction model learning method, and program |
CN109583509B (en) * | 2018-12-12 | 2020-11-03 | 南京旷云科技有限公司 | Data generation method and device and electronic equipment |
-
2020
- 2020-02-17 CN CN202010097186.2A patent/CN111325767B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2011041924A1 (en) * | 2009-10-09 | 2011-04-14 | 江苏大学 | Device and method for identifying ripe oranges in nature scene by filter spectral image technology |
WO2019062846A1 (en) * | 2017-09-28 | 2019-04-04 | 北京西格码列顿信息技术有限公司 | Medical image aided diagnosis method and system combining image recognition and report editing |
CN109711317A (en) * | 2018-12-21 | 2019-05-03 | 重庆理工大学 | The segmentation recognition method of mature orange fruit and branches and leaves based on provincial characteristics |
CN109948444A (en) * | 2019-02-19 | 2019-06-28 | 重庆理工大学 | Method for synchronously recognizing, system and the robot of fruit and barrier based on CNN |
Non-Patent Citations (1)
Title |
---|
基于形态学阈值标记分水岭算法的高分辨率影像单木树冠提取;郑鑫;王瑞瑞;靳茗茗;;中南林业调查规划(04);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN111325767A (en) | 2020-06-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109711325B (en) | Mango picking point identification method | |
CN111709489A (en) | Citrus identification method based on improved YOLOv4 | |
Liu et al. | A method of segmenting apples at night based on color and position information | |
CN108710840B (en) | Visual navigation path identification method for farmland pesticide spraying robot | |
CN108830844B (en) | Facility vegetable extraction method based on multi-temporal high-resolution remote sensing image | |
CN103324655B (en) | Image search system, image search apparatus, image search method and computer-readable storage medium | |
CN102663397B (en) | Automatic detection method of wheat seedling emergence | |
CN113191334B (en) | Plant canopy dense leaf counting method based on improved CenterNet | |
CN114612794B (en) | Remote sensing identification method for ground cover and planting structure of finely divided agricultural area | |
Bhookya et al. | Yield estimation of chilli crop using image processing techniques | |
CN112131952A (en) | Corn seedling stage plant number information extraction based on unmanned aerial vehicle remote sensing image | |
CN113255434B (en) | Apple identification method integrating fruit characteristics and deep convolutional neural network | |
CN102542560A (en) | Method for automatically detecting density of rice after transplantation | |
CN116310548A (en) | Method for detecting invasive plant seeds in imported seed products | |
CN103942555A (en) | Method for detecting nitrogen content of plant through images | |
CN111798470A (en) | Crop image entity segmentation method and system applied to intelligent agriculture | |
Setyawan et al. | Comparison of hsv and lab color spaces for hydroponic monitoring system | |
CN109300169B (en) | Semitransparent image color migration method based on linear transformation | |
CN111325767B (en) | Real scene-based citrus fruit tree image set synthesis method | |
CN109102457B (en) | Intelligent color changing system and method based on convolutional neural network | |
CN113077438B (en) | Cell nucleus region extraction method and imaging method for multi-cell nucleus color image | |
CN108176712A (en) | A kind of restorative procedure and system of alkaline land soil pollution | |
CN116883718A (en) | Sugarcane seedling missing detection positioning method based on improved YOLOV7 | |
CN110175582B (en) | Intelligent tea tree tender shoot identification method based on pixel distribution | |
CN111401121A (en) | Method for realizing citrus segmentation based on super-pixel feature extraction |
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 |