CN109919088B - Automatic extraction method for identifying individual plants of pitaya in karst region - Google Patents
Automatic extraction method for identifying individual plants of pitaya in karst region Download PDFInfo
- Publication number
- CN109919088B CN109919088B CN201910168451.9A CN201910168451A CN109919088B CN 109919088 B CN109919088 B CN 109919088B CN 201910168451 A CN201910168451 A CN 201910168451A CN 109919088 B CN109919088 B CN 109919088B
- Authority
- CN
- China
- Prior art keywords
- plants
- image
- separated
- dragon fruit
- area
- 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
Landscapes
- Image Processing (AREA)
Abstract
The invention discloses an automatic extraction method for identifying single plants of pitaya in karst regions, which comprises the following steps: (1) photographing to obtain a dragon fruit aerial photo image; (2) processing the image to obtain an orthophoto map of the measurement area; (3) preprocessing the image: (4) and (3) carrying out wave band calculation: (4) segmenting the image to obtain a threshold value; (5) obtaining a vector patch of the dragon fruit in the measuring area by segmenting a target ground object and a background value through a threshold value; (6) judging whether the pattern spots are separated; (7) counting the number of the single plants which are separated thoroughly and calculating the number of the separated connected plants; (8) and adding the number of the plants which are completely separated to the number of the connected plants to obtain the total number of the single plants in the whole measuring area. Compared with the traditional satellite remote sensing mode, the extraction method of the single dragon fruit plant can greatly improve the image calculation accuracy of the single dragon fruit plant in the karst region, realize automatic extraction and batch processing of the single dragon fruit plant, has high processing efficiency and greatly lightens the labor intensity.
Description
Technical Field
The invention belongs to the technical field of single plant identification of pitaya, and relates to an automatic extraction method for single plant identification of pitaya in a karst region.
Background
In order to relieve the contradiction between the shortage of agricultural resources and the huge environmental pressure in the karst plateau canyon region, the concept of developing the unlimited value of typical economic crops in a limited space is inherited, and the future agricultural utilization must be changed from 'extensive' to 'fine'. However, the application of the existing novel remote sensing technology for supporting precision agriculture mainly focuses on the identification and classification of ground objects. Therefore, the research and development combines special landform and landform of Guizhou to select an unmanned aerial vehicle platform to carry a visible light lens to obtain a high-resolution image, the high-resolution image only containing R, G, B three wave bands is used for carrying out wave band calculation and division to obtain a target ground object, a concept of dividing a plant cluster by the average area of plants is provided to combine with a visual programming space modeling tool model builder, and a dragon fruit single plant automatic extraction batch processing and precision verification model is built.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: provides an automatic extraction method for identifying single plants of pitaya in karst regions, and aims to solve the technical problems in the prior art.
The technical scheme adopted by the invention is as follows: a karst region dragon fruit single plant identification automatic extraction method comprises the following steps:
(1) carrying a visible light lens by using an unmanned aerial vehicle platform to plan an automatic route, setting the navigational height, the course and the lateral overlapping degree of the unmanned aerial vehicle to be 40m, 70% and 60% respectively, and taking a picture to obtain a navigation sheet;
(2) importing the aerial photo by using image splicing software, completing aerial triangle measurement calculation to generate aerial photography point cloud data, and acquiring a digital orthophoto map of a measurement area;
(3) correcting deformation, distortion, blurring and noise caused by the shaking of the unmanned aerial vehicle in the process of obtaining the aerial photo, and carrying out image enhancement, finishing, cutting and reconstruction on the aerial photo to obtain a high-resolution image of a measuring area;
(4) the method comprises the following steps of carrying out wave band calculation on the obtained high-resolution image, wherein the calculation method comprises the following steps: float (2 × b2-b1-b3/2 × b2+ b1+ b3), wherein float represents that the calculation result is a floating point type, and b1, b2 and b3 respectively represent red, green and blue three bands;
(5) dividing the image into a background part and a target part according to the gray characteristic of the image by adopting a maximum class variance method, automatically extracting and segmenting a threshold value by utilizing IDL interactive language programming, and when the threshold value T enables the class variance between the target and the background to be maximum, the threshold value T is the optimal threshold value for segmenting the target ground object;
(6) dividing the target ground object and the background value into two different image layers by using the extracted threshold value to obtain a vector patch of the pitaya in the measuring area;
(7) calculating the area of each small patch of the pattern spot by using a geometric tool of ArcGIS software, and removing a background and a fragmentary pattern spot to obtain a vector patch only containing a target ground object;
(8) judging whether the pattern spots are separated or not, counting the patches which are thoroughly separated to obtain the number of single plants of the dragon fruit, and adopting a method for dividing the connected plant clusters by the average area of the single plants of the dragon fruit which are not thoroughly separated, wherein the method comprises the following steps of: dividing the area of each connected patch by the average area of the single dragon fruit plant to obtain the number of the separated connected patches, and converting the number of the separated connected patches into an integer by using a set rule to obtain the number of the separated single plant plants of the connected plants;
(9) adding the number of the plants which are separated thoroughly to the number of the separated connected plants to obtain the total number of the single plants in the whole measuring area, and obtaining the statistical total number of the single plant separation of the target ground object in the measuring area;
(10) the human-computer interaction field verification obtains the verification accuracy of the actual single plant number of the measuring area, and the verification method comprises the following steps
In the formula, rho represents the accuracy, M represents the number of extracted plants, and N represents the actual number of plants.
The maximum class variance method in the step (5) comprises the following steps:
N 0 +N 1 ＝MN (9)
W 0 +W 1 ＝1 (10)
μ＝W 0 μ 0 +W 1 μ 1 (11)
σ＝W 0 (μ 0 -μ) 2 +W 1 (μ 1 -μ) 2 (12)
in the formula: w 0 Is the proportion of the target pixel points to the whole scene image, W 1 Is the ratio of background image element points to the whole scene image, mu 0 Is the average gray level of the target ground object pixel, mu 1 Is the average gray level of background pixels, mu is the total average gray level of the image, MN represents the image size, N 0 The number of pixels with gray scale less than T, N 1 The number of pixels with the gray scale larger than T is shown, and sigma is the inter-class variance.
The invention has the beneficial effects that: compared with the prior art, the invention has the following effects:
1) compared with the traditional satellite remote sensing mode, the extraction method of the single dragon fruit plant can greatly improve the image calculation accuracy of the single dragon fruit plant in the karst region, realize automatic extraction and batch processing of the single dragon fruit plant, has high processing efficiency and greatly reduces the labor intensity;
2) the high-definition aerial films are obtained by respectively setting the aerial height, the course and the lateral overlapping degree of the unmanned aerial vehicle to be 40m, 70% and 60%, so that the defect that the image obtaining precision of the traditional satellite in the cloudy rainy days in the karst region (such as Guizhou) is insufficient is overcome, and the purpose of obtaining the high-resolution image as required is achieved;
3) the gray scale image of the image is obtained through wave band calculation, and the vegetation index constructed by only containing red, salary and blue wave bands is used for calculation, so that the cost is lower compared with that of multispectral and hyperspectral;
4) the method is used for dividing the conjoined plants, and the conjoined plants are divided into single plants more completely. The advantages are that: the idea of dividing the plant cluster by the average area of the plant is put forward for the first time, and a certain effect is obtained through experiments;
5) the extraction result obtained by the method can provide a certain reference value for the growth monitoring, yield estimation and plant morphological information acquisition of the dragon fruits, further development of the value of the dragon fruits and accurate agricultural service in karst mountain areas.
Drawings
FIG. 1 is a technical route flow diagram;
FIG. 2 is a model diagram of automatic extraction and accuracy verification of single dragon fruit plant;
FIG. 3 is a visible light image of the unmanned aerial vehicle in the test area;
FIG. 4 is a graph of band calculation results;
FIG. 5 is a diagram showing the effect of extraction of pitaya.
Detailed Description
The invention is further described with reference to the following figures and specific embodiments.
Example 1: as shown in fig. 1-5, an automatic extraction method for identifying single plants of pitaya in karst regions includes the following steps:
(1) carrying a visible light lens by using an unmanned aerial vehicle platform to plan an automatic route, and setting the navigational height, the course and the lateral overlapping degree to be 40m, 70% and 60% respectively to obtain a navigation sheet;
(2) importing the aerial photo by using image splicing software, completing aerial triangle measurement calculation to generate aerial photography point cloud data, and acquiring a digital orthophoto map of a measurement area;
(3) correcting deformation, distortion, blurring and noise caused by the shaking of the unmanned aerial vehicle in the process of obtaining the aerial photo, and carrying out image enhancement, finishing, cutting and reconstruction on the aerial photo to obtain a high-resolution image of a measuring area;
(4) the obtained high-resolution image is subjected to wave Band calculation, the software adopts mainstream remote sensing image processing software ENVI, the tool adopts Band Math, and the calculation method comprises the following steps: float (2 × b2-b1-b3/2 × b2+ b1+ b3), wherein float represents that the calculation result is a floating point type, and b1, b2 and b3 respectively represent red, green and blue three-wave bands;
(5) dividing the image into a background part and a target part according to the gray characteristic of the image by a maximum class variance method (OTSU), and automatically extracting a segmentation threshold value by using IDL (interface description language) interactive language programming, wherein the core idea is that when the threshold value T enables the class variance between the target and the background to be maximum, the threshold value T is the optimal threshold value for segmenting the target ground object, the obtained segmentation threshold value of the image is 0.037, and the calculation speed and the accuracy are high;
the maximum class variance method comprises the following steps:
N 0 +N 1 ＝MN (9)
W 0 +W 1 ＝1 (10)
μ＝W 0 μ 0 +W 1 μ 1 (11)
σ＝W 0 (μ 0 -μ) 2 +W 1 (μ 1 -μ) 2 (12)
in the formula: w is a group of 0 Is the proportion of the target image element points to the whole scene image, W 1 Is the proportion of background image element points to the whole image, mu 0 Is the average gray level of the target ground object pixel, mu 1 Is the average gray level of background pixels, mu is the total average gray level of the image, MN represents the image size, N 0 The number of pixels with gray scale less than T, N 1 The number of pixels with the gray scale larger than T is, and sigma is the variance between classes;
(6) dividing the target ground object and the background value into two different image layers by using the extracted threshold value to obtain a vector patch of the pitaya in the measuring area, so that the target ground object can be conveniently and quickly identified;
(7) calculating the area of each small patch of pattern spot by using a geometric tool of ArcGIS software, removing background and fragmentary pattern spots to obtain a vector patch only containing a target ground object, and obtaining a completely formed vector layer of a pure target area;
(8) judging whether the pattern spots are separated: counting the patches which are thoroughly separated to obtain the number of the individual plants of the dragon fruits, putting forward an idea of dividing the connected plant clusters by the average area of the individual plants of the dragon fruits which are thoroughly separated to the plant clusters which are not thoroughly separated together, specifically, dividing the average area of the individual plants of the dragon fruits by the area of each connected patch to obtain the number of the separated individual patches which is a double-precision type, and converting the number into an integer type by using a set rule (regular according to a rounding mode) to obtain the number of the separated individual plants of the connected plants;
(9) adding the number of the plants which are separated thoroughly to the number of the plants which are separated into connected plants to obtain the total number of the single plants in the whole measuring area;
(10) the human-computer interaction field verification obtains the verification accuracy of the actual single plant number of the measuring area, and the verification method comprises the following steps
In the formula, ρ represents the accuracy, M represents the number of extracted plants, and N represents the actual number of plants. Wherein (7) - (12) are integrated in the model builder to realize automatic batch processing;
(11) and (3) evaluation of extraction results:
TABLE 1 extraction of statistics of characteristic values of each index
As shown in table 1, the total number of the automatically extracted plants is 320, the actual number of the plants obtained by the human-computer interaction field verification is 295, the number of the automatically extracted plants is 25 more than the actual number of the plants, the extraction precision is 91.7% by substituting the precision verification formula, the error rate is 8.3%, the reasons for the multiple extraction and the error score are mainly caused by the shadow of the conjoined plants and the interference of part of weeds, and the area of the conjoined plants is wrongly classified as the target ground object.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and therefore, the scope of the present invention should be subject to the protection scope of the appended claims.
Claims (2)
1. A karst region dragon fruit single plant identification automatic extraction method is characterized by comprising the following steps: the method comprises the following steps:
(1) carrying a visible light lens by using an unmanned aerial vehicle platform to plan an automatic route, setting the navigational height, the course and the lateral overlapping degree of the unmanned aerial vehicle to be 40m, 70% and 60% respectively, and taking a picture to obtain a navigation film for planting a dragon fruit measuring area;
(2) importing the aerial photo by using image splicing software, completing aerial triangle measurement calculation to generate aerial photography point cloud data, and acquiring a digital orthophoto map of a measurement area;
(3) correcting deformation, distortion, blurring and noise caused by the shaking of the unmanned aerial vehicle in the process of obtaining the aerial photo, and carrying out image enhancement, finishing, cutting and reconstruction on the aerial photo to obtain a high-resolution image of a measuring area;
(4) the method comprises the following steps of carrying out wave band calculation on the obtained high-resolution image, wherein the calculation method comprises the following steps: float (2 × b2-b1-b3/2 × b2+ b1+ b3), wherein float represents that the calculation result is a floating point type, and b1, b2 and b3 respectively represent red, green and blue three-wave bands;
(5) dividing the image into a background part and a target part according to the gray characteristic of the image by adopting a maximum class variance method, and automatically extracting a segmentation threshold value by utilizing IDL interactive language programming, wherein when the threshold value T enables the class variance between the target and the background to be maximum, the threshold value T is the optimal threshold value for segmenting the target ground object;
(6) dividing the target ground object and the background value into two different image layers by using the extracted threshold value to obtain a vector patch of the pitaya in the measuring area;
(7) calculating the area of each small patch of the pattern spot by using a geometric tool of ArcGIS software, and removing a background and a fragmentary pattern spot to obtain a vector patch only containing a target ground object;
(8) judging whether the pattern spots are separated or not, counting the patches which are thoroughly separated to obtain the number of single plants of the dragon fruit, and adopting a method for dividing the connected plant clusters by the average area of the single plants of the dragon fruit which are not thoroughly separated, wherein the method comprises the following steps of: dividing the area of each connected patch by the average area of the single dragon fruit plant to obtain the number of the separated connected patches, and converting the number of the separated connected patches into an integer by using a set rule to obtain the number of the separated single plant plants of the connected plants;
(9) adding the number of the plants which are separated thoroughly to the number of the plants which are separated into connected plants to obtain the total number of the single plants in the whole measuring area;
(10) the human-computer interaction field verification obtains the verification accuracy of the actual single plant number of the measuring area, and the verification method comprises the following steps
In the formula, ρ represents the accuracy, M represents the number of extracted plants, and N represents the actual number of plants.
2. The karst region dragon fruit single plant identification automatic extraction method as claimed in claim 1, characterized in that: the maximum class variance method in the step (5) comprises the following steps:
N 0 +N 1 ＝MN (9)
W 0 +W 1 ＝1 (10)
μ＝W 0 μ 0 +W 1 μ 1 (11)
σ＝W 0 (μ 0 -μ) 2 +W 1 (μ 1 -μ) 2 (12)
in the formula: w is a group of 0 Is the proportion of the target image element points to the whole scene image, W 1 Is the ratio of background image element points to the whole scene image, mu 0 Is the average gray level of the target ground object pixel, mu 1 Is the average gray level of background pixels, mu is the total average gray level of the image, MN represents the image size, N 0 The number of pixels with gray scale less than T, N 1 The number of pixels with the gray scale larger than T is shown, and sigma is the inter-class variance.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910168451.9A CN109919088B (en) | 2019-03-06 | 2019-03-06 | Automatic extraction method for identifying individual plants of pitaya in karst region |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910168451.9A CN109919088B (en) | 2019-03-06 | 2019-03-06 | Automatic extraction method for identifying individual plants of pitaya in karst region |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109919088A CN109919088A (en) | 2019-06-21 |
| CN109919088B true CN109919088B (en) | 2022-09-23 |
Family
ID=66963498
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910168451.9A Active CN109919088B (en) | 2019-03-06 | 2019-03-06 | Automatic extraction method for identifying individual plants of pitaya in karst region |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109919088B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110598619B (en) * | 2019-09-06 | 2023-04-07 | 中国农业科学院农业资源与农业区划研究所 | Method and system for identifying and counting fruit trees by using unmanned aerial vehicle images |
| CN111798433B (en) * | 2020-07-08 | 2024-04-26 | 贵州师范大学 | Method for identifying and counting mature dragon fruits in mountain area of plateau based on unmanned aerial vehicle remote sensing |
| CN113554675A (en) * | 2021-07-19 | 2021-10-26 | 贵州师范大学 | Edible fungus yield estimation method based on unmanned aerial vehicle visible light remote sensing |
| CN114926374B (en) * | 2022-07-21 | 2022-10-11 | 四川新迎顺信息技术股份有限公司 | Image processing method, device and equipment based on AI and readable storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017077543A1 (en) * | 2015-11-08 | 2017-05-11 | Agrowing Ltd | A method for aerial imagery acquisition and analysis |
| CN108022245A (en) * | 2017-12-06 | 2018-05-11 | 南京师范大学 | Photovoltaic panel template automatic generation method based on upper thread primitive correlation model |
-
2019
- 2019-03-06 CN CN201910168451.9A patent/CN109919088B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017077543A1 (en) * | 2015-11-08 | 2017-05-11 | Agrowing Ltd | A method for aerial imagery acquisition and analysis |
| CN108369635A (en) * | 2015-11-08 | 2018-08-03 | 阿格洛英公司 | The method with analysis is obtained for aerial image |
| CN108022245A (en) * | 2017-12-06 | 2018-05-11 | 南京师范大学 | Photovoltaic panel template automatic generation method based on upper thread primitive correlation model |
Non-Patent Citations (1)
| Title |
|---|
| 基于无人机遥感技术的玉米种植信息提取方法研究;韩文霆等;《农业机械学报》;20170125(第01期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109919088A (en) | 2019-06-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109919088B (en) | Automatic extraction method for identifying individual plants of pitaya in karst region | |
| CN109711325B (en) | Mango picking point identification method | |
| CN106384081B (en) | Slope farmland extraction method and system based on high-resolution remote sensing image | |
| CN111340826B (en) | Aerial image single tree crown segmentation algorithm based on super pixels and topological features | |
| CN109389163B (en) | Unmanned aerial vehicle image classification system and method based on topographic map | |
| CN109919083A (en) | A kind of early stage automation winter wheat drafting method based on Sentinel-2 image data | |
| Wang et al. | Side-view apple flower mapping using edge-based fully convolutional networks for variable rate chemical thinning | |
| CN109299673A (en) | The green degree spatial extraction method of group of cities and medium | |
| CN112989985B (en) | Urban built-up area extraction method integrating night light data and Landsat8OLI images | |
| CN114241326B (en) | Progressive intelligent production method and system for ground feature elements of remote sensing images | |
| CN112163639B (en) | Crop lodging grading method based on height distribution feature vector | |
| CN106875407A (en) | A kind of unmanned plane image crown canopy dividing method of combining form and marking of control | |
| CN116309670A (en) | Bush coverage measuring method based on unmanned aerial vehicle | |
| US20240290089A1 (en) | Method for extracting forest parameters of wetland with high canopy density based on consumer-grade uav image | |
| Zheng et al. | Single shot multibox detector for urban plantation single tree detection and location with high-resolution remote sensing imagery | |
| CN115115954A (en) | Intelligent identification method for pine nematode disease area color-changing standing trees based on unmanned aerial vehicle remote sensing | |
| Huang et al. | Recognition and counting of pitaya trees in karst mountain environment based on unmanned aerial vehicle RGB images | |
| CN115294482B (en) | Edible fungus yield estimation method based on unmanned aerial vehicle remote sensing image | |
| Fu et al. | Automatic detection tree crown and height using Mask R-CNN based on unmanned aerial vehicles images for biomass mapping | |
| Grigillo et al. | Classification based building detection from GeoEye-1 images | |
| CN115358991A (en) | Method and system for identifying seedling leaking quantity and position of seedlings | |
| CN112287787B (en) | Crop lodging grading method based on gradient histogram characteristics | |
| CN113554675A (en) | Edible fungus yield estimation method based on unmanned aerial vehicle visible light remote sensing | |
| CN114627064A (en) | Sparse vegetation regionalization extraction method based on two-dimensional image | |
| CN115294472A (en) | Fruit yield estimation method, model training method, equipment and storage medium |
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 |