CN113349188B - Lawn and forage precise weeding method based on cloud weeding spectrum - Google Patents
Lawn and forage precise weeding method based on cloud weeding spectrum Download PDFInfo
- Publication number
- CN113349188B CN113349188B CN202110603279.2A CN202110603279A CN113349188B CN 113349188 B CN113349188 B CN 113349188B CN 202110603279 A CN202110603279 A CN 202110603279A CN 113349188 B CN113349188 B CN 113349188B
- Authority
- CN
- China
- Prior art keywords
- weeding
- herbicide
- cloud
- weed
- spectrum
- 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
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01M—CATCHING, TRAPPING OR SCARING OF ANIMALS; APPARATUS FOR THE DESTRUCTION OF NOXIOUS ANIMALS OR NOXIOUS PLANTS
- A01M21/00—Apparatus for the destruction of unwanted vegetation, e.g. weeds
- A01M21/04—Apparatus for destruction by steam, chemicals, burning, or electricity
- A01M21/043—Apparatus for destruction by steam, chemicals, burning, or electricity by chemicals
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01G—HORTICULTURE; CULTIVATION OF VEGETABLES, FLOWERS, RICE, FRUIT, VINES, HOPS OR SEAWEED; FORESTRY; WATERING
- A01G13/00—Protecting plants
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01G—HORTICULTURE; CULTIVATION OF VEGETABLES, FLOWERS, RICE, FRUIT, VINES, HOPS OR SEAWEED; FORESTRY; WATERING
- A01G20/00—Cultivation of turf, lawn or the like; Apparatus or methods therefor
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01M—CATCHING, TRAPPING OR SCARING OF ANIMALS; APPARATUS FOR THE DESTRUCTION OF NOXIOUS ANIMALS OR NOXIOUS PLANTS
- A01M7/00—Special adaptations or arrangements of liquid-spraying apparatus for purposes covered by this subclass
- A01M7/0025—Mechanical sprayers
-
- A—HUMAN NECESSITIES
- A01—AGRICULTURE; FORESTRY; ANIMAL HUSBANDRY; HUNTING; TRAPPING; FISHING
- A01M—CATCHING, TRAPPING OR SCARING OF ANIMALS; APPARATUS FOR THE DESTRUCTION OF NOXIOUS ANIMALS OR NOXIOUS PLANTS
- A01M7/00—Special adaptations or arrangements of liquid-spraying apparatus for purposes covered by this subclass
- A01M7/0089—Regulating or controlling systems
Abstract
The invention relates to a precise lawn and pasture weeding method based on a cloud weeding spectrum, which comprises the following steps of: establishing a deep learning model according to a herbicide weed control spectrum, and identifying weed species based on the herbicide weed control spectrum; deploying the trained neural network model to a cloud end, and performing interaction between the cloud end and the weeding robot by using a 5G network; an intelligent spraying decision system capable of supporting multitask parallel is constructed, and early warning and manual intervention mechanisms are achieved through multiple terminals. The weed control spectrum based weed control spectrum recognition and spraying can effectively save the herbicide, the weeding effect is more accurate and efficient, the performance of weed recognition can be effectively improved by means of the computing power of the cloud large-scale server and the high-speed data transmission of the 5G network, and a multi-task scene is processed at the same time.
Description
Technical Field
The invention relates to the field of accurate weeding of lawns and pastures, in particular to an accurate weeding method of lawns and pastures based on a cloud weeding spectrum.
Background
Lawns are commonly found in public facilities such as park greening and stadiums. The prevention and control of lawn weeds are the key content of lawn maintenance. In China, due to the attack of weeds, the lawn is usually degraded or even wasted within a few years. For pasture, because the growth of the pasture in the seedling stage is slow, the weeding is more important, and the control of weeds seriously affects the quality of the pasture. Chemical herbicide-based weeding is a common way of controlling grass weeds in lawns. The accurate spraying of the herbicide can effectively reduce the dosage of the herbicide and reduce the environmental pollution. Weed identification is a prerequisite and key to accurate weeding. In recent years, artificial intelligence technology related to deep learning is gradually applied to the field of weed recognition, a weed recognition model is trained through a deep learning convolutional neural network, and the weed recognition model is deployed to a terminal chip and loaded into a weeding robot. The existing weed identification model and the deployment mode have the following problems:
1. only weeds are simply identified through deep learning, and no direct correlation between weed species and herbicide species is established. Different herbicides have different herbicidal spectra and weeds are tolerant differently to different herbicides. The herbicide spraying without distinguishing the weed control spectrum not only wastes the herbicide, but also is usually useless.
2. The single machine deployment mode is limited by the computing power of the deployed terminal, and the algorithm performance is difficult to further improve.
3. The updating operation of the recognition model is cumbersome. Cooperation of the chip, the weeding robot and the user is required to achieve redeployment.
4. Difficulty in manual intervention in the weeding process; data for weed identification and results of spraying were not intuitive to collect and display media.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a method for accurately weeding lawn and pasture based on a cloud weeding spectrum. The main contents are as follows:
a lawn and forage grass precise weeding method based on a cloud weeding spectrum is characterized by comprising the following steps: establishing a deep learning model according to a herbicide weed control spectrum, and identifying weed species based on the herbicide weed control spectrum; deploying the trained neural network model to a cloud end, and performing interaction between the cloud end and the weeding robot by using a 5G network; an intelligent spraying decision-making system capable of supporting multitask parallel is built, and devices related to the intelligent spraying decision-making system comprise a weeding robot, a cloud server and a user console; the weeding robot is responsible for collecting lawn and pasture field pictures, transmitting the pictures to the cloud server, receiving a weed identification result and a spraying instruction of the server, and finishing accurate weeding operation by using the loaded herbicide; a user logs in a user console to perform checking or setting operation; the cloud server receives images uploaded by the weeding robots, weed identification is completed, spraying instructions are output, massive weed data are collected and sorted for big data application, and early warning and manual intervention mechanisms are achieved through multiple terminals.
Further, a herbicide is selected, and then a neural network model is trained and built according to a herbicide weed control spectrum, wherein the trained and built model only identifies weeds sensitive to the herbicide, but not weeds tolerant to the herbicide.
Further, the training of the neural network model only includes the images of the weeds sensitive to the herbicide and the images of the weeds tolerant to the herbicide are classified as true classes, and the images of the weeds are not included in the neural network model or are included in the model and classified together with the crops as true negative classes.
Furthermore, after model training is completed, the weed control robot is deployed to the cloud, the weed control robot collects images and uploads the images to the cloud through a network to identify weeds, the cloud neural network model identifies weed species, and information is transmitted back to the weed control robot through the network after corresponding herbicides and weed positions are identified.
Furthermore, the weed identification performance is improved by means of the computing power of a cloud large-scale server and the high-speed data transmission of a 5G network; the updating of the deep learning model is only carried out at the cloud end, a weeding robot or manual intervention is not needed, and the user feels nothing in the updating process.
Furthermore, the cloud end has the capability of processing multi-task scenes simultaneously, and covers any number of weeding scenes in any area in a one-to-many mode; the cloud intelligent spraying decision-making system utilizes the user console to perform multi-terminal control interaction and real-time information interconnection with the user according to the task scene.
Preferably, the weeding modes include two modes, the first mode is used for setting the type of the herbicide loaded by the current working weeding robot for the user, and when weeds are identified at the cloud, only the weed control spectrum neural network model corresponding to the herbicide is used for identifying the weeds, namely only the weeds sensitive to the herbicide are identified; and in the second mode, the user does not specify the herbicide, the cloud traverses each model to identify the weed type and output corresponding herbicide information, and after the weeding robot receives the herbicide information, if the robot loads the herbicide, spraying operation is directly implemented, otherwise, early warning information is sent out, and the user is reminded to load the corresponding herbicide through the network terminal.
As a preferred scheme, the early warning information further comprises weeding schedule early warning, and if the weeding period is reached or a certain time is set between the weeding period and the weeding period according to the prediction, the early warning is sent to remind the user of weeding.
Compared with the prior art, the invention has the following advantages:
1. the invention can effectively save the herbicide based on the identification and spraying of the weed control spectrum, and has more accurate and efficient weeding effect.
2. By means of the computing power of the cloud large-scale server and the high-speed data transmission of the 5G network, the weed identification performance can be effectively improved.
3. The deep learning model is updated only at the cloud end without intervention of a weeding robot or manpower, and a user feels no sense in the updating process.
4. The invention can cover weeding scenes in any number in any area in a one-to-many way. The cloud has the capability of processing multi-task scenes simultaneously.
5. The cloud intelligent spraying decision-making system can perform multi-terminal control interaction and real-time information interconnection with a user by using the user console according to a task scene.
6. The cloud end can collect massive weed image data, and is convenient for application of a big data technology.
Drawings
FIG. 1: schematic diagram of weed control spectrum neural network model.
FIG. 2: weed identification flow chart.
FIG. 3: cloud intelligence spraying decision-making system schematic diagram.
Detailed Description
The above-mentioned contents of the present invention are further described in detail by way of examples below, but it should not be understood that the scope of the above-mentioned subject matter of the present invention is limited to the following examples, and any technique realized based on the above-mentioned contents of the present invention falls within the scope of the present invention.
STEP1 training and establishing neural network model according to herbicide weed-killing spectrum
And training and establishing a neural network model according to the herbicide weed control spectrum. For example, the sulfonylurea herbicides have a narrow weed control spectrum, and only weed images sensitive to the sulfonylurea herbicides can be included and classified as true when the neural network model training is carried out. Conversely, images of weeds that are tolerant to sulfonylurea herbicides may be included in the neural network model or included in the model but categorized with the crop as a true negative. The model thus trained and established only identified weeds that were sensitive to the herbicide, and not identified weeds that were tolerant to the herbicide.
Specifically, assuming that weeds a, B, and C are sensitive to herbicide 1, the neural network model 1 performs sample training only on these weeds. Thus, a plurality of neural network models are established according to the weed control spectrum, and each model only contains the weed species with the effective corresponding herbicide. A schematic diagram of a neural network model established and trained according to a herbicide weed control spectrum is shown in figure 1, STEP2, model deployment, image acquisition of a weeding robot and cloud identification
After model training is completed, the weed control robot is deployed to the cloud, the weed control robot collects images and uploads the images to the cloud through a 5G network to perform weed identification, the cloud neural network model identifies weed species, and information is transmitted back to the weed control robot through the 5G network after corresponding herbicides and weed positions are determined. A weed identification flow chart is shown in figure 2.
STEP3 cloud intelligent spraying decision-making system based on neural network model
Based on a neural network model deployed at the cloud end, an intelligent spraying decision system capable of supporting multitask parallel is constructed. The cloud intelligent spray decision system is shown in fig. 3. The device that intelligence spraying decision-making system related to includes weeding robot, high in the clouds server, user console.
Weeding robot: the weeding robot is responsible for collecting pictures of lawn and pasture fields and transmitting the pictures to the cloud server. Meanwhile, a server weed identification result and a spraying instruction are received, and the loaded herbicide is used for finishing accurate weeding operation.
Cloud server: and receiving images uploaded by the weeding robots, completing weed identification and outputting spraying instructions. Meanwhile, massive weed data are collected and sorted for big data application (such as research on weed species and coverage rate in various regions, growth condition prediction, updating and optimizing of model training data and the like).
A user console: the user can log in a Web page or a mobile phone App to enter the console, and after logging in, the user can do the following operations:
1. checking weeding work data of weeding robot (comprising running track, weed and spraying area, weeding schedule)
2. Setting a timed weeding plan
3. Setting a weeding mode
4. Receive cloud information (e.g., model update information, herbicide recommendations, etc.)
5. Receiving early warning information
The specific contents of the points 3 and 5 are as follows:
the weeding modes (point 3) comprise two modes, wherein the first mode is used for setting the types of the herbicides loaded by the current working weeding robot for the user, and when the cloud side identifies the weeds, only the weed control spectrum neural network model corresponding to the herbicides is used for identifying the weeds, namely only the weeds sensitive to the herbicides are identified. And in the second mode, the user does not specify the herbicide, the cloud traverses each model to identify the weed type and output corresponding herbicide information, after the weeding robot receives the herbicide information, if the robot loads the herbicide, spraying operation is directly implemented, otherwise, early warning (point 5) is sent, and the user is reminded to load the corresponding herbicide through a WEB or APP terminal. In addition, the early warning information (point 5) also comprises weeding schedule early warning, and if the weeding period is reached or the interval is a certain time according to the prediction, the user still does not carry out weeding work, the early warning is sent out to remind the user to carry out weeding work.
The above description is only a preferred embodiment of the present invention, and should not be taken as limiting the invention in any way, and any person skilled in the art can make any simple modification, equivalent replacement, and improvement on the above embodiment without departing from the technical spirit of the present invention, and still fall within the protection scope of the technical solution of the present invention.
Claims (4)
1. A lawn and forage grass precise weeding method based on a cloud weeding spectrum is characterized by comprising the following steps: establishing a deep learning model according to a herbicide weed control spectrum, and identifying weed species based on the herbicide weed control spectrum; deploying the trained neural network model to a cloud end, and performing interaction between the cloud end and the weeding robot by using a 5G network; an intelligent spraying decision-making system capable of supporting multitask parallel is built, and devices related to the intelligent spraying decision-making system comprise a weeding robot, a cloud server and a user console; the weeding robot is responsible for collecting lawn and pasture field pictures, transmitting the pictures to the cloud server, receiving a weed identification result and a spraying instruction of the server, and finishing accurate weeding operation by using the loaded herbicide; a user logs in a user console to perform checking or setting operation; the cloud server receives images uploaded by the weeding robots, weed identification is completed, spraying instructions are output, meanwhile, massive weed data are collected and sorted for big data application, and early warning and manual intervention mechanisms are achieved through multiple terminals;
after the model training is finished, deploying the model to the cloud end, acquiring an image by the weeding robot, uploading the image to the cloud end through a network for weed identification, identifying the weed species by the cloud end neural network model, and returning information to the weeding robot through the network after the corresponding herbicide and weed position are identified;
the weed identification performance is improved by means of computing power of a cloud large-scale server and high-speed data transmission of a 5G network; the updating of the deep learning model is only carried out at the cloud, a weeding robot or manual intervention is not needed, and the user feels nothing in the updating process;
the cloud end has the capability of simultaneously processing multi-task scenes, and one-to-many covers any number of weeding scenes in any area; the cloud intelligent spraying decision-making system performs multi-terminal control interaction and real-time information interconnection with a user by using a user console according to a task scene;
the weeding modes comprise two modes, wherein the first mode is used for setting the type of the herbicide loaded by the current working weeding robot for a user, and when weeds are identified at the cloud, only the weed control spectrum neural network model corresponding to the herbicide is used for identifying the weeds, namely only the weeds sensitive to the herbicide are identified; and in the second mode, the user does not specify the herbicide, the cloud traverses each model to identify the weed type and output corresponding herbicide information, and after the weeding robot receives the herbicide information, if the robot loads the herbicide, spraying operation is directly implemented, otherwise, early warning information is sent out, and the user is reminded to load the corresponding herbicide through the network terminal.
2. The cloud weeding spectrum-based lawn and pasture precise weeding method according to claim 1, comprising the following steps of: selecting a herbicide, and then training and building a neural network model according to a herbicide weed control spectrum, wherein the trained and built model only identifies weeds sensitive to the herbicide, but not weeds tolerant to the herbicide.
3. The cloud weeding spectrum-based lawn and pasture precise weeding method according to claim 1, comprising the following steps of: the neural network model training only includes the weed images sensitive to the herbicide and is classified into a true class, and the weed images tolerant to the herbicide are not included in the neural network model or are included in the model and classified with the crops as a true negative class.
4. The cloud weeding spectrum-based lawn and pasture precise weeding method according to claim 1, comprising the following steps of: the early warning information also comprises weeding schedule early warning, and if the weeding period is reached or a certain time is left according to the prediction, the user still does not carry out weeding work, the early warning is sent out to remind the user to carry out weeding work.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110603279.2A CN113349188B (en) | 2021-05-31 | 2021-05-31 | Lawn and forage precise weeding method based on cloud weeding spectrum |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110603279.2A CN113349188B (en) | 2021-05-31 | 2021-05-31 | Lawn and forage precise weeding method based on cloud weeding spectrum |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113349188A CN113349188A (en) | 2021-09-07 |
| CN113349188B true CN113349188B (en) | 2022-06-03 |
Family
ID=77530670
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110603279.2A Active CN113349188B (en) | 2021-05-31 | 2021-05-31 | Lawn and forage precise weeding method based on cloud weeding spectrum |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113349188B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114467900B (en) * | 2022-02-17 | 2022-12-27 | 南京林业大学 | Accurate spraying method of lawn herbicide |
| CN115251024B (en) * | 2022-08-29 | 2023-11-21 | 北京大学现代农业研究院 | Determination method and device of weeding mode, electronic equipment and weeding system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859375A (en) * | 2010-06-01 | 2010-10-13 | 南京林业大学 | Method for recognizing inline crops and weeds of seedling stage in farmland |
| CN105230224A (en) * | 2015-11-06 | 2016-01-13 | 嘉兴职业技术学院 | Intelligent weeding machine and weed removing method thereof |
| CN109961024A (en) * | 2019-03-08 | 2019-07-02 | 武汉大学 | Wheat weeds in field detection method based on deep learning |
| CN209359439U (en) * | 2018-12-27 | 2019-09-10 | 深圳宇龙机器人科技有限公司 | Automatic weed-eradicating robot |
| CN110826556A (en) * | 2019-10-15 | 2020-02-21 | 遂昌濠畅电子科技有限公司 | Broad-spectrum crop weed identification and positioning method for improving deep learning |
| JP2020529686A (en) * | 2017-10-27 | 2020-10-08 | 全北大学 産学協力団Ｉｎｄｕｓｔｒｉａｌ Ｃｏｏｐｅｒａｔｉｏｎ Ｆｏｕｎｄａｔｉｏｎ Ｃｈｏｎｂｕｋ Ｎａｔｉｏｎａｌ Ｕｎｉｖｅｒｓｉｔｙ | Devices and methods for identifying crops and weeds by learning neural networks |
| CN112699729A (en) * | 2020-12-01 | 2021-04-23 | 北京麦飞科技有限公司 | Unmanned aerial vehicle investigation and attack integrated weeding method |
-
2021
- 2021-05-31 CN CN202110603279.2A patent/CN113349188B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101859375A (en) * | 2010-06-01 | 2010-10-13 | 南京林业大学 | Method for recognizing inline crops and weeds of seedling stage in farmland |
| CN105230224A (en) * | 2015-11-06 | 2016-01-13 | 嘉兴职业技术学院 | Intelligent weeding machine and weed removing method thereof |
| JP2020529686A (en) * | 2017-10-27 | 2020-10-08 | 全北大学 産学協力団Ｉｎｄｕｓｔｒｉａｌ Ｃｏｏｐｅｒａｔｉｏｎ Ｆｏｕｎｄａｔｉｏｎ Ｃｈｏｎｂｕｋ Ｎａｔｉｏｎａｌ Ｕｎｉｖｅｒｓｉｔｙ | Devices and methods for identifying crops and weeds by learning neural networks |
| CN209359439U (en) * | 2018-12-27 | 2019-09-10 | 深圳宇龙机器人科技有限公司 | Automatic weed-eradicating robot |
| CN109961024A (en) * | 2019-03-08 | 2019-07-02 | 武汉大学 | Wheat weeds in field detection method based on deep learning |
| CN110826556A (en) * | 2019-10-15 | 2020-02-21 | 遂昌濠畅电子科技有限公司 | Broad-spectrum crop weed identification and positioning method for improving deep learning |
| CN112699729A (en) * | 2020-12-01 | 2021-04-23 | 北京麦飞科技有限公司 | Unmanned aerial vehicle investigation and attack integrated weeding method |
Non-Patent Citations (3)
| Title |
|---|
| 基于深度学习的杂草识别系统;尚建伟等;《软件导刊》;20200715(第07期);全文 * |
| 草坪杂草化学防除的研究进展;高德武等;《四川草原》;20011230(第04期);全文 * |
| 麦草光的杀草谱和对杂草群落种群演替影响与对策;邱学林等;《青海农林科技》;19981220(第04期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113349188A (en) | 2021-09-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Chamoso et al. | UAVs applied to the counting and monitoring of animals | |
| Sujaritha et al. | Weed detecting robot in sugarcane fields using fuzzy real time classifier | |
| Brunelli et al. | Energy neutral machine learning based iot device for pest detection in precision agriculture | |
| CN113349188B (en) | Lawn and forage precise weeding method based on cloud weeding spectrum | |
| Martinoli | Swarm intelligence in autonomous collective robotics: From tools to the analysis and synthesis of distributed control strategies | |
| CN107463958A (en) | Insect identifies method for early warning and system | |
| CN105843147A (en) | Smart agriculture monitoring and management system | |
| Majeed et al. | Development and performance evaluation of a machine vision system and an integrated prototype for automated green shoot thinning in vineyards | |
| CN109040302A (en) | Vacant lot information on soil moisture acquisition system | |
| CN114384826A (en) | Intelligent agricultural monitoring system based on AI simulation | |
| CN115604301A (en) | Planting environment monitoring system based on artificial intelligence | |
| CN111727457A (en) | Cotton crop row detection method and device based on computer vision and storage medium | |
| CN115661650A (en) | Farm management system based on data monitoring of Internet of things | |
| CN115390504A (en) | Wisdom farming system based on 5G thing networking | |
| CN107392398A (en) | A kind of agricultural management method, mist calculating platform and system | |
| CN115238237A (en) | Insect pest situation monitoring method, device and system | |
| CN113377141A (en) | Artificial intelligence agricultural automatic management system | |
| CN114545833B (en) | Intelligent interactive processing system of facility agriculture based on internet of things | |
| Venkatesh et al. | eagri: Smart agriculture monitoring scheme using machine learning strategies | |
| CN109343533A (en) | A kind of robot system and its control method of golf course lawn maintenance | |
| CN115115275A (en) | AI-based unattended crop disease and pest monitoring and early warning system | |
| CN112362102A (en) | Monitoring system for immediately monitoring agricultural production environment | |
| CN112019607A (en) | Intelligent farm comprehensive control method and system based on visual cloud platform | |
| CN208297367U (en) | A kind of diseases and pests of agronomic crop information detector based on Machine Vision Recognition | |
| CN116295662B (en) | Crop growth state monitoring method and device, electronic equipment and 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 |