CN113468964B - Hyperspectrum-based agricultural disease and pest monitoring method and device - Google Patents
Hyperspectrum-based agricultural disease and pest monitoring method and device Download PDFInfo
- Publication number
- CN113468964B CN113468964B CN202110600387.4A CN202110600387A CN113468964B CN 113468964 B CN113468964 B CN 113468964B CN 202110600387 A CN202110600387 A CN 202110600387A CN 113468964 B CN113468964 B CN 113468964B
- Authority
- CN
- China
- Prior art keywords
- pest
- sample
- damaged
- hyperspectral image
- detected
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The embodiment of the application discloses a hyperspectral agricultural pest monitoring method and device. The method is used for solving the problem that the prior art is difficult to accurately monitor the crop diseases and insect pests. Determining a position coordinate set of each small area to be detected; acquiring a characteristic wave band sensitive to plant diseases and insect pests corresponding to a region to be detected; collecting a hyperspectral image of a region of the crop to be measured in a characteristic wave band, and marking longitude and latitude coordinates corresponding to the hyperspectral image; determining a small area to be detected corresponding to the hyperspectral image through longitude and latitude coordinates; counting the number of damaged leaves in a small area to be detected, and when the number of the damaged leaves is larger than a first preset threshold value, inputting the hyperspectral image into a pest type identification neural network model and a pest grade estimation model to obtain the pest type and the pest grade of the crop to be detected; and monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade. By the method, the accuracy of monitoring the crop diseases and insect pests is improved.
Description
Technical Field
The application relates to the field of agricultural disease and pest monitoring, in particular to a hyperspectral agricultural disease and pest monitoring method and device.
Background
China is a traditional big agricultural country, mainly carries out agricultural cultivation for a long time, and is not a strong agricultural country in terms of agricultural technology. At present, agricultural production in China still mainly adopts a traditional production mode, and agricultural management such as fertilization, irrigation, insect killing and pest killing is mostly carried out by means of experience, so that huge waste of manpower and financial resources is caused.
Especially, the crop diseases and insect pests have the characteristics of multiple types, great influence and frequent outbreak and disaster. The mode of agricultural management is carried out by experience, so that the crops are difficult to be accurately monitored in the early period of insect damage, and large-area outbreak of the insect pests is easily caused. In addition, the prior art has inaccurate monitoring on pest types and pest flooding degrees, so that the problems of mismatching of pest types and pesticide types and more or less pesticide dosage can be caused frequently, the operation effect of precision agriculture is seriously influenced, and the agricultural yield is difficult to improve.
Disclosure of Invention
The embodiment of the application provides an agricultural disease and pest monitoring method and device based on hyperspectrum, and is used for solving the following technical problems: the prior art is difficult to accurately monitor crop diseases and insect pests.
The embodiment of the application adopts the following technical scheme:
the embodiment of the application provides an agricultural disease and pest monitoring method based on hyperspectrum. Dividing the region of the crop to be detected into a plurality of small regions to be detected, and determining a position coordinate set of each small region to be detected; according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model, obtaining a characteristic wave band sensitive to plant diseases and insect pests; collecting a hyperspectral image of a region of the crop to be measured in a characteristic waveband through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and marking longitude and latitude coordinates corresponding to the hyperspectral image; comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set to determine a small area to be detected corresponding to the hyperspectral image; counting the number of damaged leaves in the small area to be detected according to the hyperspectral image corresponding to the small area to be detected, and inputting the hyperspectral image corresponding to the small area to be detected into a pest type identification neural network model and a pest grade estimation model under the condition that the number of the damaged leaves is larger than a first preset threshold value so as to obtain the pest type and the pest grade of the crop to be detected; and monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade.
According to the embodiment of the application, the areas to be detected are divided, so that pest and disease damage monitoring can be performed on the different small areas to be detected after division, and therefore different degrees of medicine application can be performed according to pest and disease damage grades of each small area. In addition, the hyperspectral image of the to-be-measured crop area is collected in the characteristic wave band, and longitude and latitude coordinates corresponding to the hyperspectral image are marked, so that the corresponding relation between the shot hyperspectral image and the to-be-measured small area is determined. In addition, this application embodiment is through the quantity of the impaired blade in the statistics subregion that awaits measuring to distinguish the impaired reason of blade in the subregion that awaits measuring, ensure to input the hyperspectral image in pest type identification neural network model and the pest and disease damage grade estimation model, be the hyperspectral image of the impaired blade that is aroused by the pest and disease, and then reduce the condition that the crops state of an illness reason and the medicine use are unmatched, with the operation effect of improvement accurate agriculture.
In an implementation manner of the present application, before monitoring pest conditions in a crop area to be tested according to pest types and pest grades, the method further includes: randomly selecting hyperspectral images of a plurality of damaged blades within a preset interval duration to carry out multiple observations, and obtaining an average spectral reflectivity; comparing the average spectral reflectivity with the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade, and acquiring a difference value between the average spectral reflectivity and the spectral reflectivity when the average spectral reflectivity is larger than or smaller than the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade; wherein the difference comprises a positive difference and a negative difference; searching a difference value in a preset spectral reflectance difference value table, and searching a pest type corresponding to the difference value; and comparing the pest type searched according to the difference value with the pest type obtained through the pest type recognition neural network model, and re-training the pest type recognition neural network model under the condition that the error rate of the two types is greater than a second preset threshold value.
According to the embodiment of the application, the hyperspectral images of a plurality of damaged blades are observed to obtain the average spectral reflectivity of the hyperspectral images, the pest types of the damaged blades are searched according to the preset spectral reflectivity difference table, the pest types obtained by the pest type identification neural network model are detected, the pest type identification neural network model with large errors is retrained in time, and the accuracy of agricultural pest monitoring is ensured.
In an implementation manner of the application, according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic band identification network model, a characteristic band sensitive to plant diseases and insect pests is obtained, and the method specifically comprises the following steps: acquiring a satellite remote sensing image of a crop area to be detected, and extracting habitat characteristics of the satellite remote sensing image to acquire habitat characteristic parameters; acquiring one or more characteristic wave bands corresponding to the to-be-detected crop area according to the satellite remote sensing image, the habitat characteristic parameters and the preset characteristic wave band identification network model; and recombining one or more characteristic wave bands together to obtain the characteristic wave band sensitive to the plant diseases and insect pests.
According to the embodiment of the application, the habitat characteristic of the area to be detected is extracted, and the habitat characteristic parameters of the area are obtained so as to determine possible plant diseases and insect pests in the area to be detected according to the habitat characteristic parameters. The habitat characteristic parameters are input into a preset characteristic band identification network model to obtain a characteristic band corresponding to the plant diseases and insect pests suitable for living in the habitat characteristic, so that the hyperspectral image of the area to be detected is shot in the obtained characteristic band sensitive to the plant diseases and insect pests, the data volume of the obtained hyperspectral image is reduced, and the rate of plant disease and insect pest detection is improved.
In one implementation of the present application, before dividing the crop area to be tested into a plurality of small areas to be tested, the method further includes: collecting sample hyperspectral images of crops affected by plant diseases and insect pests in different sample areas through a hyperspectral meter; the corresponding characteristic wave bands of different sample regions are different when the images are acquired; acquiring a disease and pest type corresponding to the sample hyperspectral image, and constructing a disease and pest type recognition neural network model according to the sample hyperspectral image and the corresponding disease and pest type; determining the damaged leaf area of the sample in the hyperspectral image of the sample, and calculating the damaged leaf area ratio of the sample; wherein the proportion of the area of the damaged leaves of the sample is the proportion of the area of the damaged leaves of the sample in the whole area of the leaves; determining the pest grade according to the ratio of the area of the damaged leaves of the sample; wherein the pest grade is divided into three grades of mild pest, moderate pest and severe pest; and constructing a pest grade estimation model according to the hyperspectral image of the sample, the area ratio of the damaged leaves of the sample and the pest grade corresponding to the area of the damaged leaves of the sample.
In an implementation manner of the present application, calculating a ratio of damaged leaf areas of a sample specifically includes: comparing the hyperspectral image of the damaged blade of the sample with the hyperspectral image of the pre-stored healthy blade, and determining a damage position set of spots, blade fading, blade withering and blade defect in the damaged blade of the sample; obtaining a binary image corresponding to the hyperspectral image of the damaged leaf of the sample according to the hyperspectral image and the damaged position set of the damaged leaf of the sample; and calculating the proportion of the damaged blade area of the sample corresponding to the damaged position set in the whole blade area through the binary image.
The embodiment of the application determines the proportion of the damaged area in the whole blade by calculating the damaged area ratio of the sample, so as to determine the grade of the current blade suffering from the plant diseases and insect pests. And the pesticide with different concentrations is applied to the detection area according to different pest and disease damage grades, so that accurate pesticide application is realized, and the waste of pesticides is reduced.
In an implementation of the present application, after comparing the hyperspectral image of the sample damaged leaf with the hyperspectral image of the pre-stored healthy leaf and determining the spot in the sample damaged leaf, the leaf fading, the leaf withering and the damaged position set of the leaf defect, the method further comprises: comparing the brightness of the hyperspectral image corresponding to the determined damage position with the hyperspectral image of the preset non-pest damage blade; determining the damage position with the same brightness comparison result as a non-pest damage position; wherein, the non-pest damage at least comprises any one or more of soil salinization, plant water shortage, plant nutrient excess, plant nutrient deficiency, over-high temperature and over-low temperature; and in the damage position set, rejecting the non-pest damage positions.
The hyperspectral image that this application embodiment corresponds through damaging the position carries out luminance with the hyperspectral image of preset non-pest damage blade and compares to determine the damage that causes because of other reasons in damaging the blade, thereby improve the precision that accounts for the ratio to the damage blade area that causes because of the pest and disease damage, improve the rate of accuracy to the pest and disease monitoring.
In one implementation of the present application, before calculating the damaged leaf area fraction of the sample, the method further comprises: cutting off background parts except for the crop leaves in the sample hyperspectral image to obtain a sample leaf hyperspectral image; filtering the sample blade hyperspectral image through a bilateral filter, and performing principal component analysis dimensionality reduction and normalization processing on the filtered sample blade hyperspectral image to obtain a sample damaged blade; and (3) carrying out rotation processing, vertical turning processing and horizontal turning processing on damaged leaves of a part of the sample so as to increase the number of the samples.
In an implementation of the present application, a pest grade estimation model is constructed according to a sample hyperspectral image, a sample damaged leaf area ratio and a pest grade corresponding to a sample damaged leaf area, and specifically includes: inputting a test image in a sample hyperspectral image into a pest type identification neural network model to obtain a leaf damage position and a pest type corresponding to the test image; calculating the area ratio of damaged leaves corresponding to a test image output by the pest type recognition neural network model; determining the pest and disease damage grade corresponding to the damaged leaves in the test image according to the damaged leaf area ratio; and inputting a test image output by the pest type recognition neural network model, the damaged leaf area ratio corresponding to the test image and the pest grade corresponding to the damaged leaf area in the test image into a second network model for training to obtain a pest grade estimation model.
In an implementation manner of the application, a hyperspectral image corresponding to a small area to be tested is input into a pest type recognition neural network model and a pest grade estimation model to obtain the types and grades of pests to which crops to be tested are subjected, and the method specifically comprises the following steps: inputting the hyperspectral image of the crop leaf to be detected in the characteristic wave band into a pest type identification neural network model to obtain a hyperspectral image labeled with the pest type; inputting the hyperspectral image labeled with the pest type into a pest grade estimation model to obtain the pest grade of the crop leaf to be detected; and counting the number of damaged leaves of the crops to be detected corresponding to each pest grade in the small area to be detected, and taking the pest grade corresponding to the damaged leaf with the largest number as the pest grade of the small area to be detected.
The embodiment of the application provides an agricultural plant diseases and insect pests monitoring facilities based on hyperspectrum, includes:
at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to cause the at least one processor to: dividing the region of the crop to be detected into a plurality of small regions to be detected, and determining a position coordinate set of each small region to be detected; according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model, obtaining a characteristic wave band sensitive to plant diseases and insect pests; collecting a hyperspectral image of a region of the crop to be measured in a characteristic waveband through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and marking longitude and latitude coordinates corresponding to the hyperspectral image; comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set to determine a small area to be detected corresponding to the hyperspectral image; counting the number of damaged leaves in the small area to be detected according to the hyperspectral image corresponding to the small area to be detected, and inputting the hyperspectral image corresponding to the small area to be detected into a pest type identification neural network model and a pest grade estimation model under the condition that the number of the damaged leaves is larger than a first preset threshold value so as to obtain the pest type and the pest grade of the crop to be detected; and monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade.
The embodiment of the application adopts at least one technical scheme which can achieve the following beneficial effects: according to the embodiment of the application, the areas to be detected are divided, so that pest and disease damage monitoring can be performed on the different small areas to be detected after division, and therefore different degrees of medicine application can be performed according to pest and disease damage grades of each small area. In addition, the hyperspectral image of the to-be-measured crop area is collected in the characteristic wave band, and longitude and latitude coordinates corresponding to the hyperspectral image are marked, so that the corresponding relation between the shot hyperspectral image and the to-be-measured small area is determined. In addition, this application embodiment is through the quantity of the impaired blade in the statistics subregion that awaits measuring to distinguish the impaired reason of blade in the subregion that awaits measuring, ensure to input the hyperspectral image in pest type identification neural network model and the pest and disease damage grade estimation model, be the hyperspectral image of the impaired blade that is aroused by the pest and disease, and then reduce the condition that the crops state of an illness reason and the medicine use are unmatched, with the operation effect of improvement accurate agriculture.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present application, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort. In the drawings:
FIG. 1 is a flow chart of an agricultural pest monitoring method based on hyperspectrum according to an embodiment of the application;
FIG. 2 is a schematic structural diagram of an agricultural pest monitoring device based on hyperspectrum according to an embodiment of the application.
Detailed Description
The embodiment of the application provides a hyperspectral agricultural pest monitoring method and device.
In order to make those skilled in the art better understand the technical solutions in the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be obtained by a person skilled in the art without making any inventive step based on the embodiments of the present disclosure, shall fall within the scope of protection of the present application.
The crop diseases and insect pests have the characteristics of multiple varieties, large influence and frequent outbreak and disaster formation. The mode of agricultural management is carried out by experience, so that the crops are difficult to be accurately monitored in the early period of insect damage, and large-area outbreak of the insect pests is easily caused.
At present, agricultural production in China still mainly adopts a traditional production mode, and agricultural management such as fertilization, irrigation, insect killing and pest killing is mostly carried out by means of experience, so that huge waste of manpower and financial resources is caused.
In addition, the prior art has inaccurate monitoring on pest types and pest flooding degrees, so that the problems of mismatching of pest types and pesticide types and more or less pesticide dosage can be caused frequently, the operation effect of precision agriculture is seriously influenced, and the agricultural yield is difficult to improve.
In order to solve the problems, the embodiment of the application provides an agricultural disease and pest monitoring method and device based on hyperspectrum. By dividing the area to be detected, the pest and disease damage monitoring can be carried out on the divided different small areas to be detected, so that the pesticide can be applied to different degrees according to the pest and disease damage grade of each small area. In addition, the hyperspectral image of the to-be-measured crop area is collected in the characteristic wave band, and longitude and latitude coordinates corresponding to the hyperspectral image are marked, so that the corresponding relation between the shot hyperspectral image and the to-be-measured small area is determined. In addition, this application embodiment is through the quantity of the impaired blade in the statistics subregion that awaits measuring to distinguish the impaired reason of blade in the subregion that awaits measuring, ensure to input the hyperspectral image in pest type identification neural network model and the pest and disease damage grade estimation model, be the hyperspectral image of the impaired blade that is aroused by the pest and disease, and then reduce the condition that the crops state of an illness reason and the medicine use are unmatched, with the operation effect of improvement accurate agriculture.
The technical solutions proposed in the embodiments of the present application are described in detail below with reference to the accompanying drawings.
FIG. 1 is a flow chart of an agricultural pest monitoring method based on hyperspectrum according to an embodiment of the application. As shown in fig. 1, the agricultural pest monitoring method comprises the following steps:
s101, the agricultural disease and insect monitoring system conducts training on a disease and insect type recognition neural network model and a disease and insect grade estimation model.
In an embodiment of the application, through the hyperspectral meter, the sample hyperspectral images of crops affected by plant diseases and insect pests in different sample areas are collected, the collected sample hyperspectral images are divided into training set images and testing set images, the training set comprises hyperspectral images corresponding to a plurality of damaged leaves affected by the plant diseases and insect pests, the testing set comprises hyperspectral images corresponding to the damaged leaves affected by the plant diseases and insect pests, and the hyperspectral images corresponding to the damaged leaves caused by non-plant diseases and insect pests. Wherein, the corresponding characteristic wave bands of different sample regions are different when the image acquisition is carried out.
Specifically, a plurality of sample regions are preset, each sample region being provided with a different habitat characteristic. And determining the type of the pest in each area, and acquiring a hyperspectral image of the crops in each sample area by using a hyperspectral spectrometer in the characteristic wave band sensitive to the pest. The image acquired by the hyperspectral meter is an image of the canopy of the crop in the sample area.
In one embodiment of the application, a disease and pest type corresponding to the sample hyperspectral image is obtained, and a disease and pest type recognition neural network model is constructed according to the sample hyperspectral image and the corresponding disease and pest type.
Specifically, determining the type of the plant diseases and insect pests corresponding to the acquired sample hyperspectral image, and marking the damaged position of the blade in the sample hyperspectral image by using a rectangular frame. Inputting training set images in the sample hyperspectral images and the pest and disease types corresponding to the training set images into the network model for training, testing the trained model by using the sample hyperspectral images in the test set images, and obtaining the pest and disease type recognition neural network model after the test result reaches the standard.
Through the pest type recognition neural network model, the damage position in the input high-spectrum image of the damaged blade can be marked out by using a rectangular frame, and the pest type of the current blade is determined according to the damage condition of the blade.
In one embodiment of the application, the background part except the crop leaves in the sample hyperspectral image is cut out to obtain the sample leaf hyperspectral image. For example, the hyperspectral image corresponding to the leaf can be divided according to the difference between the leaf color and the background color.
In one embodiment of the application, filtering is performed on the sample blade hyperspectral image through a bilateral filter, and principal component analysis dimensionality reduction and normalization processing are performed on the filtered sample blade hyperspectral image to obtain a sample damaged blade. And (4) carrying out rotation processing, vertical turning processing and horizontal turning processing on damaged leaves of a part of the samples so as to increase the number of the samples in the test set.
In one embodiment of the application, the damaged leaf area of the sample is determined in the hyperspectral image of the sample in the test set, and the damaged leaf area ratio of the sample is calculated. Wherein, the proportion of the damaged leaf area of the sample is the proportion of the damaged leaf area of the sample in the whole leaf area.
Specifically, the hyperspectral images of the damaged leaves of the test set sample are compared with the hyperspectral images of the pre-stored healthy leaves, and a damaged position set of spots, fading, withering and defect leaves in the damaged leaves of the sample is determined.
Specifically, in the hyperspectral images of a plurality of pre-stored healthy blades, the hyperspectral images corresponding to the healthy blades with the most similar characteristics of the shapes, sizes, colors and the like of the damaged blades in the test set sample are found, and the hyperspectral images of the healthy blades and the hyperspectral images of the damaged blades are compared to determine each damaged position set in the damaged blades.
Further, a coordinate system can be established in the hyperspectral image, and a coordinate position set of each damaged blade is determined through the coordinate position. For example, the hyperspectral image coordinate system is established by taking the intersection point of a horizontal straight line where the upper edge of the hyperspectral image is located and a vertical straight line where the left edge is located as the origin of the coordinate system, taking a straight line extending horizontally to the right as the X axis, and taking a straight line extending vertically downwards as the Y axis. And determining a coordinate set of each damage position through the hyperspectral image coordinate system.
In one embodiment of the application, the hyperspectral image corresponding to the determined damage position is compared with the hyperspectral image of the preset non-pest damage blade in brightness. And determining the damage positions with the same brightness comparison result as non-pest damage positions, and rejecting the non-pest damage positions in the damage position set.
Specifically, when the hyperspectral images of the damaged leaves of the test set sample are compared with the hyperspectral images of the pre-stored healthy leaves, under the condition that the brightness of the damaged positions of the hyperspectral images and the pre-stored healthy leaves is the same, it is indicated that the damage in the damaged leaves is not caused by plant diseases and insect pests. Therefore, the damage location should not be involved in the calculation of pest monitoring.
It should be noted that the preset hyperspectral images of the non-pest damage leaves include hyperspectral images of the damaged leaves caused by various non-pest factors such as human factors, weather factors and soil.
The hyperspectral image that the position set corresponds will damage is compared with the hyperspectral image of the damaged blade of non-pest damage to determine the damage that causes because of other reasons in the damaged blade, thereby improve the accuracy of the area ratio calculation of the damaged blade that causes the pest and disease damage, improve the rate of accuracy to the pest and disease monitoring.
In one embodiment of the application, the non-pest damage includes at least any one or more of soil salinization, water deficit in the plant, nutrient surplus in the plant, nutrient deficiency in the plant, over temperature, and under temperature.
In one embodiment of the application, a binary image corresponding to the hyperspectral image of the damaged leaf of the sample is obtained according to the hyperspectral image and the damaged position set of the damaged leaf of the sample of the test set. And calculating the proportion of the damaged blade area of the test set sample corresponding to the damaged position set in the whole blade area through the binary image.
Specifically, the color of the damaged position in the damaged leaf of the test set sample is different from the color of the healthy position to a certain extent, so that a seed point is randomly selected from each damaged position coordinate according to the color difference, a coordinate point close to the seed color is filled by adopting a flood filling algorithm, and the filled coordinate points form a communicated area, so that the damaged area is divided. And converting the filled hyperspectral image of the damaged blade into a binary image, wherein the damaged area and the healthy area can be obviously divided. And calculating the total area of a plurality of damaged positions in the current binary image through the coordinate point set of the damaged region, and calculating the total area of the current damaged blade, so as to obtain the area of the sample damaged blade corresponding to the damaged position set in the current blade and the ratio of the area of the whole current blade.
In one embodiment of the present application, pest grade is determined from the ratio of damaged leaf area to test set sample. Wherein, the pest grade is divided into three grades of mild pest, moderate pest and severe pest.
Specifically, under the condition that the proportion of the area of the damaged leaves in the test set sample is less than or equal to 30% of the area of the whole leaves, determining that the damaged leaves in the test set sample correspond to mild plant diseases and insect pests. And under the condition that the ratio of the area of the damaged leaves in the test set sample is more than 30% of the area of the whole leaves and is less than or equal to 60% of the area of the whole leaves, determining that the damaged leaves in the test set sample correspond to moderate plant diseases and insect pests. And under the condition that the area ratio of the damaged leaves of the test set sample is more than 60% of the area of the whole leaves, determining that the damaged leaves of the test set sample correspond to severe plant diseases and insect pests.
In one embodiment of the application, a pest grade estimation model is constructed according to the hyperspectral image of the test set sample, the ratio of the damaged leaf area of the test set sample and the pest grade corresponding to the damaged leaf area of the test set sample.
Specifically, the test set image is input into the pest type identification neural network model, and at this time, the position of the damaged leaf is marked by a rectangular frame in the test set image, and the pest type corresponding to the damaged leaf is obtained.
Further, calculating the area ratio of damaged leaves corresponding to the test set image output by the pest type recognition neural network model, and determining the pest grade corresponding to the damaged leaves in the test set image according to the area ratio of the damaged leaves. And inputting a test set image output by the pest type recognition neural network model, the damaged leaf area ratio corresponding to the test set image and the pest grade corresponding to the damaged leaf area in the test set image into the network model for training to obtain a pest grade estimation model.
S102, dividing the to-be-detected crop area into a plurality of to-be-detected small areas, and determining a position coordinate set of each to-be-detected small area.
In one embodiment of the present application, the edge of the region to be measured is located by a locating device. For example, when the region to be measured is rectangular, the longitude and latitude corresponding to the four corners of the rectangular region to be measured are measured, so as to determine the set of position coordinates of the rectangular region to be measured.
In one embodiment of the application, the area to be measured is divided into a plurality of small areas to be measured. For example, when the region to be measured is rectangular, the distance between the length and the width of the rectangular region may be determined by the longitude and latitude of the four vertices measured in advance, so as to divide the rectangular region to be measured into a plurality of rectangular small regions.
It should be noted that the position coordinate set of each small region to be measured may be obtained by measurement of a positioning device, or may be obtained by calculation and allocation of the position coordinates of the region of the crop to be measured. The embodiments of the present application do not limit this.
S103, acquiring a characteristic wave band sensitive to plant diseases and insect pests according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model.
In one embodiment of the application, a satellite remote sensing image of a crop area to be detected is obtained, and habitat characteristic extraction is performed on the satellite remote sensing image to obtain habitat characteristic parameters.
Specifically, the habitat characteristic extraction is carried out on the satellite remote sensing image of the crop area to be detected, and the habitat characteristic parameters of the crop area to be detected, such as illumination, temperature, moisture, air, inorganic salt and the like, can be obtained. And determining the types of the diseases and insect pests suitable for survival in the crop area to be tested through the acquired habitat characteristic parameters.
In an embodiment of the application, one or more characteristic wave bands corresponding to the to-be-detected crop area are obtained according to the satellite remote sensing image, the habitat characteristic parameters and the preset characteristic wave band identification network model. And recombining one or more characteristic wave bands together to obtain the characteristic wave band sensitive to the plant diseases and insect pests.
Specifically, different pests correspond to different characteristic wave bands. Therefore, the habitat characteristic parameters of the crop area to be measured and the satellite remote sensing image are input into the preset characteristic wave band identification network model, the preset characteristic wave band identification network model determines habitat characteristic parameters corresponding to the satellite remote sensing image, and possible pests and diseases in the environment are determined according to the habitat characteristic parameters so as to determine hyperspectral characteristic wave bands corresponding to the pests and diseases. And recombining one or more characteristic wave bands together to obtain the characteristic wave band sensitive to the plant diseases and insect pests.
And S104, acquiring a hyperspectral image of the to-be-detected crop area in the characteristic wave band.
In one embodiment of the application, a hyperspectral image of an area of an agricultural work to be measured is collected in a characteristic waveband through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and longitude and latitude coordinates corresponding to the hyperspectral image are marked.
Specifically, install hyperspectral appearance and positioner on unmanned aerial vehicle, fly through unmanned aerial vehicle low latitude, carry out hyperspectral image collection to the blade of regional crops that awaits measuring. And carrying out longitude and latitude positioning on the acquisition position through a positioning device, and marking the acquired longitude and latitude coordinates on the hyperspectral image corresponding to the position.
And S105, comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set, and determining the small area to be detected corresponding to the hyperspectral image.
In an embodiment of the application, the longitude and latitude coordinates marked on the collected hyperspectral images are compared with the longitude and latitude coordinate set of each small area to be detected, and the small area to be detected to which each hyperspectral image belongs is determined.
S106, inputting the hyperspectral image corresponding to the small area to be detected into the pest type identification neural network model and the pest grade estimation model under the condition that the number of damaged leaves is larger than a first preset threshold value.
In one embodiment of the present application, statistics are performed on the number of damaged leaves in the collected hyperspectral image. And under the condition that the number of the damaged leaves is more than 10% of the collected number of all the leaves, determining that the current area to be tested suffers from plant diseases and insect pests.
It should be noted that, in the embodiment of the present application, it is preferable that, in the case that the number of damaged leaves is greater than 10% of the number of all collected leaves, it is determined that the current area to be tested is suffered from a pest, but the present application is not limited to only 10%, and the present application is not limited thereto.
In one embodiment of the application, hyperspectral images of the damaged leaves are prestored and hyperspectral images of healthy leaves are compared, and a damage position set of spots, leaf fading, leaf withering and leaf defect in the damaged leaves of the sample is determined. And comparing the hyperspectral image of the damaged blade of the sample with the hyperspectral image of the damaged blade of the preset non-pest damage, so as to determine the position of the damaged blade caused by non-pest reasons such as human factors, weather and the like in the damaged blade. And in the damaged position set, removing the positions of damaged leaves caused by non-pest and disease damage.
In one embodiment of the application, the hyperspectral image corresponding to the damaged leaf at the damaged position of the leaf caused by non-pest and disease damage is removed, and the pest and disease type identification neural network model is input to obtain the hyperspectral image marked with the pest and disease type.
In an embodiment of the application, the hyperspectral image marked with the pest type is input into a pest grade estimation model, and the pest grade of the crop leaf to be detected is obtained.
Specifically, the area ratio of the damaged leaves is calculated by inputting a pest grade estimation model, and the calculated area ratio is compared with the pest grade to determine the pest grade of the damaged leaves to be detected. For example, when the area ratio of the damaged leaves is calculated to be 20%, the current grade of the disease and insect pest corresponding to the leaves is determined to be mild disease and insect pest.
In one embodiment of the application, the number of damaged leaves of crops to be detected corresponding to each pest grade in a small area to be detected is counted, and the pest grade corresponding to the damaged leaf with the largest number is used as the pest grade of the small area to be detected.
For example, in the small area to be measured, the number of damaged leaves belonging to mild diseases and insect pests is 500, the number of damaged leaves belonging to moderate diseases and insect pests is 100, and the number of damaged leaves belonging to severe diseases and insect pests is 50. At the moment, the pest grade in the small area to be detected can be determined as the mild pest.
In one embodiment of the application, in a preset interval duration, hyperspectral images of a plurality of damaged blades are randomly selected for multiple observations, and an average spectral reflectivity is obtained. And comparing the average spectral reflectivity with the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade, and acquiring a difference value between the average spectral reflectivity and the spectral reflectivity under the condition that the average spectral reflectivity is greater than or less than the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade. Wherein the difference comprises a positive difference and a negative difference.
In an embodiment of the present application, the difference is found in a preset spectral reflectance difference table, and the type of the pest corresponding to the difference is found. And comparing the pest type searched according to the difference value with the pest type obtained through the pest type recognition neural network model, and re-training the pest type recognition neural network model under the condition that the error rate of the two types is greater than a second preset threshold value.
For example, the pest type determined by the difference is a locust, and the pest type acquired by the pest type recognition neural network model is a beetle, and at this time, it can be determined that an error exists between the two types. Comparing the pest types respectively judged by the difference method and the pest type recognition neural network model for multiple times to obtain the ratio of the times with different comparison results to the total times of comparison, so as to obtain an error rate, and retraining the pest type recognition neural network model under the condition that the error rate is more than 5%.
And S107, monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade.
In one embodiment of the application, the pest condition of each small area to be detected can be monitored according to the pest type and the pest grade. And according to the pest type and pest grade of each small area to be detected, different types and different concentrations of drugs are respectively applied, so that the waste of pesticides is reduced.
FIG. 2 is a schematic structural diagram of an agricultural pest monitoring device based on hyperspectrum according to an embodiment of the application.
As shown in figure 2, agricultural plant diseases and insect pests monitoring facilities based on hyperspectrum includes:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to:
dividing the region of the crop to be detected into a plurality of small regions to be detected, and determining a position coordinate set of each small region to be detected;
acquiring a characteristic wave band sensitive to the plant diseases and insect pests according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model;
collecting a hyperspectral image of the region of the crop to be measured in the characteristic wave band through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and marking longitude and latitude coordinates corresponding to the hyperspectral image;
comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set to determine a small area to be detected corresponding to the hyperspectral image;
according to the hyperspectral image corresponding to the small area to be detected, counting the number of damaged leaves in the small area to be detected, and inputting the hyperspectral image corresponding to the small area to be detected into a pest type identification neural network model and a pest grade estimation model under the condition that the number of the damaged leaves is larger than a first preset threshold value so as to obtain the pest type and pest grade of the crop to be detected;
and monitoring the pest condition of the region of the crop to be detected according to the pest type and the pest grade.
The embodiments in the present application are described in a progressive manner, and the same and similar parts among the embodiments can be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the embodiments of the apparatus, the device, and the nonvolatile computer storage medium, since they are substantially similar to the embodiments of the method, the description is simple, and for the relevant points, reference may be made to the partial description of the embodiments of the method.
The foregoing description of specific embodiments of the present application has been presented. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art to which the embodiments of the present application pertain. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the embodiments of the present application should be included in the scope of the claims of the present application.
Claims (7)
1. An agricultural pest and disease monitoring method based on hyperspectrum is characterized by comprising the following steps:
dividing the region of the crop to be detected into a plurality of small regions to be detected, and determining a position coordinate set of each small region to be detected;
acquiring a characteristic wave band sensitive to the plant diseases and insect pests according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model;
collecting a hyperspectral image of the region of the crop to be measured in the characteristic wave band through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and marking longitude and latitude coordinates corresponding to the hyperspectral image;
comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set to determine a small area to be detected corresponding to the hyperspectral image;
according to the hyperspectral image corresponding to the small area to be detected, counting the number of damaged leaves in the small area to be detected, and inputting the hyperspectral image corresponding to the small area to be detected into a pest type identification neural network model and a pest grade estimation model under the condition that the number of the damaged leaves is larger than a first preset threshold value so as to obtain the pest type and pest grade of the crop to be detected;
monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade;
before the dividing the region of the crop to be tested into a plurality of small regions to be tested, the method further comprises,
collecting sample hyperspectral images of crops affected by plant diseases and insect pests in different sample areas through a hyperspectral meter; the different sample regions have different corresponding characteristic wave bands when image acquisition is carried out;
acquiring a disease and insect pest type corresponding to the sample hyperspectral image, and constructing a disease and insect pest type recognition neural network model according to the sample hyperspectral image and the corresponding disease and insect pest type;
determining the damaged leaf area of the sample in the hyperspectral image of the sample, and calculating the damaged leaf area ratio of the sample; wherein the proportion of the damaged leaf area of the sample is the proportion of the damaged leaf area of the sample in the whole leaf area;
determining the pest and disease damage grade according to the ratio of the damaged leaf area of the sample; wherein the pest grades are divided into three grades of mild pest, moderate pest and severe pest;
constructing the pest grade estimation model according to the sample hyperspectral image, the sample damaged leaf area ratio and the pest grade corresponding to the sample damaged leaf area;
the damaged leaf area ratio of the sample is calculated, including,
comparing the hyperspectral image of the damaged blade of the sample with the hyperspectral image of the pre-stored healthy blade, and determining a damage position set of spots, blade fading, blade withering and blade defect in the damaged blade of the sample;
obtaining a binary image corresponding to the hyperspectral image of the damaged sample blade according to the hyperspectral image of the damaged sample blade and the damaged position set;
calculating the proportion of the damaged blade area of the sample corresponding to the damaged position set in the whole blade area through the binary image;
after comparing the hyperspectral image of the sample damaged leaf with the hyperspectral image of the pre-stored healthy leaf and determining the spot, leaf fading, leaf withering and damaged position set of the sample damaged leaf, the method also comprises,
comparing the brightness of the hyperspectral image corresponding to the determined damage position with the hyperspectral image of the preset non-pest damage blade;
determining the damage position with the same brightness comparison result as a non-pest damage position; wherein the non-pest damage at least comprises any one or more of soil salinization, plant water shortage, plant nutrient excess, plant nutrient deficiency, over-high temperature and under-low temperature;
and in the damage position set, rejecting the non-pest damage positions.
2. An agricultural pest and disease monitoring method based on hyperspectrum according to claim 1, wherein before monitoring pest and disease conditions of the crop area to be detected according to the pest type and the pest grade, the method further comprises:
randomly selecting hyperspectral images of a plurality of damaged blades within a preset interval duration to carry out multiple observations, and acquiring an average spectral reflectivity;
comparing the average spectral reflectivity with the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade, and acquiring a difference value between the average spectral reflectivity and the spectral reflectivity when the average spectral reflectivity is larger than or smaller than the spectral reflectivity corresponding to the hyperspectral image of the pre-stored healthy blade; wherein the difference comprises a positive difference and a negative difference;
searching the difference value in a preset spectral reflectance difference value table, and searching the pest and disease damage type corresponding to the difference value;
and comparing the pest type searched according to the difference with the pest type obtained through the pest type recognition neural network model, and retraining the pest type recognition neural network model under the condition that the error rate of the two types is greater than a second preset threshold value.
3. The hyperspectral agricultural pest and disease monitoring method according to claim 1 is characterized in that the obtaining of the pest and disease sensitive characteristic wave band according to the habitat characteristic parameters of the crop area to be tested and a preset characteristic wave band identification network model specifically comprises:
acquiring a satellite remote sensing image of the crop area to be detected, and extracting habitat characteristics of the satellite remote sensing image to acquire habitat characteristic parameters;
acquiring one or more characteristic wave bands corresponding to the to-be-detected crop area according to the satellite remote sensing image, the habitat characteristic parameters and a preset characteristic wave band identification network model;
and recombining the one or more characteristic wave bands together to obtain the characteristic wave band sensitive to the plant diseases and insect pests.
4. An agricultural pest and disease monitoring method based on hyperspectrum according to claim 1, wherein before calculating the damaged leaf area ratio of the sample, the method further comprises:
cutting off background parts except for the crop leaves in the sample hyperspectral image to obtain a sample leaf hyperspectral image;
filtering the sample blade hyperspectral image through a bilateral filter, and performing principal component analysis dimensionality reduction and normalization processing on the filtered sample blade hyperspectral image to obtain a sample damaged blade;
and performing rotation treatment, vertical turning treatment and horizontal turning treatment on the damaged leaves of the part of the sample to increase the number of samples.
5. The hyperspectral agricultural pest monitoring method according to claim 1 is characterized in that the pest grade estimation model is constructed according to the hyperspectral image of the sample, the ratio of the area of the damaged leaves of the sample and the pest grade corresponding to the area of the damaged leaves of the sample, and specifically comprises the following steps:
inputting a test image in the sample hyperspectral image into the pest type identification neural network model to obtain a blade damage position and a pest type corresponding to the test image;
calculating the area ratio of damaged leaves corresponding to the test image output by the pest type recognition neural network model; determining the pest and disease damage grade corresponding to the damaged leaves in the test image according to the area ratio of the damaged leaves;
inputting a test image output by the pest type recognition neural network model, the damaged leaf area ratio corresponding to the test image and the pest grade corresponding to the damaged leaf area in the test image into a second network model for training to obtain the pest grade estimation model.
6. The hyperspectral agricultural pest monitoring method based on the claim 1 is characterized in that the hyperspectral image corresponding to the small area to be detected is input into a pest type recognition neural network model and a pest grade estimation model to obtain the pest type and the pest grade of the crop to be detected, and the method specifically comprises the following steps:
inputting the hyperspectral image of the crop leaf to be detected in the characteristic wave band into the pest type identification neural network model to obtain a hyperspectral image labeled with the pest type;
inputting the hyperspectral image labeled with the pest type into the pest grade estimation model to obtain the pest grade of the crop leaf to be detected;
and counting the number of damaged leaves of the crops to be detected corresponding to each pest grade in the small area to be detected, and taking the pest grade corresponding to the damaged leaf with the largest number as the pest grade of the small area to be detected.
7. An agricultural pest monitoring device based on hyperspectrum includes:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to:
dividing the region of the crop to be detected into a plurality of small regions to be detected, and determining a position coordinate set of each small region to be detected;
acquiring a characteristic wave band sensitive to the plant diseases and insect pests according to the habitat characteristic parameters of the crop area to be detected and a preset characteristic wave band identification network model;
collecting a hyperspectral image of the region of the crop to be measured in the characteristic wave band through an unmanned aerial vehicle provided with a hyperspectral meter and a positioning device, and marking longitude and latitude coordinates corresponding to the hyperspectral image;
comparing the longitude and latitude coordinates corresponding to the hyperspectral image with the position coordinate set to determine a small area to be detected corresponding to the hyperspectral image;
according to the hyperspectral image corresponding to the small area to be detected, counting the number of damaged leaves in the small area to be detected, and inputting the hyperspectral image corresponding to the small area to be detected into a pest type identification neural network model and a pest grade estimation model under the condition that the number of the damaged leaves is larger than a first preset threshold value so as to obtain the pest type and pest grade of the crop to be detected;
monitoring the pest condition of the crop area to be detected according to the pest type and the pest grade;
before the step of dividing the region of the crop to be detected into a plurality of small regions to be detected, the method also comprises the following steps,
collecting sample hyperspectral images of crops affected by plant diseases and insect pests in different sample areas through a hyperspectral meter; the different sample regions have different corresponding characteristic wave bands when image acquisition is carried out;
acquiring a disease and insect pest type corresponding to the sample hyperspectral image, and constructing a disease and insect pest type recognition neural network model according to the sample hyperspectral image and the corresponding disease and insect pest type;
determining the damaged leaf area of the sample in the hyperspectral image of the sample, and calculating the damaged leaf area ratio of the sample; wherein the proportion of the damaged leaf area of the sample is the proportion of the damaged leaf area of the sample in the whole leaf area;
determining the pest and disease damage grade according to the ratio of the damaged leaf area of the sample; wherein the pest grades are divided into three grades of mild pest, moderate pest and severe pest;
constructing the pest grade estimation model according to the sample hyperspectral image, the sample damaged leaf area ratio and the pest grade corresponding to the sample damaged leaf area;
the damaged leaf area ratio of the sample is calculated, including,
comparing the hyperspectral image of the damaged blade of the sample with the hyperspectral image of the pre-stored healthy blade, and determining a damage position set of spots, blade fading, blade withering and blade defect in the damaged blade of the sample;
obtaining a binary image corresponding to the hyperspectral image of the damaged sample blade according to the hyperspectral image of the damaged sample blade and the damaged position set;
calculating the proportion of the damaged blade area of the sample corresponding to the damaged position set in the whole blade area through the binary image;
after the hyperspectral images of the damaged leaves of the sample are compared with the hyperspectral images of the pre-stored healthy leaves to determine the damage position set of spots, leaf fading, leaf withering and leaf defect in the damaged leaves of the sample, the method also comprises the following steps,
comparing the brightness of the hyperspectral image corresponding to the determined damage position with the hyperspectral image of the preset non-pest damage blade;
determining the damage position with the same brightness comparison result as a non-pest damage position; wherein the non-pest damage at least comprises any one or more of soil salinization, plant water shortage, plant nutrient excess, plant nutrient deficiency, over-high temperature and under-low temperature;
and in the damage position set, rejecting the non-pest damage positions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110600387.4A CN113468964B (en) | 2021-05-31 | 2021-05-31 | Hyperspectrum-based agricultural disease and pest monitoring method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110600387.4A CN113468964B (en) | 2021-05-31 | 2021-05-31 | Hyperspectrum-based agricultural disease and pest monitoring method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113468964A CN113468964A (en) | 2021-10-01 |
| CN113468964B true CN113468964B (en) | 2022-04-26 |
Family
ID=77871845
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110600387.4A Active CN113468964B (en) | 2021-05-31 | 2021-05-31 | Hyperspectrum-based agricultural disease and pest monitoring method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113468964B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114689526A (en) * | 2022-05-31 | 2022-07-01 | 北京市农林科学院智能装备技术研究中心 | Banana wilt nondestructive detection method, device and detection equipment |
| CN115187860B (en) * | 2022-07-11 | 2023-07-18 | 浙江甲骨文超级码科技股份有限公司 | Plant disease and pest evaluation method and system |
| CN115131683B (en) * | 2022-08-25 | 2022-12-09 | 金乡县林业保护和发展服务中心(金乡县湿地保护中心、金乡县野生动植物保护中心、金乡县国有白洼林场) | Forestry information identification method based on high-resolution remote sensing image |
| CN116468671B (en) * | 2023-03-21 | 2024-04-16 | 中化现代农业有限公司 | Plant disease degree detection method, device, electronic apparatus, and storage medium |
| CN117636185B (en) * | 2024-01-26 | 2024-04-09 | 安徽大学 | Pine wood nematode disease detecting system based on image processing |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105173085B (en) * | 2015-09-18 | 2017-06-16 | 山东农业大学 | Unmanned plane variable farm chemical applying automatic control system and method |
| TWI667620B (en) * | 2018-01-25 | 2019-08-01 | 國立臺灣大學 | Pest surveillance system |
| CN108669036A (en) * | 2018-03-30 | 2018-10-19 | 合肥赛为智能有限公司 | A kind of unmanned plane is to target dispenser Intelligentized regulating and controlling system |
| CN109446987A (en) * | 2018-10-29 | 2019-03-08 | 北京麦飞科技有限公司 | Method based on PCA and PNN algorithm detection rice pest grade |
| CN110009043B (en) * | 2019-04-09 | 2021-08-17 | 广东省智能制造研究所 | Disease and insect pest detection method based on deep convolutional neural network |
| CN110517311A (en) * | 2019-08-30 | 2019-11-29 | 北京麦飞科技有限公司 | Pest and disease monitoring method based on leaf spot lesion area |
| CN111339921A (en) * | 2020-02-24 | 2020-06-26 | 南京邮电大学 | Insect disease detection unmanned aerial vehicle based on lightweight convolutional neural network and detection method |
| CN111582055B (en) * | 2020-04-17 | 2023-06-09 | 清远市智慧农业研究院 | Unmanned aerial vehicle aviation pesticide application route generation method and system |
| CN111553240B (en) * | 2020-04-24 | 2021-03-16 | 四川省农业科学院农业信息与农村经济研究所 | Corn disease condition grading method and system and computer equipment |
| CN112741069B (en) * | 2021-01-14 | 2023-09-12 | 农芯科技（北京）有限责任公司 | Orchard accurate physical prevention and control plant protection device and method based on machine vision |
-
2021
- 2021-05-31 CN CN202110600387.4A patent/CN113468964B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113468964A (en) | 2021-10-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113468964B (en) | Hyperspectrum-based agricultural disease and pest monitoring method and device | |
| US10568316B2 (en) | Apparatus and methods for in-field data collection and sampling | |
| CN106971167B (en) | Crop growth analysis method and system based on unmanned aerial vehicle platform | |
| CN109948596B (en) | Method for identifying rice and extracting planting area based on vegetation index model | |
| CN104266982B (en) | A kind of large area insect pest quantifies monitoring system | |
| CN110288647B (en) | Method for monitoring irrigation area of irrigation area based on high-resolution satellite data | |
| CN110926430A (en) | Air-ground integrated mangrove forest monitoring system and control method | |
| CN108140118A (en) | For identifying the method and information system of at least one plant plant being planted on arable land | |
| Fischer et al. | Use of unmanned aircraft systems (UAS) and multispectral imagery for quantifying agricultural areas damaged by wild pigs | |
| CN102592118B (en) | Automatic detection method for time emergence of seedling of corns | |
| CN116778343B (en) | Target image feature extraction method for comprehensive identification | |
| CN111582035B (en) | Fruit tree age identification method, device, equipment and storage medium | |
| Mostafa et al. | Using LiDAR technique and modified Community Land Model for calculating water interception of cherry tree canopy | |
| CN115619286B (en) | Method and system for evaluating quality of sample plot of breeding field district | |
| Kaur | Automated nutrient deficiency detection in plants: A Review | |
| CN105184234B (en) | A kind of Wheat Straw burns the measuring method and device of pollutant discharge amount | |
| CN113326841B (en) | Hyperspectrum-based crop pest monitoring method and device | |
| Jha et al. | Wheat acreage estimation using remote sensing in tarai region of Uttarakhand | |
| Tugrul | Estimation of sugar beet biomass and yield comparing NDVI measurements and physical soil parameters | |
| CN115019205B (en) | Rape flowering phase SPAD and LAI estimation method based on unmanned aerial vehicle multispectral image | |
| CN114140695B (en) | Prediction method and system for tea tree nitrogen diagnosis and quality index determination based on unmanned aerial vehicle multispectral remote sensing | |
| CN109241866A (en) | A kind of crops automatic classification method and device based on history crop pattern map | |
| CN113588500B (en) | Method for detecting fog drop deposition effect in aviation pesticide application operation | |
| CN115719169A (en) | Arable land safety evaluation method | |
| CN113112147A (en) | Citrus huanglongbing risk assessment method based on ground actual measurement and remote sensing cooperation |
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 |