CN114782844B - Pine wood nematode disease tree identification method, system and storage medium - Google Patents
Pine wood nematode disease tree identification method, system and storage medium Download PDFInfo
- Publication number
- CN114782844B CN114782844B CN202210486161.0A CN202210486161A CN114782844B CN 114782844 B CN114782844 B CN 114782844B CN 202210486161 A CN202210486161 A CN 202210486161A CN 114782844 B CN114782844 B CN 114782844B
- Authority
- CN
- China
- Prior art keywords
- pine wood
- wood nematode
- nematode disease
- target area
- tree
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Cultivation Of Plants (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the technical field of plant detection, and discloses a pine wood nematode disease tree identification method, a system and a storage medium.
Description
Technical Field
The invention relates to the technical field of plant detection, in particular to a pine wood nematode disease tree identification method, a pine wood nematode disease tree identification system and a storage medium.
Background
Pine wood nematode disease (Bursaphelenchus xylophilus) is a disease and pest caused by pine wood nematodes, which can rapidly spread in forest to cause large-area infection of trees and die, pine trees are taken as main infection objects, and are one of the most dangerous forest biological disasters in China, and the disease causes over 5000 tens of thousands of plants of the dead pine trees in 35 years from 1982 to 2017, so that the economic loss reaches billions of yuan, and the ecology and economy of Chinese forestry are greatly lost. The outside of the disease tree shows the visual characteristics of needle leaf fading and browning and water loss wilting and sagging, the traditional pine wood nematode disease monitoring mainly comprises the steps that a monitor periodically inspects and monitors the color-changing pine tree showing the symptoms of the pine wood nematode and timely cleans up infected and dead disease trees, but due to the fact that relevant monitoring stations and corresponding monitoring staff are not established in part of forest areas, the pine wood nematode disease monitoring is incomplete, the disease tree of the pine wood nematode disease cannot be timely and accurately detected, and the spread of epidemic situation is accelerated.
The prior art discloses an intelligent identification method for a pine wood nematode disease occurrence area, which applies remote sensing image data and deep learning technology to the field of pine wood nematode disease monitoring. And constructing a semantic segmentation sample data set of the pine wood nematode disease occurrence area based on the remote sensing image, constructing a UNet semantic segmentation model, and training and optimizing the model to realize intelligent identification of the pine wood nematode disease occurrence area. However, when the label picture is manufactured, only pictures containing the dead wood of the pine wood nematode disease are marked, and pictures without the dead wood of the pine wood nematode disease are marked as background types, so that the characteristics of the background type pictures are not obvious, the semantic segmentation model is difficult to accurately identify the area which does not contain the pine wood nematode disease tree as the background type, and the precision of identifying the pine wood nematode disease tree is low.
Disclosure of Invention
The invention aims to disclose a method for identifying pine wood nematode disease trees and predicting positions with higher accuracy.
In order to achieve the above object, the present invention provides a pine wood nematode disease tree identification method, comprising the steps of:
s0: acquiring a remote sensing image of a target area, and cutting the remote sensing image into a plurality of images to be identified with the same size;
s1: acquiring an image to be identified;
s2: inputting an image to be identified into a preliminary identification model of the pine wood nematode disease tree, and outputting a distribution diagram of a suspected pine wood nematode disease tree;
s3: inputting the image to be identified into a ground object classification model, and outputting a ground object distribution map; the species of the ground object comprise pine wood nematode disease trees, healthy pine trees, fir trees, bamboo, weeds, bare soil, construction land and cultivated land, wherein the pine wood nematode disease trees are specifically pine wood nematode disease infected pine wood nematode disease trees;
s4: superposing a distribution diagram of the suspected pine wood nematode disease tree and a ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematodes to obtain a spatial distribution diagram of the pine wood nematode disease tree;
s5: repeating the steps S1-S4, and splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified to obtain the spatial distribution diagram of the pine wood nematode disease trees in the target area.
Further, the preliminary identification model of pine wood nematode disease tree is determined by the following method:
constructing a preliminary identification model of pine wood nematode disease tree;
acquiring a data set A, wherein the data set A comprises typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode trees in the target area, and the number of the typical pictures of the ground objects in the target area is equal to the number of the typical pictures of the suspected pine wood nematode trees in the target area;
and training the preliminary identification model of the pine wood nematode disease tree through the data set A to obtain a trained preliminary identification model of the pine wood nematode disease tree.
Further, the data set a is obtained by:
obtaining typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode disease trees in the target area, wherein the number of the typical pictures of the ground objects in the target area is larger than that of the typical pictures of the suspected pine wood nematode disease trees in the target area;
and carrying out random up-down overturn, random left-right overturn or random brightness change on the typical pictures of the suspected pine wood nematode disease tree in the target area, so that the number of the typical pictures of the ground object in the target area is equal to the number of the typical pictures of the suspected pine wood nematode disease tree in the target area.
Further, the ground object classification model is determined by:
constructing a ground object classification model;
acquiring a data set B, wherein the data set B comprises typical pictures of ground features which can not be transmitted by the pine wood nematodes in a target area and typical pictures of ground features which can be transmitted by the pine wood nematodes in the target area, and the number of the typical pictures of the ground features which can not be transmitted by the pine wood nematodes in the target area is equal to the number of the typical pictures of the ground features which can be transmitted by the pine wood nematodes in the target area;
and training the ground object classification model through the data set B to obtain a trained ground object classification model.
Further, the pine wood nematode disease tree preliminary identification model is a deep_v3+ semantic segmentation model based on ResNet.
Further, the ground object classification model is an image classification model constructed by a ResNet network.
In addition, the invention also provides a pine wood nematode disease tree identification system, which comprises:
the acquisition module is used for acquiring the image to be identified;
the first identification module is used for inputting an image to be identified into the preliminary identification model of the pine wood nematode disease tree and outputting a distribution diagram of a suspected pine wood nematode disease tree;
the second recognition module is used for inputting the image to be recognized into the ground feature classification model and outputting a ground feature distribution map;
and the fusion module is used for superposing the distribution diagram of the suspected pine wood nematode disease tree and the ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematodes to obtain the spatial distribution diagram of the pine wood nematode disease tree.
Further, the method further comprises the following steps:
the clipping module is used for acquiring a remote sensing image of the target area and clipping the remote sensing image into a plurality of images to be identified with the same size;
the splicing module is used for acquiring the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, and acquiring the spatial distribution diagrams of the pine wood nematode disease trees in the target area.
Furthermore, the present invention provides a computer-readable storage medium, on which a computer program comprising a pine wood nematode disease tree identification method is stored, which when executed by a processor, implements the steps of the pine wood nematode disease tree identification method.
Compared with the prior art, the beneficial effects of the method are as follows:
according to the invention, through identifying the suspected pine wood nematode disease tree and the ground object species in the image, the suspected pine wood nematode disease tree overlapped with the ground object which can not be transmitted by the pine wood nematode is subjected to reverse masking, so that the suspected pine wood nematode disease tree overlapped with the ground object which can not be transmitted by the pine wood nematode is determined to be not the pine wood nematode disease tree in fact, other ground objects similar to the image characteristics of the pine wood nematode disease tree are removed, the probability that other ground objects are mispredicted as the suspected pine wood nematode disease tree is reduced, and the identification precision and the accuracy of the pine wood nematode disease tree are improved.
Drawings
FIG. 1 is a flow chart of a pine wood nematode disease tree identification method of an embodiment of the present invention;
FIG. 2 is a block diagram of a system for identifying pine wood nematode disease tree according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a ResNet network in an embodiment of the present invention;
fig. 4 is a schematic structural diagram of a deep_v3+ semantic segmentation model in an embodiment of the present invention.
Detailed Description
The following describes in further detail the embodiments of the present invention with reference to the drawings and examples. The following examples are illustrative of the invention and are not intended to limit the scope of the invention.
Embodiment one:
as shown in fig. 1, a pine wood nematode disease tree identification method according to a preferred embodiment of the present invention includes the following steps:
s1: acquiring an image to be identified;
s2: inputting an image to be identified into a preliminary identification model of the pine wood nematode disease tree, and outputting a distribution diagram of a suspected pine wood nematode disease tree;
s3: inputting the image to be identified into a ground object classification model, and outputting a ground object distribution map;
s4: and superposing the distribution diagram of the suspected pine wood nematode disease tree and the ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematode to obtain the spatial distribution diagram of the pine wood nematode disease tree.
According to the method, the suspected pine wood nematode disease tree which is overlapped with the ground object which cannot be transmitted by the pine wood nematodes is subjected to reverse masking by identifying the suspected pine wood nematode disease tree and the ground object type in the image, so that the suspected pine wood nematode disease tree which is overlapped with the ground object which cannot be transmitted by the pine wood nematodes can be determined to be not the suspected pine wood nematode disease tree in fact, the probability that other ground objects are mispredicted to be the suspected pine wood nematode disease tree is reduced, and the identification precision of the pine wood nematode disease tree is improved.
Further, the preferred embodiment of the invention provides a pine wood nematode disease tree identification method, which comprises the following steps:
s0: acquiring a remote sensing image of a target area, and cutting the remote sensing image into a plurality of images to be identified with the same size;
it should be noted that, by taking a plurality of photographs at a plurality of places in the target area by the unmanned aerial vehicle, calibration and mosaic processing are performed on the plurality of photographs, and a remote sensing image of the target area is obtained. The pine wood nematode disease tree is scattered and scattered on the high-resolution satellite remote sensing image, so that the pine wood nematode disease tree can be only represented by a plurality of to tens of unequal pixels, the possibility of identifying the pine wood nematode disease tree through the high-resolution satellite remote sensing image is low, the unmanned aerial vehicle shooting photo has the advantages of large scale, high resolution, low cost and the like, and the unmanned aerial vehicle shooting photo is far less influenced by the shielding of cloud layers than the satellite remote sensing image, so that the obtained image scale is larger, the resolution is larger, the single plant disease tree can be identified, the identification effect is better, the cost for acquiring the remote sensing image of the target area is lower, the remote sensing image can be completely identified by taking the fact that the model of the remote sensing image is difficult to train, the identification speed is slow, the remote sensing image is cut into a plurality of images to be identified with the same size, the model to be identified can be trained quickly, and the identification speed is high, and the training effect is improved.
In this embodiment, an unmanned aerial vehicle with a model number of Phantom 4 PRO and a model number of FC6520 is adopted, the unmanned aerial vehicle has a flying relative height of 154 meters, a heading overlapping rate and a side overlapping rate of 80%, the unmanned aerial vehicle maintains an equidistant shooting state, after shooting is finished, difference solution data after POS is associated to each photo by using Pix 4d software, so that the photo has longitude and latitude coordinates, height, unmanned aerial vehicle heading and attitude angle information when the photo is shot, calibration and mosaic processing are performed on the photo according to the information, and a remote sensing image with a coordinate system of WGS-84 and a spatial resolution of 0.033738273 m is output after the remote sensing image is spliced, wherein the remote sensing image is a digital orthographic image. And cutting the remote sensing image into a plurality of images to be identified with the same size and the image resolution of 288 x 288 so as to improve the training speed and the identification speed of the depth model.
S1: acquiring an image to be identified;
it should be noted that, in this embodiment, an image to be identified with a resolution of 288×288 after a remote sensing image is cut is obtained.
S2: inputting an image to be identified into a preliminary identification model of the pine wood nematode disease tree, and outputting a distribution diagram of a suspected pine wood nematode disease tree;
in this embodiment, the preliminary identification model of the pine wood nematode disease is a deep_v3+ semantic segmentation model based on a ResNet, as shown in fig. 3 and fig. 4, the ResNet network has a residual learning structure which does not degrade with the increase of the network layer number, gradient dispersion or gradient explosion caused by the increase of the convolutional neural network layer number in the model of the ground object classification model is avoided, the deep_v3+ semantic segmentation model has excellent multi-scale feature capturing capability and boundary detail processing capability, and the deep_v3+ semantic segmentation model based on the ResNet has excellent feature extraction effect, so that the identification precision of the preliminary identification model of the pine wood nematode disease can be improved.
S3: inputting the image to be identified into a ground object classification model, and outputting a ground object distribution map;
in this embodiment, the ground object classification model is an image classification model constructed by a res net network, so that the recognition accuracy of the preliminary recognition model of the pine wood nematode disease tree can be improved. The species of the ground object comprise pine wood nematode disease trees, healthy pine trees, fir trees, bamboo, weeds, bare soil, construction land and cultivated land, wherein the pine wood nematode disease trees are specifically pine wood nematode disease infected pine wood.
S4: and superposing the distribution diagram of the suspected pine wood nematode disease tree and the ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematode to obtain the spatial distribution diagram of the pine wood nematode disease tree.
In this embodiment, the ground objects that the pine nematodes cannot spread include fir, bamboo, weeds, bare soil, construction land, and cultivated land.
S5: repeating the steps S1-S4, and splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified to obtain the spatial distribution diagram of the pine wood nematode disease trees in the target area.
According to the invention, through identifying the suspected pine wood nematode disease tree and the ground object type in the image, the suspected pine wood nematode disease tree overlapped with the fir tree, the bamboo, the weed, the bare soil, the construction land and the cultivated land six types of ground objects is subjected to reverse masking, so that the suspected pine wood nematode disease tree overlapped with the fir tree, the bamboo, the weed, the bare soil, the construction land and the cultivated land six types of ground objects can be determined to be not the suspected pine wood nematode disease tree in fact, especially the suspected pine wood nematode disease tree overlapped with the fir tree is more likely to be dead fir tree, the dead fir tree is similar to the pine tree infected with the pine wood nematode disease in appearance, the probability that other ground objects are mispredicted as the suspected pine wood nematode disease tree is reduced, and the identification precision of the pine wood nematode disease tree is improved.
The preliminary identification model of pine wood nematode disease is determined by the following method:
step 301: constructing a preliminary identification model of pine wood nematode disease tree;
in this embodiment, the preliminary identification model of the pine wood nematode disease tree is a deep_v3+ semantic segmentation model based on ResNet, so that the identification accuracy of the preliminary identification model of the pine wood nematode disease tree can be improved.
Step 302: acquiring a data set A, wherein the data set A comprises typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode trees in the target area, and the number of the typical pictures of the ground objects in the target area is equal to the number of the typical pictures of the suspected pine wood nematode trees in the target area;
in the present embodiment, the data set a is acquired by:
obtaining typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode disease trees in the target area, wherein the number of the typical pictures of the ground objects in the target area is larger than that of the typical pictures of the suspected pine wood nematode disease trees in the target area;
the photographs obtained by the unmanned aerial vehicle were visually interpreted, and typical pictures of eight types of land features, i.e., pine, healthy pine, fir, bamboo, weed, bare soil, construction land, and cultivated land, were selected, and typical pictures of healthy pine, fir, bamboo, weed, bare soil, construction land, and cultivated land were taken as typical pictures of land features in the target area, and typical pictures of pine were taken as typical pictures of suspected pine in the target area. In this embodiment, a total of 2000 pictures are selected, wherein the pictures comprise 400 pine tree diseases, 300 healthy pine trees, 500 fir trees, 100 Zhang Zhu, 100 weeds, 200 bare soil, 200 construction lands and 200 cultivated lands, namely, a total of 400 typical pictures of suspected pine tree diseases in a target area and 1600 typical pictures of land features in a target area.
The typical pictures of the 400 pine wood nematode disease trees are processed based on an object-oriented image segmentation method, image segmentation vector graphics corresponding to the typical pictures of the 400 pine wood nematode disease trees are obtained, pixel values of segmentation areas corresponding to the pine wood nematode disease trees on the image segmentation vector graphics are marked as 1 according to the typical pictures of the pine wood nematode disease trees, other ground object types are marked as 0, vector grid conversion operation is carried out on the marked vector graphics, and tag pictures with the resolution of 288 x 288 are output. And generating tag pictures with the resolution of 288 x 288 for typical pictures of ground objects in 1600 target areas, wherein the pixel values of the tag pictures are all 0.
And carrying out random up-down overturn, random left-right overturn or random brightness change on the typical pictures of the suspected pine wood nematode disease tree in the target area, so that the number of the typical pictures of the ground object in the target area is equal to the number of the typical pictures of the suspected pine wood nematode disease tree in the target area.
It should be noted that, the typical pictures of 400 pine wood nematode trees are randomly turned up and down, randomly turned left and right or randomly changed in brightness, so that the typical pictures of the pine wood nematode trees are amplified to 1600, and the typical pictures of ground objects in other 1600 target areas are kept unchanged, wherein the maximum absolute value of the random brightness change is 0.1. Typical pictures of suspected pine wood nematode trees in the 1600 target areas after amplification and typical pictures of ground objects in the 1600 target areas together form a data set A, and tag pictures corresponding to the 3200 pictures. The number of typical pictures of suspected pine wood nematode trees in a target area is amplified by means of data enhancement such as random up-down overturn, random left-right overturn or random brightness change, the number of pictures processed by an object-oriented image segmentation method can be greatly reduced, the workload of manual visual interpretation is greatly reduced, and the effect of processing and manually identifying thousands of typical pictures of suspected pine wood nematode trees in the target area can be achieved only by processing and manually identifying fewer typical pictures of suspected pine wood nematode trees in the target area.
Step 303: and training the preliminary identification model of the pine wood nematode disease tree through the data set A, and outputting the trained preliminary identification model of the pine wood nematode disease tree.
The method comprises the steps of inputting an image to be identified into a trained preliminary identification model of the pine wood nematode disease tree, outputting the image to be identified and a label picture corresponding to the image to be identified, wherein the label picture is a distribution map of the suspected pine wood nematode disease tree, and a region with a pixel value of 1 on the label picture is a distribution region of the suspected pine wood nematode disease tree.
The ground object classification model is determined by the following method:
step 401: constructing a ground object classification model;
in this embodiment, the ground object classification model is an image classification model constructed by a res net network.
Step 402: acquiring a data set B, wherein the data set B comprises typical pictures of ground features which can not be transmitted by the pine wood nematodes in a target area and typical pictures of ground features which can be transmitted by the pine wood nematodes in the target area, and the number of the typical pictures of the ground features which can not be transmitted by the pine wood nematodes in the target area is equal to the number of the typical pictures of the ground features which can be transmitted by the pine wood nematodes in the target area;
the pictures obtained by the unmanned aerial vehicle are visually interpreted, and typical pictures of eight types of ground features, namely pine tree disease, healthy pine tree, fir tree, bamboo, weeds, bare soil, construction land and cultivated land, are selected, wherein the typical pictures of the two types of ground features of pine tree disease and healthy pine tree are typical pictures of ground features which can be transmitted by pine nematodes in a target area, and the typical pictures of six types of ground features of fir tree, bamboo, weeds, bare soil, construction land and cultivated land are typical pictures of ground features which can not be transmitted by pine nematodes in the target area. In this embodiment, the typical pictures of the ground objects which can be transmitted by the pine wood nematodes in the 700 target areas and the ground objects which can not be transmitted by the pine wood nematodes in the 1300 target areas are included.
And cutting the central area of the 2000 pictures to obtain the labels of the typical pictures of the eight types of ground features with the size of 2000 pixels 240 by 240 and the ground feature types corresponding to the typical pictures. The cut typical pictures can more highlight typical characteristics of the typical pictures, the classification effect of the image classification model is improved, and experiments show that the typical pictures with the size can further improve the classification effect of the image classification model.
And carrying out random up-down overturn, random left-right overturn or random brightness change on typical pictures of the ground objects which can be transmitted by the pine wood nematodes in the 700 target areas, so that the typical pictures of the ground objects which can be transmitted by the pine wood nematodes in the target areas are amplified to 1300, the ground objects which cannot be transmitted by the pine wood nematodes in other 1300 target areas are kept unchanged, and the maximum absolute value of the random brightness change is 0.1. The amplified land feature which can be transmitted by the pine wood nematodes in the 1300 target areas and the land feature which can not be transmitted by the pine wood nematodes in the 1300 target areas form a data set B together with the tags of the land feature types corresponding to the 2600 pictures. The quantity of the ground objects which can be transmitted by the pine wood nematodes in the target area is amplified by means of data enhancement such as random up-down overturn, random left-right overturn or random brightness change, the workload of manual visual interpretation can be greatly reduced, and the effect of manually identifying the typical pictures of thousands of ground objects which can be transmitted by the pine wood nematodes can be achieved only by manually identifying the typical pictures of the fewer ground objects which can be transmitted by the pine wood nematodes.
Step 403: and training the ground object classification model through the data set B to obtain a trained ground object classification model.
It should be noted that, the image to be identified is input to the trained ground object classification model, and the image to be identified with the size of 240×240 pixels and the ground object type corresponding to the image to be identified are output. The feature distribution map is 288 x 288 of the pixel size, the image to be identified and the feature type corresponding to the image to be identified.
In the prior art, the evaluation method of remote sensing image classification is mostly constructed based on whether single pixels are accurately predicted in category, in the actual production application of identifying pine wood nematode disease on high-resolution unmanned aerial vehicle images, the accurate position of the pine wood nematode disease on the pixel level is not required to be determined, whether the pine wood nematode disease exists in the grid range is not required, according to the identification and positioning of the grid level area, prevention and control means such as deforestation or unmanned aerial vehicle application are implemented, in the embodiment, each image to be identified is taken as one grid in the grid, through classifying the ground object category of each image to be identified, whether the possibility of the pine wood nematode disease exists in the grid where the image to be identified exists is judged, if the possibility of the pine wood nematode disease exists, even if the suspected pine wood nematode disease exists in the grid, the suspected pine wood nematode disease does not need to be continuously focused on the grid, if the possibility of the pine wood nematode disease exists in the grid is not required, and the suspected pine wood nematode disease identified in the grid is positioned, or unmanned aerial vehicle application is implemented. The invention can improve the model training speed by reducing the unnecessary recognition accuracy of image classification, and can rapidly and accurately realize the recognition of the pine wood nematode disease tree in the target area.
Embodiment two:
as shown in fig. 2, a preferred embodiment of the present invention provides a pine wood nematode disease tree identification system, comprising:
the acquisition module is used for acquiring the image to be identified;
the first identification module is used for inputting an image to be identified into the preliminary identification model of the pine wood nematode disease tree and outputting a distribution diagram of a suspected pine wood nematode disease tree;
the second recognition module is used for inputting the image to be recognized into the ground feature classification model and outputting a ground feature distribution map;
and the fusion module is used for superposing the distribution diagram of the suspected pine wood nematode disease tree and the ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematodes to obtain the spatial distribution diagram of the pine wood nematode disease tree.
According to the method, the suspected pine wood nematode disease tree which is overlapped with the ground object which cannot be transmitted by the pine wood nematodes is subjected to reverse masking by identifying the suspected pine wood nematode disease tree and the ground object type in the image, so that the suspected pine wood nematode disease tree which is overlapped with the ground object which cannot be transmitted by the pine wood nematodes can be determined to be not the suspected pine wood nematode disease tree in fact, the probability that other ground objects are mispredicted to be the suspected pine wood nematode disease tree is reduced, and the identification precision of the pine wood nematode disease tree is improved.
Further, the method further comprises the following steps:
the clipping module is used for acquiring a remote sensing image of the target area and clipping the remote sensing image into a plurality of images to be identified with the same size;
the splicing module is used for acquiring the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, and acquiring the spatial distribution diagrams of the pine wood nematode disease trees in the target area.
It should be noted that, considering that when the area of the target area is larger, the remote sensing image contains much information, a model capable of completely identifying the remote sensing image is not only difficult to train, but also has a low identification speed, and the remote sensing image is cut into a plurality of images to be identified with the same size, so that the model capable of identifying the images to be identified can be quickly trained, and the identification speed is also high.
Embodiment III:
the preferred embodiment of the present invention provides a computer-readable storage medium having stored thereon a computer program comprising a pine wilt tree identification method, which when executed by a processor, implements the steps of the pine wilt tree identification method.
In summary, the embodiments of the present invention provide a method, a system, and a storage medium for identifying a suspected pine wood nematode disease tree and a ground object species in an image, where a reverse mask is performed on a suspected pine wood nematode disease tree that coincides with a ground object that cannot be transmitted by a pine wood nematode, so that it can be determined that a suspected pine wood nematode disease tree that coincides with a ground object that cannot be transmitted by a pine wood nematode is not actually a suspected pine wood nematode disease tree, the probability that other ground objects are mispredicted as a suspected pine wood nematode disease tree is reduced, and the identification accuracy of the pine wood nematode disease tree is improved.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and substitutions can be made by those skilled in the art without departing from the technical principles of the present invention, and these modifications and substitutions should also be considered as being within the scope of the present invention.
Claims (6)
1. The pine wood nematode disease tree identification method is characterized by comprising the following steps:
s0: acquiring a remote sensing image of a target area, and cutting the remote sensing image into a plurality of images to be identified with the same size;
s1: acquiring an image to be identified;
s2: inputting an image to be identified into a preliminary identification model of the pine wood nematode disease tree, and outputting a distribution diagram of a suspected pine wood nematode disease tree; the preliminary identification model of pine wood nematode disease is determined by the following method:
constructing a preliminary identification model of pine wood nematode disease tree;
acquiring a data set A, wherein the data set A comprises typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode trees in the target area, and the number of the typical pictures of the ground objects in the target area is equal to the number of the typical pictures of the suspected pine wood nematode trees in the target area;
training the preliminary identification model of the pine wood nematode disease tree through the data set A to obtain a trained preliminary identification model of the pine wood nematode disease tree;
the species of the ground object comprise pine wood nematode disease trees, healthy pine trees, fir trees, bamboo, weeds, bare soil, construction land and cultivated land, wherein the pine wood nematode disease trees are specifically pine wood nematode disease infected pine wood nematode disease trees;
s3: inputting the image to be identified into a ground object classification model, and outputting a ground object distribution map; the ground object classification model is determined by the following method:
constructing a ground object classification model;
acquiring a data set B, wherein the data set B comprises typical pictures of ground features which can not be transmitted by the pine wood nematodes in a target area and typical pictures of ground features which can be transmitted by the pine wood nematodes in the target area, and the number of the typical pictures of the ground features which can not be transmitted by the pine wood nematodes in the target area is equal to the number of the typical pictures of the ground features which can be transmitted by the pine wood nematodes in the target area;
training the ground object classification model through the data set B to obtain a trained ground object classification model;
s4: superposing a distribution diagram of the suspected pine wood nematode disease tree and a ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematodes to obtain a spatial distribution diagram of the pine wood nematode disease tree;
s5: repeating the steps S1-S4, and splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified to obtain the spatial distribution diagram of the pine wood nematode disease trees in the target area.
2. The method for identifying pine wood nematode disease tree according to claim 1, wherein the data set a is obtained by:
obtaining typical pictures of ground objects in a target area and typical pictures of suspected pine wood nematode disease trees in the target area, wherein the number of the typical pictures of the ground objects in the target area is larger than that of the typical pictures of the suspected pine wood nematode disease trees in the target area;
and carrying out random up-down overturn, random left-right overturn or random brightness change on the typical pictures of the suspected pine wood nematode disease tree in the target area, so that the number of the typical pictures of the ground object in the target area is equal to the number of the typical pictures of the suspected pine wood nematode disease tree in the target area.
3. The method for identifying pine wood nematode disease tree according to claim 1, wherein the preliminary identification model of pine wood nematode disease tree is a deep_v3+ semantic segmentation model based on ResNet.
4. The method of claim 1, wherein the ground object classification model is an image classification model constructed with a res net network.
5. A pine wood nematode disease tree identification system, comprising:
the clipping module is used for acquiring a remote sensing image of the target area and clipping the remote sensing image into a plurality of images to be identified with the same size;
the acquisition module is used for acquiring the image to be identified;
the first identification module is used for inputting an image to be identified into the preliminary identification model of the pine wood nematode disease tree and outputting a distribution diagram of a suspected pine wood nematode disease tree;
the second recognition module is used for inputting the image to be recognized into the ground feature classification model and outputting a ground feature distribution map;
the fusion module is used for superposing a distribution diagram of the suspected pine wood nematode disease tree and a ground object distribution diagram, and carrying out reverse masking on the suspected pine wood nematode disease tree overlapped with the ground object which cannot be transmitted by the pine wood nematodes to obtain a spatial distribution diagram of the pine wood nematode disease tree;
the splicing module is used for acquiring the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, splicing the spatial distribution diagrams of the pine wood nematode disease trees corresponding to all the images to be identified, and acquiring the spatial distribution diagrams of the pine wood nematode disease trees in the target area.
6. A computer readable storage medium having stored thereon a computer program, wherein the computer program, when executed by a processor, implements the pine wood nematode disease tree identification method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210486161.0A CN114782844B (en) | 2022-05-06 | 2022-05-06 | Pine wood nematode disease tree identification method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210486161.0A CN114782844B (en) | 2022-05-06 | 2022-05-06 | Pine wood nematode disease tree identification method, system and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114782844A CN114782844A (en) | 2022-07-22 |
| CN114782844B true CN114782844B (en) | 2023-05-12 |
Family
ID=82435764
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210486161.0A Active CN114782844B (en) | 2022-05-06 | 2022-05-06 | Pine wood nematode disease tree identification method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114782844B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118314478B (en) * | 2024-04-22 | 2024-10-11 | 中国烟草总公司湖南省公司 | Method, medium and system for identifying aphid infectious viruses in large scale |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180010718A (en) * | 2016-07-22 | 2018-01-31 | 주식회사 엔젠소프트 | Apparatus for detecting of bursaphelenchus xylophilus, method and system using thereof |
| CN108460760A (en) * | 2018-03-06 | 2018-08-28 | 陕西师范大学 | A kind of Bridge Crack image discriminating restorative procedure fighting network based on production |
| CN108535193A (en) * | 2018-03-19 | 2018-09-14 | 电子科技大学 | A kind of forestry typical case pest and disease damage remote-sensing monitoring method |
| CN109766744A (en) * | 2018-11-21 | 2019-05-17 | 北京农业智能装备技术研究中心 | A kind of identification of Bursaphelenchus xylophilus sick tree and localization method and system |
| CN109948563A (en) * | 2019-03-22 | 2019-06-28 | 华南农业大学 | A kind of withered tree detection localization method of the pine nematode based on deep learning |
| CN114373140A (en) * | 2022-01-13 | 2022-04-19 | 国家林业和草原局生物灾害防控中心 | Intelligent identification method for pine wood nematode disease occurrence area |
| CN114387528A (en) * | 2021-12-29 | 2022-04-22 | 浙江同创空间技术有限公司 | Pine nematode disease monitoring space-air-ground integrated monitoring method |
-
2022
- 2022-05-06 CN CN202210486161.0A patent/CN114782844B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180010718A (en) * | 2016-07-22 | 2018-01-31 | 주식회사 엔젠소프트 | Apparatus for detecting of bursaphelenchus xylophilus, method and system using thereof |
| CN108460760A (en) * | 2018-03-06 | 2018-08-28 | 陕西师范大学 | A kind of Bridge Crack image discriminating restorative procedure fighting network based on production |
| CN108535193A (en) * | 2018-03-19 | 2018-09-14 | 电子科技大学 | A kind of forestry typical case pest and disease damage remote-sensing monitoring method |
| CN109766744A (en) * | 2018-11-21 | 2019-05-17 | 北京农业智能装备技术研究中心 | A kind of identification of Bursaphelenchus xylophilus sick tree and localization method and system |
| CN109948563A (en) * | 2019-03-22 | 2019-06-28 | 华南农业大学 | A kind of withered tree detection localization method of the pine nematode based on deep learning |
| CN114387528A (en) * | 2021-12-29 | 2022-04-22 | 浙江同创空间技术有限公司 | Pine nematode disease monitoring space-air-ground integrated monitoring method |
| CN114373140A (en) * | 2022-01-13 | 2022-04-19 | 国家林业和草原局生物灾害防控中心 | Intelligent identification method for pine wood nematode disease occurrence area |
Non-Patent Citations (4)
| Title |
|---|
| identitying pine wood nematode disease using UAV images and deep learning algorithms;Jun Qin et al.;《arXiv》;第1-6页 * |
| 基于决策树的无人机高光谱遥感影像地物分类研究;万欢等;《河北农业科学》;第23卷(第01期);第101-104页 * |
| 基于无人机影像的一种农田景观小尺度地物的分类方法研究;陈柳;《中国优秀硕士学位论文全文数据库农业科技辑》(第第08期期);第D043-23页 * |
| 松材线虫病变色松树遥感监测研究进展;陶欢等;《林业科学研究》;第33卷(第03期);第172-183页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114782844A (en) | 2022-07-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ampatzidis et al. | Agroview: Cloud-based application to process, analyze and visualize UAV-collected data for precision agriculture applications utilizing artificial intelligence | |
| Ampatzidis et al. | Citrus rootstock evaluation utilizing UAV-based remote sensing and artificial intelligence | |
| Guo et al. | Aerial imagery analysis–quantifying appearance and number of sorghum heads for applications in breeding and agronomy | |
| Safonova et al. | Detection of Norway spruce trees (Picea abies) infested by bark beetle in UAV images using YOLOs architectures | |
| CN109033937B (en) | Method and system for counting plant number through unmanned aerial vehicle image | |
| Röder et al. | Application of optical unmanned aerial vehicle-based imagery for the inventory of natural regeneration and standing deadwood in post-disturbed spruce forests | |
| CN110070571B (en) | Phyllostachys pubescens morphological parameter detection method based on depth camera | |
| CN111582234A (en) | UAV and deep learning-based large-range oil tea forest fruit intelligent detection and counting method | |
| Hu et al. | UAV remote sensing monitoring of pine forest diseases based on improved Mask R-CNN | |
| CN114239756B (en) | Insect pest detection method and system | |
| CN115115954A (en) | Intelligent identification method for pine nematode disease area color-changing standing trees based on unmanned aerial vehicle remote sensing | |
| CN114782844B (en) | Pine wood nematode disease tree identification method, system and storage medium | |
| CN113379738A (en) | Method and system for detecting and positioning epidemic trees based on images | |
| CN117392382A (en) | Single tree fruit tree segmentation method and system based on multi-scale dense instance detection | |
| CN112084989A (en) | Unmanned aerial vehicle and CNN-based large-range pine wood nematode withered vertical wood intelligent detection method | |
| Lyu et al. | Development of phenotyping system using low altitude UAV imagery and deep learning | |
| CN109657540B (en) | Withered tree positioning method and system | |
| CN113807309B (en) | Deep learning-based orchard machine walking route planning method | |
| Paul et al. | Estimation of shade tree density in tea garden using remote sensing images and deep convolutional neural network | |
| CN113033386B (en) | High-resolution remote sensing image-based transmission line channel hidden danger identification method and system | |
| Ozguven et al. | The technology uses in the determination of sugar beet diseases | |
| Miao et al. | Crop weed identification system based on convolutional neural network | |
| Huang et al. | Recognition and counting of pitaya trees in karst mountain environment based on unmanned aerial vehicle RGB images | |
| CN116721385A (en) | Machine learning-based RGB camera data cyanobacteria bloom monitoring method | |
| Yano et al. | Weed identification in sugarcane plantation through images taken from remotely piloted aircraft (RPA) and KNN classifier |
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 |