CN117392668A - Wheat scab state evaluation method and system and electronic equipment - Google Patents
Wheat scab state evaluation method and system and electronic equipment Download PDFInfo
- Publication number
- CN117392668A CN117392668A CN202311399450.8A CN202311399450A CN117392668A CN 117392668 A CN117392668 A CN 117392668A CN 202311399450 A CN202311399450 A CN 202311399450A CN 117392668 A CN117392668 A CN 117392668A
- Authority
- CN
- China
- Prior art keywords
- wheat
- ear
- scab
- disease
- real
- 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.)
- Pending
Links
- 241000209140 Triticum Species 0.000 title claims abstract description 354
- 235000021307 Triticum Nutrition 0.000 title claims abstract description 353
- 206010039509 Scab Diseases 0.000 title claims abstract description 177
- 238000011156 evaluation Methods 0.000 title abstract description 16
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 115
- 241000495841 Oenanthe oenanthe Species 0.000 claims abstract description 92
- 238000001514 detection method Methods 0.000 claims abstract description 86
- 238000000034 method Methods 0.000 claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 39
- 201000010099 disease Diseases 0.000 claims description 83
- 241001024327 Oenanthe <Aves> Species 0.000 claims description 67
- 238000003062 neural network model Methods 0.000 claims description 62
- 210000005069 ears Anatomy 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 11
- 238000002372 labelling Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 208000032625 disorder of ear Diseases 0.000 claims description 5
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000003860 storage Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 10
- 230000004927 fusion Effects 0.000 description 9
- 238000005070 sampling Methods 0.000 description 9
- 238000004364 calculation method Methods 0.000 description 8
- 239000000284 extract Substances 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000000750 progressive effect Effects 0.000 description 7
- 238000011160 research Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 6
- 238000011176 pooling Methods 0.000 description 6
- 230000001629 suppression Effects 0.000 description 6
- 238000012545 processing Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 208000024891 symptom Diseases 0.000 description 4
- 238000002054 transplantation Methods 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 230000011218 segmentation Effects 0.000 description 2
- 241000196324 Embryophyta Species 0.000 description 1
- 241000282414 Homo sapiens Species 0.000 description 1
- 208000005374 Poisoning Diseases 0.000 description 1
- 206010047700 Vomiting Diseases 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 230000036528 appetite Effects 0.000 description 1
- 235000019789 appetite Nutrition 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 235000013339 cereals Nutrition 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013305 food Nutrition 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 208000021245 head disease Diseases 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 230000002688 persistence Effects 0.000 description 1
- 231100000572 poisoning Toxicity 0.000 description 1
- 230000000607 poisoning effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/68—Food, e.g. fruit or vegetables
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
- G06V10/451—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters with interaction between the filter responses, e.g. cortical complex cells
- G06V10/454—Integrating the filters into a hierarchical structure, e.g. convolutional neural networks [CNN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Biodiversity & Conservation Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a wheat scab state evaluation method and system and electronic equipment for wheat fields, and relates to the technical field of wheat scab detection. The method comprises the steps of inputting a real-time wheat image into a wheat head identification model to obtain a plurality of real-time single wheat head images; and respectively inputting the real-time single wheat ear images into a wheat head scab detection model to obtain a wheat head scab detection result of each real-time single wheat ear image corresponding to the wheat ear, and further determining the wheat head scab disease state of the area of the wheat Tian Daice. According to the wheat head scab detection method, the wheat head scab detection efficiency, accuracy and rationality can be improved by constructing and training the wheat head identification model and the wheat head scab detection model.
Description
Technical Field
The invention relates to the technical field of wheat scab detection, in particular to a wheat scab state evaluation method and system in a wheat field and electronic equipment.
Background
Wheat is one of the most widely planted crops worldwide, wheat diseases are one of important factors affecting wheat quality and yield, and wheat scab is one of common diseases in the wheat planting process. Wheat scab causes spike rot symptoms in the wheat head portion and thus reduces wheat yield. More seriously, consumers can cause poisoning symptoms such as appetite reduction, nausea and vomiting after carelessly eating the wheat with diseases, and the health and food safety of human beings are endangered. Therefore, the development of early nondestructive, rapid and portable detection method research on wheat scab is of great significance in guaranteeing agricultural production and grain safety.
The traditional wheat spike and wheat scab detection method is usually manual visual screening, and whether wheat is infected with scab and scab incidence are visually investigated in fields. The manual screening method is time-consuming and labor-consuming, has low efficiency, is easily influenced by factors such as visual fatigue and the like in actual operation, and has the characteristics of low efficiency and low persistence in the detection process.
At present, the hyperspectral technology is adopted to detect wheat ears and scab in related researches. The hyperspectral image technology combines the traditional imaging technology and the spectrum technology, and can realize high-efficiency, green and high-resolution detection on various samples. The hyperspectral image technology can shoot different positions of wheat ears, small ears and the like of wheat plants, and the onset condition of scab can be judged by analyzing related data. However, when the hyperspectral correlation technique is used, the data amount is large, complex data analysis is required, and the equipment cost is high. The field data collection work cannot be completed rapidly in a large range. When the digital image processing and machine vision technology is used for processing the crop disease detection problem, the damage-free and rapid disease positioning can be realized, the data processing is relatively simple, and the calculation cost is low. Related technologies have been widely used in the field of crop disease detection, and have been developed to some extent. At present, related researches are used for completing the identification of the wheat scab indoors, but the related researches for identifying the scab of the wheat naturally growing in the field are less. In addition, in the related research, when the severity degree of wheat scab is identified, the research is carried out from the viewpoint of dividing the lesion area, however, according to the current national standard, the severity degree of the wheat scab should be judged from the spikelet disease condition of wheat spikes. Therefore, the related research at present also has the technical problems of limited recognition range, inaccurate recognition result and the like.
Disclosure of Invention
The invention aims to provide a wheat scab state evaluation method, a system and electronic equipment, which can improve the efficiency, accuracy and rationality of wheat scab detection and wheat scab state evaluation in a wheat field.
In order to achieve the above object, the present invention provides the following solutions:
a wheat scab status assessment method in wheat fields, comprising:
acquiring a real-time wheat image of a wheat Tian Daice area;
inputting the real-time wheat image into a wheat head identification model to obtain a real-time wheat head identification result of the region to be detected; the wheat head identification model is obtained by training a first lightweight neural network model by utilizing a plurality of marked historical wheat images;
dividing the real-time wheat image according to the real-time wheat head identification result to obtain a plurality of real-time single wheat head images;
respectively inputting a plurality of real-time single wheat ear images into a wheat head scab detection model to obtain a wheat head scab detection result of each wheat ear corresponding to each real-time single wheat ear image; the small ear scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat ear images; the detection result of the head blight of the wheat ears comprises whether each small ear on the wheat ears has head blight or not;
And determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the scab of the multiple spikes.
Optionally, before the acquiring the real-time wheat image of the area of the wheat Tian Daice, the method further includes:
constructing a first lightweight neural network model; the first lightweight neural network model is obtained by adding an EMA attention module before a small target recognition layer and a medium target recognition layer of the lightweight neural network;
acquiring a plurality of historical wheat images;
marking each wheat ear in the historical wheat image respectively to obtain a plurality of marked historical wheat images;
and training the first lightweight neural network model by taking the historical wheat image as input and the marked historical wheat image as output to obtain the wheat ear recognition model.
Optionally, before the acquiring the real-time wheat image of the area of the wheat Tian Daice, the method further includes:
constructing a second lightweight neural network model; the second lightweight neural network model is obtained by adding a front attention module and a rear attention module in front of a small target recognition layer of the lightweight neural network;
cutting the historical wheat images according to the marked frames in the marked historical wheat images to obtain a plurality of historical single wheat ear images;
Labeling whether each spike in the historical single spike image has scab or not to obtain a labeled historical single spike image;
and training the second lightweight neural network model by taking the historical single wheat ear image as input and the marked historical single wheat ear image as output to obtain the wheat head scab detection model.
Optionally, determining the wheat scab disease state of the area of wheat Tian Daice based on the plurality of spike scab detection results includes:
acquiring the number of the real-time single wheat ear images as the total number of wheat ears in the region to be detected;
determining the number of diseased wheat ears according to the detection results of the scab of a plurality of small ears;
determining the ratio of the diseased wheat ears to the total number of the wheat ears as the disease ear rate of the region to be detected; the spike disease rate is used for measuring the wheat scab state of the wheat field.
Optionally, determining the wheat scab disease state of the area of wheat Tian Daice according to the detection results of the plurality of wheat scab further includes:
determining the disease grade of each diseased wheat ear according to the detection results of the scab of a plurality of small ears;
determining the quantity of diseased wheat ears corresponding to different diseased grades;
determining the disease index of the region to be detected according to the total number of wheat ears of the region to be detected and the quantity of diseased wheat ears corresponding to a plurality of disease grades; the disease index is used for measuring the wheat scab state of the wheat field.
Optionally, the disease index is:
wherein I represents the index of the disease, h i The i-th diseased level corresponds to the number of diseased wheat ears, i represents the i-th diseased level, and H represents the total number of wheat ears in the area to be detected.
Optionally, the determining the disease grade of each diseased wheat ear according to the detection result of the scab of a plurality of small ears comprises:
determining any one of the real-time single ear images as a current real-time single ear image, and determining the ear corresponding to the current real-time single ear image as a current ear;
acquiring the number of diseased wheat ears and the total number of the wheat ears of the current wheat ears according to the detection result of the wheat head scab of the current wheat ears;
determining the ratio of the number of diseased small ears of the current wheat ear to the total number of small ears as the ratio of the diseased small ears of the current wheat ear;
and determining the diseased grade of the current wheat spike according to the diseased small spike ratio of the current wheat spike.
Optionally, the determining the disease grade of the current wheat spike according to the disease spike ratio of the current wheat spike includes:
judging whether the disease small ear proportion of the current wheat ear is equal to 0 or not, and obtaining a first judging result;
if the first judgment result is yes, judging that the diseased grade of the current wheat ear is 0;
if the first judgment result is negative, judging whether the current wheat ear disease small ear ratio is smaller than a first disease small ear ratio threshold value or not, and obtaining a second judgment result;
If the second judgment result is yes, judging that the diseased grade of the current wheat ear is 1;
if the second judgment result is negative, judging whether the disease spike ratio of the current wheat spike is smaller than a second disease spike ratio threshold value or not, and obtaining a third judgment result; the second disease spike duty cycle threshold is greater than the first disease spike duty cycle threshold;
if the third judgment result is yes, judging that the diseased grade of the current wheat ear is 2;
if the third judging result is negative, judging whether the current wheat ear disease small ear ratio is smaller than a third disease small ear ratio threshold value or not, and obtaining a fourth judging result; the third disease spike duty cycle threshold is greater than the second disease spike duty cycle threshold;
if the fourth judgment result is yes, judging that the diseased grade of the current wheat ear is 3;
and if the fourth judgment result is negative, judging that the diseased grade of the current wheat ear is 4.
A wheat scab status assessment method in wheat fields, comprising:
the real-time wheat image acquisition module is used for acquiring a real-time wheat image of the wheat Tian Daice area;
the wheat head identification module is used for inputting the real-time wheat image into a wheat head identification model to obtain a real-time wheat head identification result of the region to be detected; the wheat head identification model is obtained by training a first lightweight neural network model by utilizing a plurality of marked historical wheat images;
The real-time single wheat ear image segmentation module is used for segmenting the real-time wheat images according to the real-time wheat ear identification result to obtain a plurality of real-time single wheat ear images;
the small ear scab detection module is used for respectively inputting a plurality of real-time single wheat ear images into a small ear scab detection model to obtain a small ear scab detection result of a wheat ear corresponding to each real-time single wheat ear image; the small ear scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat ear images; the detection result of the head blight of the wheat ears comprises whether each small ear on the wheat ears has head blight or not;
and the wheat scab disease state determining module is used for determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the plurality of wheat scab.
An electronic device comprising a memory for storing a computer program and a processor that runs the computer program to cause the electronic device to perform the wheat scab status assessment method; the memory is a readable storage medium.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects:
According to the wheat scab state evaluation method, the wheat scab state evaluation system and the electronic equipment provided by the invention, a real-time wheat image is input into a wheat head recognition model to obtain a real-time wheat head recognition result of a region to be detected; the wheat ear recognition model is obtained by training a first lightweight neural network model by utilizing a plurality of marked historical wheat images; dividing the real-time wheat image according to the real-time wheat head identification result to obtain a plurality of real-time single wheat head images; respectively inputting a plurality of real-time single wheat ear images into a wheat head scab detection model to obtain a wheat head scab detection result of each real-time single wheat ear image corresponding to a wheat ear; the wheat head scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat head images; the wheat head scab detection result comprises whether each wheat head on the wheat head has scab or not; and determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the scab of the multiple spikes. According to the wheat head scab detection method, the wheat head scab detection efficiency, accuracy and rationality can be improved by constructing and training the wheat head identification model and the wheat head scab detection model.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions of the prior art, the drawings that are needed in the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present invention, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart showing a wheat scab status evaluation method in wheat field in example 1 of the present invention;
FIG. 2 is a flowchart showing the whole wheat scab status evaluation method in wheat field in example 1 of the present invention;
FIG. 3 is a schematic diagram of a wheat scab status evaluation method in wheat field in example 1 of the present invention;
FIG. 4 is a schematic view of real-time wheat image in example 1 of the present invention;
FIG. 5 is a schematic view of a spike identification model in example 1 of the present invention;
FIG. 6 is a diagram showing the real-time spike identification result in embodiment 1 of the present invention;
FIG. 7 is a first real-time single ear image of embodiment 1 of the present invention;
FIG. 8 is a second real-time single ear image of example 1 of the present invention;
FIG. 9 is a schematic diagram of the structure of the test model for scab of tassel in example 1 of the present invention;
FIG. 10 is a schematic diagram showing the detection result of scab on a first real-time single ear image in example 1 of the present invention;
FIG. 11 is a graph showing the results of the wheat head blight detection of the second real-time single ear image in example 1 of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
The invention aims to provide a wheat scab state evaluation method, a system and electronic equipment, which can improve the efficiency, accuracy and rationality of wheat scab detection and wheat scab state evaluation in a wheat field.
In order that the above-recited objects, features and advantages of the present invention will become more readily apparent, a more particular description of the invention will be rendered by reference to the appended drawings and appended detailed description.
Example 1
As shown in fig. 1-3, the embodiment provides a wheat scab state evaluation method in a wheat field, including:
step 101: real-time wheat images of the wheat Tian Daice area were acquired. A schematic of the real-time wheat image is shown in fig. 4.
Step 102: and inputting the real-time wheat image into the wheat head identification model to obtain a real-time wheat head identification result of the region to be detected (as shown in figure 6). The wheat ear recognition model (as shown in fig. 5) is obtained by training a first lightweight neural network model by using a plurality of marked historical wheat images.
Step 103: and dividing the real-time wheat image according to the real-time wheat head identification result to obtain a plurality of real-time single wheat head images. The real-time single ear image is shown in figures 7-8.
Step 104: and respectively inputting the real-time single wheat ear images into a wheat head scab detection model to obtain a wheat head scab detection result (shown in fig. 10-11) of the wheat ear corresponding to each real-time single wheat ear image. The head blight detection model (as shown in fig. 9) is obtained by training a second lightweight neural network model by using a plurality of marked historical single ear images. The wheat head scab detection result comprises whether each wheat head on the wheat head has scab.
Step 105: and determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the scab of the multiple spikes.
Prior to step 101, further comprising:
step 106: and constructing a first lightweight neural network model. The first lightweight neural network model is obtained by adding an EMA attention module before a small target recognition layer and a medium target recognition layer of the lightweight neural network.
Step 107: a plurality of historical wheat images are acquired.
Step 108: and marking each wheat ear in the historical wheat image respectively to obtain a plurality of marked historical wheat images.
Step 109: and training the first lightweight neural network model by taking the historical wheat image as input and the marked historical wheat image as output to obtain the wheat head identification model.
Prior to step 101, further comprising:
step 1010: and constructing a second lightweight neural network model. The second lightweight neural network model is obtained by adding a front attention module and a rear attention module in front of a small target recognition layer of the lightweight neural network.
Step 1011: and cutting the historical wheat images according to the marked frames in the marked historical wheat images to obtain a plurality of historical single wheat ear images.
Step 1012: labeling whether each spike in the historical single spike image has scab or not to obtain a labeled historical single spike image.
Step 1013: and training the second lightweight neural network model by taking the historical single wheat ear image as input and the marked historical single wheat ear image as output to obtain the wheat head scab detection model.
Step 105: comprising:
Step 105-1: and acquiring the number of the real-time single wheat ear images as the total number of wheat ears in the region to be detected.
Step 105-2: and determining the number of diseased wheat ears according to the detection results of the scab of the multiple wheat ears.
Step 105-3: and determining the ratio of the diseased wheat spike number to the total number of the wheat spikes as the disease spike rate of the region to be detected. The ear disease rate is used for measuring the wheat scab state in the wheat field.
Step 105, further includes:
step 105-4: and determining the disease grade of each diseased wheat spike according to the detection results of the scab of the multiple spikelets.
Step 105-5: and determining the number of diseased wheat ears corresponding to different diseased grades.
Step 105-6: and determining the disease index of the region to be detected according to the total number of wheat ears of the region to be detected and the quantity of diseased wheat ears corresponding to a plurality of disease grades. The disease index is used for measuring the wheat scab state in the wheat field.
The disease index is as follows:
wherein I represents the index of the disease, h i The i-th diseased level corresponds to the number of diseased wheat ears, i represents the i-th diseased level, and H represents the total number of wheat ears in the area to be detected.
Step 105-4, comprising:
step 105-4-1: and determining any one of the real-time single ear images as the current real-time single ear image, and determining the ear corresponding to the current real-time single ear image as the current ear.
Step 105-4-2: and acquiring the number of diseased spikes and the total number of spikes of the current wheat according to the detection result of the head blight of the current wheat.
Step 105-4-3: and determining the ratio of the number of diseased spikes to the total number of spikes of the current wheat spike as the ratio of the diseased spikes of the current wheat spike.
Step 105-4-4: and determining the diseased grade of the current wheat spike according to the diseased small spike ratio of the current wheat spike.
Step 105-4-4, comprising:
step 105-4-4-1: judging whether the current wheat head small ear proportion is equal to 0 or not, and obtaining a first judging result. If the first judgment result is yes, executing step 105-4-4-2; if the first determination result is no, step 105-4-4-3 is performed.
Step 105-4-4-2: and judging that the disease grade of the current wheat head is 0.
Step 105-4-4-3: judging whether the current wheat head small ear duty ratio is smaller than the first disease small ear duty ratio threshold value or not, and obtaining a second judging result. If the second judgment result is yes, executing the step 105-4-4-4; if the second determination result is no, step 105-4-4-5 is performed.
Step 105-4-4-4: and judging that the disease grade of the current wheat head is 1.
Step 105-4-4-5: and judging whether the disease small ear duty ratio of the current wheat ear is smaller than a second disease small ear duty ratio threshold value, and obtaining a third judging result. The second disease spike fraction threshold is greater than the first disease spike fraction threshold. If the third judgment result is yes, executing the step 105-4-4-6; if the third determination result is no, step 105-4-4-7 is performed.
Step 105-4-4-6: and judging that the diseased grade of the current wheat ear is 2.
Step 105-4-4-7: and judging whether the current small ear duty ratio of the wheat ears is smaller than a third small ear duty ratio threshold value, and obtaining a fourth judgment result. The third disease spike fraction threshold is greater than the second disease spike fraction threshold. If the fourth judgment result is yes, executing the step 105-4-4-8; if the fourth determination result is no, step 105-4-4-9 is performed.
Step 105-4-4-8: and judging that the disease grade of the current wheat head is 3.
Step 105-4-4-9: and judging that the disease grade of the current wheat head is 4.
As shown in fig. 2, the overall flow of the present invention includes: step 1: acquiring a color image of field wheat in a region to be detected; step 2: performing wheat ear labeling on the field wheat color image, and performing recognition and segmentation processing based on a lightweight neural network model to obtain a wheat ear color image; step 3: labeling the scab disease condition of wheat based on a color image of the wheat ear, and performing identification treatment by using a lightweight neural network model to obtain the scab disease identification result of the wheat ear. Step 4: based on the wheat head scab disease identification result of the wheat ears, the severity, the ear rate and the disease index of the wheat head scab in the acquisition area are calculated and determined.
Step 2 further comprises:
s21: based on the color image of the wheat in the field, a wheat head identification data set and a wheat head identification neural network are constructed, so that the complete wheat head identification precision without shielding in the field is realized, meanwhile, the model is light, the terminal transplantation is convenient to realize, and the wheat head identification neural network model is trained.
S22: based on the wheat ear recognition neural network model, a wheat ear recognition result of the field wheat color image and the wheat ear color image are obtained.
Step 3 further comprises:
s31: based on the wheat ear color image, a wheat head scab identification data set is constructed, a wheat head scab identification neural network structure is constructed, the identification precision of the wheat ears is improved, the model is light, and terminal transplantation is facilitated. Training a small spike scab recognition neural network model.
S32: and obtaining the head blight disease recognition result of the wheat ears based on the head blight recognition neural network model.
Step 4 further comprises:
s41: based on the identification result of wheat head scab disease of wheat ears, the number of wheat ears corresponding to each grade of severity of scab disease and the total number of wheat ears in a detection area are obtained according to a national standard evaluation method.
S42: and calculating the wheat scab disease spike rate and the disease index of the image acquisition area according to the total number of wheat spikes and the scab disease grade of each wheat spike.
Step S21 further includes:
and marking complete and clear wheat ears in the field wheat images by using image marking software LabelImg, and constructing a wheat ear identification data set.
Designing a lightweight wheat head recognition neural network structure.
The model backbone network part firstly extracts initial characteristic information in the image through convolution operation.
And then, the characteristic information is transmitted into a FaterNet Block (Faster Neural Networks Block) module to obtain a characteristic diagram, the progressive channel normalization technology is used for reducing the parameter quantity and the calculated quantity of the network model, and simultaneously, two convolution layers are used as inverted residual blocks to further improve the performance and the efficiency of the model.
The method comprises the steps of using convolution operation to fuse the extracted feature graphs, then transmitting the feature graphs into an EMA (effect Multi-scale attribute) attention module, dividing an input channel into a plurality of groups to reduce calculation complexity, capturing cross-channel relation, calculating global information codes by an average pooling method, further calculating channel weight of each group, calculating cross-latitude interaction among the groups by a matrix multiplication method, capturing pixel level pairwise relation, and finally remolding weighted features back to original size to capture cross-channel and cross-space information so as to improve performance of a model.
And finally, carrying out global average pooling and fusion on the feature images, and reducing the parameter number of the model so as to improve the training efficiency and the reasoning speed of the model.
The head of the lightweight wheat head recognition neural network model firstly performs two downsampling on the feature map extracted by the main network, and then adopts a progressive feature pyramid network AFPN (Asymptotic Feature Pyramid Network) to perform feature fusion. Generating a high-level feature map through a top-down path, wherein the top-down path consists of a plurality of convolution layers; then upsampling the low-level feature map to the same resolution as the high-level feature map through a bottom-up path; the bottom-up path consists of an up-sampling layer and a convolution layer, wherein the up-sampling layer up-samples the low-level feature map to the same resolution as the high-level feature map, and the convolution layer fuses the up-sampled feature map with the high-level feature map; and finally, fusing the features of different levels by using an element-by-element addition mode to obtain a feature map. An EMA attention module is added in front of the small target recognition layer and the medium target recognition layer, so that the problem that the wheat head targets are small and difficult to recognize is solved, and the recognition precision is improved; and transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame.
Training the designed identification network model by using the wheat head identification data set to obtain the lightweight wheat head identification neural network model.
Step S22 further includes:
and sequentially transmitting the field wheat color images into the wheat head recognition neural network model. After identification, a wheat head identification result of the image is obtained, and a wheat head color image is obtained after segmentation based on the identification result.
Specifically, step S31 further includes:
and marking the disease condition of the wheat head scab by using image marking software LabelImg on the front, side and 45-degree wheat head color images respectively, and constructing a wheat head scab identification data set.
And designing a lightweight small spike scab recognition neural network model structure.
The model backbone network part firstly extracts initial characteristic information in an image through convolution operation; then, the characteristic information is transmitted into a small target recognition module SPD-Conv (Space-to-Depth convolution layer) to extract a characteristic image, the small target recognition module converts the input characteristic information into a depth image, and each pixel point corresponds to a characteristic channel; and then downsampling the depth map by using step-free convolution, and finally converting the downsampled characteristic information back to the original space dimension to obtain the characteristic map, thereby solving the problem that the small spike target is difficult to identify.
The feature map is then passed into a depth separable convolution consisting of two convolution layers, the first using a 3 x 3 convolution kernel to extract local information of the target and the second using a 1 x 1 convolution kernel to concatenate the information of the different channels to generate a richer feature.
The feature map is transmitted into a Mobile Block (Mobile-Friendly Vision Transformer Block) module to further extract global information, the module is composed of a convolution layer, a self-attention layer and a full-connection layer, the convolution layer is used for extracting local information of the feature map, the self-attention layer is used for learning global dependency relations among the feature maps, the full-connection layer is used for generating a final feature map, the weight of the model is reduced, and the model is suitable for terminal transplanting.
The head of the light-weight small spike scab recognition neural network model firstly carries out up-sampling on the feature map by using a bilinear interpolation method; then further extracting different features through the convolution layer; the feature map fusion of different levels is realized by using a splicing method; an attention module BiFormer (Vision Transformer with Bi-Level Routing Attention) is added in front of the small target recognition layer to improve the recognition accuracy of the model to the small spike small target, the attention module consists of a front attention module and a rear attention module, the front attention module is used for capturing long-distance dependency relations between features, and the rear attention module is used for capturing short-distance dependency relations between features; and finally, transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame.
Training the designed target recognition network model by using the small head blight recognition data set to obtain the lightweight small head scab recognition neural network model.
Step S32 further includes:
and sequentially transmitting the wheat head color images into a wheat head scab recognition neural network model. And obtaining the identification result of the scab disease of the wheat ears after identification.
Step S41 further includes:
based on the wheat head scab disease identification result of the wheat head, the ratio of the number of the diseased wheat heads to the total wheat heads is calculated according to the following calculation formula:
wherein L is the ratio of the number of diseased spikes to the total number of spikes, E is the number of diseased spikes, and E is the total number of spikes. According to the ratio of wheat head disease spike number to all spikes, the disease severity of wheat head can be determined, and the judgment standard is the ratio of the disease spike number to all spikes indicating spike rot symptoms (or white spike symptoms caused by stalk rot) according to the specification of national standard 'wheat head forecast technical Specification', and is divided into five grades:
level 0: has no disease.
Stage 1: the number of diseased spikelets is less than 1/4 of the total spikelets.
2 stages: the number of diseased spikelets accounts for 1/4 to 1/2 of the total spikelets.
3 stages: the number of diseased spikelets accounts for 1/2-3/4 of the total spikelets.
4 stages: the number of diseased spikelets is more than 3/4 of the total spikelets.
And obtaining the number of the wheat ears and the total number of the wheat ears corresponding to each level of severity of the scab disease condition in the detection area based on the calculated result of the scab disease severity of all the wheat ears.
Step S42 further includes:
and calculating the wheat scab disease spike rate and disease index of the image acquisition area based on the number of spikes corresponding to the severity of each level of scab disease and the total number of wheat spikes in the detection area. The disease spike rate refers to the ratio of the number of diseased wheat spikes to the total number of spikes investigated, and the calculation formula is as follows:
wherein X represents the head ratio of wheat scab in the collecting region, H represents the total head number of the wheat suffering from the disease, and H represents the total head number of the wheat in the collecting region.
The disease index is a comprehensive index for measuring the prevalence and severity of scab disease, and is used for representing the average level of disease occurrence, and the calculation formula is as follows:
wherein I represents the index of the disease, h i The number of ears corresponding to each level of severity, i represents the severity of the illness, and H represents the investigationTotal spike number.
The method can quickly and effectively identify the wheat scab in the field and accurately calculate the disease index, and is helpful for the accurate management of wheat planting in the field and the disease detection and control. And photographing the wheat in the area to be detected by using a mobile phone or a camera, acquiring a sufficient quantity of field wheat color images, and transmitting the images to a computer for processing and storing. And dividing the marked pictures into a training set, a verification set and a test set according to the proportion of 7:2:1 by combining the wheat head identification public data set, and constructing the wheat head identification data set.
Designing a lightweight wheat head recognition neural network structure. The model backbone network part firstly extracts initial characteristic information in an image through convolution operation; and then, the characteristic information is transmitted into a FaterNetBlock module to obtain a characteristic diagram, the progressive channel normalization technology is used for reducing the parameter quantity and the calculated quantity of the network model, and simultaneously, two convolution layers are used as inverted residual blocks to further improve the performance and the efficiency of the model. The method comprises the steps of using convolution operation to fuse the extracted feature images, then transmitting the feature images into an EMA attention module, dividing an input channel into a plurality of groups to reduce calculation complexity, capturing cross-channel relation at the same time, calculating global information codes through an average pooling method, further calculating channel weight of each group, calculating cross-latitude interaction among the groups through a matrix multiplication method, capturing pixel-level pairwise relation, and finally reshaping weighted features back to the original size to capture cross-channel and cross-space information. And finally, carrying out global average pooling and fusion on the feature images. The head of the lightweight wheat head recognition neural network model firstly carries out downsampling twice on a feature map extracted by a main network, and then adopts a progressive feature pyramid network AFPN to carry out feature fusion. Generating a high-level feature map through a top-down path, wherein the top-down path consists of a plurality of convolution layers; then upsampling the low-level feature map to the same resolution as the high-level feature map through a bottom-up path; the bottom-up path consists of an up-sampling layer and a convolution layer, wherein the up-sampling layer up-samples the low-level feature map to the same resolution as the high-level feature map, and the convolution layer fuses the up-sampled feature map with the high-level feature map; and finally, fusing the features of different levels by using an element-by-element addition mode to obtain a feature map. An EMA attention module is added in front of the small target recognition layer and the medium target recognition layer, so that recognition accuracy is improved; and transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame. Training the designed identification network model by using the wheat head identification data set to obtain the lightweight wheat head identification neural network model. Training a lightweight ear recognition neural network model based on the ear recognition dataset; and transmitting the field wheat color image into a wheat head recognition neural network model to obtain a recognition result and a wheat head color image.
And marking the wheat head color image by using a LabelImg marking tool, and marking the scab disease condition of the wheat head spike. Dividing the marked pictures into a training set, a verification set and a test set according to the ratio of 7:2:1, and constructing a head blight identification data set. And designing a lightweight small spike scab recognition neural network model structure.
The model backbone network part firstly extracts initial characteristic information in an image through convolution operation; then, feature information is transmitted into a small target recognition module SPD-Conv to extract a feature image, the small target recognition module converts the input feature information into a depth image, and each pixel point corresponds to a feature channel; and then downsampling the depth map by using step-free convolution, and finally converting the downsampled feature information back to the original space dimension to obtain the feature map. The feature map is then passed into a depth separable convolution consisting of two convolution layers, the first using a 3 x 3 convolution kernel to extract local information of the target and the second using a 1 x 1 convolution kernel to concatenate the information of the different channels to generate a richer feature. The feature map is transmitted into a MobileViT Block module to further extract global information, the module is composed of a convolution layer, a self-attention layer and a full-connection layer, the convolution layer is used for extracting local information of the feature map, the self-attention layer is used for learning global dependency relations among the feature maps, and the full-connection layer is used for generating a final feature map. The head of the light-weight small spike scab recognition neural network model firstly carries out up-sampling on the feature map by using a bilinear interpolation method; then further extracting different features through the convolution layer; the feature map fusion of different levels is realized by using a splicing method; adding an attention module BiFormer in front of the small target recognition layer to improve the recognition accuracy of the model to the small spike small target, wherein the attention module consists of a front attention module and a rear attention module, the front attention module is used for capturing long-distance dependency among features, and the rear attention module is used for capturing close-distance dependency among features; and finally, transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame. Training the designed target recognition network model by using the small head blight recognition data set to obtain the lightweight small head scab recognition neural network model. Training a light-weight small head scab recognition neural network model based on the small head scab recognition data set; and transmitting the wheat head color image into a wheat head scab recognition neural network model to obtain a recognition result.
And sequentially reading the identification result of the scab of the wheat corresponding to each wheat color image, and calculating the disease spike rate and the disease index.
The color image collected in this embodiment is collected in a field wheat field by a camera, as shown in fig. 3, and 2143 color images are collected in total. And marking the wheat color image, namely marking each unoccluded and complete wheat ear, combining 3373 wheat ears identification data sets, totaling 5516 wheat ears, and dividing the marked pictures into 3861 training sets, 1104 verification sets and 551 test sets according to the ratio of 7:2:1.
The embodiment designs a lightweight wheat head recognition neural network model structure, and initial characteristic information in an image is extracted by a main network part of the model through convolution operation.
And then, the characteristic information is transmitted into a FaterNet Block module to obtain a characteristic diagram, the progressive channel normalization technology is used for reducing the parameter quantity and the calculated quantity of the network model, and simultaneously, two convolution layers are used as inverted residual blocks to further improve the performance and the efficiency of the model.
The method comprises the steps of using convolution operation to fuse the extracted feature graphs, then transmitting the feature graphs into an EMA attention module, dividing an input channel into a plurality of groups to reduce calculation complexity, capturing cross-channel relation at the same time, calculating global information codes through an average pooling method, further calculating channel weight of each group, calculating cross-latitude interaction among the groups through a matrix multiplication method, capturing pixel-level pairwise relation, and finally reshaping weighted features back to the original size to capture cross-channel and cross-space information so as to improve performance of a model.
And finally, carrying out global average pooling and fusion on the feature images, and improving the training efficiency and the reasoning speed of the model.
The head of the lightweight wheat head recognition neural network model firstly carries out downsampling twice on a feature map extracted by a main network, and then adopts a progressive feature pyramid network AFPN to carry out feature fusion. Generating a high-level feature map through a top-down path, wherein the top-down path consists of a plurality of convolution layers; then upsampling the low-level feature map to the same resolution as the high-level feature map through a bottom-up path; the bottom-up path consists of an up-sampling layer and a convolution layer, wherein the up-sampling layer up-samples the low-level feature map to the same resolution as the high-level feature map, and the convolution layer fuses the up-sampled feature map with the high-level feature map; and finally, fusing the features of different levels by using an element-by-element addition mode to obtain a feature map. An EMA attention module is added in front of the small target recognition layer and the medium target recognition layer, so that the problem that the wheat head targets are small and difficult to recognize is solved, and the recognition precision is improved; and transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame.
Training the designed identification network model by using the wheat head identification data set to obtain the lightweight wheat head identification neural network model. The model structure parameter is 680 ten thousand, 39% compared with the original Yolov8 model network structure, 22 hundred million calculated amount is reduced by 24% compared with the original Yolov8 model network structure, the wheat spike identification precision reaches 96.2%, and the method is suitable for terminal transplantation.
Based on the wheat head identification data set, training a light wheat head identification neural network model, and transmitting the wheat color image acquired in the field into the network model to obtain a wheat head identification result and a wheat head color image.
Labeling small ears of the wheat ear image, labeling scab disease conditions of each small ear, dividing the labeled pictures into a training set, a verification set and a test set according to the ratio of 7:2:1, and constructing a small ear scab identification data set.
The embodiment designs a lightweight small spike scab recognition neural network model structure, wherein a main network part of the model firstly extracts initial characteristic information in an image through convolution operation; then, feature information is transmitted into a small target recognition module SPD-Conv to extract a feature image, the small target recognition module converts the input feature information into a depth image, and each pixel point corresponds to a feature channel; and then downsampling the depth map by using step-free convolution, and finally converting the downsampled feature information back to the original space dimension to obtain the feature map.
The feature map is then passed into a depth separable convolution consisting of two convolution layers, the first using a 3 x 3 convolution kernel to extract local information of the target and the second using a 1 x 1 convolution kernel to concatenate the information of the different channels to generate a richer feature.
The feature map is transmitted into a MobileViT Block module to further extract global information, the module is composed of a convolution layer, a self-attention layer and a full-connection layer, the convolution layer is used for extracting local information of the feature map, the self-attention layer is used for learning global dependency relations among the feature maps, and the full-connection layer is used for generating a final feature map.
The head of the light-weight small spike scab recognition neural network model firstly carries out up-sampling on the feature map by using a bilinear interpolation method; then further extracting different features through the convolution layer; the feature map fusion of different levels is realized by using a splicing method; adding an attention module BiFormer in front of the small target recognition layer to improve the recognition accuracy of the model to the small spike small target, wherein the attention module consists of a front attention module and a rear attention module, the front attention module is used for capturing long-distance dependency among features, and the rear attention module is used for capturing close-distance dependency among features; and finally, transmitting the feature map into a target recognition layer to obtain a recognition result, removing redundant detection results by using a non-maximum suppression algorithm, and reserving the most probable target frame.
Training the designed target recognition network model by using the small head blight recognition data set to obtain the lightweight small head scab recognition neural network model. The model structure parameter is 120 ten thousand, compared with the original YOLOv8 model network structure, 89% of the model structure parameter is reduced, the calculated amount is 2 hundred million, compared with the original YOLOv8 model network structure, 93% of the model structure parameter is reduced, the identification precision of the head blight of the small spike reaches 94.9%, and the model structure parameter is suitable for terminal transplantation.
Training a light-weight small head scab recognition neural network model based on the data set, and transmitting the wheat head color image into a small head scab recognition result of a high network model; based on the head blight recognition result, 10 wheat ears are shared in this example, wherein the number of 0-grade wheat ears is 5; the number of the wheat ears of level 1 is 4; the number of the 2-grade wheat ears is 1 plant; the number of the wheat ears of the grade 3 is 0; the number of the wheat ears of grade 4 is 0. According to the total number of wheat ears and the quantity of wheat ears at each diseased level of the scab, the wheat scab disease spike rate of the collecting area in the embodiment is calculated to be 50%, the scab disease index is 15, and finally the identification and disease index calculation of the wheat scab in the field are realized.
Example 2
In order to perform the corresponding method of the above embodiment 1 to achieve the corresponding functions and technical effects, a wheat scab state evaluation method in wheat fields is provided below, including:
And the real-time wheat image acquisition module is used for acquiring real-time wheat images of the region of the wheat Tian Daice.
The wheat head identification module is used for inputting the real-time wheat image into the wheat head identification model to obtain a real-time wheat head identification result of the region to be detected. The wheat ear recognition model is obtained by training a first lightweight neural network model by using a plurality of marked historical wheat images.
The real-time single wheat ear image segmentation module is used for segmenting the real-time wheat images according to the real-time wheat ear identification result to obtain a plurality of real-time single wheat ear images.
The small ear scab detection module is used for respectively inputting a plurality of real-time single wheat ear images into the small ear scab detection model to obtain small ear scab detection results of wheat ears corresponding to each real-time single wheat ear image. The wheat head scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat head images. The wheat head scab detection result comprises whether each wheat head on the wheat head has scab.
And the wheat scab disease state determining module is used for determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the plurality of wheat scab.
Example 3
The present embodiment provides an electronic device, including a memory and a processor, where the memory is configured to store a computer program, and the processor is configured to execute the computer program to cause the electronic device to execute a wheat scab status assessment method described in embodiment 1. The memory is a readable storage medium.
In the present specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different point from other embodiments, and identical and similar parts between the embodiments are all enough to refer to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant points refer to the description of the method section.
The principles and embodiments of the present invention have been described herein with reference to specific examples, the description of which is intended only to assist in understanding the methods of the present invention and the core ideas thereof; also, it is within the scope of the present invention to be modified by those of ordinary skill in the art in light of the present teachings. In view of the foregoing, this description should not be construed as limiting the invention.
Claims (10)
1. A wheat scab state assessment method in wheat fields, comprising:
Acquiring a real-time wheat image of a wheat Tian Daice area;
inputting the real-time wheat image into a wheat head identification model to obtain a real-time wheat head identification result of the region to be detected; the wheat head identification model is obtained by training a first lightweight neural network model by utilizing a plurality of marked historical wheat images;
dividing the real-time wheat image according to the real-time wheat head identification result to obtain a plurality of real-time single wheat head images;
respectively inputting a plurality of real-time single wheat ear images into a wheat head scab detection model to obtain a wheat head scab detection result of each wheat ear corresponding to each real-time single wheat ear image; the small ear scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat ear images; the detection result of the head blight of the wheat ears comprises whether each small ear on the wheat ears has head blight or not;
and determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the scab of the multiple spikes.
2. The method for evaluating the scab status of wheat in a wheat field of claim 1, further comprising, prior to the acquiring the real-time wheat image of the region Tian Daice:
Constructing a first lightweight neural network model; the first lightweight neural network model is obtained by adding an EMA attention module before a small target recognition layer and a medium target recognition layer of the lightweight neural network;
acquiring a plurality of historical wheat images;
marking each wheat ear in the historical wheat image respectively to obtain a plurality of marked historical wheat images;
and training the first lightweight neural network model by taking the historical wheat image as input and the marked historical wheat image as output to obtain the wheat ear recognition model.
3. The method for evaluating the scab status of wheat in a wheat field of claim 2, further comprising, prior to the acquiring the real-time wheat image of the region Tian Daice:
constructing a second lightweight neural network model; the second lightweight neural network model is obtained by adding a front attention module and a rear attention module in front of a small target recognition layer of the lightweight neural network;
cutting the historical wheat images according to the marked frames in the marked historical wheat images to obtain a plurality of historical single wheat ear images;
labeling whether each spike in the historical single spike image has scab or not to obtain a labeled historical single spike image;
And training the second lightweight neural network model by taking the historical single wheat ear image as input and the marked historical single wheat ear image as output to obtain the wheat head scab detection model.
4. The method for evaluating the wheat scab status of a wheat field according to claim 1, wherein determining the wheat scab disease status of the area of wheat Tian Daice based on the plurality of spike scab detection results comprises:
acquiring the number of the real-time single wheat ear images as the total number of wheat ears in the region to be detected;
determining the number of diseased wheat ears according to the detection results of the scab of a plurality of small ears;
determining the ratio of the diseased wheat ears to the total number of the wheat ears as the disease ear rate of the region to be detected; the spike disease rate is used for measuring the wheat scab state of the wheat field.
5. The method for evaluating the wheat scab status of wheat in a wheat field of claim 4, wherein the determining the wheat scab status of the area of wheat Tian Daice based on the plurality of spike scab detection results, further comprises:
determining the disease grade of each diseased wheat ear according to the detection results of the scab of a plurality of small ears;
determining the quantity of diseased wheat ears corresponding to different diseased grades;
determining the disease index of the region to be detected according to the total number of wheat ears of the region to be detected and the quantity of diseased wheat ears corresponding to a plurality of disease grades; the disease index is used for measuring the wheat scab state of the wheat field.
6. The method for evaluating the scab status of wheat in a wheat field of claim 5, wherein the disease index is:
wherein I represents the index of the disease, h i The i-th diseased level corresponds to the number of diseased wheat ears, i represents the i-th diseased level, and H represents the total number of wheat ears in the area to be detected.
7. The method for evaluating the scab status of wheat in a wheat field according to claim 5, wherein said determining the disease level of each diseased ear based on the plurality of spike scab detection results comprises:
determining any one of the real-time single ear images as a current real-time single ear image, and determining the ear corresponding to the current real-time single ear image as a current ear;
acquiring the number of diseased wheat ears and the total number of the wheat ears of the current wheat ears according to the detection result of the wheat head scab of the current wheat ears;
determining the ratio of the number of diseased small ears of the current wheat ear to the total number of small ears as the ratio of the diseased small ears of the current wheat ear;
and determining the diseased grade of the current wheat spike according to the diseased small spike ratio of the current wheat spike.
8. The method for evaluating the scab status of wheat in a wheat field according to claim 7, wherein said determining the disease level of the current ear according to the disease ear ratio of the current ear comprises:
Judging whether the disease small ear proportion of the current wheat ear is equal to 0 or not, and obtaining a first judging result;
if the first judgment result is yes, judging that the diseased grade of the current wheat ear is 0;
if the first judgment result is negative, judging whether the current wheat ear disease small ear ratio is smaller than a first disease small ear ratio threshold value or not, and obtaining a second judgment result;
if the second judgment result is yes, judging that the diseased grade of the current wheat ear is 1;
if the second judgment result is negative, judging whether the disease spike ratio of the current wheat spike is smaller than a second disease spike ratio threshold value or not, and obtaining a third judgment result; the second disease spike duty cycle threshold is greater than the first disease spike duty cycle threshold;
if the third judgment result is yes, judging that the diseased grade of the current wheat ear is 2;
if the third judging result is negative, judging whether the current wheat ear disease small ear ratio is smaller than a third disease small ear ratio threshold value or not, and obtaining a fourth judging result; the third disease spike duty cycle threshold is greater than the second disease spike duty cycle threshold;
if the fourth judgment result is yes, judging that the diseased grade of the current wheat ear is 3;
and if the fourth judgment result is negative, judging that the diseased grade of the current wheat ear is 4.
9. A wheat scab state assessment method in wheat fields, comprising:
the real-time wheat image acquisition module is used for acquiring a real-time wheat image of the wheat Tian Daice area;
the wheat head identification module is used for inputting the real-time wheat image into a wheat head identification model to obtain a real-time wheat head identification result of the region to be detected; the wheat head identification model is obtained by training a first lightweight neural network model by utilizing a plurality of marked historical wheat images;
the real-time single wheat ear image segmentation module is used for segmenting the real-time wheat images according to the real-time wheat ear identification result to obtain a plurality of real-time single wheat ear images;
the small ear scab detection module is used for respectively inputting a plurality of real-time single wheat ear images into a small ear scab detection model to obtain a small ear scab detection result of a wheat ear corresponding to each real-time single wheat ear image; the small ear scab detection model is obtained by training a second lightweight neural network model by using a plurality of marked historical single wheat ear images; the detection result of the head blight of the wheat ears comprises whether each small ear on the wheat ears has head blight or not;
and the wheat scab disease state determining module is used for determining the wheat scab disease state of the wheat Tian Daice area according to the detection results of the plurality of wheat scab.
10. An electronic device comprising a memory and a processor, the memory for storing a computer program, the processor running the computer program to cause the electronic device to perform a wheat scab status assessment method according to any one of claims 1 to 8; the memory is a readable storage medium.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311399450.8A CN117392668A (en) | 2023-10-26 | 2023-10-26 | Wheat scab state evaluation method and system and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311399450.8A CN117392668A (en) | 2023-10-26 | 2023-10-26 | Wheat scab state evaluation method and system and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117392668A true CN117392668A (en) | 2024-01-12 |
Family
ID=89464503
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311399450.8A Pending CN117392668A (en) | 2023-10-26 | 2023-10-26 | Wheat scab state evaluation method and system and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117392668A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746163A (en) * | 2024-01-23 | 2024-03-22 | 哈尔滨工程大学 | Radar working mode identification method based on multi-scale vision transducer |
-
2023
- 2023-10-26 CN CN202311399450.8A patent/CN117392668A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117746163A (en) * | 2024-01-23 | 2024-03-22 | 哈尔滨工程大学 | Radar working mode identification method based on multi-scale vision transducer |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113065558B (en) | Lightweight small target detection method combined with attention mechanism | |
| CN112949565B (en) | Single-sample partially-shielded face recognition method and system based on attention mechanism | |
| CN109800736B (en) | Road extraction method based on remote sensing image and deep learning | |
| CN111680706B (en) | Dual-channel output contour detection method based on coding and decoding structure | |
| CN113673590B (en) | Rain removing method, system and medium based on multi-scale hourglass dense connection network | |
| CN115099297B (en) | Soybean plant phenotype data statistical method based on improved YOLO v5 model | |
| CN113139489B (en) | Crowd counting method and system based on background extraction and multi-scale fusion network | |
| CN114972208B (en) | YOLOv 4-based lightweight wheat scab detection method | |
| CN115272828A (en) | Intensive target detection model training method based on attention mechanism | |
| CN114943893B (en) | Feature enhancement method for land coverage classification | |
| CN117392668A (en) | Wheat scab state evaluation method and system and electronic equipment | |
| CN111967464A (en) | Weak supervision target positioning method based on deep learning | |
| CN113435254A (en) | Sentinel second image-based farmland deep learning extraction method | |
| CN114120359A (en) | Method for measuring body size of group-fed pigs based on stacked hourglass network | |
| CN118379288B (en) | Embryo prokaryotic target counting method based on fuzzy rejection and multi-focus image fusion | |
| CN116630700A (en) | Remote sensing image classification method based on introduction channel-space attention mechanism | |
| CN114663769B (en) | Fruit identification method based on YOLO v5 | |
| CN115410081A (en) | Multi-scale aggregated cloud and cloud shadow identification method, system, equipment and storage medium | |
| CN116524189A (en) | High-resolution remote sensing image semantic segmentation method based on coding and decoding indexing edge characterization | |
| CN116168235A (en) | Hyperspectral image classification method based on double-branch attention network | |
| CN115546187A (en) | Agricultural pest and disease detection method and device based on YOLO v5 | |
| CN115526852A (en) | Molten pool and splash monitoring method in selective laser melting process based on target detection and application | |
| CN115496891A (en) | Wheat lodging degree grading method and device | |
| CN115035381A (en) | Lightweight target detection network of SN-YOLOv5 and crop picking detection method | |
| CN114998375A (en) | Live fish weight estimation method and system based on example segmentation |
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 |