CN115690585B - Method and system for extracting wheat tillering number based on digital photo - Google Patents
Method and system for extracting wheat tillering number based on digital photo Download PDFInfo
- Publication number
- CN115690585B CN115690585B CN202211414929.XA CN202211414929A CN115690585B CN 115690585 B CN115690585 B CN 115690585B CN 202211414929 A CN202211414929 A CN 202211414929A CN 115690585 B CN115690585 B CN 115690585B
- Authority
- CN
- China
- Prior art keywords
- training set
- label
- photo
- digital
- tillering
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A40/00—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production
- Y02A40/10—Adaptation technologies in agriculture, forestry, livestock or agroalimentary production in agriculture
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a method and a system for extracting wheat tillering number based on a digital photo. The method comprises the following steps: collecting winter wheat digital photos to obtain a training set and a verification set; performing data enhancement on the training set to obtain a training set after data enhancement; taking the actual tiller number as a label, and making sample labels for the training set and the verification set after the data enhancement; constructing a tillering number extraction model; training and parameter setting are carried out on the tillering number extraction model by applying the training set after data enhancement and the corresponding label; and quantitatively evaluating the accuracy of the tillering number extraction model in extracting the tillering number of the photo of the verification set. According to the scheme provided by the invention, the tillering number of wheat can be extracted from the digital photo, the relative error is ensured to be about 10% by the precision, the mode of shooting the digital photo in the future can be met, and the tillering number of winter wheat is extracted based on deep learning to replace manual field counting.
Description
Technical Field
The invention belongs to the field of plant growth condition estimation, and particularly relates to a method and a system for extracting wheat tillering number based on a digital photo.
Background
Wheat tillering refers to branching of wheat from a wheat plant occurring below or near the ground, and the number of tillers per unit area is a key agronomic feature for growth observation and crop management in wheat production. Especially, the tillering is used as a main mechanism of the adaptation of the wheat to the existing resources, and the planting density in the current stage determines the effective tillering and directly influences the yield and quality of the wheat. Meanwhile, the tillering number can provide accurate emergence rate for farmers and provide basis for water and fertilizer management. Thus, the real-time assessment of tiller number at the tillering stage helps to monitor wheat population growth or as a major phenotypic indicator of crop breeding selection varieties. However, the traditional tillering quantity measuring method mostly depends on complex manual counting, and the observation method has low efficiency and poor precision, is time-consuming and labor-consuming, is easy to cause human errors, and is easy to damage crops.
Computer vision is a potential automatic solution, for example, in 2018, li Qiongyan and the like, detection of the tiller number of the single plant wheat is completed through image segmentation technology and morphological processing (Li Qiongyan, gao Yunpeng, weng Yuchen. An automatic detection method for the tiller number of the wheat based on RGB images is CN107993243A, 2018.) in 2019, du Mengmeng and the like, a regression model is established through image segmentation technology, and finally, the tiller density of a manual counting sampling area is obtained through interpolation. (Du Mengmeng, ji Jiangtao, du Xinwu, etc.) A wheat tillering density measuring and calculating method based on the multispectral remote sensing image of the unmanned aerial vehicle is CN110163138A [ P ].2019 ]. In 2019, yan Hua is obtained by extracting the wheat pixels in the frame in real time by a deep learning module of the unmanned plane, dividing the extracted pixels by the empirical parameters to obtain an estimated value, and finally interpolating to obtain a spatial distribution map (Yan Hua, wei Yancong, liu Long, etc. the real-time counting method of wheat plants based on deep learning image segmentation is CN110503647A [ P ] 2019 ]. However, the current method is limited by the mutual shielding among wheat tillers, can only adapt to the tillering counting of wheat in the seedling stage, and can be influenced by the problems of insufficient illumination, appearance change, inconsistent scale and the like during image acquisition. With the long-term development of counting technology based on deep learning, the research work covers each link of wheat yield measurement, especially the counting of wheat ears, but less reports of counting dense tillers in the tillering stage.
The traditional method for acquiring the information of the wheat stem tillering density through manual field investigation has the defects of insufficient timeliness and precision, large workload, low efficiency and incapability of accurately reflecting the space difference condition of the wheat stem tillering density in the field by sparse point source statistical data. The existing digital image extraction method cannot be used for densely tillering, but can only be used for tillering counting of wheat in seedling stage.
Disclosure of Invention
In order to solve the technical problems, the invention provides a method and a system for extracting the tiller number of wheat based on a digital photo, so as to solve the technical problems.
The invention discloses a method for extracting wheat tillering number based on a digital photo, which comprises the following steps:
s1, acquiring a winter wheat digital photo to obtain a training set and a verification set;
s2, carrying out data enhancement on the training set to obtain a training set with enhanced data;
s3, taking a real tiller number as a label, and making sample labels for the training set and the verification set after data enhancement;
s4, constructing a tillering number extraction model;
s5, training and parameter setting are carried out on the tillering number extraction model by applying the training set after data enhancement and the corresponding labels;
and S6, quantitatively evaluating the accuracy of the tillering number extraction model in extracting the photo tillering number of the verification set.
According to the method of the first aspect of the present invention, in the step S1, the method for acquiring the winter wheat digital photos to obtain the training set and the verification set includes:
according to the following steps of 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo.
According to the method of the first aspect of the present invention, in the step S1, the method for acquiring the winter wheat digital photos to obtain the training set and the verification set further includes:
each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
According to the method of the first aspect of the present invention, in the step S2, the method for enhancing the training set includes:
carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; and (3) carrying out multi-sample synthesis on the digital photos of the training set, namely respectively and averagely cutting different photo frames in the digital photos of the training set into four parts, and then disturbing and re-splicing the photo frames to form a new photo frame.
According to the method of the first aspect of the present invention, in the step S3, with the actual tiller number as a label, the method for making a sample label for the training set after data enhancement includes:
for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label.
According to the method of the first aspect of the present invention, in the step S3, with the actual tiller number as a label, the method for making a sample label for the training set after data enhancement further includes:
for digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tillering numbers of the corresponding original photos.
According to the method of the first aspect of the present invention, in the step S3, the method for making a sample label for the verification set by using the actual tiller number as a label includes:
and (3) the ground actual measurement value of the tiller number used for the verification set label.
The invention discloses a system for extracting wheat tillering number based on digital photos, which comprises:
the first processing module is configured to collect winter wheat digital photos to obtain a training set and a verification set;
the second processing module is configured to perform data enhancement on the training set to obtain a training set after data enhancement;
the third processing module is configured to take the actual tiller number as a label and make sample labels on the training set and the verification set after the data enhancement;
a fourth processing module configured to construct a tiller number extraction model;
a fifth processing module configured to train and set parameters of the tillering number extraction model by applying the training set after data enhancement and the corresponding label;
a sixth processing module configured to quantitatively evaluate accuracy of the tiller number extraction model for photo tiller number extraction of the verification set.
According to the system of the second aspect of the present invention, the first processing module is configured to collect a digital photograph of winter wheat, and the obtaining a training set and a verification set includes:
according to the following steps of 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo.
According to the system of the second aspect of the present invention, the second processing module is configured to collect a digital photograph of winter wheat, and the obtaining a training set and a verification set includes:
each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
According to the system of the second aspect of the present invention, the second processing module is configured to enhance the training set, and the enhancing includes:
carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; and (3) carrying out multi-sample synthesis on the digital photos of the training set, namely respectively and averagely cutting different photo frames in the digital photos of the training set into four parts, and then disturbing and re-splicing the photo frames to form a new photo frame.
According to the system of the second aspect of the present invention, the third processing module is configured to label the training set after the data enhancement with a real tiller number, and the sample labeling includes:
for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label.
According to the system of the second aspect of the present invention, the third processing module is configured to label the training set after the data enhancement with a real tiller number, and the sample labeling further includes:
for digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tillering numbers of the corresponding original photos.
According to the system of the second aspect of the present invention, the third processing module is configured to label the verification set with a true tiller number, and the sample labeling includes:
and (3) the ground actual measurement value of the tiller number used for the verification set label.
A third aspect of the invention discloses an electronic device. The electronic device comprises a memory and a processor, the memory storing a computer program, the processor implementing the steps in a method of extracting wheat tillers based on a digital photo of any one of the first aspects of the present disclosure when the computer program is executed.
A fourth aspect of the invention discloses a computer-readable storage medium. A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps in a method of extracting wheat tillers based on a digital photo of any one of the first aspects of the present disclosure.
Therefore, the proposal provided by the invention can extract the tillering number of wheat from the digital photo, ensures the relative error to be about 10 percent, can meet the requirement of taking the digital photo in the future, extracts the tillering number of winter wheat based on deep learning, and replaces manual field counting.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are needed in the description of the embodiments or the prior art will be briefly described, and it is obvious that the drawings in the description below are some embodiments of the present invention, and other drawings can be obtained according to the drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for extracting wheat tiller number based on digital photos according to an embodiment of the invention;
FIG. 2 is a digital photograph of a wheat plant with a label value 224 according to an embodiment of the present invention;
FIG. 3 is a digital photograph of a wheat plant with a label value of 240 according to an embodiment of the present invention;
FIG. 4 is a digital photograph of wheat plants with a label value of 248 according to an embodiment of the present invention;
FIG. 5 is a block diagram of a system for extracting wheat tiller number based on digital photos according to an embodiment of the present invention;
fig. 6 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments of the present invention. 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 discloses a method for extracting wheat tillering number based on digital photos. Fig. 1 is a flowchart of a method for extracting wheat tillering number based on a digital photo according to an embodiment of the present invention, as shown in fig. 1, the method includes:
s1, acquiring a winter wheat digital photo to obtain a training set and a verification set;
s2, carrying out data enhancement on the training set to obtain a training set with enhanced data;
s3, taking a real tiller number as a label, and making sample labels for the training set and the verification set after data enhancement;
s4, constructing a tillering number extraction model;
s5, training and parameter setting are carried out on the tillering number extraction model by applying the training set after data enhancement and the corresponding labels;
and S6, quantitatively evaluating the accuracy of the tillering number extraction model in extracting the photo tillering number of the verification set.
In step S1, a winter wheat digital photo is collected to obtain a training set and a verification set.
In some embodiments, in the step S1, the method for acquiring the winter wheat digital photos to obtain the training set and the verification set includes:
according to the following steps of 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo.
The method for acquiring the winter wheat digital photos to obtain the training set and the verification set further comprises the following steps:
each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
Specifically, according to 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo. The mode of determining the area can be through placing square frame on ground, also can mark length on the tape measure, put it on ground, according to 1:1, a square area of the ground is photographed vertically. Each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
At a small Shang Shanguo accurate agricultural base, the tape was placed on the ground by marking the length of 0.5m on the tape, according to 1:1, 6-7 vertical photos are taken near each sampling point, and finally 2600 winter wheat digital photos with the size ranging from 0.5m to 0.5m are obtained according to the following steps: the scale of 3 is divided into training and validation sets.
And in step S2, data enhancement is carried out on the training set, and the training set with the enhanced data is obtained.
In some embodiments, in the step S2, the method for enhancing the training set includes:
because the digital photos of the wheat are taken, the situation of shielding exists, and small differences of conditions such as shooting angles, illumination and the like exist among different photos. Carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; and (3) carrying out multi-sample synthesis on the digital photos of the training set, namely respectively and averagely cutting different photo frames in the digital photos of the training set into four parts, and then disturbing and re-splicing the photo frames to form a new photo frame. The number of digital photos of the training set after data enhancement is 5 times that of the training set.
And in step S3, taking the actual tiller number as a label, and making sample labels for the training set and the verification set after data enhancement.
In some embodiments, in the step S3, with the actual tiller number as a label, the method for making a sample label for the training set after data enhancement includes:
for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label.
For digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tillering numbers of the corresponding original photos.
The method for making the sample label of the verification set by taking the actual tiller number as the label comprises the following steps:
and (3) the ground actual measurement value of the tiller number used for the verification set label.
In step S4, a tiller number extraction model is constructed.
Specifically, based on a deep learning framework (pytorch is taken as an example here), a classical Convolutional Neural Network (CNN) is packaged in a class form, and is used as a backbone network for extracting features, so that the call (such as ResNet, VGGNet, denseNet and the like) in the subsequent training is facilitated; the convolution layer, pooling layer, activation layer, full connection layer are still built according to the original network structure, the Dense connection layer (i.e. nn. Linear in pytorch) is replaced in the output layer (i.e. last full connection layer), the number of output nodes is set to 1, the activation function is replaced from softmax by other activation functions (e.g. linear, sigmoid, etc.), and the loss function is replaced by MSELoss. The rest of the structure and parameters remain unchanged.
And in step S5, training and parameter setting are carried out on the tillering number extraction model by applying the training set after data enhancement and the corresponding label.
Specifically, training is performed based on a DenseNet extraction model, and the hardware environment for the training, verification and test process is as follows:
gold 6226R 2.9GHz 16 core processor, NVIDIA GeForce RTX 3090 24GB graphics card, 256GB Hailishi DDR4 memory. The software environment is Python 3.8; pyTorch 1.8; torchvision 0.6.
The initial learning rate is set to 0.01, and the learning rate is adaptively updated by calling the reduce lronplateau in the pytorch, i.e. when the verified loss function is no longer reduced after 10 eopchs, the learning rate is adjusted to be one half of the original. The training number (epoch) is set as: 200, the number of training samples per Batch (Batch Size) is set to: the network optimizer (optimizer) is set to: adam, L2 regularization coefficient is set to: the exponential decay rate of the first moment estimate was set to 0.00005 to 0.9-0.99, and the exponential decay rate of the second moment estimate was set to 0.999..
In step S6, the accuracy of the tillering number extraction model in extracting the photo tillering number of the verification set is quantitatively evaluated.
Specifically, an optimal tillering number extraction model based on a DenseNet backbone network which is finally trained is called, the tillering number of the digital photo of the verification set is extracted, the tillering number is compared with the label of the verification set, and indexes such as average absolute error (Mean Absolute Error, MAE), average relative error (Mean Relative Error, MRE), root mean square error (Root Mean Squard Error, RMSE) and correlation coefficient r are used for quantitatively evaluating the extraction accuracy of the digital photo of the image of the test set.
Wherein y is i In order to be able to predict the value,
and m is the number of samples.
In the case of Batch size=8, the test results of the model at different learning rates are shown in table 1 below:
table 1 results of accuracy verification on different learning rate test sets
Finally, obtaining a tillering number extraction model with highest precision on the learning rate of 5e-5, extracting the digital photo of the verification set by using the obtained tillering number extraction model, wherein a final obtained tillering number extraction result sample is shown in, for example, fig. 2-4, and fig. 2 is a digital photo of wheat planting with a label value of 224 according to an embodiment of the invention, and the final extraction result is 224.12; FIG. 3 is a digital photograph of a wheat plant with a label value of 240 according to an embodiment of the present invention, with a final extraction result of 239.96; fig. 4 is a digital photograph of wheat plants with a label value of 248 according to an embodiment of the present invention, and the final extraction result is 248.65.
In summary, the proposal provided by the invention can extract the tillering number of wheat from the digital photo, ensures the relative error to be about 10 percent, can meet the requirement of taking the digital photo in the future, extracts the tillering number of winter wheat based on deep learning, and replaces manual field counting.
The invention discloses a system for extracting wheat tillering number based on digital photos. FIG. 5 is a block diagram of a system for extracting wheat tiller number based on digital photos according to an embodiment of the present invention; as shown in fig. 5, the system 100 includes:
a first processing module 101 configured to collect a winter wheat digital photograph, resulting in a training set and a validation set;
the second processing module 102 is configured to perform data enhancement on the training set to obtain a training set after data enhancement;
a third processing module 103, configured to perform sample label making on the training set and the verification set after the data enhancement by taking the actual tiller number as a label;
a fourth processing module 104 configured to construct a tiller number extraction model;
a fifth processing module 105 configured to train and set parameters for the tillering number extraction model by applying the training set and the corresponding label after data enhancement;
a sixth processing module 106 configured to quantitatively evaluate the accuracy of the tiller number extraction model for photo tiller number extraction of the verification set.
According to the system of the second aspect of the present invention, the first processing module 101 is configured to collect digital photos of winter wheat, and the obtaining the training set and the verification set includes:
according to the following steps of 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo.
According to the system of the second aspect of the present invention, the second processing module 102 is configured to collect digital photos of winter wheat, and the obtaining the training set and the verification set includes:
each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
According to the system of the second aspect of the present invention, the second processing module 102 is configured to enhance the training set, including:
carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; and (3) carrying out multi-sample synthesis on the digital photos of the training set, namely respectively and averagely cutting different photo frames in the digital photos of the training set into four parts, and then disturbing and re-splicing the photo frames to form a new photo frame.
According to the system of the second aspect of the present invention, the third processing module 103 is configured to label the training set after the data enhancement with the actual tiller number, and the sample labeling includes:
for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label.
According to the system of the second aspect of the present invention, the third processing module 103 is configured to label the training set after the data enhancement with a real tiller number, and the sample labeling further includes:
for digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tillering numbers of the corresponding original photos.
According to the system of the second aspect of the present invention, the third processing module 103 is configured to label the verification set with a true tiller number, and the labeling of the verification set includes:
and (3) the ground actual measurement value of the tiller number used for the verification set label.
A third aspect of the invention discloses an electronic device. The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor executes the computer program to realize the steps in the method for extracting the wheat tillering number based on the digital photo in any one of the first aspect of the disclosure.
Fig. 6 is a block diagram of an electronic device according to an embodiment of the present invention, and as shown in fig. 6, the electronic device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Wherein the processor of the electronic device is configured to provide computing and control capabilities. The memory of the electronic device includes a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the electronic device is used for conducting wired or wireless communication with an external terminal, and the wireless communication can be achieved through WIFI, an operator network, near Field Communication (NFC) or other technologies. The display screen of the electronic equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the electronic equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the structure shown in fig. 6 is merely a structural diagram of a portion related to the technical solution of the present disclosure, and does not constitute a limitation of the electronic device to which the present application is applied, and that a specific electronic device may include more or less components than those shown in the drawings, or may combine some components, or have different component arrangements.
A fourth aspect of the invention discloses a computer-readable storage medium. The computer readable storage medium has stored thereon a computer program which, when executed by a processor, performs the steps in a method for extracting wheat tillers based on a digital photo according to any one of the first aspects of the present disclosure.
Note that the technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be regarded as the scope of the description. The foregoing examples represent only a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.
Claims (6)
1. A method for extracting wheat tillering number based on digital photos, which is characterized by comprising the following steps:
s1, acquiring a winter wheat digital photo to obtain a training set and a verification set;
s2, carrying out data enhancement on the training set to obtain a training set with enhanced data;
in the step S2, the method for enhancing the training set includes:
carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; the digital photos of the training set are synthesized by multiple samples, namely, different photo frames in the digital photos of the training set are respectively cut into four parts on average and then are mixed and spliced again to form new photo frames;
s3, taking a real tiller number as a label, and making sample labels for the training set and the verification set after data enhancement;
in the step S3, the method for making the sample label for the training set after the data enhancement by using the actual tiller number as the label includes: for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label;
in the step S3, with the actual tiller number as a label, the method for making a sample label for the training set after the data enhancement further includes: for the digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tiller number of the corresponding original photo;
in the step S3, the method for making the sample label for the verification set by using the actual tiller number as the label includes: ground actual measurement values of tiller numbers used for the verification set labels;
s4, constructing a tillering number extraction model;
based on a deep learning framework, packaging a classical convolutional neural network-CNN in a class form to serve as a backbone network for extracting features; the convolution layer, the pooling layer, the activation layer and the full connection layer are still built according to the original network structure, a Dense connection layer is replaced in an output layer, the number of output nodes is set to be 1, the activation function is replaced by other activation functions from softmax, and a loss function is replaced by MSE loss; the rest structure and parameters remain unchanged;
s5, training and parameter setting are carried out on the tillering number extraction model by applying the training set after data enhancement and the corresponding labels;
s6, quantitatively evaluating the accuracy of the tillering number extraction model in extracting the photo tillering number of the verification set;
invoking a final trained optimal tillering number extraction model based on a DenseNet backbone network, extracting the tillering number of the digital photo of the verification set, comparing the tillering number with the label of the verification set, and quantitatively evaluating the extraction accuracy of the digital photo of the image of the test set by using an average absolute error MAE, an average relative error MRE, a root mean square error RMSE and a correlation coefficient r index;
wherein y is i In order to be able to predict the value,
and m is the number of samples.
2. The method for extracting wheat tillering number based on digital photo according to claim 1, wherein in step S1, the method for collecting winter wheat digital photo, obtaining training set and verification set comprises:
according to the following steps of 1:1, taking a square area of the ground vertically, and collecting a winter wheat digital photo.
3. The method according to claim 2, wherein in the step S1, the method for collecting the winter wheat digital photos to obtain the training set and the validation set further comprises:
each digital photo is uniformly stored as JPG pictures according to an RGB format of 1024 x 3, and each JPG picture has corresponding geographic position coordinate information when being stored.
4. A system for extracting wheat tillering numbers based on digital photos, the system comprising:
the first processing module is configured to collect winter wheat digital photos to obtain a training set and a verification set;
the second processing module is configured to perform data enhancement on the training set to obtain a training set after data enhancement;
carrying out space geometric transformation on the digital photos of the training set, namely horizontal overturning and vertical overturning; performing pixel color conversion on the digital photo of the training set, namely adding Gaussian noise and sharpening an image; the digital photos of the training set are synthesized by multiple samples, namely, different photo frames in the digital photos of the training set are respectively cut into four parts on average and then are mixed and spliced again to form new photo frames;
the third processing module is configured to take the actual tiller number as a label and make sample labels on the training set and the verification set after the data enhancement;
taking the actual tiller number as a label, and making a sample label for the training set after data enhancement comprises the following steps:
for the digital photo subjected to space geometric transformation and pixel color transformation, the ground actual measurement value of the tiller number used by the label;
taking the actual tiller number as a label, and making a sample label on the training set after the data enhancement further comprises the following steps:
for the digital photos synthesized by multiple samples, the label uses the sum of actual measurement values of the tiller number of the corresponding original photo;
taking the actual tiller number as a label, and carrying out sample label making on the verification set comprises the following steps:
ground actual measurement values of tiller numbers used for the verification set labels;
a fourth processing module configured to construct a tiller number extraction model;
based on a deep learning framework, packaging a classical convolutional neural network-CNN in a class form to serve as a backbone network for extracting features; the convolution layer, the pooling layer, the activation layer and the full connection layer are still built according to the original network structure, a Dense connection layer is replaced in an output layer, the number of output nodes is set to be 1, the activation function is replaced by other activation functions from softmax, and a loss function is replaced by MSE loss; the rest structure and parameters remain unchanged;
a fifth processing module configured to train and set parameters of the tillering number extraction model by applying the training set after data enhancement and the corresponding label;
a sixth processing module configured to quantitatively evaluate accuracy of the tiller number extraction model for photo tiller number extraction of the verification set;
invoking a final trained optimal tillering number extraction model based on a DenseNet backbone network, extracting the tillering number of the digital photo of the verification set, comparing the tillering number with the label of the verification set, and quantitatively evaluating the extraction accuracy of the digital photo of the image of the test set by using an average absolute error MAE, an average relative error MRE, a root mean square error RMSE and a correlation coefficient r index;
wherein y is i In order to be able to predict the value,
and m is the number of samples.
5. An electronic device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps in a method of extracting wheat tiller number based on a digital photo as claimed in any one of claims 1 to 3 when the computer program is executed.
6. A computer readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, implements the steps of a method for extracting wheat tiller number based on digital photos according to any of claims 1 to 3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211414929.XA CN115690585B (en) | 2022-11-11 | 2022-11-11 | Method and system for extracting wheat tillering number based on digital photo |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211414929.XA CN115690585B (en) | 2022-11-11 | 2022-11-11 | Method and system for extracting wheat tillering number based on digital photo |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115690585A CN115690585A (en) | 2023-02-03 |
| CN115690585B true CN115690585B (en) | 2023-06-06 |
Family
ID=85052543
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211414929.XA Active CN115690585B (en) | 2022-11-11 | 2022-11-11 | Method and system for extracting wheat tillering number based on digital photo |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115690585B (en) |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10713484B2 (en) * | 2018-05-24 | 2020-07-14 | Blue River Technology Inc. | Semantic segmentation to identify and treat plants in a field and verify the plant treatments |
| CN109978047B (en) * | 2019-03-22 | 2022-05-27 | 南京农业大学 | Method for extracting tiller number of field wheat |
| CN111462058B (en) * | 2020-03-24 | 2024-01-05 | 华南农业大学 | Method for rapidly detecting effective rice ears |
| GB202009146D0 (en) * | 2020-06-16 | 2020-07-29 | Dark Horse Tech Ltd | System and method for crop monitoring |
| CN112733622B (en) * | 2020-12-25 | 2023-07-04 | 广西慧云信息技术有限公司 | Gramineae plant tillering quantity counting method |
| CN114463637B (en) * | 2022-02-07 | 2023-04-07 | 中国科学院空天信息创新研究院 | Winter wheat remote sensing identification analysis method and system based on deep learning |
| CN114494882B (en) * | 2022-02-07 | 2023-04-18 | 中国科学院空天信息创新研究院 | Winter wheat remote sensing identification analysis method and system based on random forest |
-
2022
- 2022-11-11 CN CN202211414929.XA patent/CN115690585B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN115690585A (en) | 2023-02-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108961269B (en) | Pig weight measuring and calculating method and system based on image | |
| CN111340141A (en) | Crop seedling and weed detection method and system based on deep learning | |
| CN111950396B (en) | Meter reading neural network identification method | |
| CN109141248A (en) | Pig weight measuring method and system based on image | |
| CN109886928A (en) | A kind of target cell labeling method, device, storage medium and terminal device | |
| CN109492665A (en) | Detection method, device and the electronic equipment of growth period duration of rice | |
| Roth et al. | Repeated multiview imaging for estimating seedling tiller counts of wheat genotypes using drones | |
| CN115481368A (en) | Vegetation coverage estimation method based on full remote sensing machine learning | |
| Sun et al. | Remote estimation of grafted apple tree trunk diameter in modern orchard with RGB and point cloud based on SOLOv2 | |
| CN115494007A (en) | Random forest based high-precision rapid detection method and device for soil organic matters | |
| Guo et al. | Panicle Ratio Network: streamlining rice panicle measurement by deep learning with ultra-high-definition aerial images in the field | |
| Xu et al. | An automatic wheat ear counting model based on the minimum area intersection ratio algorithm and transfer learning | |
| CN115690585B (en) | Method and system for extracting wheat tillering number based on digital photo | |
| CN109523509B (en) | Method and device for detecting heading stage of wheat and electronic equipment | |
| CN112116595A (en) | End-to-end automatic plant root system characteristic segmentation system | |
| Patel et al. | Deep Learning-Based Plant Organ Segmentation and Phenotyping of Sorghum Plants Using LiDAR Point Cloud | |
| CN115346212A (en) | Apple flower growth state detection method based on improved YOLOv5 | |
| CN114882385A (en) | Method for counting wheat ears in field based on unmanned aerial vehicle platform | |
| CN113222887A (en) | Deep learning-based nano-iron labeled neural stem cell tracing method | |
| CN111582246A (en) | Method and system for estimating grazing rate based on alpine meadow grassland grass yield | |
| CN104866897A (en) | Aphid situation investigation method and device | |
| CN114743108B (en) | Grassland mouse condition identification and quantification method and mouse condition recorder | |
| CN116052141B (en) | Crop growth period identification method, device, equipment and medium | |
| CN116740351A (en) | Crop leaf area index estimation method based on depth image semantic segmentation | |
| CN117935250A (en) | Phenotype extraction method for natural-morphology rice ears |
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 |