CN113936254B - Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction - Google Patents
Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction Download PDFInfo
- Publication number
- CN113936254B CN113936254B CN202111549697.4A CN202111549697A CN113936254B CN 113936254 B CN113936254 B CN 113936254B CN 202111549697 A CN202111549697 A CN 202111549697A CN 113936254 B CN113936254 B CN 113936254B
- Authority
- CN
- China
- Prior art keywords
- rice
- accumulated temperature
- field image
- stage
- rice field
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- G06N3/048—Activation functions
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a method, a method and a device for training a rice development period recognition model by combining accumulated temperature correction, which quotes rice field images and accumulated temperature data as recognition parameters, judges the development period of rice by combining image characteristics and accumulated temperature characteristics, fully considers the accumulated temperature rule followed by the growth and development of rice and obtains an accurate automatic recognition result of the development period. The deep learning neural network for identifying the rice development period is trained, the mapping from the rice field image and the accumulated temperature data to the rice development period identification result is constructed, the automatic judgment of the rice development period identification by computer vision is realized, and the identification speed is greatly improved. Meanwhile, the rice field image is subjected to strengthening treatment to improve the generalization capability, the rice field image is subjected to graying treatment through an improved ExG factor ultra-green feature algorithm, and the rice field image is segmented through the maximum inter-class variance, so that the green leaf features of the rice field image can be efficiently extracted, the identification precision is greatly improved, and the error of manual observation is avoided.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method, a method and a device for training a rice development period recognition model by combining accumulated temperature correction.
Background
Rice is the most important grain crop in China and is staple food for more than 60% of people in China. The timely acquisition of accurate rice development period information has great significance for rice growth monitoring, field management, quality and yield estimation and the like, the existing rice development period observation of a basic agricultural observation station still mainly depends on manual work, and due to the fact that some observation sites are far away from the station, observation personnel cannot guarantee that development periods or proper observation time cannot be missed, the development time recording is inaccurate, and the problem that the growth of local areas is delayed cannot be found. Meanwhile, the judgment of the rice development period per se has larger human errors due to subjectivity of different observers. Therefore, an efficient and accurate automatic identification method for the development period of rice is urgently needed.
Disclosure of Invention
The embodiment of the invention provides a method, a method and a device for training a rice development period recognition model by combining accumulated temperature correction, so as to eliminate or improve one or more defects in the prior art and solve the problems of insufficient timeliness, large error and the like in the manual recognition of the rice development period.
The technical scheme of the invention is as follows:
in one aspect, the invention provides a method for training a rice development period recognition model by combining accumulated temperature correction, which comprises the following steps:
acquiring a plurality of sample data, wherein each sample data comprises a rice field image of a shooting day in the whole growth cycle of the rice and daily accumulated temperature data comprising daily average temperature of the rice from a transplanting day to the shooting day;
and performing enhancement treatment on the rice field image in each sample, wherein the enhancement treatment at least comprises the following steps: size cutting, random transformation, normalization, graying treatment and plant segmentation treatment; calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data;
acquiring the rice field image and the effective accumulated temperature after the enhancement treatment of each sample, adding the corresponding development period as a label, and constructing a training sample set;
acquiring a preset neural network model, inputting the rice field image subjected to enhancement processing in a sample into a convolutional neural network by the preset neural network model to extract image characteristics, inputting the effective accumulated temperature in the sample into a first fully-connected neural network to extract accumulated temperature characteristics, connecting and combining the image characteristics and the accumulated temperature characteristics, and inputting the combined image characteristics and the accumulated temperature characteristics into a second fully-connected neural network to obtain a rice growth period identification result;
and training the preset neural network by adopting the training sample set to obtain a rice development period recognition model.
In some embodiments, the size cropping comprises cropping the padfield image in each sample into a three-channel RGB map of a set size; the random transformation comprises the step of horizontally turning, vertically turning or angularly transforming the rice field images in the samples; and the normalization comprises the step of respectively carrying out mean value removal normalization or standard normalization on the three-channel image layers of the paddy field image.
In some embodiments, the graying process includes: carrying out gray processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows:
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing;
in the enhancement processing of the rice field images in the samples, the plant segmentation processing comprises:
calculating a gray level histogram of the rice field image subjected to graying processing;
setting a gray value t, dividing the gray histogram into foreground color and background color, calculating the gray average value M of all pixels, calculating the gray average value MA of the foreground color and the gray average value MB of the background color, wherein the proportion of the number of foreground color pixels to the total number of pixels is PA, the proportion of the number of background color pixels to the total number of pixels is PB, calculating an inter-class variance ICV, and the calculation formula is as follows:
and traversing all the gray values t, and acquiring a t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
In some embodiments, the effective accumulated temperature of the rice from the transplanting day to the shooting day is calculated according to the day-by-day accumulated temperature data, and the calculation formula is as follows:
wherein A iseFor effective accumulation of temperature, TiThe average temperature of the i-th day from the transplanting day, B is the lower limit temperature of the growth and development of the rice, the lower limit temperature of the growth period from the transplanting initial stage to the jointing stage is 12 ℃, and the lower limit temperature of the growth period from the booting initial stage to the harvesting stage is 15 ℃.
In some embodiments, after calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data, the method further includes: and carrying out normalization processing on the effective accumulated temperature, wherein the calculation formula is as follows:
wherein the content of the first and second substances,
normalized for the effective integrated temperature of the ith sample, AeiFor the effective integrated temperature of the ith sample,
is the average value of the effective accumulated temperature of each sample,
the standard deviation of the effective integrated temperature of each sample is obtained.
In some embodiments, the developmental stage comprises at least: transplanting initial stage, transplanting common stage, turning green initial stage, turning green common stage, tillering initial stage, tillering common stage, jointing initial stage, jointing common stage, booting initial stage, booting common stage, heading initial stage, heading common stage, heading terminal stage, milk ripening initial stage, milk ripening common stage, ripening initial stage, ripening common stage and harvesting stage.
In some embodiments, the convolutional neural network is a ResNet network, the first fully-connected neural network comprises two layers, each layer comprising a fully-connected layer, a batch normalization layer, a ReLU activation function layer, and a random deactivation layer; the second fully-connected neural network comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer.
In another aspect, the present invention further provides a method for identifying a rice development period by combining accumulated temperature correction, comprising:
acquiring a rice field image to be identified and the daily average temperature of the rice field image to be identified in the period from a transplanting date to a shooting date;
preprocessing the rice field image to be identified, wherein the preprocessing at least comprises size cutting, normalization, graying processing and plant segmentation processing;
calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data;
inputting the preprocessed rice field image to be recognized and the effective accumulated temperature into the rice development period recognition model obtained by training in the rice development period recognition model training method combined with accumulated temperature correction, and obtaining a rice development period recognition result.
Preferably, in the method for identifying a rice development period by integrating temperature accumulation correction, the graying process includes: carrying out gray processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows:
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing;
the plant segmentation process includes:
calculating a gray level histogram of the rice field image subjected to graying processing;
setting a gray value t, dividing the gray histogram into foreground color and background color, calculating the gray average value M of all pixels, calculating the gray average value MA of the foreground color and the gray average value MB of the background color, wherein the proportion of the number of foreground color pixels to the total number of pixels is PA, the proportion of the number of background color pixels to the total number of pixels is PB, calculating an inter-class variance ICV, and the calculation formula is as follows:
and traversing all the gray values t, and acquiring a t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
In another aspect, the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of the method are implemented.
The invention has the beneficial effects that:
according to the method and the device for training the rice development period identification model by combining the accumulated temperature correction, the rice field image and the corresponding accumulated temperature data are simultaneously quoted as identification parameters, the development period of the rice is judged by combining the image characteristic and the accumulated temperature characteristic, the effective accumulated temperature is taken as a new characteristic quantity to be combined with the image data, the accumulated temperature rule followed by the growth and development of the rice is fully considered, and a more accurate identification result is obtained. The deep learning neural network for identifying the rice development period is trained, the mapping from the rice field image and the accumulated temperature data to the identification result of the rice development period is constructed, the automatic judgment of the identification by computer vision is realized, and the identification speed and the identification precision are greatly improved. Meanwhile, the rice field image is subjected to strengthening treatment, the generalization capability of the trained rice development period identification model can be enhanced through size cutting, random transformation and normalization treatment, the rice field image is subjected to graying treatment through an improved ExG factor ultragreen feature algorithm, the rice field image is segmented through the maximum inter-class variance, the green leaf features of the rice field image can be extracted efficiently, and the identification precision is greatly improved.
Furthermore, the method comprises the steps of configuring a model comprising a convolution neural network, a first fully-connected neural network and a second fully-connected neural network aiming at the paddy field image; setting a convolutional neural network as a ResNet network, and setting a first fully-connected neural network to comprise two layers, wherein each layer comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random deactivation layer; the second fully-connected neural network is arranged to comprise a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer, so that image characteristics and accumulated temperature characteristics of rice to be identified are fully mined, the initial stage, the common stage and the final stage of a unified development stage can be identified on a smaller granularity, and high-precision identification is realized.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims hereof as well as the appended drawings.
It will be appreciated by those skilled in the art that the objects and advantages that can be achieved with the present invention are not limited to the specific details set forth above, and that these and other objects that can be achieved with the present invention will be more clearly understood from the detailed description that follows.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the principles of the invention. In the drawings:
FIG. 1 is a schematic flow chart of a method for training a rice growth stage recognition model by combining temperature accumulation correction according to an embodiment of the present invention;
FIG. 2 is a schematic flow chart of a method for identifying a rice development period by combining temperature accumulation correction according to an embodiment of the present invention;
fig. 3 is a schematic diagram illustrating a paddy field image obtained by a camera in the training method of the rice growth stage recognition model combined with temperature accumulation correction according to an embodiment of the present invention;
FIG. 4 is a comparison of the rice field images before and after segmentation;
FIG. 5 is a schematic diagram illustrating a preset neural network structure in the method for training a rice growth stage recognition model by integrating temperature accumulation correction according to an embodiment of the present invention;
fig. 6 is a schematic diagram of a training process of the method for training a rice growth period recognition model by combining with temperature accumulation correction according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the following embodiments and accompanying drawings. The exemplary embodiments and descriptions of the present invention are provided to explain the present invention, but not to limit the present invention.
It should be noted that, in order to avoid obscuring the present invention with unnecessary details, only the structures and/or processing steps closely related to the scheme according to the present invention are shown in the drawings, and other details not so relevant to the present invention are omitted.
It should be emphasized that the term "comprises/comprising" when used herein, is taken to specify the presence of stated features, elements, steps or components, but does not preclude the presence or addition of one or more other features, elements, steps or components.
It is also noted herein that the term "coupled," if not specifically stated, may refer herein to not only a direct connection, but also an indirect connection in which an intermediate is present.
The existing technical method for identifying the rice development stage basically judges the rice development stage based on single two-dimensional image feature extraction and identification, and mainly depends on manual identification, and although the development stage capable of being identified is the main development stage in which most crops in the field are remarkably changed in morphology, such as the green turning stage, the tillering stage, the jointing stage, the booting stage, the heading stage, the milk stage and the mature stage of rice, larger errors exist. In fact, the crossing of the developmental stages of a crop, in addition to a distinct change in morphology (mutation), also has a change in internal physiology (gradual change). Theoretical research and practical application of agricultural meteorology prove that the plants require a certain amount of temperature accumulation from sowing to maturity. The accumulated temperature is the sum of the temperatures in a certain temperature range in a certain development period. Therefore, the agricultural meteorological observation data and the crop image data are combined to carry out combined deep learning, and the method is more reliable and accurate than the method which only depends on image identification and better accords with the physiological development rule of plants.
In actual agricultural weather service and agricultural activity guidance work, workers need to know more detailed development stages so as to accurately master the growth and development states of crops. For example, the tillering stage alone is divided into a beginning stage and a general stage, and the heading stage is divided into a beginning stage, a general stage and an end stage (full heading stage). In the prior art, no technical method for identifying the development stage of crops can provide a scheme for identifying the fine development stage of rice and cannot meet the requirement of observation business service in the development stage of rice. The temperature accumulation theory is introduced, which is helpful to identify the development period of rice more finely and accurately.
The invention provides a method for training a rice development period recognition model by combining accumulated temperature correction, which comprises the following steps of S101-S105:
step S101: and acquiring a plurality of sample data, wherein each sample data comprises a rice field image of a shooting day in the whole growth cycle of the rice and day-by-day accumulated temperature data comprising the average temperature of the rice day-by-day from the transplanting day to the shooting day period.
Step S102: and performing enhancement treatment on the rice field images in the samples, wherein the enhancement treatment at least comprises the following steps: size cutting, random transformation, normalization, graying treatment and plant segmentation treatment; and calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data.
Step S103: and acquiring the rice field image and the effective accumulated temperature after the enhancement treatment of each sample, adding the corresponding development period as a label, and constructing a training sample set.
Step S104: and acquiring a preset neural network model, inputting the rice field image subjected to enhancement processing in the sample into a convolutional neural network by the preset neural network model to extract image characteristics, inputting the effective accumulated temperature in the sample into a first fully-connected neural network to extract accumulated temperature characteristics, connecting and combining the image characteristics and the accumulated temperature characteristics, and inputting the combined image characteristics and the accumulated temperature characteristics into a second fully-connected neural network to obtain a rice growth period recognition result.
Step S105: and training the preset neural network by adopting a training sample set to obtain a rice development period recognition model.
In step S101, in the method for training a rice development period recognition model by integrating accumulated temperature correction in this embodiment, in order to construct a mapping from a rice field image, accumulated temperature data, and a development period, a plurality of samples are obtained based on existing data, each sample includes a rice field image, and a daily average temperature of rice in a period from a transplanting date to a shooting date. Specifically, in order to obtain better identification capability, the development stage corresponding to each sample should include all development stage types, and further detailed to the beginning stage, the common stage and the end stage of each development stage, so as to realize identification of finer granularity of the development stage by means of accumulated temperature data.
The sample data is obtained from the rice development period data historical data of observation stations in different years, and all the pictures are classified in standard development periods.
The paddy field image can be shot through the field camera, and the shooting angle can be divided into the level and shoot, be 30 degrees angles with the level and shoot and be 60 degrees angles with the level and shoot. The daily average air temperature is obtained by averaging the hourly air temperatures observed by the automatic field meteorological station.
In step S102, sample data is preprocessed, including enhancement processing of rice field images and accumulated temperature calculation of daily average temperature, to obtain more accurate characteristics.
Specifically, for image data in a sample, a field camera can directly acquire an image in an RGB format of a rice field, and the image comprises three channels of red, green and blue. In some embodiments, the size cropping comprises cropping the padfield image in each sample into a three-channel RGB map of a set size; the random transformation comprises the step of horizontally turning, vertically turning or angularly transforming the rice field images in the samples; and the normalization comprises the step of respectively carrying out mean value removal normalization or standard normalization on the three-channel image layers of the paddy field image.
The rice field images are firstly cut into standard images according to preset sizes, and preferably, all shot rice field images can be converted into 64 pixel × 3 channel (RGB) images for subsequent standardization processing.
And carrying out random transformation on the cut image, specifically, the random transformation comprises horizontal turning, vertical turning and angle transformation, and for limited data, through data enhancement processing, more samples can be formed to improve the generalization capability of model training. During the angle transformation process, the angle-transformed image is further processed into a 64 pixel × 3 channel (RGB) image.
Further, paddy field images are denormalizedProcessing makes the image resistant to attacks of geometric transformation, which can find those invariants in the image to know that the images are originally the same or a series. Specifically, in this embodiment, the mean value removing normalization processing is adopted, and the R, G, B channel values of the paddy field image are x respectivelyr、xg、xbThe calculation formula for the R channel conversion is:
the calculation formula after G channel conversion is as follows:
the calculation formula after B channel conversion is as follows:
furthermore, the images of the rice field are subjected to gray processing, and the plant part is segmented, so that background information of soil, sundries, people and the like in the rice field is eliminated, and characteristic information of the rice growth period is extracted to the maximum extent.
In some embodiments, in the step S102, in the enhancing the padfield image in each sample, the graying includes: carrying out gray processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows:
； （4）
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing; in addition, the phasesCompared with a general extreme green feature algorithm of the ExG factor, the parameters are corrected in the embodiment to adapt to the rice field image, so that the features are better extracted.
In the rice field image enhancement processing of each sample, the plant segmentation processing comprises the following steps of S1021 to S1023:
step S1021: and calculating a gray level histogram of the rice field image subjected to the graying treatment.
Step S1022: setting a gray value t, dividing a gray histogram into foreground colors and background colors, calculating the average value M of gray of all pixels, calculating the average value MA of gray of foreground colors and the average value MB of gray of background colors, calculating the inter-class variance ICV, wherein the ratio of the number of foreground color pixels to the total number of pixels is PA, the ratio of the number of background color pixels to the total number of pixels is PB, and the calculation formula is as follows:
； （5）
step S1023: and traversing all the gray values t, and acquiring the t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
In the steps S1021 to S1023 of the embodiment, the optimal segmentation threshold value of the rice field image is obtained through the inter-class variance with the maximum past value, so that the foreground color and the background color are effectively distinguished, and the extraction effect of the rice plant pixels is improved.
In some embodiments, in step S102, the effective accumulated temperature of the rice from the transplanting day to the shooting day is calculated according to the day-by-day accumulated temperature data, and the calculation formula is:
wherein A iseFor effective accumulation of temperature, TiThe average temperature of the i-th day from the transplanting day, and B is the lower limit temperature of the rice growth and development. The lower limit growth temperature from the transplanting initial stage to the jointing common stage is 12 ℃, and the lower limit growth temperature from the booting initial stage to the harvesting stage is 15 ℃.
In some embodiments, in step S102, after calculating the effective accumulated temperature of the rice from the transplanting day to the shooting day according to the day-by-day accumulated temperature data, the method further includes performing normalization processing on the effective accumulated temperature, and the calculation formula is as follows:
wherein the content of the first and second substances,
normalized for the effective integrated temperature of the ith sample, AeiFor the effective integrated temperature of the ith sample,
is the average value of the effective accumulated temperature of each sample,
the standard deviation of the effective integrated temperature of each sample is obtained.
In step S103, based on the rice field images after the enhancement processing of each sample and the corresponding effective accumulated temperature, the development period to which each sample belongs is labeled as a label, and a training sample set is formed for training.
In some embodiments, the method further comprises: randomly dividing samples in the training sample set into a training set with a first set proportion for training, a verification set with a second set proportion for verifying a rice development period recognition model obtained by training, and a test set with a third set proportion for testing; the sum of the first set ratio, the second set ratio and the third set ratio is 1. The training sample set is batched and used for training, verifying and testing respectively, so that the identification quality of the obtained model can be guaranteed.
Further, in the process of adding the label, the development period as the label at least comprises: transplanting initial stage, transplanting common stage, turning green initial stage, turning green common stage, tillering initial stage, tillering common stage, jointing initial stage, jointing common stage, booting initial stage, booting common stage, heading initial stage, heading common stage, heading terminal stage, milk ripening initial stage, milk ripening common stage, ripening initial stage, ripening common stage and harvesting stage. After training is carried out based on enough sample data, the development stage of the rice and the stage corresponding to the development stage can be identified with finer granularity.
In step S104, a preset neural network model is constructed for training, in this embodiment, the parameters for identifying the development period of the rice include two parameters, namely, a rice field image and an effective accumulated temperature, and for feature extraction of the rice field image, a Convolutional Neural Network (CNN) is adopted to obtain image features; for the features of the effective accumulated temperature, a first fully-connected neural network can be adopted for extraction. And finally, combining and connecting the image characteristics extracted by the convolutional neural network and the effective accumulated temperature characteristics extracted by the first fully-connected neural network, inputting values into the second fully-connected neural network for processing, and outputting a final recognition result.
In some embodiments, the convolutional neural network is a ResNet network, the first fully-connected neural network comprising two layers, each layer comprising a fully-connected layer, a batch normalization layer, a ReLU activation function layer, and a random deactivation layer; the second fully-connected neural network comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer.
In step S105, the preset neural network is trained using a training sample set, and supervised learning is performed using a general loss function, such as a cross entropy loss function.
On the other hand, the invention also provides a method for identifying the rice development period by combining the accumulated temperature correction, as shown in fig. 2, the method comprises the following steps of S201 to S203:
step S201: and acquiring the rice field image to be identified and the daily average temperature of the rice field image to be identified in the period from the transplanting date to the shooting date.
Step S202: and preprocessing the rice field image to be identified, wherein the preprocessing at least comprises size cutting, normalization, graying processing and plant segmentation processing.
Step S203: and calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data.
Step S204: inputting the preprocessed rice field image to be recognized and the effective accumulated temperature into the rice development period recognition model obtained by training in the rice development period recognition model training method combined with accumulated temperature correction in the steps from S101 to S105, and obtaining a rice development period recognition result.
Specifically, in step S202, the graying process includes: carrying out graying processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows according to a formula 4:
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing;
the plant segmentation process includes:
step S2021: and calculating a gray level histogram of the grayed rice field image.
Step S2022: setting a gray value t, dividing a gray histogram into foreground colors and background colors, calculating the average value M of gray of all pixels, calculating the average value MA of gray of foreground colors and the average value MB of gray of background colors, calculating an inter-class variance ICV, wherein the ratio of the number of foreground color pixels to the total number of pixels is PA, the ratio of the number of background color pixels to the total number of pixels is PB, and the calculation formula refers to formula 5:
step S2023: and traversing all the gray values t, and acquiring a t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
In another aspect, the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the steps of the method are implemented.
The invention is illustrated below with reference to a specific example:
the embodiment provides a rice fine development period identification method based on the combination of a CNN algorithm and a crop accumulated temperature theory, so that more accurate and more detailed identification of the rice development period is realized, and the defect of artificial observation is overcome. And acquiring the observation data of the rice in the past year, and processing sample data. The technical scheme of the embodiment is as follows:
1. and (3) rice field image data processing:
1.1 obtaining the image sequence of a paddy field which is longer than a full growth period.
1.2 classifying all the pictures in the development period according to the shooting time information of each image sample and the history data of the development period of the rice of the observation station of the corresponding year.
1.3 carry out data enhancement processing with the paddy field image, the purpose is to increase sample diversity, is favorable to neural network more comprehensive understanding to the crop characteristic.
1.4, carrying out plant segmentation treatment on the rice field image, eliminating background information such as soil, sundries, people and the like in the rice field, and extracting characteristic information of the rice growth period to the maximum extent.
2. Accumulated temperature data processing:
2.1 acquiring a daily average temperature data sequence corresponding to the rice field image data shooting time.
And 2.2 calculating the effective accumulated temperature corresponding to the shooting date of each picture.
And 2.3 integrating the image information and the accumulated temperature data to form data.
3. Constructing a deep learning neural network model:
3.1 for rice field image data, a Convolutional Neural Network (CNN) model is constructed.
3.2 for accumulated temperature data, constructing a full-connection neural network model.
4. Training a neural network model:
4.1 load image and form data for each sample.
4.2 input the image and form data into CNN model and fully connected neural network model respectively.
And 4.3, combining the outputs of the two network models, and inputting the outputs into another fully-connected neural network to generate a final developmental stage recognition result.
For the technical solution of this embodiment, the following is specific:
step 1: and acquiring a rice field image sequence of a certain rice field in a whole growth period.
The image acquisition mode is shooting by a field camera, and the camera can be provided with a shooting angle and a cruising path. The camera mounting method is shown in fig. 3, the main camera position is shot in three vertical angles, the first angle is basically parallel to the horizontal rod, the second angle is 30 degrees with the horizontal rod, and the third angle is 60 degrees with the horizontal rod. Each angle horizontally rotates 360 degrees, 1 piece of picture is shot every 45 degrees, and 8 pieces of pictures are shot in total. Every hour, 24 pictures are taken from angles one, two and three (3 rounds of scanning). The total number of theoretical images in the whole year is more than 70000.
Step 2: and classifying all the pictures in the development period according to the shooting time information of each picture sample and the historical data of the rice development period of the observation station of the corresponding year.
Removing image data before transplantation, and classifying other pictures according to development stage, wherein the total number of the pictures comprises 18 types: transplanting initial stage, transplanting common stage, turning green initial stage, turning green common stage, tillering initial stage, tillering common stage, jointing initial stage, jointing common stage, booting initial stage, booting common stage, heading final stage, milk ripening initial stage, milk ripening common stage, ripening initial stage, ripening common stage, and harvesting. Each developmental stage is further divided into a beginning stage, a general stage and an end stage.
And step 3: the rice field image is subjected to data enhancement processing, so that the diversity of samples is increased, and the understanding of the neural network on crop characteristics is more comprehensive.
Re-sizing: all pictures were converted to 64 pixel by 3 channels (RGB).
And (4) random overturning: and randomly extracting a certain proportion of pictures in the sample to turn over horizontally or vertically.
And (3) mean value removal normalization: r, G, B three channel valuesIs other than xr、xg、xbThe calculation formula for the R channel conversion is:
the calculation formula after G channel conversion is as follows:
the calculation formula after B channel conversion is as follows:
and 4, step 4: and (3) carrying out plant segmentation treatment on the rice field image, eliminating background information such as soil, sundries, people and the like in the rice field, and extracting characteristic information of the rice growth period to the maximum extent.
1) And performing graying processing on the rice field image by adopting an ExG factor ultragreen characteristic algorithm to obtain a gray scale image of the rice field.
R, G, B the three channel values are xr, xg and xb respectively, the theoretical calculation formula of the ExG factor is:
the ExG theoretical calculation method mainly aims at extracting green plants in the image, and has a good effect of extracting rice seedlings. However, when the rice begins to heading, the extraction effect of the theoretical algorithm on the yellow ear is not ideal. Through tests, the ExG factor algorithm is properly adjusted, and the extraction of seedlings and rice ears in the whole growth period of rice can be met. Therefore, the following modified ExG factor calculation formula is adopted in this embodiment, referring to equation 4:
2) and (5) carrying out plant segmentation treatment on the rice gray-scale map by adopting an Otsu method.
The Otsu method is a method for automatically calculating a threshold value for a foreground and a background of a gray image. Because the green characteristics of rice plants are obvious, the difference between the gray scale factor ExG value of the rice plants and the ExG values of other background objects in the image is obvious, and the optimal segmentation threshold can be conveniently found by using an Otsu method.
The Otsu method selects the criterion of the segmentation threshold as the maximum inter-class variance (intra-class variance), and the algorithm principle is as follows:
for a gray-scale image, its gray-scale histogram is computed:
wherein the content of the first and second substances,
is provided with a gray value
The number of the pixels of (a) is,
is the total number of pixels in the image. According to the definition formula, the compound can be obtained,
give out a pair
An estimate of the probability of occurrence, and therefore the histogram, indicates the distribution of the image gray values.
By arbitrarily selecting a gray value t, the gray histogram can be divided into a front part and a rear part, namely A and B, which correspond to the foreground color and the background color. Let the average value of the gray scales of all pixels of the image be M, and the average value of the gray scales of the pixels of A, B be MA and MB, the ratio of the number of pixels of the a part to the total number of pixels be PA, and the ratio of the number of pixels of the B part to the total number of pixels be PB. The inter-class variance definition given by Otsu is given by equation 5:
the optimal segmentation threshold t is the value that maximizes the ICV. For the image, various values of t can be traversed, the corresponding ICV is calculated, the optimal segmentation threshold value is obtained, when the gray level of the pixel in the image is larger than the threshold value, the pixel point is considered to be a rice plant, and otherwise, the pixel point is the background. For example, fig. 4 is a schematic diagram showing a comparison between before and after segmentation of the padfield image.
And 5: and organizing and generating comprehensive form data of image information and effective accumulated temperature.
And calculating the effective accumulated temperature of each picture from the rice transplanting date to the picture shooting date by using the rice development period data and the daily average temperature data corresponding to the image data sample time period of the agricultural meteorological observation station closest to the geographic position of the rice field. The calculation method is as follows:
extracting the average temperature of the pictures in the period from the transplanting day to the picture shooting day (n days in total)
；
Calculating the effective accumulated temperature of the time interval
Referring to formula 6:
wherein the content of the first and second substances,
the average daily temperature of the i-th day in the development stage, and B is the lower limit temperature of the rice growth and development in the development stage, and the values in this example are shown in table 1.
Watch (A)
Lower limit temperature of growth (B value, DEG C) of rice in each development stage
And organizing the sample picture name, the development period and the effective accumulated temperature corresponding to the picture shooting date into form data and storing the form data.
Step 6: and carrying out normalization processing on the accumulated temperature data.
Similar to the mean value removing normalization processing of the image data, the normalization processing method of the integrated temperature data refers to equation 7:
wherein the content of the first and second substances,
normalized for the effective integrated temperature of the ith sample, AeiFor the effective integrated temperature of the ith sample,
is the average value of the effective accumulated temperature of each sample,
the standard deviation of the effective integrated temperature of each sample is obtained.
And 7: and constructing a neural network model.
The neural network consists of two parts: for image data, a Convolutional Neural Network (CNN) is used; for accumulated temperature data, a fully connected neural network was used. The network model structure and flow are shown in fig. 5. Image and form data are first loaded for each sample and input into the CNN model and the fully connected neural network, respectively. And then combining the output characteristic quantities of the two networks, and inputting the combined output characteristic quantities into another fully-connected neural network to generate a final developmental stage recognition result.
In the embodiment, ResNet is selected as the CNN model, preprocessed rice field image data are input, and 256 characteristic quantities are output; the fully-connected neural network model for training accumulated temperature data comprises two layers, wherein each layer consists of a fully-connected network, Batch Normalization (Batch Normalization), a ReLU activation function and random inactivation (Dropout), effective accumulated temperature data sequences corresponding to images one by one are input, and 8 characteristic quantities are output. Combining the feature quantities output by the two networks into 264 feature quantities, inputting the feature quantities into a layer of fully-connected network formed by a fully-connected network, Batch Normalization, a ReLU activation function and random inactivation (Dropout), and outputting a final identification result of the developmental stage.
And 8: and (5) training a neural network model.
Referring to fig. 6, all the processed rice field image data samples and corresponding effective accumulated temperature data are randomly divided into 80% of training data sets and 20% of testing data sets (test sets); then, the training data set is divided into a training set (train set) and a verification set (valid set) by using a K-fold cross validation method (5-fold is used in the embodiment) for model training and model training.
And inputting the divided training set and verification set data into a constructed neural network model, and adjusting the model hyper-parameters (learning rate, batch size, iteration cycle and the like). And testing the development period recognition effect of the optimal model on the test data set to obtain a recognition result.
In the embodiment, the rice field image segmentation technology combining the ExG factor and the Otsu algorithm is used in the image preprocessing part, so that the background noise information of the rice field is eliminated to the maximum extent, and the rice characteristics can be learned by CNN (CNN); when calculating the ExG factor, the coefficient in the theoretical algorithm formula is adjusted, so that the method is more suitable for the segmentation of rice crop images; the effective accumulated temperature is used as a new characteristic quantity to be combined with the image characteristics for deep learning, and the accumulated temperature rule followed by the growth and development of rice is utilized to correct the development period identification of the rice field image, so that the method has more advantages than the method which only utilizes the image characteristics; by utilizing the development period data and the temperature data, the method not only can identify the main rice development period information, but also can identify the initial stage, the common stage and the final stage of the same development period on smaller granularity, and the identification result is more in line with the requirements of agricultural weather service; the invention provides a solution for overcoming the defect of artificial observation in the rice development period.
The technical scheme provided by the method is applied to the rice development period identification in 2018-2020 by the M3705 rice test station in Zhenjiang city, Jiangsu province, and the result shows that under the condition of selecting the CNNs with different structural complexity, the accuracy of the rice development period identification is improved to different degrees by adding the accumulated temperature correction compared with the simple CNN image identification. For example, when ResNet18 is selected as CNN, the accuracy of CNN identification in the development period is about 83.0%, and after the correction of accumulated temperature is added, the accuracy is improved to 89.4%; when ResNet50 is selected as CNN, the accuracy of CNN identification in the development period is about 88.5%, and after the correction by adding accumulated temperature, the accuracy is increased to 91.8%. Therefore, the invention plays an important role in the identification of the rice in the fine development stage.
In the embodiment, the rice field image segmentation technology combining the ExG factor and the Otsu algorithm is used in the image preprocessing part, so that the background noise information of the rice field is eliminated to the maximum extent, and the rice feature learning by the CNN is facilitated. When calculating the ExG factor, a theoretical algorithm formula is used
Is adjusted to
So that the method is more suitable for the segmentation of rice crop images. The effective accumulated temperature is taken as a new characteristic quantity to be combined with image data, and the accumulated temperature rule followed by the growth and development of rice is fully considered. By using a comprehensive network model containing CNN and a fully connected network, the rice field image data and the accumulated temperature form data can be synchronously and jointly trained. Capable of identifying the same developmental stage at a smaller granularityThe identification results at the beginning, the common period and the end period are more in line with the requirements of agricultural weather service.
In summary, in the method, the device and the apparatus for training the rice development period identification model by combining the accumulated temperature correction, the rice field image and the corresponding accumulated temperature data are simultaneously used as identification parameters, the development period of the rice is judged by combining the image characteristic and the accumulated temperature characteristic, the effective accumulated temperature is used as a new characteristic quantity to be combined with the image data, the accumulated temperature rule followed by the growth and development of the rice is fully considered, and a more accurate identification result is obtained. The deep learning neural network for identifying the rice development period is trained, the mapping from the rice field image and the accumulated temperature data to the identification result of the rice development period is constructed, the automatic judgment of the identification by computer vision is realized, and the identification speed and the identification precision are greatly improved. Meanwhile, the rice field image is subjected to strengthening treatment, the generalization capability of the trained rice development period identification model can be enhanced through size cutting, random transformation and normalization treatment, the rice field image is subjected to graying treatment through an improved ExG factor ultragreen feature algorithm, the rice field image is segmented through the maximum inter-class variance, the green leaf features of the rice field image can be extracted efficiently, and the identification precision is greatly improved.
The method includes the steps that a model comprising a convolutional neural network, a first fully-connected neural network and a second fully-connected neural network is configured for a paddy field image; setting a convolutional neural network as a ResNet network, and setting a first fully-connected neural network to comprise two layers, wherein each layer comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random deactivation layer; the second fully-connected neural network is arranged to comprise a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer, so that image characteristics and accumulated temperature characteristics of rice to be identified are fully mined, the initial stage, the common stage and the final stage of a unified development stage can be identified on a smaller granularity, and high-precision identification is realized.
Those of ordinary skill in the art will appreciate that the various illustrative components, systems, and methods described in connection with the embodiments disclosed herein may be implemented as hardware, software, or combinations of both. Whether this is done in hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, plug-in, function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine-readable medium or transmitted by a data signal carried in a carrier wave over a transmission medium or a communication link. A "machine-readable medium" may include any medium that can store or transfer information. Examples of a machine-readable medium include electronic circuits, semiconductor memory devices, ROM, flash memory, Erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, Radio Frequency (RF) links, and so forth. The code segments may be downloaded via computer networks such as the internet, intranet, etc.
It should also be noted that the exemplary embodiments mentioned in this patent describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above-described steps, that is, the steps may be performed in the order mentioned in the embodiments, may be performed in an order different from the order in the embodiments, or may be performed simultaneously.
Features that are described and/or illustrated with respect to one embodiment may be used in the same way or in a similar way in one or more other embodiments and/or in combination with or instead of the features of the other embodiments in the present invention.
The above description is only a preferred embodiment of the present invention, and is not intended to limit the present invention, and various modifications and changes may be made to the embodiment of the present invention by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A method for training a rice development period recognition model by combining accumulated temperature correction is characterized by comprising the following steps:
acquiring a plurality of sample data, wherein each sample data comprises a rice field image of a shooting day in the whole growth cycle of the rice and daily accumulated temperature data comprising daily average temperature of the rice from a transplanting day to the shooting day;
and performing enhancement treatment on the rice field image in each sample, wherein the enhancement treatment at least comprises the following steps: size cutting, random transformation, normalization, graying treatment and plant segmentation treatment; calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data;
acquiring the rice field image and the effective accumulated temperature after the enhancement treatment of each sample, adding the corresponding development period as a label, and constructing a training sample set;
acquiring a preset neural network model, inputting the rice field image subjected to enhancement processing in a sample into a convolutional neural network by the preset neural network model to extract image characteristics, inputting the effective accumulated temperature in the sample into a first fully-connected neural network to extract accumulated temperature characteristics, connecting and combining the image characteristics and the accumulated temperature characteristics, and inputting the combined image characteristics and the accumulated temperature characteristics into a second fully-connected neural network to obtain a rice growth period identification result;
training the preset neural network by adopting the training sample set to obtain a rice development period recognition model;
wherein, in the enhancement processing of the rice field images in each sample, the graying processing comprises: carrying out gray processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows:
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing;
in the enhancement processing of the rice field images in the samples, the plant segmentation processing comprises:
calculating a gray level histogram of the rice field image subjected to graying processing;
setting a gray value t, dividing the gray histogram into foreground color and background color, calculating the gray average value M of all pixels, calculating the gray average value MA of the foreground color and the gray average value MB of the background color, wherein the proportion of the number of foreground color pixels to the total number of pixels is PA, the proportion of the number of background color pixels to the total number of pixels is PB, calculating an inter-class variance ICV, and the calculation formula is as follows:
and traversing all the gray values t, and acquiring a t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
2. The method for training a rice development period recognition model combined with accumulated temperature correction according to claim 1, wherein in the enhancement processing of the rice field image in each sample, the size cutting comprises cutting the rice field image in each sample into a three-channel RGB map with a set size; the random transformation comprises the step of horizontally turning, vertically turning or angularly transforming the rice field images in the samples; and the normalization comprises the step of respectively carrying out mean value removal normalization or standard normalization on the three-channel image layers of the paddy field image.
3. The method for training a rice development period recognition model combined with accumulated temperature correction according to claim 1, wherein the effective accumulated temperature of the rice from the transplanting day to the shooting day is calculated according to the day-by-day accumulated temperature data, and the calculation formula is as follows:
wherein A iseFor effective accumulation of temperature, TiTo start from the transplanting dateThe average temperature of the ith day, B is the lower limit temperature of rice growth and development, the lower limit temperature of growth from the transplanting initial stage to the jointing stage is 12 ℃, and the lower limit temperature of growth from the booting initial stage to the harvesting stage is 15 ℃.
4. The method for training a rice development period recognition model by combining accumulated temperature correction according to claim 3, wherein after calculating the effective accumulated temperature of the rice from the transplanting day to the shooting day according to the day-by-day accumulated temperature data, the method further comprises: and carrying out normalization processing on the effective accumulated temperature, wherein the calculation formula is as follows:
wherein the content of the first and second substances,
normalized for the effective integrated temperature of the ith sample, AeiFor the effective integrated temperature of the ith sample,
is the average value of the effective accumulated temperature of each sample,
the standard deviation of the effective integrated temperature of each sample is obtained.
5. The method for training a rice development stage recognition model based on temperature correction as claimed in claim 1, wherein the development stage comprises at least: transplanting initial stage, transplanting common stage, turning green initial stage, turning green common stage, tillering initial stage, tillering common stage, jointing initial stage, jointing common stage, booting initial stage, booting common stage, heading initial stage, heading common stage, heading terminal stage, milk ripening initial stage, milk ripening common stage, ripening initial stage, ripening common stage and harvesting stage.
6. The method for training a rice development period recognition model combined with accumulated temperature correction according to claim 1, wherein the convolutional neural network is a ResNet network, the first fully-connected neural network comprises two layers, each layer comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer; the second fully-connected neural network comprises a fully-connected layer, a batch normalization layer, a ReLU activation function layer and a random inactivation layer.
7. A method for identifying a rice development period by combining accumulated temperature correction is characterized by comprising the following steps:
acquiring a rice field image to be identified and the daily average temperature of the rice field image to be identified in the period from a transplanting date to a shooting date;
preprocessing the rice field image to be identified, wherein the preprocessing at least comprises size cutting, normalization, graying processing and plant segmentation processing;
calculating the effective accumulated temperature of the rice from the transplanting date to the shooting date according to the day-by-day accumulated temperature data;
inputting the preprocessed rice field image to be recognized and the effective accumulated temperature into a rice development period recognition model obtained by training in the rice development period recognition model training method combined with accumulated temperature correction according to any one of claims 1 to 6, and obtaining a rice development period recognition result.
8. The method for identifying a rice development period in combination with temperature accumulation correction according to claim 7, wherein the graying process comprises: carrying out gray processing on the rice field image by adopting a corrected ExG factor ultragreen characteristic algorithm, wherein the calculation formula is as follows:
wherein x isrIs the R channel value, xgIs the value of G channel, xbThe value of the B channel is obtained, and the ExG is a gray value obtained after processing;
the plant segmentation process includes:
calculating a gray level histogram of the rice field image subjected to graying processing;
setting a gray value t, dividing the gray histogram into foreground color and background color, calculating the gray average value M of all pixels, calculating the gray average value MA of the foreground color and the gray average value MB of the background color, wherein the proportion of the number of foreground color pixels to the total number of pixels is PA, the proportion of the number of background color pixels to the total number of pixels is PB, calculating an inter-class variance ICV, and the calculation formula is as follows:
and traversing all the gray values t, and acquiring a t value with the maximum corresponding ICV value as an optimal segmentation threshold value for segmenting the plants in the rice field image.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method according to any of claims 1 to 8 are implemented when the processor executes the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111549697.4A CN113936254B (en) | 2021-12-17 | 2021-12-17 | Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111549697.4A CN113936254B (en) | 2021-12-17 | 2021-12-17 | Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113936254A CN113936254A (en) | 2022-01-14 |
| CN113936254B true CN113936254B (en) | 2022-03-01 |
Family
ID=79289227
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111549697.4A Active CN113936254B (en) | 2021-12-17 | 2021-12-17 | Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113936254B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115836639A (en) * | 2022-11-11 | 2023-03-24 | 四川省农业科学院园艺研究所 | Water and fertilizer supply method and device for tomato protected soilless substrate cultivation and storage medium |
| CN115797771A (en) * | 2022-12-07 | 2023-03-14 | 北大荒信息有限公司 | Method and device for judging leaf age of rice in cold region, computer and storage medium |
| CN116052141B (en) * | 2023-03-30 | 2023-06-27 | 北京市农林科学院智能装备技术研究中心 | Crop growth period identification method, device, equipment and medium |
| CN116453003B (en) * | 2023-06-14 | 2023-09-01 | 之江实验室 | Method and system for intelligently identifying rice growth vigor based on unmanned aerial vehicle monitoring |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304973A (en) * | 2018-02-11 | 2018-07-20 | 中国农业大学 | Area crops maturity period prediction technique based on accumulated temperature, radiation and soil moisture content |
| CN109492665A (en) * | 2018-09-28 | 2019-03-19 | 江苏省无线电科学研究所有限公司 | Detection method, device and the electronic equipment of growth period duration of rice |
-
2021
- 2021-12-17 CN CN202111549697.4A patent/CN113936254B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304973A (en) * | 2018-02-11 | 2018-07-20 | 中国农业大学 | Area crops maturity period prediction technique based on accumulated temperature, radiation and soil moisture content |
| CN109492665A (en) * | 2018-09-28 | 2019-03-19 | 江苏省无线电科学研究所有限公司 | Detection method, device and the electronic equipment of growth period duration of rice |
Non-Patent Citations (2)
| Title |
|---|
| 基于大尺度海温因子的江苏省水稻适宜移栽期预报模型研究;高苹等;《气象》;20151231;第41卷(第12期);全文 * |
| 水稻农业气候资源变化特征及影响分析;徐敏等;《中国农学通报》;20161231;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113936254A (en) | 2022-01-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113936254B (en) | Rice development period recognition model training method, recognition method and device combined with accumulated temperature correction | |
| CN110751019B (en) | High-resolution image crop automatic extraction method and device based on deep learning | |
| CN111461052A (en) | Migration learning-based method for identifying lodging regions of wheat in multiple growth periods | |
| CN109086826B (en) | Wheat drought identification method based on image deep learning | |
| RU2018143339A (en) | RECOGNITION OF WEEDS IN THE NATURAL ENVIRONMENT | |
| CN111898503B (en) | Crop identification method and system based on cloud coverage remote sensing image and deep learning | |
| JPWO2020044480A1 (en) | Server device of crop growth stage judgment system, growth stage judgment method and program | |
| CN108073947B (en) | Method for identifying blueberry varieties | |
| CN109063660B (en) | Crop identification method based on multispectral satellite image | |
| CN112734083A (en) | Rice harvester path planning control system based on machine vision | |
| CN114627467B (en) | Rice growth period identification method and system based on improved neural network | |
| CN113011221A (en) | Crop distribution information acquisition method and device and measurement system | |
| CN112330672B (en) | Crop leaf area index inversion method based on PROSAIL model and canopy coverage optimization | |
| CN111582035B (en) | Fruit tree age identification method, device, equipment and storage medium | |
| CN117197595A (en) | Fruit tree growth period identification method, device and management platform based on edge calculation | |
| CN110781865A (en) | Crop growth control system | |
| CN115828181A (en) | Potato disease category identification method based on deep learning algorithm | |
| CN114663791A (en) | Branch recognition method for pruning robot in unstructured environment | |
| CN114612794A (en) | Remote sensing identification method for land covering and planting structure in finely-divided agricultural area | |
| CN114299379A (en) | Shadow area vegetation coverage extraction method based on high dynamic image | |
| CN116052141B (en) | Crop growth period identification method, device, equipment and medium | |
| CN114463642A (en) | Cultivated land plot extraction method based on deep learning | |
| CN116310846B (en) | Disease and pest identification method and system based on image identification and data mining | |
| CN117151477B (en) | Crop anomaly identification method, device, electronic equipment and storage medium | |
| TWI709111B (en) | Method for rapidly positioning crops |
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 |