CN113008742B - Method and system for detecting deposition amount of fog drops - Google Patents
Method and system for detecting deposition amount of fog drops Download PDFInfo
- Publication number
- CN113008742B CN113008742B CN202110203887.4A CN202110203887A CN113008742B CN 113008742 B CN113008742 B CN 113008742B CN 202110203887 A CN202110203887 A CN 202110203887A CN 113008742 B CN113008742 B CN 113008742B
- Authority
- CN
- China
- Prior art keywords
- deposition amount
- layer
- aerial vehicle
- unmanned aerial
- data
- 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
- 230000008021 deposition Effects 0.000 title claims abstract description 202
- 238000000034 method Methods 0.000 title claims abstract description 61
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 73
- 238000001514 detection method Methods 0.000 claims abstract description 51
- 238000012549 training Methods 0.000 claims abstract description 51
- 230000007613 environmental effect Effects 0.000 claims description 46
- 238000005507 spraying Methods 0.000 claims description 35
- 240000007651 Rubus glaucus Species 0.000 claims description 25
- 235000011034 Rubus glaucus Nutrition 0.000 claims description 25
- 235000009122 Rubus idaeus Nutrition 0.000 claims description 25
- 230000006870 function Effects 0.000 claims description 19
- 238000011176 pooling Methods 0.000 claims description 13
- 239000000575 pesticide Substances 0.000 claims description 12
- 238000004590 computer program Methods 0.000 claims description 11
- 239000007921 spray Substances 0.000 claims description 10
- 238000009826 distribution Methods 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 9
- 238000003860 storage Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 7
- 239000002245 particle Substances 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 230000005540 biological transmission Effects 0.000 claims description 5
- 239000003595 mist Substances 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 3
- 238000000151 deposition Methods 0.000 description 172
- 238000002474 experimental method Methods 0.000 description 16
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 15
- 238000005070 sampling Methods 0.000 description 14
- 238000010586 diagram Methods 0.000 description 12
- 238000004891 communication Methods 0.000 description 9
- 238000012360 testing method Methods 0.000 description 9
- 238000012795 verification Methods 0.000 description 8
- 238000004364 calculation method Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000010606 normalization Methods 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 241000196324 Embryophyta Species 0.000 description 3
- 230000006378 damage Effects 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 244000068988 Glycine max Species 0.000 description 2
- 235000010469 Glycine max Nutrition 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000013401 experimental design Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000007788 liquid Substances 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000006855 networking Effects 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 239000000700 radioactive tracer Substances 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 241000238631 Hexapoda Species 0.000 description 1
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 241000607479 Yersinia pestis Species 0.000 description 1
- 230000003213 activating effect Effects 0.000 description 1
- 238000012271 agricultural production Methods 0.000 description 1
- 238000009529 body temperature measurement Methods 0.000 description 1
- 239000013256 coordination polymer Substances 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000001035 drying Methods 0.000 description 1
- 238000001704 evaporation Methods 0.000 description 1
- 230000008020 evaporation Effects 0.000 description 1
- 230000010006 flight Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 239000011347 resin Substances 0.000 description 1
- 229920005989 resin Polymers 0.000 description 1
- 230000008054 signal transmission Effects 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N15/00—Investigating characteristics of particles; Investigating permeability, pore-volume or surface-area of porous materials
- G01N15/04—Investigating sedimentation of particle suspensions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Chemical & Material Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Dispersion Chemistry (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a method and a system for detecting deposition amount of fog drops, wherein the method comprises the following steps: carrying out data acquisition on crop environment information to obtain target environment information data; and inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fogdrop deposition amount prediction model to obtain a detection result of the fogdrop deposition amount, wherein the trained fogdrop deposition amount prediction model is obtained by training a convolutional neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data. The invention reduces the operation time of the model while keeping the advantage of the convolutional neural network algorithm in extracting the data characteristics, and is convenient for remotely detecting the deposition amount of the fog drops in real time.
Description
Technical Field
The invention relates to the technical field of agricultural pesticide detection, in particular to a method and a system for detecting the deposition amount of fog drops.
Background
Unmanned aerial vehicle is an important field of fine agriculture research, and the application of unmanned aerial vehicle in spraying pesticides for preventing and treating plant diseases and insect pests has important significance in agriculture. Compared with the traditional manual backpack sprayer and self-propelled ground spraying equipment, the unmanned aerial vehicle has higher spraying height and smaller droplet size, so the droplet deposition of the unmanned aerial vehicle is easy to drift and is difficult to reach a target. The amount of deposited droplets is an important indicator of the volume of droplets deposited per unit area. The accurate fogdrop deposition amount is obtained, so that a reasonable spraying decision is made, the spraying quality is improved, and the pesticide consumption is reduced.
The most common method for detecting the droplet deposition is water-sensitive paper detection, and after the water-sensitive paper collects droplets, the droplet deposition information is analyzed by combining an image processing method. The method can directly observe the deposition effect. However, the accuracy of droplet deposition detection can be limited by the overlap of droplets on water-sensitive paper and the resolution of the scanner. In some studies, droplet deposition was detected indirectly by calculating and analyzing the amount of tracer added to the pesticide. The tracer method can sensitively detect the deposition of the fog drops, and the raw material cost is low. However, in order to obtain, but acquire, the droplet deposition data, subsequent data processing and complex analysis are required. There are also commercial products for detecting the quality of droplet deposits, such as laser particle size analyzers and leaf moisture sensors. Furthermore, analytical balances have been used to directly measure the mass of droplet deposition. This device is accurate in measuring the amount of deposited droplets. However, the method is expensive, is mostly used for academic research, and is not suitable for agricultural production application.
In the prior art, some fog drop deposition image detection systems based on yolo network are provided, the system scheme still uses water-sensitive paper to collect fog drops, the defect of water-sensitive paper application cannot be avoided, and the system is lack of portability; some provide a portable aviation application fog droplet deposition amount detection device, adopt the device to detect when the crop is intensive, can cause the injury to the crop, in addition, get into the field after applying medicine and gather the fog droplet deposition amount, can cause the harm to the health; some provide a plant protection machinery spraying operation droplet deposition on-line measuring system, because the electric capacity sensor output voltage who uses is relevant with the conductivity of liquid, when liquid conductivity changes, can influence the accuracy that the system detected droplet deposition, simultaneously, aviation operation droplet particle diameter is less, and the volume of spraying is less, and adoption electric capacity sensor can be difficult to support the collection of trace droplet deposition because of its sensitivity is not enough. Therefore, a method and a system for detecting deposition amount of mist droplets are needed to solve the above problems.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a method and a system for detecting the deposition amount of fog drops.
The invention provides a method for detecting deposition amount of fog drops, which comprises the following steps:
carrying out data acquisition on crop environment information to obtain target environment information data;
and inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fogdrop deposition amount prediction model to obtain a detection result of the fogdrop deposition amount, wherein the trained fogdrop deposition amount prediction model is obtained by training a convolutional neural network by using sample unmanned aerial vehicle operation parameter data and sample environment information data which are marked with input parameter type labels.
According to the method for detecting the deposition amount of the fog drops, the trained prediction model of the deposition amount of the fog drops is obtained by training through the following steps:
constructing a training sample set according to the sample environment information data, the sample unmanned aerial vehicle operation parameters and the corresponding sample droplet deposition amount;
and inputting the training sample set into the convolutional neural network for training, and if the training sample set meets a preset condition, acquiring a trained droplet deposition amount prediction model.
According to the method for detecting the deposition amount of the fogdrop, provided by the invention, the training sample set is input into the convolutional neural network for training, and if a preset condition is met, a trained fogdrop deposition amount prediction model is obtained, and the method comprises the following steps:
inputting the training sample set into the convolutional neural network for training to obtain a predicted fog droplet deposition amount, obtaining a loss function value of the convolutional neural network according to the predicted fog droplet deposition amount and the sample fog droplet deposition amount, and obtaining a trained fog droplet deposition amount prediction model if the loss function value meets a convergence condition through judgment.
According to the method for detecting the deposition amount of the fog drops, provided by the invention, the data acquisition is carried out on the environmental information to obtain the target environmental information data, and the method comprises the following steps:
the method comprises the steps that environmental information is collected through a raspberry serving node, target environmental information data are obtained, the raspberry serving node is used for processing information data collected by a temperature and humidity sensor, an air speed sensor and a GPS sensor, and the raspberry serving node sends the target environmental information to a cloud server end in a wireless transmission mode.
According to the fog droplet deposition amount detection method provided by the invention, the fog droplet deposition amount prediction model is a one-dimensional convolution neural network model, the one-dimensional convolution neural network model comprises 3 layers of convolution neural networks, 2 layers of pooling layers, 2 layers of full-link layers and 1 layer of output layers, and the activation function of the output layer of the fog droplet deposition amount prediction model is RELU.
According to the method for detecting the deposition amount of the fog drops, provided by the invention, 50% of neurons in the convolutional neural network have the weight value of 0.
According to the method for detecting the deposition amount of the fogdrop, before the preset unmanned aerial vehicle operation parameter data and the target environment information data are input into the trained fogdrop deposition amount prediction model, the method further comprises the following steps:
and preprocessing the preset unmanned aerial vehicle operation parameter data and the target environment information to obtain scaled data, wherein the preprocessing comprises standard normal distribution and statistical distribution standardization.
The invention also provides a droplet deposition amount detection system, comprising:
the environmental information data acquisition module is used for acquiring the crop environmental information to acquire target environmental information data;
and the droplet deposition amount detection module is used for inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained droplet deposition amount prediction model to obtain a detection result of the droplet deposition amount, wherein the trained droplet deposition amount prediction model is obtained by training a convolutional neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data.
The present invention also provides an electronic device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method for detecting deposition amount of droplets as described in any one of the above.
The present invention also provides a non-transitory computer-readable storage medium having stored thereon a computer program that, when executed by a processor, implements the steps of the method for detecting a deposition amount of droplets as described in any one of the above.
According to the method and the system for detecting the deposition amount of the fog drops, the target environment information data is obtained by acquiring the environment information, and then the preset unmanned aerial vehicle operation parameter data and the target environment information data are input into the trained prediction model of the deposition amount of the fog drops, so that the detection result of the deposition amount of the fog drops is obtained.
Drawings
In order to more clearly illustrate the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
Fig. 1 is a schematic flow chart of a method for detecting deposition amount of droplets according to the present invention;
FIG. 2 is a diagram illustrating comparison between a predicted value and a true value of the droplet deposition amount of the one-dimensional convolutional neural network model provided by the present invention;
FIG. 3 is a schematic diagram of the calculation of a one-dimensional convolution network according to the present invention;
FIG. 4 is a diagram of a one-dimensional convolutional neural network model architecture provided in the present invention;
FIG. 5 is a schematic structural view of a droplet deposition amount detection system according to the present invention;
FIG. 6 is a schematic diagram of an actual application of the droplet deposition amount detection system provided by the present invention;
FIG. 7 is a schematic diagram showing the comparison result between the fog drop deposition amount detection system provided by the present invention and the detection of the fog drop deposition amount by using the water-sensitive paper;
fig. 8 is a schematic structural diagram of an electronic device provided in the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings, and it is obvious that the described embodiments are some, but not all embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a method for detecting deposition amount of droplets according to the present invention, and as shown in fig. 1, the present invention provides a method for detecting deposition amount of droplets, including:
Because field operation environment is complicated, unmanned aerial vehicle is when the spraying operation, and the droplet not only can receive the environmental impact, also can exert an influence to environmental index in turn. Unmanned aerial vehicle sprays the back, along with spraying regional steam concentration increase, the vapor pressure risees, and relative humidity also changes. In addition, in the process of depositing the fog drops on the target crops, the phenomenon of water evaporation exists, so that the radiant heat reaching the crop canopies is reduced, and the change of the environmental temperature is caused.
In the invention, in step 101, after the unmanned aerial vehicle sprays pesticide according to preset unmanned aerial vehicle operation parameter data, a wireless sensor is adopted to acquire environmental temperature information and wind speed information to acquire target environmental information data. Wherein, wind speed information is the information of average wind speed, and ambient temperature information includes temperature before the spraying, humidity before the spraying, maximum humidity after the spraying and minimum temperature after the spraying, and predetermined unmanned aerial vehicle operation parameter includes that shower nozzle droplet particle diameter, unmanned aerial vehicle spray speed, unmanned aerial vehicle spray height and pesticide quantity.
Further, in step 102, the preset unmanned aerial vehicle operation parameter data and the target environment information data are input into a droplet deposition amount prediction model containing a Convolutional Neural Network (CNN) to obtain a detection result of the droplet deposition amount. The CNN model structure is designed based on a Keras deep learning framework sequence model, and comprises a convolution layer, a pooling layer, a full-connection layer and an output layer.
According to the method for detecting the deposition amount of the fogdrops, the target environment information data is obtained by collecting the crop environment information, and then the preset unmanned aerial vehicle operation parameter data and the target environment information data are input into the trained fogdrops deposition amount prediction model to obtain the detection result of the deposition amount of the fogdrops.
On the basis of the embodiment, the trained fogdrop deposition amount prediction model is obtained by training through the following steps:
constructing a training sample set according to the sample environment information data, the sample unmanned aerial vehicle operation parameters and the corresponding sample droplet deposition amount;
and inputting the training sample set into the convolutional neural network for training, and if the preset conditions are met, acquiring a trained fogdrop deposition amount prediction model.
In the invention, a large amount of data is required for establishing the fog drop deposition amount prediction model. Four different growth periods of soybeans were selected for 4 experiments, respectively, for collecting modeling data. The invention adopts full factor experimental design, selects 4 influencing factors and determines numerical levels of all the factors as sample unmanned aerial vehicle operation parameters according to unmanned aerial vehicle test environment and key technical parameters, namely spray nozzle droplet particle size (105 mu m, 120 mu m, 135 mu m, 150 mu m and 165 mu m), unmanned aerial vehicle spraying speed (1m/s, 2m/s, 3m/s, 4m/s, 5m/s and 6m/s), unmanned aerial vehicle spraying height (2.5m, 3m, 3.5m, 4m and 4.5m) and pesticide dosage (6L/ha, 9L/ha, 12L/ha, 15L/ha, 18L/ha and 21L/ha). Spray operations were performed using the XAG P20 plant protection drone according to the sample drone operational parameters. The method comprises the steps that a fog drop deposition amount detection system is applied to an experiment, the detection system collects environmental information of crops by using a wireless sensor, and sample environmental information data are obtained; and collecting deposited fog drops by using 2.6X 7.6cm of water-sensitive paper to obtain the deposition amount of the sample fog drops.
Optionally, the sample unmanned aerial vehicle operation parameters recorded to correspond to the droplet deposition amount include a nozzle droplet particle size, an unmanned aerial vehicle spray speed, an unmanned aerial vehicle spray height, and a pesticide usage amount. Meanwhile, the sample environment information required to be recorded in each spraying experiment comprises the temperature and humidity before the experiment, the maximum humidity within 5 minutes after the spraying, the minimum temperature within 5 minutes after the spraying and the average wind speed within 5 minutes after the experiment. After each experiment, the water-sensitive paper is immediately collected and sealed for storage. After being brought back to the laboratory for drying, the water sensitive paper is scanned into a digital image with 600dpi × 600dpi pixels by using a high resolution scanner. And (4) performing image processing by using DepositScan software to obtain the deposition amount of the sample fog drops.
Alternatively, 900 combinations were designed for each experiment to be tested, based on the idea of full-factorial experimental design. In order to reduce errors caused by the environment, 5 sampling points are set for each group of tests and used for calculating the average value of sampled data, and all the sampling points are located under the air route of the unmanned aerial vehicle. 3600 groups of data are obtained in 4 experiments, and the data comprise unmanned aerial vehicle operation parameters, environmental information data and corresponding droplet deposition amount and are used for subsequent analysis and modeling. 3200 group data is selected as a training sample set and used for training a model; selecting 200 groups of data as a verification set for verifying the generalization capability and the optimization model of the model; and selecting 200 groups of data as a test set, obtaining a predicted value of the droplet deposition amount, and comparing the predicted value of the droplet deposition amount with the sample droplet deposition amount value for evaluating the prediction capability of the model. And gradually adjusting the performance of the CNN prediction model through the training set and the verification set.
Optionally, inputting the training sample set into a convolutional neural network for training, and gradually converging a loss function in the convolutional neural network along with the increase of the iteration times until a preset condition is met, so that the fact that the fog drop deposition amount prediction model completes a preliminary training process is shown; and further, inputting the verification set into a preliminarily trained convolutional neural network model, and gradually converging a loss function in the convolutional neural network until a preset condition is met, so that the training process of the droplet deposition amount prediction model is finished. And the generalization capability and the prediction performance of the fog drop deposition amount prediction model are improved through multiple training iterations of the training set and the verification set.
On the basis of the above embodiment, the inputting the training sample set into the convolutional neural network for training, and if a preset condition is satisfied, obtaining a trained droplet deposition amount prediction model, includes:
inputting the training sample set into the convolutional neural network for training to obtain a predicted fog drop deposition amount, obtaining a loss function value of the convolutional neural network according to the predicted fog drop deposition amount and the sample fog drop deposition amount, and obtaining a trained fog drop deposition amount prediction model if the loss function value meets a convergence condition through judgment.
In the invention, the loss in the training process and the verification process changes along with the increase of the iteration times, the loss function curve smoothly decreases, the loss functions of the training set and the verification set gradually converge along with the increase of the iteration times, and the fog drop deposition amount prediction model can be used for an experimental data set. After 150 training iterations, the loss function change rate of the training set and the verification set tends to be 0, which indicates that the droplet deposition amount model has completed the training process.
Furthermore, a test set is selected to measure the prediction capability of the model, and the prediction accuracy and the regression fitting capability of the test set are important bases for evaluating the performance of the model.
FIG. 2 is a diagram illustrating comparison between a predicted value and a true value of droplet deposition amount of a one-dimensional convolutional neural network model provided by the present invention, and as shown in FIG. 2, R of a CNN model is found according to a comparison between predicted data and actual data 2 The model has better prediction performance, and the CNN can comprehensively extract the abstract features of the data through convolution operation. From this, the CNN model is suitable for predicting the deposition amount of the mist droplets.
In addition, the coefficient R was measured 2 For comparing the consistency of the predicted value and the actual value of the deposition amount of the fog dropsAnd the RMSE represents the root-mean-square error of the predicted value and the true value of the deposition amount of the fog drops and serves as a standard for measuring the model prediction result.
On the basis of the above embodiment, the acquiring the environmental information to acquire the target environmental information data includes:
the environmental information is collected through the raspberry pi node, target environmental information data are obtained, the environmental information is sent to the cloud server end through the 4G data transmission module, and the raspberry pi node is used for processing information data collected by the temperature and humidity sensor, the wind speed sensor and the GPS sensor.
In the invention, a Raspberry Pi (Raspberry Pi) -based platform is used as an acquisition node for processing data acquired by a wireless sensor, wherein the wireless sensor comprises a GPS (global positioning system) sensor, a temperature and humidity sensor and an air speed sensor. Acquiring the position information of each sampling point according to a GPS sensor; according to the temperature and humidity sensor and the wind speed sensor, environment temperature and humidity information and wind speed information of current position information of each sampling point are collected, the environment temperature and humidity information and the wind speed information are sent to a database of a cloud server based on a 4G network, and target environment information data are obtained according to the database of the cloud server.
On the basis of the above embodiment, the droplet deposition amount prediction model is a one-dimensional convolutional neural network model, the one-dimensional convolutional neural network model includes a 3-layer convolutional neural network, a 2-layer pooling layer, a 2-layer fully-connected layer, and a 1-layer output layer, and an output layer activation function of the droplet deposition amount prediction model is RELU.
In the invention, a one-dimensional CNN algorithm is adopted to establish a fog drop deposition amount prediction model, and CNN is utilized to accurately extract relevant characteristics of the fog drop deposition amount. By reconstructing the CNN structure, feature extraction is integrated into the multilayer perceptron. The CNN has good self-adaptive performance and is suitable for occasions with complex environmental information and ambiguous inference rules. The input layer of the CNN model is one-dimensional data, and the CNN model is also one-dimensional data output after convolution layer regression analysis. Therefore, the invention adopts the one-dimensional CNN structure, compared with the two-dimensional CNN, the one-dimensional CNN effectively utilizes the advantages of the convolution algorithm, and has the advantages of small calculation amount, less hardware requirement and higher calculation speed.
Fig. 3 is a schematic calculation diagram of a one-dimensional convolution network provided by the present invention. As shown in fig. 3, for a convolutional layer of the CNN model, the input of the previous layer is convolved with a convolution kernel to obtain the input of the layer:
wherein the content of the first and second substances,
the mth input of the l +1 th layer,
is the nth input to the layer l neurons,
is the convolution kernel weight between the nth input of the l layer neurons and the mth input of the l +1 layer,
is the offset corresponding to the mth input of layer l +1, N l Is the total input number of the l-th layer, f 1D Representing the convolution calculation of the one-dimensional CNN.
The convolution kernel is the core of CNN and is an important factor affecting feature extraction. The one-dimensional convolution calculation result of each layer of neuron convolution kernel is defined as follows:
where i ≦ 0 ≦ u-1, j ≦ 0 ≦ v-1, k ≦ 0 ≦ u-v, u is the length of each input, v is the length of each convolution kernel, and u-v +1 is the length of each output.
Fig. 4 is a schematic diagram of a one-dimensional convolutional neural network model architecture provided by the present invention, and as shown in fig. 4, the specific structure of the one-dimensional CNN model is as follows: 1. the first layer is a convolution layer, the number of filters is set to be 64, and the size of a convolution kernel is 3 multiplied by 1; 2. the second layer is a maximum pooling layer, the size of a convolution kernel is 3 multiplied by 1, and the number of filters is 64 so as to reduce the complexity of output and overfitting of a model; 3. two layers of convolution characteristics are added for learning characteristics with higher dimensionality, the third layer is a convolution layer with 32 filters in total, the size of a convolution kernel is 3 multiplied by 1, the fourth layer is a convolution layer with 16 filters, and the size of the convolution kernel is 3 multiplied by 1; 4. adding an average pooling layer at the fifth layer to avoid overfitting, wherein the size of a convolution kernel of the filter of the layer is 3 multiplied by 1, the number of the filters is 16, taking the average value of three weights to combine, and only one weight is left after each filter passes through the fifth layer; 5. the sixth layer and the seventh layer are all full connection layers with the unit of 16; 6. and the final output layer obtains an output value, namely the fog drop deposition output by the 1D-CNN model. Since the regression problem is to be solved by the model, the activation function of the output layer in the CNN structure is RELU, and the activation functions of other hidden layers are RELU.
On the basis of the above embodiment, the weight of 50% of the neurons in the convolutional neural network is 0.
In the invention, the overfitting problem of the CNN model is considered because the number of the neuron and the network layers is large. dropout is used for randomly setting the weight value to 0, and the method can reduce the sensitivity of the network to small changes of data and prevent overfitting. The model tests five parameters of dropout, and finds that when dropout is set to 0.5, 50% of the neuron weights in the CNN model are randomly set to 0, i.e., half of the neuron weights in the convolutional neural network are 0. In this case, the training loss of the validation set does not increase much compared to the set point, and the sensitivity of the CNN model to small perturbations can be reduced to avoid overfitting.
On the basis of the above embodiment, before inputting the preset unmanned aerial vehicle operation parameter data and the target environment information data into the trained droplet deposition amount prediction model, the method further includes:
and preprocessing the preset unmanned aerial vehicle operation parameter data and the target environment information data to obtain scaled data, wherein the preprocessing comprises standard normal distribution and statistical distribution standardization.
In the invention, when a one-dimensional CNN is used for establishing a model, input data needs to be preprocessed, and normalization and stardardization are adopted for zooming the data. normalization converts the data to [0, 1] and makes the data satisfy a standard normal distribution through normalization. If the order of magnitude of the sample input of the layer after the input layer is too large, the output of the layer will enter the saturation region in advance after activating the function. This situation may result in the activation function being insensitive to the characteristics. Thus, the input to CNN should be scaled, introducing batch normalization. batch normalization normalizes the statistical distribution of all samples. Meanwhile, each sample has own statistical distribution, so that the generalization capability of the CNN can be improved.
Fig. 5 is a schematic structural diagram of the droplet deposition amount detection system provided by the present invention, and as shown in fig. 5, the present invention provides a droplet deposition amount detection system, which includes an environmental information data acquisition module 501 and a droplet deposition amount detection module 502, where the environmental information data acquisition module 501 is configured to perform data acquisition on crop environmental information to obtain target environmental information data; the droplet deposition amount detection module 502 is configured to input preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained droplet deposition amount prediction model to obtain a detection result of droplet deposition amount, where the trained droplet deposition amount prediction model is obtained by training a convolutional neural network according to sample unmanned aerial vehicle operation parameter data and sample environment information data.
In the present invention. The environmental information data acquisition module 501 can acquire crop environmental information through raspberry pi nodes, and each node is composed of a temperature and humidity sensor, a wind speed sensor, a GPS sensor, a 4G data transmission module and a processing unit. Acquiring the position information of each sampling point according to a GPS sensor; according to the temperature and humidity sensor and the wind speed sensor, environment temperature and humidity information and wind speed information of current position information of each sampling point are collected, the environment temperature and humidity information and the wind speed information are sent to a database of a cloud server based on a 4G network, and target environment information data are obtained according to the database of the cloud server.
Wherein, the temperature and humidity sensor and the raspberry pi node pass through I 2 C, communication is carried out by a communication protocol; the wind speed sensor carries out analog-to-digital signal conversion through a PCF8591T module, and the converted digital signal is processed through I 2 C, the communication protocol is communicated with the raspberry sending node; the GPS sensor obtains the position of a sampling point through an SIM7600CE 4G expansion board, and performs signal transmission with a raspberry dispatching node based on a USB to UART interface of a CP 2102.
Furthermore, each raspberry pi node comprises a temperature and humidity sensor, a wind speed sensor, a GPS sensor, a 4G data transmission module and a processing unit. Raspberry Pi type 4B is used as a processing unit. SHT35 temperature and humidity sensor pass I 2 The protocol C is communicated with Raspberry Pi, and the temperature measurement error of the temperature and humidity sensor is +/-0.2 ℃, and the humidity measurement error is +/-1.5%. The measurement error of the wind speed sensor is +/-0.2 m/s, the output is an analog signal, and the PCF8591T module is used as an external A/D module of the raspberry pi to receive a voltage signal generated by the wind speed sensor because the raspberry pi does not have an analog-to-digital conversion module. The PCF8591T module is I-based 2 And the A/D conversion module of the C has 8bit precision. The module receives voltage signals of the wind speed sensor, and the voltage signals pass through the module I after analog-to-digital conversion 2 The C protocol is passed into the raspberry pie. In order to obtain the position information of the sampling point, a SIM7600CE 4G expansion board is additionally arranged on the raspberry derivative, the expansion board is designed based on a 40pin GPIO interface of the raspberry derivative, and is suitable for a raspberry derivative systemAnd a column main board. And 4G dial-up networking is supported, and GPS, Beidou and Glonass base station positioning is supported. The expansion board is provided with a GNSS antenna interface and an SIM card supporting DDAS and LTE main antennas so as to realize the acquisition of LBS information and 4G signals. The invention utilizes the expansion board to complete the acquisition of longitude and latitude positions of sampling points, transmits data into a raspberry group through a USB to UART interface based on a CP2102, and realizes the interaction of system hardware and a server database through a 4G network.
Optionally, the raspberry pi can also acquire a 4G signal through an expansion board by using an AT instruction, so as to meet the field networking requirement.
In the invention, the droplet deposition amount detection module 502 reads environmental temperature and humidity information, wind speed information and preset unmanned aerial vehicle operation parameter data from a database of a cloud server, and calculates the droplet deposition amount by using a droplet deposition amount prediction model of 1D-CNN in droplet deposition amount detection software, so as to obtain a final droplet deposition amount detection result. The 1D-CNN-based fog drop deposition amount prediction model can be stored as a tflite-format model and applied to fog drop deposition amount detection software added with a TensorFlow Lite deep learning framework, and the fog drop deposition amount detection software is installed in an intelligent android mobile phone. And loading fog drop deposition amount detection software of the 1D-CNN-containing fog drop deposition amount prediction model by taking an android smart phone as a platform, and realizing portable checking of the fog drop deposition amount.
Optionally, the droplet deposition amount detection system of the present invention is provided in a mobile phone as detection software. Fig. 6 is a schematic view of an actual application of the droplet deposition amount detection system provided by the present invention. As shown in fig. 6, environmental temperature and humidity information and wind speed information are collected by a temperature and humidity sensor and a wind speed sensor, the environmental information is sent to a database of a cloud server through a 4G network, and pesticide application parameter data of the unmanned aerial vehicle are manually input into the database. And reading and displaying data in a cloud server database by using droplet deposition amount detection software in the Android smart phone, and calculating a droplet deposition amount value by using an embedded 1D-CNN droplet deposition amount prediction model to realize portable and remote viewing of droplet deposition amount information.
In the invention, a server used by the system is built based on Tencent cloud service, and the cloud server is a standard S2 and is specifically configured to be a CPU1 core, a memory 2GB and a network speed of 2 Mbps. The system of the Server is Ubuntu Server 18.04.1LTS 64 bits, a MySQL database is configured in the Server and serves as a data storage center of the whole system, the data writing part in the database is mainly completed by a raspberry sending end, and the data reading and displaying are realized by droplet deposition amount detection software at an Android smart phone end. Under a Raspbian system, a MySQLdb library in python language is used as a drive of a raspberry group connected with MySQL, on the premise that the raspberry group is connected with a 4G network, environmental information data acquired by a sensor is automatically written into an MySQL data table established by a server, and meanwhile, operation parameter data of an unmanned aerial vehicle is manually written into the same data table.
In the invention, Android smart phone end fog drop acquisition software of the system is installed on a smart phone, on the premise that the smart phone is connected with a network, data access to a database under a cloud server under an Android language is realized by using a MySQL connector Java package, and the acquired database data is processed by a proper adapter to obtain a required data type. And storing the fog drop deposition amount prediction model based on the 1D-CNN as a tflite-format model, and applying the model to fog drop deposition amount detection software added with a TensorFlow Lite deep learning framework to realize the calculation of the fog drop deposition amount. The Android client of the system takes a Material Design language as an interface Design principle. The Android interface display depends on Activity in the Android component, and MainActivity, LonActivity, SignupAActivity, GroupAActivity and DataActivity and corresponding xml files are respectively created to realize a main interface, a login interface, a registration interface, a category interface and a data interface of the client. And related Material controls are used during interface design, and comprise a Toolbar, a sliding menu, a suspension button, a card type layout and a foldable title bar, and tools such as Baidu map Android SDK are jointly used for realizing visualization of droplet deposition data at an Android phone end.
Optionally, the intelligent mobile phone end fogdrop obtaining software can detect and display the fogdrop deposition amount data in real time on line, and can also read data stored in a cloud server database in a historical manner to detect the fogdrop deposition amount.
Optionally, the fog drop deposition amount detection system provided by the invention can be applied to a field environment, the GPS sensor is fixed at the top of a sampling crop canopy, and the raspberry pi module is sealed in the resin shell, so that the damage of pesticide fog drops of the unmanned aerial vehicle to electronic elements in the module can be prevented. In order to enable the obtained fogdrops to be closer to a real deposition state, a temperature and humidity sensor in the device is fixed near the blade at each sampling point, meanwhile, the wind speed sensor is installed on the supporting rod and inserted into soil near the sampling points, the system is powered on, automatic collection of the fogdrops deposition amount can be achieved, and collected fogdrops deposition amount data can be checked in real time through fogdrops deposition amount detection software at an Android end.
In order to compare the detection effect of the invention on the deposition amount of the fogdrop with that of the prior art, the detection system for the deposition amount of the fogdrop and the test result of the water-sensitive paper are adopted to carry out a comparison experiment. Experiments are carried out in the mature period of the soybeans, and the consistency degree of the fogdrop deposition obtained by comparing the fogdrop deposition amount detection system with the water sensitive paper is verified. The experiment is carried out for 5 flights, the grain diameter of the fogdrop of the unmanned aerial vehicle in each flight is set to be 135 microns, the flight height is 3m, and the flight speed is 3 m/s. The spraying amount of the unmanned aerial vehicle is changed in each flight, and is respectively 6L/ha, 9L/ha, 12L/ha, 15L/ha and 18L/ha. 10 sampling points are selected for each flight and are uniformly and randomly arranged below the unmanned plane flight line. The temperature range during the experiment was: 22.9-32.1 ℃, the air humidity range is 38-72%, and the wind speed range is as follows: 2.2m/s-3 m/s. By measuring the coefficient R 2 The consistency of the fog drop deposition amounts of 50 sampling points obtained by the two methods is compared, and the error of the system is evaluated by using RMSE (RMSE), wherein the specific formula is as follows:
wherein, y i Is the true value of the deposition amount of fog drops, h (x) i ) And m is the total number of samples.
In order to further evaluate the practical application effect of the model in the fog drop deposition amount detection system. And embedding the fog drop deposition amount prediction model based on 1D-CNN into a fog drop deposition amount detection system, and comparing the fog drop deposition amount situation obtained by the fog drop deposition amount detection system and the water-sensitive paper through the experimental method.
FIG. 7 is a schematic diagram showing the comparison result of the fog drop deposition amount detection system and the method for detecting the fog drop deposition amount by using the water-sensitive paper, and referring to FIG. 7, R of the fog drop deposition amount is obtained by the two methods 2 0.924 and RMSE 0.026, the accuracy obtained for the amount of deposited droplets was high. The accuracy of the model for predicting the deposition amount of the fog drops in a field experiment by using the detection system for the deposition amount of the fog drops is reduced for the prediction result of the test set, because the test set and the model training set are from the same experimental environment, and the experimental period of the verification experiment is later than the data acquisition period of the training set. Although the experimental plot and the experimental crop are the same, the accuracy of the model in the system for predicting the deposition amount of the fog drops is influenced due to different experimental time. For the present experiment, although R 2 The fog drop deposition amount is reduced but still larger than 0.9, so the consistency of the system and the water-sensitive paper for detecting the fog drop deposition amount is better, and the system is feasible for field real-time detection of the fog drop deposition amount.
According to the fog drop deposition amount detection system, the environmental information of crops is subjected to data acquisition through the environmental information data acquisition module 501 to obtain target environmental information data, then the preset unmanned aerial vehicle operation parameter data and the target environmental information data are input into the trained fog drop deposition amount prediction model through the fog drop deposition amount detection module 502 to obtain the detection result of the fog drop deposition amount.
The system provided by the present invention is used for executing the above method embodiments, and for the specific processes and details, reference is made to the above embodiments, which are not described herein again.
Fig. 8 is a schematic structural diagram of an electronic device provided in the present invention, and as shown in fig. 8, the electronic device may include: a processor (processor)801, a communication interface (communication interface)802, a memory (memory)803 and a communication bus 804, wherein the processor 801, the communication interface 802 and the memory 803 complete communication with each other through the communication bus 804. The processor 801 may call logic instructions in the memory 803 to perform a method of droplet deposition amount detection, the method comprising: carrying out data acquisition on crop environment information to obtain target environment information data; inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fogdrop deposition amount prediction model to obtain a detection result of fogdrop deposition amount, wherein the trained fogdrop deposition amount prediction model is obtained by training a convolutional neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data.
In addition, the logic instructions in the memory 803 may be implemented in the form of software functional units and stored in a computer readable storage medium when the logic instructions are sold or used as a stand-alone product. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-only memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.
In another aspect, the present invention also provides a computer program product comprising a computer program stored on a non-transitory computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, enable the computer to perform the method for detecting deposition amount of droplets provided by the above methods, the method comprising: carrying out data acquisition on crop environment information to obtain target environment information data; inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fogdrop deposition amount prediction model to obtain a detection result of fogdrop deposition amount, wherein the trained fogdrop deposition amount prediction model is obtained by training a convolutional neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data.
In yet another aspect, the present invention also provides a non-transitory computer-readable storage medium, on which a computer program is stored, the computer program being implemented by a processor to perform the method for detecting deposition amount of fog drops provided in the above embodiments, the method comprising: carrying out data acquisition on crop environment information to obtain target environment information data; and inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fogdrop deposition amount prediction model to obtain a detection result of the fogdrop deposition amount, wherein the trained fogdrop deposition amount prediction model is obtained by training a convolutional neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data.
The above-described embodiments of the apparatus are merely illustrative, and the units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one position, or may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.
Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware. Based on the understanding, the above technical solutions substantially or otherwise contributing to the prior art may be embodied in the form of a software product, which may be stored in a computer-readable storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method according to the various embodiments or some parts of the embodiments.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (9)
1. A method for detecting a deposition amount of mist droplets, comprising:
carrying out data acquisition on crop environment information to obtain target environment information data;
inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained droplet deposition amount prediction model to obtain a detection result of droplet deposition amount, wherein the target environment information data comprises pre-spraying temperature, pre-spraying humidity, post-spraying maximum humidity and post-spraying minimum temperature; the trained fogdrop deposition amount prediction model is obtained by training a one-dimensional convolution neural network by using sample unmanned aerial vehicle operation parameter data and sample environment information data which are marked with input parameter type labels;
the sample unmanned aerial vehicle operation parameter data comprises spray nozzle droplet particle size, unmanned aerial vehicle spraying speed, unmanned aerial vehicle spraying height and pesticide consumption; the sample environment information data comprises temperature and humidity information before spraying, temperature and humidity information after spraying and air speed information;
the one-dimensional convolutional neural network comprises 3 layers of convolutional neural networks, 2 layers of pooling layers, 2 layers of full-connection layers and 1 layer of output layer, the output layer activation function of the droplet deposition amount prediction model is RELU, and the one-dimensional convolutional neural network has the specific structure that:
the first layer is a convolution layer, the number of filters is 64, and the size of the convolution kernel is 3 multiplied by 1;
the second layer is a maximum pooling layer, the size of a convolution kernel is 3 multiplied by 1, and the number of filters is 64;
the third layer is a convolution layer with 32 filters in total, and the size of the convolution kernel is 3 multiplied by 1;
the fourth layer is a convolution layer, the number of the filters is 16, and the size of a convolution kernel is 3 multiplied by 1;
adding an average pooling layer in the fifth layer, wherein the convolution kernel size of the filters in the average pooling layer is 3 multiplied by 1, and the number of the filters is 16;
the sixth layer and the seventh layer are all fully connected layers and are used for feature combination, and the number of the filters of each fully connected layer is 16;
the eighth layer is an output layer.
2. The method for detecting the deposition amount of fog drops according to claim 1, wherein the trained model for predicting the deposition amount of fog drops is trained by the following steps:
constructing a training sample set according to the sample environment information data, the sample unmanned aerial vehicle operation parameters and the corresponding sample droplet deposition amount;
and inputting the training sample set into the one-dimensional convolutional neural network for training, and if a preset condition is met, acquiring a trained droplet deposition amount prediction model.
3. The method for detecting the deposition amount of fogdrops according to claim 2, wherein the step of inputting the training sample set into the one-dimensional convolution neural network for training, and if a preset condition is met, obtaining a trained prediction model of the deposition amount of fogdrops comprises:
inputting the training sample set into the one-dimensional convolution neural network for training to obtain a predicted fog droplet deposition amount, obtaining a loss function value of the one-dimensional convolution neural network according to the predicted fog droplet deposition amount and the sample fog droplet deposition amount, and obtaining a trained fog droplet deposition amount prediction model if the loss function value meets a convergence condition through judgment.
4. The method for detecting the deposition amount of the fogdrops according to claim 1, wherein the crop collects environmental information and acquires target environmental information data, and the method comprises the following steps:
environmental information is collected through a raspberry serving node, target environmental information data are obtained, the raspberry serving node is used for processing information data collected by a temperature and humidity sensor, a wind speed sensor and a GPS sensor, and the raspberry serving node sends the target environmental information to a cloud server side in a wireless transmission mode.
5. The method of claim 2, wherein the weight of 50% of neurons in the one-dimensional convolutional neural network is 0.
6. The method of claim 1, wherein prior to inputting the pre-set drone operation parameter data and the target environment information data into the trained predictive model of droplet deposition, the method further comprises:
and preprocessing the preset unmanned aerial vehicle operation parameter data and the target environment information data to obtain scaled data, wherein the preprocessing comprises standard normal distribution and statistical distribution standardization.
7. A droplet deposition amount detection system, comprising:
the environmental information data acquisition module is used for acquiring the crop environmental information to acquire target environmental information data;
the device comprises a fog drop deposition amount detection module, a target environment information prediction module and a control module, wherein the fog drop deposition amount detection module is used for inputting preset unmanned aerial vehicle operation parameter data and the target environment information data into a trained fog drop deposition amount prediction model to obtain a detection result of the fog drop deposition amount, and the target environment information data comprise a temperature before spraying, a humidity before spraying, a maximum humidity after spraying and a minimum temperature after spraying; the trained fogdrop deposition amount prediction model is obtained by training a one-dimensional convolution neural network through sample unmanned aerial vehicle operation parameter data and sample environment information data marked with input parameter type labels;
the sample unmanned aerial vehicle operation parameter data comprises spray nozzle droplet particle size, unmanned aerial vehicle spraying speed, unmanned aerial vehicle spraying height and pesticide consumption; the sample environment information data comprises temperature and humidity information before spraying, temperature and humidity information after spraying and air speed information;
the one-dimensional convolutional neural network comprises a 3-layer convolutional neural network, a 2-layer pooling layer, a 2-layer full-connection layer and a 1-layer output layer, the output layer activation function of the droplet deposition amount prediction model is RELU, and the specific structure of the one-dimensional convolutional neural network is as follows:
the first layer is a convolution layer, the number of filters is 64, and the size of the convolution kernel is 3 multiplied by 1;
the second layer is a maximum pooling layer, the size of a convolution kernel is 3 multiplied by 1, and the number of filters is 64;
the third layer is a convolution layer, 32 filters are totally arranged, and the size of a convolution kernel is 3 multiplied by 1;
the fourth layer is a convolution layer, the number of the filters is 16, and the size of a convolution kernel is 3 multiplied by 1;
adding an average pooling layer in the fifth layer, wherein the convolution kernel size of the filters in the average pooling layer is 3 x 1, and the number of the filters is 16;
the sixth layer and the seventh layer are all full connection layers and are used for feature combination, and the number of filters of each full connection layer is 16;
the eighth layer is an output layer.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of the method of any one of claims 1 to 6 when executing the computer program.
9. A non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program when executed by a processor implements the steps of the method for detecting deposition amount of droplets according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110203887.4A CN113008742B (en) | 2021-02-23 | 2021-02-23 | Method and system for detecting deposition amount of fog drops |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110203887.4A CN113008742B (en) | 2021-02-23 | 2021-02-23 | Method and system for detecting deposition amount of fog drops |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113008742A CN113008742A (en) | 2021-06-22 |
| CN113008742B true CN113008742B (en) | 2022-08-19 |
Family
ID=76408607
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110203887.4A Active CN113008742B (en) | 2021-02-23 | 2021-02-23 | Method and system for detecting deposition amount of fog drops |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113008742B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113588500B (en) * | 2021-07-06 | 2023-05-26 | 沈阳农业大学 | Method for detecting fog drop deposition effect in aviation pesticide application operation |
| CN113628174A (en) * | 2021-07-21 | 2021-11-09 | 上海交通大学 | Subjective and objective ultrasonic medical image quality evaluation method and system |
| CN115034276A (en) * | 2022-08-15 | 2022-09-09 | 农业农村部南京农业机械化研究所 | Method, device and system for detecting droplet distribution of high-flow spray head |
| CN115451877A (en) * | 2022-09-29 | 2022-12-09 | 天津大学 | Supersonic separator droplet deposition amount and deposition position acoustic prediction method |
| CN116087042B (en) * | 2023-02-16 | 2023-09-08 | 中国矿业大学 | Real-time monitoring system and method for dust deposition volume spatial distribution |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108469403B (en) * | 2018-01-16 | 2021-02-12 | 江苏大学 | Plant protection machinery spraying operation fog droplet deposition amount on-line detection system and method |
| CN108362744B (en) * | 2018-04-04 | 2023-06-30 | 农业部南京农业机械化研究所 | Indoor measuring device and method for dynamic spray deposition distribution characteristics of spray rod type sprayer |
| CN109507070A (en) * | 2018-10-22 | 2019-03-22 | 中国农业大学 | Detect the method and device of Droplet deposition |
| CN209525221U (en) * | 2019-01-23 | 2019-10-22 | 河北农业大学 | A kind of pesticide deposition characteristic waves are according to acquisition and pesticide deposition amount detection device |
| CN109557003A (en) * | 2019-01-23 | 2019-04-02 | 河北农业大学 | A kind of pesticide deposition quantity measuring method, device and data acquisition combination unit |
| CN110188720A (en) * | 2019-06-05 | 2019-08-30 | 上海云绅智能科技有限公司 | A kind of object detection method and system based on convolutional neural networks |
| CN110378431A (en) * | 2019-07-24 | 2019-10-25 | 中国人民解放军国防科技大学 | Convolutional neural network-based supersonic combustion chamber combustion mode detection method |
| CN110689519B (en) * | 2019-08-21 | 2022-06-17 | 华南农业大学 | Fog drop deposition image detection system and method based on yolo network |
| CN110838286B (en) * | 2019-11-19 | 2024-05-03 | 腾讯科技（深圳）有限公司 | Model training method, language identification method, device and equipment |
| CN111008733B (en) * | 2019-11-25 | 2023-11-28 | 华南农业大学 | Crop growth control method and system |
| CN111280151B (en) * | 2020-02-06 | 2021-10-26 | 山东农业大学 | Variable pesticide application control method based on cotton growth period recognition |
| CN112241951A (en) * | 2020-10-19 | 2021-01-19 | 深圳大学 | Agricultural monitoring method, system and computer equipment based on raspberry pi and LORA |
-
2021
- 2021-02-23 CN CN202110203887.4A patent/CN113008742B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN113008742A (en) | 2021-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113008742B (en) | Method and system for detecting deposition amount of fog drops | |
| Feng et al. | A comprehensive review on recent applications of unmanned aerial vehicle remote sensing with various sensors for high-throughput plant phenotyping | |
| Amarasingam et al. | A review of UAV platforms, sensors, and applications for monitoring of sugarcane crops | |
| CN108693119B (en) | Intelligent pest and disease damage investigation and printing system based on unmanned aerial vehicle hyperspectral remote sensing | |
| Yang et al. | Unmanned aerial vehicle remote sensing for field-based crop phenotyping: current status and perspectives | |
| Tardaguila et al. | Smart applications and digital technologies in viticulture: A review | |
| CN111582055B (en) | Unmanned aerial vehicle aviation pesticide application route generation method and system | |
| Sassu et al. | Advances in unmanned aerial system remote sensing for precision viticulture | |
| CN113920474B (en) | Internet of things system and method for intelligently supervising citrus planting situation | |
| CN111666855B (en) | Animal three-dimensional parameter extraction method and system based on unmanned aerial vehicle and electronic equipment | |
| CN112990262A (en) | Integrated solution system for monitoring and intelligent decision of grassland ecological data | |
| Postolache et al. | IoT-based systems for soil nutrients assessment in horticulture | |
| Hao et al. | Evaluation of UAV spraying quality based on 1D-CNN model and wireless multi-sensors system | |
| CN114898238A (en) | Wild animal remote sensing identification method and device | |
| Bendig | Unmanned aerial vehicles (UAVs) for multi-temporal crop surface modelling. A new method for plant height and biomass estimation based on RGB-imaging | |
| Nathgosavi | A survey on crop yield prediction using machine learning | |
| Singh et al. | Usage of Internet of Things based devices in smart agriculture for monitoring the field and pest control | |
| Sumnall et al. | Effect of varied unmanned aerial vehicle laser scanning pulse density on accurately quantifying forest structure | |
| CN113030903B (en) | Seaweed nutrition level inversion method based on unmanned aerial vehicle hyperspectrum and laser radar | |
| Gundim et al. | Precision irrigation trends and perspectives: a review | |
| Puletti et al. | Enhancing wall-to-wall forest structure mapping through detailed co-registration of airborne and terrestrial laser scanning data in Mediterranean forests | |
| CN113256567B (en) | Banana leaf area index detection method and system | |
| Bongomin et al. | UAV image acquisition and processing for high‐throughput phenotyping in agricultural research and breeding programs | |
| Livanos et al. | Extraction of reflectance maps for smart farming applications using unmanned aerial vehicles | |
| Arunmozhiselvi et al. | A systematic approach to agricultural drones using a machine learning model |
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 |