CN115643874A - Agricultural automatic accurate control variable rate fertilization method - Google Patents
Agricultural automatic accurate control variable rate fertilization method Download PDFInfo
- Publication number
- CN115643874A CN115643874A CN202211362912.4A CN202211362912A CN115643874A CN 115643874 A CN115643874 A CN 115643874A CN 202211362912 A CN202211362912 A CN 202211362912A CN 115643874 A CN115643874 A CN 115643874A
- Authority
- CN
- China
- Prior art keywords
- fertilizer
- network
- ditching
- fixed connection
- variable rate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Abstract
The invention discloses an agricultural automatic accurate control variable rate fertilization method, which is characterized by comprising the following steps: the method comprises the following steps: the fertilizing device is installed on the moving mechanism, and the moving mechanism is operated to enter an orchard in a hilly and mountainous area; step two: measuring and calculating the volume of the crown of the fruit tree; step three: determining the fertilizing amount according to the volume of the crown; step four: controlling the moving mechanism to drive the fertilizing device to move to a proper position; step five: and controlling the fertilizer applying device to output fertilizer for applying fertilizer. The invention relates to the field of variable rate fertilization methods, in particular to an agricultural automatic and accurate variable rate fertilization control method. The invention aims to provide an agricultural automatic and accurate variable rate fertilization control method, which is convenient for variable rate fertilization.
Description
Technical Field
The invention relates to the field of variable rate fertilization methods, in particular to an agricultural automatic and accurate variable rate fertilization control method.
Background
The fertilizer is made by chemical and/or physical methods and contains one or more nutrient elements needed by the growth of crops. Also called inorganic fertilizer, including nitrogenous fertilizer, phosphate fertilizer, potash fertilizer, trace element fertilizer, compound fertilizer, etc. In China, due to the fact that chemical fertilizers are used too extensively, the problems of unbalanced soil nutrition, reduced fertility, environmental pollution and the like are caused. To solve the above problems, precise variable control of the fertilizer is required.
If the method for automatically and accurately controlling the variable rate fertilization can be provided, the solid particle variable rate fertilization operation can be realized according to the relation among the relative variables such as the rotating speed of the fertilizer wheel, the fertilization speed, the volume of the fruit tree and the like. It is beneficial to realize variable rate fertilization.
Disclosure of Invention
The invention aims to provide an agricultural automatic and accurate variable rate fertilization control method, which is convenient for variable rate fertilization.
The invention adopts the following technical scheme to realize the purpose of the invention:
an agricultural automatic accurate control variable rate fertilization method is characterized by comprising the following steps:
the method comprises the following steps: the fertilizing device is arranged on the moving mechanism, and the moving mechanism is operated to the orchard in the hilly and mountainous areas;
step two: measuring and calculating the volume of the crown of the fruit tree;
step three: determining the fertilizing amount according to the volume of the crown;
step four: controlling the moving mechanism to drive the fertilizing device to move to a proper position;
step five: and controlling the fertilizer applying device to output fertilizer for applying fertilizer.
As a further limitation of the technical solution, the specific process of the second step is as follows:
step two is as follows: collecting a data set, shooting and collecting pictures of the fruit trees by using a camera, marking the pictures into an xml file in a VOC (volatile organic compound) format by using marking software LabelImg, and marking the fruit trees by using a rectangular frame, wherein the frame displays the name and position information marked by the pictures;
step two: processing images, namely, before inputting images into a convolutional network to extract different semantic information, carrying out a series of processing such as image normalization processing, image standardization processing and the like on the acquired images;
step two and step three: extracting network characteristics, wherein in the characteristic extraction process, low-level characteristics such as textures, frames, colors and the like are generally extracted by a shallow network, high-level and abstract characteristics are extracted by a deep network, and the main purpose is to fit a picture target with a marked real area so as to achieve the minimum error rate;
step two: the network weight is stored, after the network training is finished, a better weight value can be obtained between layers, the optimal prediction effect can be achieved, and the weight is reserved and used for detecting fruit trees in the follow-up process;
step two, five: fruit tree detection, namely processing pictures obtained in an actual environment and transmitting the processed pictures into a network on the basis of the previously obtained network weight, wherein the weight parameters between layers reach a better level in the training process, and the input images only need to be transmitted forward under the parameters to finally obtain the actual target class probability;
step two, step six: the whole graph is directly put in the input end of the YOLOv5 network, and then the position of a bounding box (a bounding box) and the category of the bounding box are regressed in the output layer.
As a further limitation of the technical solution, the specific process of the second step is as follows:
step two, step one: the method comprises the following steps that (1) a data set is amplified, a convolutional neural network has the capacity of processing brand new data, when samples are few, a model is easy to be over-fitted, the weight distribution is not reasonable enough, the network cannot have enough generalization capacity and feature summarization capacity, the original data set is expanded by changing the brightness, the chroma, the contrast, the sharpness and the like of an original image based on basic colors, and the original labeling result can be still kept due to the related image scale change, so that the adaptability of the network model to the environment is improved;
step two and step two: image normalization processing: in order to accelerate the network training, the pixel points of the acquired RGB image range from 0 to 255, and the pixel is converted into the pixel points from 0 to 1 by using the formula 2 to carry out normalization operation:
step two, step three: image normalization processing: the main principle of the method is that the data are subjected to mean value removal to realize centralized processing, the mean value and the variance of pixel points are calculated through (formula 3) and (formula 4), and then the data are subjected to standardization processing through (formula 5);
as a further limitation of the technical solution, the specific process of the second step and the third step is as follows:
step two, step three and step one: extracting the neural network characteristic information of the processed picture through a formula 6;
step two, step three, step two: the image size after convolution operation changes, mainly the number of channels increases and the length and width change, as (formula 7), the information of the spatial layer becomes more, and the extracted features are more beneficial to network training;
step two, step three: standardized network layer: the neural network training process is a process of continuously updating parameter fitting training data back and forth, and is used for standardizing a characteristic diagram and performance in order to improve the network optimization speed, wherein the standardization process is shown as (formula 8) (formula 10) and (formula 11), and the convergence capacity of the network is enhanced;
wherein: m is the number of images selected by one iteration;
μ B is the mean of one batch;
wherein:
is the variance;
wherein:
is a normalized value;
epsilon is a minimum number which prevents the denominator from being 0;
wherein: the input of the normalization layer is B = { x = 1，2…，m }；
y i Is the normalized output;
γ and β are two parameters learned in training.
As a further limitation of the technical scheme, the fertilizer device comprises a frame, the frame is fixedly connected with a fertilizer box, an outer sheave fertilizer discharger is fixedly connected with a discharge port of the fertilizer box, a driving motor reducer is fixed on the frame through a transmission main shaft and a pair of bearing seats, the outer sheave fertilizer discharger is fixed under the fertilizer box, an electric push rod is fixedly connected with the frame, one end of a compression spring is fixedly connected with a push rod end of the electric push rod, the other end of the compression spring is fixedly connected with the main shaft, the frame bearing is connected with the transmission main shaft, the transmission main shaft is fixedly connected with the driving motor, the driving motor is fixedly connected with a reversing driver, an output shaft of the driving motor is fixedly connected with an input shaft of the reversing driver, an output shaft of the reversing driver is fixedly connected with a driving transmission chain wheel, and the reversing driver is fixedly connected with a ditching and fertilizing component.
As a further limitation of the technical scheme, the ditching and fertilizing component comprises a material receiving hopper and a ditching spindle, wherein the material receiving hopper is connected with the ditching spindle through an expansion sleeve coupler, the material receiving hopper is fixedly connected with an expansion sleeve, the expansion sleeve is fixedly connected with a transmission chain wheel, the material receiving hopper is fixedly connected with a deep groove ball bearing, an outer ring of the deep groove ball bearing is fixedly connected with a suspension device, the suspension device is fixedly connected with a shell of the reversing driver, an anti-blocking spiral is arranged in the ditching spindle, a segmented spiral blade is fixedly connected with the outer side of the ditching spindle, the ditching spindle is provided with a fertilizer outlet, and the ditching spindle is fixedly connected with a soil entering blade.
As a further limitation of this aspect, one end of the drive chain encircles the drive sprocket and one end of the drive chain encircles the drive sprocket.
As a further limitation of the technical solution, the specific process of the step five is as follows: when the fertilizer box is in an initial state, the electric push rod is contracted to enable the ditching and fertilizing component to incline, so that the fertilizer box is convenient to transport, and when the fertilizer box is fertilized, the electric push rod is controlled to extend out, so that the transmission main shaft rotates, the ditching and fertilizing component enters soil in an inclined mode, the ditching and fertilizing component is prevented from being blocked in the soil entering stage, the driving motor reducer and the driving motor are controlled to be opened, the driving motor reducer drives the outer grooved wheel fertilizer discharger to rotate, so that the fertilizer in the fertilizer box is discharged into the receiving hopper, the driving motor drives the reversing driver to move, the reversing driver drives the transmission chain wheel to rotate through the transmission chain, the transmission chain wheel drives the expansion sleeve to rotate, the expansion sleeve drives the receiving hopper to rotate, the receiving hopper drives the expansion sleeve coupler, the ditching main shaft, the segmented spiral blades, the soil entering blades and the anti-blocking spiral blades to rotate, the segmented spiral blades realize rotary slotting, the fertilizer enters the main shaft along the receiving hopper to rotate and be sprayed out from the fertilizer outlet, and fertilizer particles and soil particles broken at the bottom layer are lifted by the segmented spiral blades and are stirred by the segmented spiral blades, so that the fertilizer can be effectively mixed with the fertilizer.
As a further limitation of the technical scheme, the receiving hopper is matched with the outer grooved wheel fertilizer distributor.
As a further limitation of the present technical solution, the frame is fixedly connected with a three-point suspension device, and the three-point suspension device is used for connecting the moving mechanism.
Compared with the prior art, the invention has the advantages and positive effects that:
1. the anti-blocking spiral of the device can rotate along with the rotating main shaft, but the spiral direction is opposite to the rotating direction, so that fertilizer particles and soil entering the hollow shaft can be smoothly discharged. From fertilizer outlet exhaust fertilizer granule and the broken soil granule of bottom by segmentation helical blade promotion to being mixed by the stirring of segmentation helical blade, can realizing the effectual mixture of soil and fertilizer, improving the utilization efficiency of fertilizer, and the broken soil of operation direction of advance cutting can be shed to the rear in the slot of opening, accomplishes the earthing operation. The annular ditching, fertilizing and earthing operations are completed at one time.
2. When this device normally fertilizies, compression spring is in the state of extending, can be compressed when the ditching fertilization is revolved the part and is met thick root system or stone, crosses the barrier, ensures that the device can not die or overload and damage.
3. The control system of the method realizes the variable fertilization operation of the solid particles according to the relation among the related variables such as the rotating speed of the fertilizer discharging wheel, the fertilizing vehicle speed, the volume of the fruit trees and the like and the control rule algorithm, and a certain amount of solid particle fertilizer can be uniformly spread in the target ditch.
Drawings
Fig. 1 is a general structure diagram of the ditching and fertilizing device of the invention.
Fig. 2 is a structural diagram of the vertical spiral ditching and fertilizing component of the invention.
FIG. 3 is a schematic flow chart of the present invention.
FIG. 4 is a schematic view of the installation of the tachometer encoder of the present invention.
In the figure: 1. the fertilizer box, 2, a frame, 3, a three-point suspension device, 4, a ditching and fertilizing component, 5, an outer grooved wheel fertilizer discharger, 6, a driving motor reducer, 7, a driving motor, 8, an electric push rod, 9, a driving chain, 10, a compression spring, 11, a driving main shaft, 12, a reversing driver, 13, a receiving hopper, 14, a driving chain wheel, 15, a deep groove ball bearing, 16, an expansion sleeve coupler, 17, a ditching main shaft, 18, a segmented spiral blade, 19, a fertilizer outlet, 20, an earth entering blade, 21, an expansion sleeve, 22, a suspension device, 23 and an anti-blocking spiral.
Detailed Description
One embodiment of the present invention will be described in detail below with reference to the accompanying drawings, but it should be understood that the scope of the invention is not limited to the embodiment.
The invention comprises the following steps:
the method comprises the following steps: the fertilizing device is installed on the moving mechanism, and the moving mechanism is operated to enter an orchard in a hilly and mountainous area;
step two: measuring and calculating the volume of the crown of the fruit tree;
step three: determining the fertilizing amount according to the volume of the crown;
step four: controlling the moving mechanism to drive the fertilizing device to move to a proper position;
step five: and controlling the fertilizer applying device to output fertilizer for applying fertilizer.
The specific process of the second step is as follows:
step two is as follows: collecting a data set, shooting and collecting pictures of the fruit trees by using a camera, marking the pictures as xml files in a VOC (volatile organic compound) format by using marking software LabelImg, and marking the fruit trees by using a rectangular frame, wherein the frame displays the information of the names and the positions marked by the pictures;
step two: processing images, namely, before inputting images into a convolutional network to extract different semantic information, carrying out a series of processing such as image normalization processing, image standardization processing and the like on the acquired images;
step two and step three: network feature extraction, in the process of feature extraction, generally shallow layer network extracts low-level features such as textures, frames and colors, deep layer network extracts high-level and abstract features, and the main purpose is to fit a picture target with a marked real area so as to achieve the minimum error rate;
step two, four: after the network training is finished, a better weight value can be obtained between each layer, the optimal prediction effect can be achieved, and the weight is reserved for detecting fruit trees subsequently;
step two and step five: fruit tree detection, namely processing pictures acquired under an actual environment and transmitting the processed pictures into a network on the basis of the previously acquired network weight, wherein the weight parameters among all layers reach a better level in the training process, and the input images only need to be transmitted forward under the parameters to finally acquire the actual target class probability;
step two, step six: the whole graph is directly put into the input end of the Yolov5 network, and then the position of a bounding box (bounding box) and the category to which the bounding box belongs are returned in the output layer.
The YOLOv5 has the main advantages that a Pythrch frame is used, the method is very user-friendly, a data set of the method can be conveniently trained, each batch of training data is transmitted through a data loader, and meanwhile the training data are enhanced; moreover, the model has a very light model size, the model training is very fast, and batch reasoning produces real-time results. The method can directly and effectively reason the input of the port of the network camera for single images, batch processing images, videos and even network cameras. The weight file can be converted into a format which can be used by the mobile phone and directly deployed to a mobile phone application end.
The second step comprises the following specific processes:
step two, step one: the data set is amplified, the convolutional neural network has the capacity of processing brand new data, but when the samples are few, the model is easy to over-fit, the weight distribution is not reasonable enough, the network cannot have enough generalization capacity and feature summarization capacity, the original data set is expanded by changing the brightness, the chroma, the contrast, the sharpness and the like of the original image based on basic colors, and the original labeling result can be still kept due to the scale change of the related image, so that the adaptability of the network model to the environment is improved;
step two and step two: image normalization processing: in order to accelerate the network training, the pixel points of the acquired RGB image range from 0 to 255, and the pixel is converted into the pixel points from 0 to 1 by using the formula 2 to carry out normalization operation:
step two and three: image normalization processing: the main principle of the method is that the data are subjected to mean value removal to realize centralized processing, the mean value and the variance of pixel points are calculated through (formula 3) and (formula 4), and then the data are subjected to standardization processing through (formula 5);
the specific process of the second step three is as follows:
step two, step three and step one: extracting the neural network characteristic information of the processed picture through a formula 6;
step two, step three, step two: the image size after convolution operation changes, mainly the number of channels increases and the length and width change, as shown in (formula 7), the information of the spatial layer becomes more, and the extracted features are more beneficial to network training;
step two, step three: standardized network layer: the neural network training process is a process of continuously updating parameter fitting training data back and forth, and is used for standardizing a characteristic diagram and performance in order to improve the network optimization speed, wherein the standardization process is shown as (formula 8) (formula 10) and (formula 11), so that the convergence capacity of the network is enhanced;
wherein: m is the number of images selected by one iteration;
μ B is the mean of one batch;
wherein:
is the variance;
wherein:
is a normalized value;
epsilon is a minimum number for preventing denominator from being 0;
wherein: the input of the normalization layer isB＝{x 1，2…，m }；
y i Is a normalized output;
γ and β are two parameters learned in training.
The fertilizer applicator comprises a frame 2, the frame 2 is fixedly connected with a fertilizer box 1, a discharge port of the fertilizer box 1 is fixedly connected with an outer grooved wheel fertilizer distributor 5, a driving motor reducer 6 is fixed on the frame 2 through a transmission main shaft 11 and a pair of bearing seats, the outer grooved wheel fertilizer distributor 5 is installed and fixed below the fertilizer box 1, the frame 2 is fixedly connected with an electric push rod 8, the push rod end of the electric push rod 8 is fixedly connected with one end of a compression spring 10, the other end of the compression spring 10 is fixedly connected with a main shaft 11, the frame 2 is connected with the transmission main shaft 11 through a bearing, the transmission main shaft 11 is fixedly connected with a driving motor 7, the driving motor 7 is fixedly connected with a reversing driver 12, the output shaft of the driving motor 7 is fixedly connected with an input shaft of the reversing driver 12, the output shaft of the reversing driver 12 is fixedly connected with a driving transmission chain wheel, and the reversing driver 12 is fixedly connected with a ditching and fertilizing component 4.
The ditching fertilization part 4 is including connecing hopper 13 and ditching main shaft 17, connect the hopper 13 with ditching main shaft 17 connects through the cover shaft coupling 16 that expands, connect hopper 13 fixed connection cover 21 that expands, cover 21 fixed connection drive sprocket 14 expands, connect hopper 13 fixed connection deep groove ball bearing 15, deep groove ball bearing 15's outer lane fixed connection linkage 22, linkage 22 fixed connection the shell of switching-over actuator 12, be provided with in the ditching main shaft 17 and prevent blockking up spiral 23, the outer fixed connection segmentation helical blade 18 of ditching main shaft 17, ditching main shaft 17 is provided with out fertile mouth 19, ditching main shaft 17 fixed connection buries native blade 20.
One end of the drive chain 9 encircles the drive sprocket and one end of the drive chain 9 encircles the drive sprocket 14.
The concrete flow of the step five is as follows: in an initial state, the electric push rod 8 is contracted to enable the ditching and fertilizing component 4 to incline so as to facilitate transportation of the fertilizing device, during fertilizing, the electric push rod 8 is controlled to extend out to enable the transmission main shaft 11 to rotate to enable the ditching and fertilizing component 4 to enter soil in an inclined mode, the ditching and fertilizing component 4 is prevented from being blocked in a soil entering stage, the driving motor reducer 6 and the driving motor 7 are controlled to be opened, the driving motor reducer 6 drives the outer grooved wheel fertilizer discharger 5 to rotate, fertilizer in the fertilizer box 1 is discharged into the receiving hopper 13, the driving motor 7 drives the reversing driver 12 to move, the reversing driver 12 drives the transmission chain wheel 14 to rotate through the transmission chain 9, the transmission chain wheel 14 drives the expansion sleeve 21 to rotate, the expansion sleeve 21 drives the receiving hopper 13 to rotate, the receiving hopper 13 drives the expansion sleeve coupler 16, the ditching main shaft 17, the segmented spiral blades 18, the soil entering blade 20 and the spiral screw 23 to rotate, the segmented spiral blades 18 realize rotation, fertilizer enters the ditching and fertilizer discharging from the fertilizer outlet port 19 along the receiving hopper 13, the segmented spiral blades 18 and the spiral blades are rotated, the segmented spiral blades 18 can effectively improve the efficiency of fertilizer and the fertilizer mixing and the fertilizer particles can be effectively utilized.
The receiving hopper 13 is matched with the outer sheave fertilizer apparatus 5.
The frame 2 is fixedly connected with a three-point suspension device 3, and the three-point suspension device 3 is used for being connected with the moving mechanism.
The working process of the invention is as follows: the rotary speed encoder support is arranged on a shell of an outer grooved wheel fertilizer distributor 5, a fertilizer discharging shaft of the outer grooved wheel fertilizer distributor 5 is provided with a gear ring, a mounting support is arranged on the shell of the outer grooved wheel fertilizer distributor 5, the rotary speed encoder support is fixedly connected with a rotary speed encoder shell, a rotary speed encoder input shaft bearing is connected with a support, a gear is arranged at the end part of a rotary speed encoder input shaft, and the gear is meshed with the gear ring.
And a camera is arranged on the moving mechanism or the fertilizer box, so that the camera is opposite to the crown of the fruit tree.
The rotating speed encoder, the camera, the electric push rod 8, the outer grooved wheel fertilizer discharger 5 and the driving motor 7 are respectively and electrically connected with the controller. The measurement and the variable fertilization of the volume of the crown of the fruit tree are realized.
The rotating speed encoder calculates the current rotating speed of the outer grooved wheel fertilizer distributor 5, converts the output signal into a pulse signal, and inputs the pulse signal to the controller so as to obtain the current fertilizing speed, facilitate display and real-time management, and participate in one of fertilizing amount regulating factors.
The fertilizing device is installed on the moving mechanism, and the moving mechanism is operated to enter an orchard in a hilly and mountainous area;
and (4) measuring and calculating the volume of the crown of the fruit tree.
And determining the fertilizing amount according to the volume of the crown.
And controlling the moving mechanism to drive the fertilizing device to move to a proper position.
And controlling the fertilizer applying device to output fertilizer for applying fertilizer.
In an initial state, the electric push rod 8 contracts to enable the ditching and fertilizing component 4 to incline and facilitate transportation of the fertilizing device, during fertilizing, the electric push rod 8 is controlled to stretch out, the transmission main shaft 11 is made to rotate, the ditching and fertilizing component 4 is made to obliquely enter soil, the ditching and fertilizing component 4 is prevented from being blocked in the soil entering stage, the outer grooved wheel fertilizer discharger 5 and the driving motor 7 are controlled to be opened, the outer grooved wheel fertilizer discharger 5 works to discharge fertilizer in the fertilizer box 1 into the receiving hopper 13, the driving motor 7 drives the reversing driver 12 to move, the reversing driver 12 drives the transmission chain wheel 14 to rotate through the transmission chain 9, the transmission chain wheel 14 drives the expansion sleeve 21 to rotate, the expansion sleeve 21 drives the receiving hopper 13 to rotate, the receiving hopper 13 drives the expansion sleeve coupler 16, the ditching main shaft 17, the segmented spiral blades 18, the soil entering blade 20 and the anti-blocking spiral 23 to rotate, the segmented spiral blades 18 achieve rotary slotting, the fertilizer enters the main shaft 17 along the receiving hopper 13 and rotates and is sprayed out from the fertilizer outlet 19, fertilizer particles and the soil particles broken at the bottom layer are lifted by the segmented spiral blades 18 and are stirred and mixed, and the fertilizer can be effectively mixed with the fertilizer.
The anti-blocking spiral 23 of the device can rotate along with the rotating main shaft, but the spiral rotating direction is opposite to the rotating direction, so that fertilizer particles and soil entering the hollow shaft can be smoothly discharged. The fertilizer granules discharged from the fertilizer outlet 19 and the soil granules crushed at the bottom layer are lifted by the segmented helical blades 18 and are stirred and mixed by the segmented helical blades 18, so that the effective mixing of the soil and the fertilizer can be realized, the utilization efficiency of the fertilizer is improved, and the soil crushed in the operation advancing direction can be thrown into a groove formed at the rear part to finish the earthing operation. The operations of annular ditching, fertilizing and earthing are finished at one time.
When this device normally fertilizies, compression spring 10 is in the state of extending, can be compressed when ditching fertilization revolves the part and runs into thicker root system or stone, crosses the barrier, ensures that the device can not die or overload and damage.
The control system of the method realizes the variable fertilization operation of the solid particles according to the relation among the related variables such as the rotating speed of the fertilizer discharging wheel, the fertilizing vehicle speed, the volume of the fruit trees and the like and the control rule algorithm, and a certain amount of solid particle fertilizer can be uniformly spread in the target ditch.
The above disclosure is only for the specific embodiment of the present invention, but the present invention is not limited thereto, and any variations that can be made by those skilled in the art should fall within the scope of the present invention.
Claims (10)
1. An agricultural automatic accurate control variable rate fertilization method is characterized by comprising the following steps:
the method comprises the following steps: the fertilizing device is installed on the moving mechanism, and the moving mechanism is operated to enter an orchard in a hilly and mountainous area;
step two: measuring and calculating the volume of the crown of the fruit tree;
step three: determining the fertilizing amount according to the volume of the crown;
step four: controlling the moving mechanism to drive the fertilizing device to move to a proper position;
step five: and controlling the fertilizer applying device to output fertilizer for applying fertilizer.
2. The agricultural automatic and accurate control variable rate fertilization method according to claim 1, wherein: the specific process of the second step is as follows:
step two, firstly: collecting a data set, shooting and collecting pictures of the fruit trees by using a camera, marking the pictures as xml files in a VOC (volatile organic compound) format by using marking software LabelImg, and marking the fruit trees by using a rectangular frame, wherein the frame displays the information of the names and the positions marked by the pictures;
step two: processing images, namely, before inputting images into a convolutional network to extract different semantic information, carrying out a series of processing such as image normalization processing, image standardization processing and the like on the acquired images;
step two and step three: extracting network characteristics, wherein in the characteristic extraction process, low-level characteristics such as textures, frames, colors and the like are generally extracted by a shallow network, high-level and abstract characteristics are extracted by a deep network, and the main purpose is to fit a picture target with a marked real area so as to achieve the minimum error rate;
step two, four: after the network training is finished, a better weight value can be obtained between each layer, the optimal prediction effect can be achieved, and the weight is reserved for detecting fruit trees subsequently;
step two and step five: fruit tree detection, namely processing pictures acquired under an actual environment and transmitting the processed pictures into a network on the basis of the previously acquired network weight, wherein the weight parameters among all layers reach a better level in the training process, and the input images only need to be transmitted forward under the parameters to finally acquire the actual target class probability;
step two, step six: the whole graph is directly put into the input end of the Yolov5 network, and then the position of a bounding box (bounding box) and the category to which the bounding box belongs are returned in the output layer.
3. The agricultural automatic and accurate control variable rate fertilization method according to claim 2, characterized in that: the second step comprises the following specific processes:
step two, step one: the data set is amplified, the convolutional neural network has the capacity of processing brand new data, but when the samples are few, the model is easy to over-fit, the weight distribution is not reasonable enough, the network cannot have enough generalization capacity and feature summarization capacity, the original data set is expanded by changing the brightness, the chroma, the contrast, the sharpness and the like of the original image based on basic colors, and the original labeling result can be still kept due to the scale change of the related image, so that the adaptability of the network model to the environment is improved;
step two and step two: image normalization processing: in order to accelerate the network training, the pixel points of the acquired RGB image range from 0 to 255, and the pixel is converted into the pixel points from 0 to 1 by using the formula 2 to carry out normalization operation:
step two, step three: image normalization processing: the main principle of the method is that the data are subjected to mean value removal to realize centralized processing, the mean value and the variance of pixel points are solved through (formula 3) and (formula 4), and then the standardization processing is carried out through (formula 5);
4. the agricultural automatic precision control variable rate fertilization method of claim 2, wherein: the specific process of the second step and the third step is as follows:
step two, step three and step one: extracting neural network characteristic information from the processed picture through a formula 6;
step two, step three, step two: the image size after convolution operation changes, mainly the number of channels increases and the length and width change, as shown in (formula 7), the information of the spatial layer becomes more, and the extracted features are more beneficial to network training;
step two and step three: standardized network layer: the neural network training process is a process of continuously updating parameter fitting training data back and forth, and is used for standardizing a characteristic diagram and performance in order to improve the network optimization speed, wherein the standardization process is shown as (formula 8) (formula 10) and (formula 11), so that the convergence capacity of the network is enhanced;
wherein: m is the number of images selected by one iteration;
μ B is the mean of one batch;
wherein:
is the variance;
wherein:
is a normalized value;
epsilon is a minimum number for preventing denominator from being 0;
wherein: the input of the normalization layer is B = { x = 1，2…，m }；
y i Is the normalized output;
γ and β are two parameters learned in training.
5. The agricultural automatic and accurate control variable rate fertilization method according to claim 1, wherein: fertilizer injection unit includes frame (2), frame (2) fixed connection fertilizer can (1), the outer sheave fertilizer apparatus of discharge gate fixed connection (5) of fertilizer can (1), driving motor reduction gear (6) are fixed in on frame (2) through transmission main shaft (11) and a pair of bearing frame, outer sheave fertilizer apparatus (5) installation is fixed in under fertilizer can (1), frame (2) fixed connection electric putter (8), the one end of the push rod fixed connection compression spring (10) of electric putter (8), the other end fixed connection of compression spring (10) main shaft (11), transmission main shaft (11) is connected to frame (2) bearing, transmission main shaft (11) fixed connection driving motor (7), driving motor (7) fixed connection switching-over actuator (12), the output shaft fixed connection of driving motor (7) the input shaft of switching-over actuator (12), the output shaft fixed connection driving sprocket of switching-over actuator (12), switching-over actuator (12) fixed connection ditching fertilization part (4).
6. The agricultural automatic precision control variable rate fertilization method of claim 5, wherein: the ditching fertilization part (4) is including connecing hopper (13) and ditching main shaft (17), connect hopper (13) with ditching main shaft (17) are connected through expanding set shaft coupling (16), connect hopper (13) fixed connection cover (21) that expand, cover (21) fixed connection driving sprocket (14) expand, connect hopper (13) fixed connection deep groove ball bearing (15), outer lane fixed connection linkage (22) of deep groove ball bearing (15), linkage (22) fixed connection the shell of reversing actuator (12), be provided with in ditching main shaft (17) and prevent blockking up spiral (23), ditching main shaft (17) outer fixed connection segmentation helical blade (18), ditching main shaft (17) are provided with fertilizer outlet (19), ditching main shaft (17) fixed connection income native blade (20).
7. The agricultural automatic and accurate control variable rate fertilization method according to claim 6, wherein: one end of the transmission chain (9) surrounds the driving transmission chain wheel, and one end of the transmission chain (9) surrounds the transmission chain wheel (14).
8. The agricultural automatic and accurate control variable rate fertilization method according to claim 7, wherein: the concrete flow of the step five is as follows: in an initial state, the electric push rod (8) is contracted to enable the ditching and fertilizing component (4) to incline so as to facilitate transportation of the fertilizing device, during fertilizing, the electric push rod (8) is controlled to extend out to enable the transmission main shaft (11) to rotate so as to enable the ditching and fertilizing component (4) to incline to enter the soil, the ditching and fertilizing component (4) is prevented from being blocked in the soil entering stage, the driving motor reducer (6) and the driving motor (7) are controlled to be opened, the driving motor reducer (6) drives the outer grooved wheel fertilizer discharger (5) to rotate, so that fertilizer in the fertilizer box (1) is discharged into the receiving hopper (13), the driving motor (7) drives the reversing driver (12) to move, the reversing driver (12) drives the transmission chain wheel (14) to rotate through the transmission chain (9), the transmission chain wheel (14) drives the expansion sleeve (21) to rotate, the expansion sleeve (21) drives the receiving hopper (13) to rotate, the receiving hopper (13) drives the expansion sleeve (17), the main shaft (17), the ditching and the spiral blade (18) to rotate in a subsection manner, and the ditching and the fertilizer outlet blade (13) enters the spiral blade (18) and the anti-blocking spiral blade (17) to rotate along the soil outlet blade (13), fertilizer particles and the broken soil particles in bottom are promoted by the sectional helical blade (18), and are stirred and mixed by the sectional helical blade (18), so that the effective mixing of soil and fertilizer can be realized, and the utilization efficiency of the fertilizer is improved.
9. The agricultural automatic and accurate control variable rate fertilization method according to claim 6, wherein: the receiving hopper (13) is matched with the outer grooved wheel fertilizer distributor (5).
10. The agricultural automatic precision control variable rate fertilization method of claim 5, wherein: the rack (2) is fixedly connected with a three-point suspension device (3), and the three-point suspension device (3) is used for being connected with the moving mechanism.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211362912.4A CN115643874A (en) | 2022-11-02 | 2022-11-02 | Agricultural automatic accurate control variable rate fertilization method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211362912.4A CN115643874A (en) | 2022-11-02 | 2022-11-02 | Agricultural automatic accurate control variable rate fertilization method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115643874A true CN115643874A (en) | 2023-01-31 |
Family
ID=84994596
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211362912.4A Pending CN115643874A (en) | 2022-11-02 | 2022-11-02 | Agricultural automatic accurate control variable rate fertilization method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115643874A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116897668A (en) * | 2023-07-17 | 2023-10-20 | 辽宁省农业机械化研究所 | Electric-drive crop sowing and fertilizing control method and system |
| CN117730655A (en) * | 2024-02-20 | 2024-03-22 | 云南省农业科学院质量标准与检测技术研究所 | Quantitative analysis method, device, equipment and storage medium for vigor of rice seeds |
| CN117730655B (en) * | 2024-02-20 | 2024-05-14 | 云南省农业科学院质量标准与检测技术研究所 | Quantitative analysis method, device, equipment and storage medium for vigor of rice seeds |
-
2022
- 2022-11-02 CN CN202211362912.4A patent/CN115643874A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116897668A (en) * | 2023-07-17 | 2023-10-20 | 辽宁省农业机械化研究所 | Electric-drive crop sowing and fertilizing control method and system |
| CN116897668B (en) * | 2023-07-17 | 2024-01-23 | 辽宁省农业机械化研究所 | Electric-drive crop sowing and fertilizing control method and system |
| CN117730655A (en) * | 2024-02-20 | 2024-03-22 | 云南省农业科学院质量标准与检测技术研究所 | Quantitative analysis method, device, equipment and storage medium for vigor of rice seeds |
| CN117730655B (en) * | 2024-02-20 | 2024-05-14 | 云南省农业科学院质量标准与检测技术研究所 | Quantitative analysis method, device, equipment and storage medium for vigor of rice seeds |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| AU2018346351B2 (en) | Disease recognition from images having a large field of view | |
| CN111316896B (en) | Irrigation control system of cultivation technique that becomes more meticulous | |
| US20200051180A1 (en) | Crop disease recognition and yield estimation | |
| CN115643874A (en) | Agricultural automatic accurate control variable rate fertilization method | |
| CN113228047A (en) | Plant disease detection using multi-stage, multi-scale deep learning | |
| CN1726750A (en) | Virtual GPS accurate agricultural variable subsoil application system | |
| US20200329632A1 (en) | Spraying support system for working machine | |
| CN114916298B (en) | Variable broadcasting and fertilizing device and method for agricultural unmanned aerial vehicle | |
| CN112136423A (en) | Multifunctional agricultural robot | |
| CN107097958A (en) | The UAS and method of a kind of variable fertilization | |
| EP3932167A1 (en) | Implement mounted sensors sensing surface/furrow characteristics and control | |
| CN115362811A (en) | Mountain crop intelligent cultivation system based on digital twinning | |
| CN1404709A (en) | Precision variable fertilizing machine | |
| CN109892078A (en) | A kind of fertilizer film paving sowing multiple working integrated equipment | |
| CN115643872A (en) | Agricultural automatic accurate control variable fertilization control system | |
| CN109379930B (en) | Fertilizer spreading control system and method | |
| CN113228910B (en) | Rice field water-saving fertilizer-saving intelligent regulation and control device | |
| CN213784122U (en) | Multifunctional agricultural robot | |
| CN113056996A (en) | Agricultural fertilizer injection unit based on thing networking | |
| CN110679238A (en) | Orchard ditching and fertilizing device and method | |
| CN116548103A (en) | Cultivation and seeding and pesticide spraying integrated equipment and intelligent pesticide spraying method | |
| CN110210375A (en) | Adaptive farm crops fertilizer deficiency region detection unmanned plane and detection method | |
| CN115349338B (en) | Variable topdressing control system and method based on crop growth vigor | |
| CN115581135A (en) | Intelligent unmanned sowing and fertilizing device and method based on machine vision | |
| CN219146138U (en) | Fertilizer box for agricultural machinery |
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 |