Special crop fertilizer granulation method based on machine vision

Abstract

The invention belongs to the field of visual control, and discloses a special crop fertilizer granulation method based on machine vision, which is applied to a granulation system comprising a rotary drum granulator, a raw material supply device, a steam supply device, a control device and a shooting device, and the method comprises the following steps: S1, starting the rotary drum granulator by the control device; s2, the control device controls valves of the raw material supply device and the water vapor supply device to be opened; s3, shooting the dried fertilizer particles by a shooting device to obtain a to-be-identified image; s4, the control device recognizes the to-be-recognized image, the diameter of the fertilizer particles is obtained, and the valve opening degree of the steam supply device is calculated; s5, a valve of the steam supply device is adjusted according to the valve opening degree; s6, the control device judges whether granulation stop conditions are met or not, if yes, a valve of the raw material supply device and a valve of the steam supply device are closed, and the rotary drum granulator is closed; and if not, entering S3. According to the invention, the diameter difference of fertilizer particles is smaller.

Description

Technical Field

[0001] The invention relates to the field of vision control, in particular to a method for granulating crop-specific fertilizer based on machine vision. Background Art

[0002] Crop-specific fertilizer refers to fertilizer that is specially used for crops in a certain region by scientifically mixing nutrients such as nitrogen, phosphorus, potassium, and trace elements according to regional soil conditions and the fertilizer requirements of the crops.

[0003] Common crop fertilizer granulation processes include: drum granulation, disc granulation, spray granulation, high tower granulation, etc. Among them, drum granulation is also called roller granulation. Its main working mode is wet granulation. A certain amount of water or steam is used to make the basic fertilizer fully react chemically after being humidified in the cylinder. Under certain liquid phase conditions, the rotating motion of the cylinder generates extrusion force between the material particles to agglomerate into balls.

[0004] The advantages of the drum granulation production process are relatively small formula restrictions, high output, low investment, and short construction period. At present, in a single drum granulator, the steam flow and pressure are manually adjusted on site. However, the manual adjustment of steam is determined by observing the size of the granulation particles in the granulation barrel. Excessive steam will cause more agglomeration and larger particle size, while too little steam will cause the granulation particle size to be low. The steam flow directly affects the particle size and moisture of the granulation. However, visual inspection by the human eye is subjective and uncertain. Therefore, the particle size of the crop-specific fertilizer produced by the drum granulator varies greatly, which is not conducive to accurately controlling the amount of fertilizer in subsequent fertilization. Summary of the invention

[0005] The purpose of the present invention is to disclose a method for granulating crop-specific fertilizer based on machine vision to solve the technical problems raised in the background technology.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention provides a method for granulating crop-specific fertilizer based on machine vision, which is applied to a granulation system including a rotary drum granulator, a raw material supply device, a steam supply device, a control device, and a camera device, and is characterized in that it includes:

[0008] S1, the control device starts the drum granulator, so that the drum of the drum granulator rotates;

[0009] S2, the control device controls the valve of the raw material supply device to open and controls the valve of the water vapor supply device to open, so that the granulation raw materials and water vapor enter the drum of the rotary drum granulator;

[0010] S3, a photographing device photographs the dried fertilizer particles discharged from the outlet of the drum granulator to obtain an image to be identified;

[0011] S4, the control device identifies the image to be identified, obtains the diameter of the fertilizer particles, and calculates the valve opening of the steam supply device based on the diameter of the fertilizer particles;

[0012] S5, the control device adjusts the valve of the steam supply device according to the valve opening;

[0013] S6, the control device determines whether the granulation stop condition is reached. If so, the valve of the raw material supply device and the valve of the steam supply device are closed, and the rotary drum granulator is closed; if not, enter S3.

[0014] Preferably, starting the drum granulator comprises:

[0015] The relay of the rotary drum granulator is controlled so that the relay is connected to the power supply circuit of the rotary drum granulator.

[0016] Preferably, the raw material supply device includes a slurry supply pipeline, a sulfuric acid pipeline and an ammonia pipeline.

[0017] Preferably, it also includes a discharge conveyor belt;

[0018] The discharge conveyor belt is used to transfer the fertilizer particles discharged from the discharge port of the drum granulator to the drying device;

[0019] The drying device is used to dry the fertilizer granules.

[0020] Preferably, the photographing device comprises a sampling device, a vibrating device and a photographing device;

[0021] The sampling device is used to sample the dried fertilizer particles from the discharge port of the drying device and transfer the fertilizer particles to the vibration device;

[0022] The photographing device is used to photograph the fertilizer particles on the vibrating device to obtain an image to be identified.

[0023] Preferably, the vibration device comprises a tray body and a vibration assembly arranged below the tray body, and the vibration assembly is used to drive the tray body to vibrate, thereby driving the fertilizer particles on the tray body to vibrate to disperse the fertilizer particles.

[0024] Preferably, photographing the fertilizer particles on the vibrating device to obtain an image to be identified includes:

[0025] The fertilizer particles on the tray body are photographed to obtain an image to be identified.

[0026] Preferably, transferring the fertilizer granules to a vibrating device comprises:

[0027] Transfer the fertilizer granules to the tray body.

[0028] Preferably, identifying the image to be identified and obtaining the diameter of the fertilizer particles includes:

[0029] Grayscale the image to be recognized to obtain a grayscale image;

[0030] Denoising the grayscale image to obtain a denoised image;

[0031] Binarization is performed on the denoised image to obtain a binary image;

[0032] Performing connected domain detection on the binary image to obtain the connected domain in the binary image;

[0033] Calculate the diameter of each connected domain separately;

[0034] Remove the maximum and minimum values ​​of all diameters and calculate the average value of the remaining diameters;

[0035] The average value was taken as the diameter of the fertilizer granules.

[0036] Preferably, determining whether the granulation stop condition is reached includes:

[0037] It is judged whether the cumulative running time of the drum granulator reaches the preset time. If so, it means that the granulation stop condition is reached. If not, it means that the granulation stop condition is not reached.

[0038] Beneficial effects:

[0039] Compared with the traditional method of judging the size of fertilizer particles based on the human eye, the present invention uses machine vision technology to calculate the diameter of dried fertilizer particles during the granulation process of crop-specific fertilizers, and then controls the input amount of water vapor based on the diameter, thereby making the error between the diameter of the fertilizer particles and the target diameter smaller, and the difference in the diameter of the manufactured fertilizer particles is smaller. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0041] Figure 1 It is a schematic diagram of the crop-specific fertilizer granulation method based on machine vision of the present invention.

[0042] Figure 2 The process of identifying the image to be identified and obtaining the diameter of the fertilizer particles. DETAILED DESCRIPTION

[0043] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0044] like Figure 1 The present invention provides a method for granulating crop-specific fertilizer based on machine vision, which is applied to a granulation system including a rotary drum granulator, a raw material supply device, a steam supply device, a control device, and a camera device, and is characterized in that it includes:

[0045] S1, the control device starts the drum granulator, so that the drum of the drum granulator rotates;

[0046] S2, the control device controls the valve of the raw material supply device to open and controls the valve of the water vapor supply device to open, so that the granulation raw materials and water vapor enter the drum of the rotary drum granulator;

[0047] S3, a photographing device photographs the dried fertilizer particles discharged from the outlet of the drum granulator to obtain an image to be identified;

[0048] S4, the control device identifies the image to be identified, obtains the diameter of the fertilizer particles, and calculates the valve opening of the steam supply device based on the diameter of the fertilizer particles;

[0049] S5, the control device adjusts the valve of the steam supply device according to the valve opening;

[0050] S6, the control device determines whether the granulation stop condition is reached. If so, the valve of the raw material supply device and the valve of the steam supply device are closed, and the rotary drum granulator is closed; if not, enter S3.

[0051] In the above scheme, during the granulation process of crop-specific fertilizer, machine vision technology is used to calculate the diameter of the dried fertilizer particles, and then the input amount of water vapor is controlled based on the diameter, so that the error between the diameter of the fertilizer particles and the target diameter (i.e., the minimum diameter that meets the factory standards) is smaller, and the difference in the diameter of the manufactured fertilizer particles is smaller.

[0052] Preferably, starting the drum granulator comprises:

[0053] The relay of the rotary drum granulator is controlled so that the relay is connected to the power supply circuit of the rotary drum granulator.

[0054] Specifically, by turning on the switch of the relay, the power supply circuit of the rotary drum granulator is turned on.

[0055] Preferably, the raw material supply device includes a slurry supply pipeline, a sulfuric acid pipeline and an ammonia pipeline.

[0056] Specifically, the slurry includes ammonium sulfate and urea. The slurry generation process can refer to the content disclosed in claim 1 of the patent publication number CN102503622A.

[0057] Specifically, the water vapor supply device may include a pipeline for transporting water vapor and a valve disposed on the pipeline for controlling the amount of steam provided to the drum per unit time.

[0058] Preferably, it also includes a discharge conveyor belt;

[0059] The discharge conveyor belt is used to transfer the fertilizer particles discharged from the discharge port of the drum granulator to the drying device;

[0060] The drying device is used to dry the fertilizer granules.

[0061] The drying process enables the fertilizer particles to be effectively separated by a vibration device. If not dried, the fertilizer particles may break during the vibration process, which makes it impossible to effectively measure the diameter of the fertilizer particles.

[0062] Preferably, the photographing device comprises a sampling device, a vibrating device and a photographing device;

[0063] The sampling device is used to sample the dried fertilizer particles from the discharge port of the drying device and transfer the fertilizer particles to the vibration device;

[0064] The photographing device is used to photograph the fertilizer particles on the vibrating device to obtain an image to be identified.

[0065] Specifically, the vibration device is mainly used to disperse the overlapping fertilizer particles by vibration, thereby reducing the probability that multiple fertilizer particles form a connected domain in the image to be identified, thereby improving the accuracy of the diameter of the fertilizer particles obtained based on the image to be identified.

[0066] The sampling device can be arranged above the conveyor belt below the discharge port of the drying device, so as to sample the fertilizer particles on the conveyor belt.

[0067] Specifically, the sampling device takes a sample once every fixed time interval, and the fixed time interval may be 1 minute.

[0068] Preferably, the vibration device comprises a tray body and a vibration assembly arranged below the tray body, and the vibration assembly is used to drive the tray body to vibrate, thereby driving the fertilizer particles on the tray body to vibrate to disperse the fertilizer particles.

[0069] The structure of the vibration device can refer to the structure disclosed in patent CN217577291U, and the tray body of the present invention is equivalent to the vibration tray in CN217577291U. The vibration assembly is equivalent to the other parts of CN217577291U except the vibration tray.

[0070] The sampling device may be a mechanical arm with a porous suction cup. The fertilizer particles may be sucked by the suction cup through the adsorption capacity of the suction cup and moved by the mechanical arm into the tray body.

[0071] Preferably, photographing the fertilizer particles on the vibrating device to obtain an image to be identified includes:

[0072] The fertilizer particles on the tray body are photographed to obtain an image to be identified.

[0073] After the vibration component vibrates the fertilizer particles on the tray body, the fertilizer particles on the tray body can be photographed to obtain an image to be identified.

[0074] Preferably, transferring the fertilizer granules to a vibrating device comprises:

[0075] Transfer the fertilizer granules to the tray body.

[0076] When the sampling device includes a suction cup, after the suction cup reaches above the tray body, the suction cup stops providing suction, thereby causing the fertilizer particles to fall into the tray body.

[0077] Preferably, if Figure 2 , identify the image to be identified and obtain the diameter of the fertilizer particles, including:

[0078] Grayscale the image to be recognized to obtain a grayscale image;

[0079] Denoising the grayscale image to obtain a denoised image;

[0080] Binarization is performed on the denoised image to obtain a binary image;

[0081] Performing connected domain detection on the binary image to obtain the connected domain in the binary image;

[0082] Calculate the diameter of each connected domain separately;

[0083] Remove the maximum and minimum values ​​of all diameters and calculate the average value of the remaining diameters;

[0084] The average value was taken as the diameter of the fertilizer granules.

[0085] By excluding the maximum and minimum values ​​of the diameter, the influence of the data that is too large or too small on the final calculated diameter can be reduced, and a more accurate diameter can be obtained.

[0086] The diameter of the connected domain can be obtained by the minimum circumscribed circle algorithm.

[0087] Preferably, performing noise reduction on the grayscale image to obtain the noise-reduced image includes:

[0088] Use the median filter algorithm to reduce the noise of the grayscale image and obtain a median filtered image;

[0089] Calculate the cluster value of each pixel in the median filtered image;

[0090] Use the clustering algorithm to cluster all cluster values ​​and divide all cluster values ​​into two clusters;

[0091] Calculate the average of the clustering values ​​for each cluster separately;

[0092] The pixel points corresponding to the cluster value in the cluster with the larger average value among the two clusters are stored in the set PU;

[0093] In the grayscale image, an adaptive noise reduction algorithm is used to reduce the noise of the pixel points in the PU, and the grayscale value of the pixel points in the PU after the noise reduction by the adaptive noise reduction algorithm is obtained;

[0094] The grayscale values ​​of the pixels in the median filtered image are optimized based on the grayscale values ​​after denoising to obtain a denoised image.

[0095] The present invention improves the existing denoising method, so that the present invention can not only retain more edge information after denoising, but also effectively improve the efficiency of denoising. Specifically, unlike the prior art, the present invention does not directly use two methods to perform denoising in succession, that is, first use a denoising algorithm to perform denoising, and then use another denoising algorithm to perform denoising on the image obtained by the previous denoising algorithm. Since the other denoising algorithm is performed on the basis of the denoised image, it is easy to cause excessive denoising, resulting in excessive loss of edge information, making it impossible to effectively calculate the diameter of the connected domain of the denoised image, which is obviously not in line with the application scenario of the present invention.

[0096] Therefore, after using the median filter algorithm for noise reduction, the present invention calculates the cluster values, then clusters the cluster values, and then uses the adaptive noise reduction algorithm to reduce noise only for the pixels corresponding to the cluster values ​​in the clusters with larger average values, and finally uses the results obtained by the adaptive noise reduction algorithm on the median filtered image to optimize the pixels in the PU. This processing method can not only reduce the number of pixels that need to be denoised using the adaptive noise reduction algorithm, but also because the adaptive noise reduction is performed on the original un-noised image, the grayscale value obtained after the adaptive noise reduction algorithm contains more edge information, so that in the final noise reduction image, not only the noise is effectively suppressed, but also the edge information is more retained. Compared with the existing noise reduction methods, the speed of this noise reduction method is effectively improved.

[0097] Preferably, the calculation formula of the clustering value is:

[0098]

[0099] sumval p Represents the clustering value of pixel p, Gy p,1 Represents the gray value of p in the gray image, Gy p,2 Represents the gray value of p after noise reduction by the median filter algorithm.

[0100] In the present invention, the clustering value is calculated by the grayscale change amplitude of the pixel before and after the median filter algorithm performs noise reduction. The larger the grayscale change amplitude, the larger the corresponding clustering value, indicating that the degree of noise pollution of such pixel is higher. Therefore, the present invention subsequently obtains a set of pixel points with a larger clustering value through a clustering algorithm, so that only such pixel points can be processed by another noise reduction algorithm, reducing the number of pixel points involved in the noise reduction process of another noise reduction algorithm.

[0101] Preferably, a clustering algorithm is used to cluster all cluster values, and all cluster values ​​are divided into two clusters, including:

[0102] Use K-means clustering algorithm to cluster all cluster values ​​and divide all cluster values ​​into two clusters.

[0103] K-means clustering is a center-based unsupervised learning clustering algorithm that iteratively groups samples into k clusters so that the sum of the distances between each sample and the center or mean of its class is minimized. The basic principle is: first, randomly select k samples from the original data set as the initial cluster centers, then analyze the remaining samples one by one, calculate the distance between the sample and the k cluster centers, and assign the sample to the closest cluster; then recalculate the center of each cluster (usually the mean of all samples in the cluster), and repeat this process until the stopping condition is met (for example, the cluster center no longer changes or the preset number of iterations is reached).

[0104] The present invention does not directly stipulate a fixed clustering standard to classify the pixels in the median filter image into two types. Because the fixed clustering standard, such as setting a clustering value as a threshold for classification, cannot be well applied to images with different grayscale distributions, and easily leads to an excessive number of pixels participating in another noise reduction algorithm. Therefore, the present invention adaptively classifies the pixels in the median filter image through a clustering algorithm, so that the present invention can be applied to more images with different grayscale distributions. Thereby effectively controlling the number of pixels participating in the calculation process of another noise reduction algorithm.

[0105] Preferably, the calculation formula for the average value of the cluster value of the cluster is:

[0106]

[0107] clust represents the clusters composed of cluster values, sumval ave represents the average value of the cluster, Nclust represents the total number of cluster values ​​in the cluster, and i represents the cluster value in the cluster.

[0108] Specifically, a cluster contains multiple cluster values. By calculating the average value of these cluster values, it can be reflected how much the pixel points corresponding to the cluster values ​​in the cluster are affected by noise. The greater the degree of influence, the greater the average value of the cluster value.

[0109] Preferably, in the grayscale image, an adaptive noise reduction algorithm is used to reduce the noise of the pixels in the PU, and the grayscale value of the pixels in the PU after the noise reduction by the adaptive noise reduction algorithm is obtained, including:

[0110] Use the following formula to calculate the grayscale value of the pixel in the PU after the adaptive noise reduction algorithm is used:

[0111]

[0112] Gyadpt s Represents the gray value of the pixel in PU after denoising by the adaptive denoising algorithm, Gy s and Gy jThey represent the grayscale values ​​of pixel s and pixel j in the grayscale image, stoj represents the total number of pixels on the line segment between pixel s and pixel j, and sumval s and sumval j They respectively represent the clustering values ​​of the pixels with the same coordinates as the pixels s and j in the median filtered image, Ga represents the grayscale control parameter, Gb represents the quantity control parameter, Gc represents the clustering control parameter, and us represents the set of pixels with points whose distance from the pixel s is less than 5 in the grayscale image.

[0113] The denoising process of the present invention is obtained by comprehensively calculating the difference in grayscale values ​​between two pixel points s and j, the total number of pixel points on the line segment between the two pixel points s and j, and the clustering value of j. Therefore, if the absolute value of the difference in grayscale values ​​between the two pixel points s and j is larger, the total number of pixel points on the line segment between the two pixel points s and j is larger, and the clustering value of j is larger, the influence of j on the denoising result of s is larger, so that the distribution structure of the pixel values ​​of the image can be maintained during the denoising process. Moreover, when s is near the edge, the edge information of s can be retained due to the large difference in grayscale values ​​between the non-edge pixel points and s. In addition, the present invention also introduces a parameter of clustering value. Since the clustering value is calculated on the basis of median denoising, it includes the degree of influence of j on the noise, so that the pixel points with a greater degree of influence on the denoising result can have a smaller influence, thereby effectively suppressing the noise while retaining more edge information. This is because if we only consider the difference in grayscale values ​​between two pixels s and j, then when the absolute value of this difference is small, but in fact the clustering values ​​of the corresponding pixels of s and j in the median filtered image are quite different, the noise in us will easily have a greater impact on the denoising result of s.

[0114] Preferably, the line segment between the pixel point s and the pixel point j is a line segment with the shortest length between the pixel point s and the pixel point j.

[0115] When a line segment passes through a pixel, it means that the pixel is located on the line segment.

[0116] Preferably, Gym represents the maximum value of the grayscale value in the grayscale image;

[0117] widnum represents the number of columns of pixels in the grayscale image;

[0118] sumvalm represents the maximum value of the clustering value of the pixel points of the median filter image; λ represents a preset positive integer.

[0119] Specifically, the value range of λ is [30,50]. Preferably, the value of λ may be 35.

[0120] Preferably, the grayscale values ​​of pixels in the median filtered image are optimized based on the grayscale values ​​after noise reduction to obtain the noise reduced image, including:

[0121] Each pixel in the PU is processed as follows:

[0122] Let pzu represent the pixel in PU. After optimization, the gray value of pzu is:

[0123]

[0124] Gm pzu,2 Indicates the gray value of pzu after optimization, Gm pzu,1 Indicates the gray value of pzu before optimization, Gyadpt pzu Represents the grayscale value of pzu after denoising by the adaptive denoising algorithm.

[0125] Through the above formula, the filtering result of another median filtering algorithm can be introduced while retaining part of the results of the median filtering, so that the final denoised image contains more edge information.

[0126] Preferably, calculating the valve opening of the steam supply device based on the diameter of the fertilizer particles comprises:

[0127] Use the following formula to calculate the valve opening of the steam supply device:

[0128]

[0129] R1 represents the valve opening of the steam supply device, D1 represents the diameter of the fertilizer particles, Ds represents the preset diameter, and Rs represents the preset valve opening.

[0130] The calculation formula of the valve opening of the present invention can reduce the valve opening accordingly when the diameter of the fertilizer particles is too large, so that the diameter of the fertilizer particles produced subsequently can be increased; and when the diameter is too small, the valve opening is increased accordingly, so that the diameter of the fertilizer particles produced subsequently can be reduced, thereby improving the accuracy of the diameter of the produced fertilizer particles.

[0131] Specifically, the preset diameter is the maximum value of the diameter of the fertilizer particles within the diameter range that meets the factory standards.

[0132] Preferably, the preset diameter is 5 mm.

[0133] Preferably, the preset valve opening is 90%.

[0134] Preferably, adjusting the valve of the steam supply device according to the valve opening comprises:

[0135] Adjust the opening of the valve of the steam supply device to R1.

[0136] Preferably, determining whether the granulation stop condition is reached includes:

[0137] It is judged whether the cumulative running time of the drum granulator reaches the preset time. If so, it means that the granulation stop condition is reached. If not, it means that the granulation stop condition is not reached.

[0138] Specifically, the preset duration is a manually set duration, for example, the duration of a shift, to avoid continuing granulation during a shift change.

[0139] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for granulating crop-specific fertilizers based on machine vision, applied to a granulation system including a rotary drum granulator, a raw material supply device, a steam supply device, a control device, and a camera device, characterized in that: include: S1, the control device starts the drum granulator, so that the drum of the drum granulator rotates; S2, the control device controls the valve of the raw material supply device to open and controls the valve of the water vapor supply device to open, so that the granulation raw materials and water vapor enter the drum of the rotary drum granulator; S3, a photographing device photographs the dried fertilizer particles discharged from the outlet of the drum granulator to obtain an image to be identified; S4, the control device identifies the image to be identified, obtains the diameter of the fertilizer particles, and calculates the valve opening of the steam supply device based on the diameter of the fertilizer particles; S5, the control device adjusts the valve of the steam supply device according to the valve opening; S6, the control device determines whether the granulation stop condition is reached. If so, the valve of the raw material supply device and the valve of the steam supply device are closed, and the rotary drum granulator is closed; if not, enter S3.

2. The method for granulating crop-specific fertilizer based on machine vision according to claim 1, characterized in that: Start up the drum granulator, including: The relay of the rotary drum granulator is controlled so that the relay is connected to the power supply circuit of the rotary drum granulator.

3. The method for granulating crop-specific fertilizer based on machine vision according to claim 2, characterized in that: The raw material supply device includes a slurry supply pipeline, a sulfuric acid pipeline and an ammonia pipeline.

4. The method for granulating crop-specific fertilizer based on machine vision according to claim 1, characterized in that: Also included is an outfeed conveyor belt; The discharge conveyor belt is used to transfer the fertilizer particles discharged from the discharge port of the drum granulator to the drying device; The drying device is used to dry the fertilizer granules.

5. The method for granulating crop-specific fertilizer based on machine vision according to claim 4, characterized in that: The photographing device includes a sampling device, a vibration device and a photographing device; The sampling device is used to sample the dried fertilizer particles from the discharge port of the drying device and transfer the fertilizer particles to the vibration device; The photographing device is used to photograph the fertilizer particles on the vibrating device to obtain an image to be identified.

6. The method for granulating crop-specific fertilizer based on machine vision according to claim 5, characterized in that: The vibration device comprises a tray body and a vibration component arranged below the tray body, and the vibration component is used to drive the tray body to vibrate, thereby driving the fertilizer particles on the tray body to vibrate to disperse the fertilizer particles.

7. The method for granulating crop-specific fertilizer based on machine vision according to claim 6, characterized in that: The fertilizer particles on the vibrating device are photographed to obtain images to be identified, including: The fertilizer particles on the tray body are photographed to obtain an image to be identified.

8. The method for granulating crop-specific fertilizer based on machine vision according to claim 6, characterized in that: Transfer fertilizer granules to a vibrating device, including: Transfer the fertilizer granules to the tray body.

9. The method for granulating crop-specific fertilizer based on machine vision according to claim 1, characterized in that: Identify the image to be identified and obtain the diameter of the fertilizer particles, including: Grayscale the image to be recognized to obtain a grayscale image; Denoising the grayscale image to obtain a denoised image; Binarization is performed on the denoised image to obtain a binary image; Performing connected domain detection on the binary image to obtain the connected domain in the binary image; Calculate the diameter of each connected domain separately; Remove the maximum and minimum values ​​of all diameters and calculate the average value of the remaining diameters; The average value was taken as the diameter of the fertilizer granules.

10. The method for granulating crop-specific fertilizer based on machine vision according to claim 1, characterized in that: Determine whether the granulation stop conditions are met, including: It is judged whether the cumulative running time of the drum granulator reaches the preset time. If so, it means that the granulation stop condition is reached. If not, it means that the granulation stop condition is not reached.

Images