Granulated fertilizer flow sensing device and control system and control method thereof

Abstract

The invention discloses a granular fertilizer flow sensing device machine and a control system and a control method thereof. The granular fertilizer flow sensing device comprises a camera obscura shell, an industrial camera, an LED lamp, a black background plate and a core controller; the core controller adopts an improved YOLOv5s-seg model to carry out segmentation detection on a granular fertilizer image collected by the industrial camera, mask contour information is extracted, and the real-time fertilizer discharge flow is calculated according to the mask contour information. The control system comprises the sensing device, a fertilizer box, a Hall speed measurement module, an aperture type fertilizer feeding device and a microprocessor; and the microprocessor drives the aperture type fertilizer feeding device to adjust the opening degree of the fertilizer feeding port through a fuzzy PID controller algorithm based on the target fertilizer feeding flow and the real-time fertilizer feeding flow. Through high-precision visual sensing, advanced image processing and closed-loop control, precise monitoring and precise regulation and control of the flow of the granular fertilizer are realized, and the fertilization uniformity and the fertilizer utilization rate are improved. The precise fertilizing device is mainly used for precise fertilizing operation of the centrifugal fertilizer distributor.

Description

Technical Field

[0001] The invention belongs to the technical field of fertilization control, and in particular relates to a granular fertilizer flow sensor device and a control system and a control method thereof. Background Art

[0002] In the agricultural production process, scientific and reasonable fertilization not only helps to increase crop yields and improve soil nutrient structure, but also effectively reduces fertilizer resource waste and promotes the sustainable and stable development of agriculture. As an important path to achieve efficient use of agricultural resources and reduce production input losses, precision agriculture technology has received widespread attention in recent years. Among them, variable fertilization technology, as one of the core technologies in precision agriculture, can integrate variable control strategies into fertilization equipment. According to the soil nutrient status, crop growth status and pre-fertilization plan of different plots, the amount of fertilizer applied can be dynamically adjusted to accurately match the nutrient needs of crops, thereby significantly improving fertilizer utilization. Centrifugal fertilizer spreaders have been widely used in base fertilizer and topdressing operations in large fields due to their advantages such as wide spreading width, high operating efficiency, simple structure and uniform spreading.

[0003] At present, centrifugal fertilizer spreaders mainly control the fertilizer discharge flow rate by adjusting the opening of the discharge port of the fertilizer discharge device. In actual operation, the flow rate fluctuates greatly during the actual fertilizer discharge process. Current equipment generally lacks real-time and accurate fertilizer discharge flow detection devices and effective closed-loop control mechanisms, resulting in difficulty in accurately controlling the discharge amount. Therefore, in order to achieve real-time and accurate monitoring and automatic adjustment and control of the fertilizer discharge flow rate, it is urgently necessary to equip it with a high-precision fertilizer flow detection device and a fast-response control system. In response to the above technical problems, the present invention proposes a centrifugal fertilizer spreader fertilizer flow sensor, detection method and closed-loop fertilizer flow control system based on deep learning. Summary of the Invention

[0004] Purpose of the invention: The purpose of the present invention is to address the deficiencies in the prior art and provide a granular fertilizer flow sensing device, a centrifugal fertilizer spreader fertilizer flow control system and a method for using the system, so as to solve the problems of low fertilizer flow detection accuracy, inaccurate control, slow response speed and so on in the background technology, and to achieve accurate, real-time monitoring and closed-loop control of fertilizer flow.

[0005] Technical solution: To achieve the above-mentioned purpose, the present invention provides a granular fertilizer flow sensing device, comprising: a dark box shell for forming an at least partially closed image acquisition environment; an industrial camera, arranged in the dark box shell, for acquiring images of the granular fertilizer; an LED light, arranged in the dark box shell, for providing illumination for the granular fertilizer; a black background board, arranged in the dark box shell, serving as a high-contrast background when the granular fertilizer is photographed by the industrial camera; a core controller, data-connected to the industrial camera, configured to: receive the image captured by the industrial camera; use an image processing algorithm to segment and detect the granular fertilizer in the image, and extract the mask contour information of the particles; and calculate the fertilizer discharge flow in real time based on the mask contour information to obtain a real-time fertilizer discharge flow signal.

[0006] Furthermore, the image processing algorithm is an improved YOLOv5s-seg model, including: an input layer (Input) for inputting a granular fertilizer image; a backbone network (Backbone), adopting an HGNetV2 network structure, including a Stem layer for extracting initial features and at least one HG Stage layer, wherein the HG Stage layer internally includes a downsampling layer for reducing the spatial dimension of the feature map to extract multi-scale features of the granular fertilizer image; a neck network (Neck), adopting an AKConv structure, wherein the AKConv structure includes: an offset layer for adjusting the convolution sampling shape by learning an offset Offset, and a resampling layer Resample for resampling the feature map according to the adjusted sampling shape, and the resampled feature map is output after being re-convolved, normalized, and processed with a SiLU activation function; a detection head (Detect), connected to the neck network Neck, wherein the detection head includes a small particle detection branch for processing a high-resolution feature map with a spatial resolution of 160×160, and detection branches for processing 80×80 and 40×40 feature maps, respectively.

[0007] Furthermore, the core controller calculates the granular fertilizer flow rate by the following steps: approximating the granular fertilizer to a spherical model using a rounding fitting method, estimating the mass of a single granular fertilizer and calculating the total mass of the granules in each frame of the image;

[0008] The granular fertilizer flow rate per unit time is calculated according to the following formula:

[0009]

[0010] Where Q represents the fertilizer flow rate, N represents the total number of granular fertilizer image frames captured by the camera, i represents the i-th frame image used to calculate the fertilizer flow rate, M represents the number of granular fertilizer detected in each frame image, j represents the j-th granular fertilizer in each frame image; D represents the equivalent diameter of each granular fertilizer, p represents the fertilizer density, and t represents the total time for capturing granular fertilizer image frames.

[0011] Furthermore, the industrial camera is configured to capture images of the falling granular fertilizer in a timed interval manner, and the timed interval T of the industrial camera is determined by the following formula:

[0012]

[0013] Among them, H1 is the distance from the top of the camera field of view to the fertilizer drop port, H2 is the distance from the bottom of the camera field of view to the fertilizer drop port, and g is the acceleration of gravity.

[0014] Furthermore, the black background plate is fixed to the inner side of the dark box shell through a slot, and the black background plate is also used to divert the granular fertilizer falling from the fertilizer box drop port.

[0015] Another aspect of the present invention provides a centrifugal fertilizer spreader fertilizer flow control system composed of the above-mentioned granular fertilizer flow sensor device, comprising: a fertilizer box provided with a bottom discharge port for storing granular fertilizer; the granular fertilizer flow sensor device; a Hall effect speed measurement module for monitoring the travel speed of the fertilizer spreader; an aperture-type fertilizer discharge device installed at the bottom discharge port of the fertilizer box, having a blade group with adjustable opening and a microprocessor; the microprocessor is configured to: calculate a control output through a controller algorithm based on a preset target fertilizer flow rate and a real-time fertilizer discharge flow signal output by the core controller, and adjust the fertilizer discharge port opening of the blade group according to the control output to achieve closed-loop precise control of the fertilizer flow rate; I 2 C communication bus for implementing I 2 Bus communication between C parsing module, Hall speed measurement module, microprocessor and core controller; 2 The C parsing module is connected to the integrated human-computer interaction terminal and is used for system operation control, operation status monitoring and parameter setting.

[0016] Furthermore, the aperture-type fertilizer discharge device also includes a fixed seat, a fixed ring, a rotating disk, gears and a stepping motor, wherein the fixed seat is fixed to the bottom of the fertilizer box, the blade group forms a circular fertilizer discharge port with adjustable opening by stacking multiple blades, the rotating disk is sleeved on the outside of the blade group and connected to the stepping motor through the gear, the stepping motor is controlled by a microprocessor to drive the rotating disk to rotate, and the blade group rotates synchronously with the rotating disk to adjust the relative positions of multiple blades, thereby forming circular fertilizer discharge ports with different openings.

[0017] Furthermore, the system also includes a fertilizer spreading device, which includes a fertilizer spreading disc and fertilizer spreading blades, and is arranged below the fertilizer flow sensor device, for evenly spreading the granular fertilizer into the field through centrifugal force.

[0018] The present invention also provides a method for using the above-mentioned centrifugal fertilizer spreader fertilizer flow control system, comprising the following steps:

[0019] Step 1: Start the fertilizer flow sensor device, the LED light starts to emit a stable light source, and the industrial camera presets the collection field of view and timing interval parameters;

[0020] Step 2: When the granular fertilizer is discharged from the aperture-type fertilizer discharge device and passes through the detection area of ​​the fertilizer flow sensor device, the industrial camera captures images of the falling granular fertilizer at preset time intervals under the conditions of LED lighting and a black background, and transmits the granular fertilizer images to the core controller in real time;

[0021] Step 3: The core controller uses the improved YOLOv5s-seg segmentation model to identify and segment the received image, extracting the mask contour information of the granular fertilizer in the image. Based on the extracted mask contour information, the core controller uses a rounding fitting method to approximate the particles into a spherical model, estimate the mass of each particle, and calculate the total mass of all particles in the frame image. The real-time fertilizer discharge flow rate per unit time is calculated based on the sampling time parameter.

[0022] Step 4: Based on the target fertilization flow rate and the real-time fertilizer discharge flow rate, the microprocessor calculates the control output through the controller algorithm;

[0023] Step 5: driving the aperture-type fertilizer discharge device according to the control output to adjust the fertilizer flow rate.

[0024] Furthermore, the controller algorithm is a fuzzy PID control algorithm, which compares the fertilizer discharge flow detected in real time by the fertilizer flow sensor with the target fertilizer flow calculated based on the set target fertilizer amount, machine travel speed and working width, calculates the control output, adjusts the fertilizer discharge port opening of the aperture-type fertilizer discharge device, and realizes closed-loop precise control of the fertilizer flow.

[0025] Beneficial effects: Compared with the existing technology, the advantages of the present invention are: high detection accuracy: high-quality image acquisition is guaranteed through a dark box environment, a high-contrast background plate and a stable light source; the improved YOLOv5s-seg model is used to accurately segment and extract masks of granular fertilizers, and combined with the subsequent flow calculation method, high-precision real-time monitoring of granular fertilizer flow can be achieved.

[0026] Precise closed-loop control: Based on real-time monitoring of fertilizer discharge flow and target fertilization flow, the aperture-type fertilizer discharge device is driven by a microprocessor and controller algorithm (such as fuzzy PID) for closed-loop control, which can quickly respond to flow changes, accurately control the amount of fertilizer applied, and improve fertilizer utilization.

[0027] High system integration and easy operation: through I 2 C bus realizes the communication between modules and 2 The C parsing module is connected to the human-computer interaction terminal to facilitate users to perform system operation control, status monitoring and parameter setting.

[0028] Strong adaptability: The improved YOLOv5s-seg model is optimized for small particle detection and can adapt to different types and sizes of fertilizer granules. The camera's timing acquisition strategy takes into account the physical characteristics of falling fertilizer, ensuring effective sampling. The diversion effect of the black background also helps to evenly distribute particles within the field of view. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A schematic diagram of the granular fertilizer passing through the detection area according to the present invention;

[0030] Figure 2 This is a schematic structural diagram of the fertilizer flow sensor of the present invention;

[0031] Figure 3 This is a structural diagram of the improved YOLOv5s-seg model described in the present invention;

[0032] Figure 4 This is a structural diagram of the Backbone module in the improved YOLOv5s-seg model described in the present invention;

[0033] Figure 5 This is the AKConv structure diagram in the improved YOLOv5s-seg model described in the present invention;

[0034] Figure 6 This is a schematic diagram of the detection and segmentation effect of the granular fertilizer according to the present invention;

[0035] Figure 7 This is a structural schematic diagram of the centrifugal fertilizer spreader of the present invention;

[0036] Figure 8 This is a structural diagram of the aperture type fertilizer discharge device of the present invention;

[0037] Figure 9 This is a schematic diagram of the integrated human-computer interaction terminal interface of the present invention;

[0038] Figure 10 The present invention is based on I 2Schematic diagram of closed-loop fertilizer flow control system based on C-bus communication;

[0039] Figure 11 This is a flow chart of the fertilizer flow control system of the centrifugal fertilizer spreader described in the present invention.

[0040] In the figure: fertilizer box 1, aperture-type fertilizer discharge device 2, fertilizer flow sensor device 3, fertilizer spreading blade 4, fertilizer spreading disc 5, fixing base 6, fixing ring 7, rotating disc 8, blade group 9, gear 10, stepping motor 11, microprocessor 12, dark box housing 13, LED light 14, black background board 15, industrial camera 16, camera support plate 17, core controller 18, 101, fertilizer outlet; 102, granular fertilizer; 1401, camera field of view; 1402, target granular fertilizer flow. DETAILED DESCRIPTION

[0041] The technical solution of the present invention is described in detail below with reference to the accompanying drawings, but the protection scope of the present invention is not limited to the embodiments.

[0042] Example 1: See Figure 1 、 Figure 2 The granular fertilizer flow sensor device provided by the present invention is installed below the bottom discharge port of the fertilizer box 1 of the centrifugal fertilizer spreader. The device includes a dark box shell 13, an LED light 14, a black background board 15, an industrial camera 16, a camera support plate 17, and a core controller 18. Among them, the dark box shell 13 is fixedly installed at the bottom of the fertilizer box by bolt connection, which is used to cover the entire aperture-type fertilizer discharge device to form a semi-closed imaging environment, effectively avoid external light interference, and improve the stability and accuracy of image acquisition; the camera support plate 17 is used to fix the industrial camera 16 to ensure its constant viewing angle and clear imaging; the industrial camera 16 is used to collect images of granular fertilizers passing through the field of view and falling from the fertilizer discharge port; the LED light 14 is installed inside the shell to provide a uniform and stable lighting source to illuminate the falling granular fertilizers to ensure imaging quality; the black background plate 15 is installed inside the dark box shell through a card slot and is located in the middle position on both sides of it. It not only has a diversion effect on the falling fertilizer particles, but also improves the edge clarity of the fertilizer particle image by providing a high-contrast background, which facilitates subsequent image segmentation; the core controller 18 receives the image data collected by the industrial camera, and calls the deployed image processing algorithm to segment and identify the granular fertilizers in the image, extract its mask contour information, thereby realizing real-time flow calculation and output of the granular fertilizers.

[0043] Figure 1 The schematic diagram of the present invention shows that the granular fertilizer passes through the detection area, which passes through the field of view set by the industrial camera, that is, the detection area for image acquisition. In this process, a single granular fertilizer is affected by gravity F. g and air resistance F dDue to the small mass and volume of fertilizer particles and the short falling path, under normal operating conditions, the air resistance F c Much smaller than gravity F g , which can be ignored. Therefore, assuming that the granular fertilizer is in free fall, the following formula can be used to deduce that the time it takes for the granular fertilizer to enter the camera's field of view and completely leave it is the timing interval:

[0044]

[0045] Where T represents the timing interval, H1 represents the distance from the top of the camera's field of view to the fertilizer drop port, H2 represents the distance from the bottom of the camera's field of view to the fertilizer drop port, and g represents the acceleration of gravity.

[0046] In this example, the YOLOv5s-seg model structure is optimized and improved for the granular fertilizer image segmentation detection task, improving the model's detection accuracy and deployment efficiency. Specifically, the image processing algorithm is an improved YOLOv5s-seg model, consisting of an input layer (Input), a backbone network (Backbone), a neck network (Neck), and a detection head (Detect).

[0047] The input layer (Input) is used to receive the granular fertilizer image to be processed. In this embodiment, the size of the input image is uniformly processed to 640×640×3, where 640×640 represents the spatial resolution of the image and 3 represents the three RGB color channels.

[0048] Figure 4 This is a structural diagram of the backbone network in the improved YOLOv5s-seg model of the present invention, which is used to extract multi-scale feature information from the input granular fertilizer image. In order to achieve high efficiency and lightweight of the model, the backbone network in this embodiment adopts the HGNetV2 network structure, including a Stem layer, four HG Stage layers, a global average pooling layer (GAP), a convolutional layer (Conv 1x1) and a fully connected layer (FC). The Stem layer serves as the initial preprocessing layer of the HGNetV2 network to extract features from the granular fertilizer image. The HG bock network structure of the HG Stage layer performs hierarchical processing on the image data, so that the network learns from low-level to high-level features. The LDS layer of the HG Stage layer reduces the spatial dimension of the granular feature map, thereby reducing the number of parameters and calculations. The global average pooling layer can reduce the spatial dimension of the feature map to each feature. Figure 1 The final convolutional layer integrates the extracted features and maps them to the corresponding categories, and outputs the results through the fully connected layer.

[0049] Neck network Neck uses AKConv structure to replace the traditional convolution structure. Figure 5 Figure 2 shows the AKConv architecture in the improved YOLOv5s-seg model. The convolution layer (Conv2d) performs convolution on the input image and uses the learned offset layer (Offset) to adjust the initial sampling shape to accommodate the image's features. The resampling layer (Resample) resamples the feature map based on the adjusted sampling shape. The resampled feature map is reshaped, reconvolved, and normalized before being output using the SiLU activation function. This flexible convolution operation dynamically adjusts the receptive field to capture the distinct characteristics of large and small particles, thereby improving detection accuracy.

[0050] The detection head (Detect) is connected to the neck network, and uses the fused feature map to finally generate the bounding box and mask segmentation results of the granular fertilizer. In order to accurately match the size characteristics of the granular fertilizer, this embodiment customizes the structure of the detection head. The detection head of this embodiment is composed of a set of preset detection branches that process feature maps of different resolutions. Specifically, the detection head includes three detection branches: a high-resolution detection branch for processing the feature map with a spatial resolution of 160×160 output by the neck network. This branch is specifically responsible for detecting and segmenting small target particles in the image. Because it utilizes the feature map with the highest resolution and the richest spatial details, it can significantly enhance the underlying feature extraction capabilities of small particles; a medium-resolution detection branch for processing 80×80 feature maps, which is responsible for detecting medium-sized particles; a low-resolution detection branch for processing 40×40 feature maps, which is responsible for detecting larger particles.

[0051] The constructed granular fertilizer image dataset was fed into the improved YOLOv5s-seg model for training, obtaining model weight parameters for granular image segmentation and detection. During the actual detection process, the granular fertilizer image to be detected was fed into the model loaded with trained weights to output a granular mask segmentation. The backbone network was used to extract multi-scale feature information from the input image; the Neck function was used to fuse the multi-scale features extracted by the backbone network to enhance the contextual semantic information for target detection. The detection head, Detect, generated the segmentation results of the granular fertilizer mask and then output the mask's contour information. In the strategy for improving the YOLOv5s-seg model, the lightweight HGNetV2 network is selected to replace the model's backbone network, reducing computational complexity and memory consumption, making the trained model more suitable for deployment on mobile and embedded devices. In the neck network, the Conv structure is replaced with the AKConv structure, a flexible dynamic convolution method that can adaptively adjust the receptive field according to the target size, thereby more effectively capturing the different characteristics of large and small particles and improving the model's ability to segment targets of different scales. The present invention adds a set of detection heads specifically for small-particle fertilizer segmentation detection to the detection head, strengthening the underlying feature extraction capabilities and effectively improving the detection accuracy of small target particles. At the same time, unnecessary large-target detection heads are removed, further reducing the model's redundant computational complexity and parameter size.

[0052] Structural Optimization: Compared to the standard YOLOv5s model, the detection head in this example does not include a large object detection branch for processing feature maps with lower resolutions (e.g., 20×20). Because extremely large objects do not exist in the granular fertilizer task, this redundant branch was removed.

[0053] Figure 6 This is a schematic diagram comparing the detection and segmentation effects of the YOLOv5s-seg network model for granular fertilizer used in the present invention before and after improvement. It can be clearly seen that the recognition and detection effect of the granular fertilizer by the improved YOLOv5s-seg network model is significantly better than that of the YOLOv5s-seg network model before improvement; Figure 8 In the figure, (a), (b), and (c) represent three different flow conditions: no overlap, no overlap, and dense overlap between granular fertilizers. It can be clearly seen from the figure that no matter which case, the segmentation results (white mask) of the improved YOLOv5s-seg model are more accurate and complete than the YOLOv5s-seg model in the middle. Especially in the dense overlap scenario shown in (c), the improved model can accurately segment more and smaller particles, proving the effectiveness of the technical solution of the present invention.

[0054] Example 2: Figure 7The centrifugal fertilizer spreader shown in the figure consists of a fertilizer bin 1, an aperture-type fertilizer discharging device 2, a fertilizer flow sensor 3, fertilizer spreading blades 4, and a fertilizer spreading disc 5. The fertilizer bin 1 is composed of two adjacent inverted conical shells, with a dual-channel bottom discharge opening. The aperture-type fertilizer discharging device 2 is mounted at the bottom discharge opening of the fertilizer bin 1 and secured via a mounting bracket 13. The aperture-type fertilizer discharging device 2 adjusts the opening of its internal blades to create different discharge openings, allowing granular fertilizer to be continuously and evenly discharged under the action of gravity. The fertilizer flow sensor 3 is mounted at the bottom of the fertilizer bin 1, enclosing the aperture-type fertilizer discharging device, and is used to monitor the discharged fertilizer flow in real time. The fertilizer spreading disc 5 is located approximately 100 mm directly below the fertilizer flow sensor 3. It is responsible for evenly spreading the falling granular fertilizer onto the field through the centrifugal force of the fertilizer spreading blades 4, achieving high-efficiency fertilization over a large area.

[0055] Figure 8 This is a schematic diagram of the aperture-type fertilizer discharge device of the present invention, consisting of a fixed base 6, a fixed ring 7, a rotating disk 8, a blade assembly 9, a gear 10, a stepper motor 11, and a microprocessor 12. The fixed base 6 is bolted to the bottom feed opening of the fertilizer bin 1. The blade assembly is assembled from multiple blades with positioning posts on both sides, forming a circular fertilizer discharge opening with adjustable opening. The blades are made of a metal with a tough material to ensure their durability during operation. They are fixed to the fixed base by a bottom stopper to ensure a controlled range of motion. The rotating disk covers the upper portion, and the positioning posts are used to accurately embed the blade assembly on the rotating disk, allowing for uniform rotational adjustment. The fixed ring is positioned above the rotating disk, enhancing the compactness and stability of the device. The output shaft of the stepper motor is connected to a gear structure, which meshes with teeth arranged on the circumference of the rotating disk. The microprocessor controls the rotation direction and speed of the stepper motor by outputting pulse signals, driving the rotating disk to rotate, causing the blade assembly to rotate synchronously, thereby adjusting the opening of the fertilizer discharge opening and achieving precise regulation of the fertilizer discharge flow rate.

[0056] Example 2: Figure 9 This is a schematic diagram of a control system for a granular fertilizer flow sensor device according to Example 1 of the present invention, which also includes a fertilizer box 1, an aperture-type fertilizer discharge device 2, a Hall speed measurement module, and an aperture-type fertilizer discharge device 2. 2 C communication bus, I 2 C analysis module and fertilizer spreading device, the fertilizer spreading device includes fertilizer blades 4 and fertilizer spreading discs 5.

[0057] The Hall effect speed measurement module is used to collect the real-time speed information of the fertilizer spreading machine, and the fertilizer flow sensor device is used to detect the current actual fertilizer flow rate. 2 C communication bus to I 2 C parsing module.

[0058] I2 The C parsing module is responsible for identifying and parsing I 2 The data frame information transmitted by the C communication bus includes the vehicle speed collected by the Hall effect speed measurement module, the control quantity output by the microprocessor, and the fertilizer discharge flow input by the core controller. It also classifies and processes the transmitted data to ensure that different data frames are accurately received and parsed. It also serves as a temporary buffer for data and identifies communication errors.

[0059] by I 2 After completing data analysis and integration, the C analysis module passes it to the control system. The control system dynamically calculates the required fertilizer flow rate based on the preset target fertilizer amount and operating width, and compares it with the actual fertilizer discharge flow rate to form a closed-loop control.

[0060] Figure 10 This is a schematic diagram of the integrated human-computer interaction terminal interface of the present invention. The control interface is used to set the target fertilizer amount for the operating field according to the fertilization mode, and can realize the start and stop operation of the controller and the adjustment of the fertilizer discharge port opening; the monitoring interface is used to display in real time the current fertilizer flow value calculated by the fertilizer flow detection device, the operating speed of the fertilizer spreading machine collected by the Hall effect speed measurement module, the number of granular fertilizer image frames collected by the industrial camera, and the quality information of the granular fertilizer in each frame image, so as to facilitate the subsequent cumulative calculation of the total fertilization mass; the system operation interface is used to save data and terminate the system program.

[0061] Figure 11 This is a flow chart of the variable fertilizer control system for the centrifugal fertilizer spreader of the present invention. The control system calculates the target fertilizer flow rate q(t) based on the fertilizer amount Q of the input unit target plot, the real-time machine forward speed v collected by the Hall effect speed measurement module, and the working width W. The fuzzy PID controller collects the real-time fertilizer flow rate q'(t) and the target fertilizer flow rate q(t) detected by the fertilizer flow sensor, and calculates the state of the input actual flow deviation e and the flow deviation change rate ec. By analyzing the relationship between the two, a fuzzy level ("negative large", "negative medium", "negative small", "zero", "positive small", "positive medium", "positive large") is formulated, and the fuzzy control quantity is converted into a specific PID gain adjustment value ΔK. p , ΔK i , ΔK d , to achieve fast and accurate control of fertilizer discharge flow.

[0062] The integrated human-computer interaction terminal uses the fertilization prescription map to obtain the target fertilizer amount B for the fertilization plot. The Hall effect speed measurement module collects and generates the current real-time speed v of the fertilizer spreader. Combined with the fertilizer spreader's operating width W, the controller calculates the fertilizer flow rate Q of the target plot in real time according to the following flow mathematical formula:

[0063]

[0064] The feedback link of the closed-loop fertilizer flow control system is: the fertilizer flow sensor collects the real-time fertilizer flow information through I 2 The C communication bus feeds back to the integrated human-computer interaction terminal. The flow control system interpolates and calculates the fertilizer flow error based on the real-time fertilizer flow information fed back and the required fertilizer flow of the target plot. The microprocessor calculates the control output using the fuzzy PID algorithm and adjusts the fertilizer discharge port opening of the aperture-type fertilizer discharge device to complete the control of the fertilizer flow. The integrated human-computer interaction terminal uses the control interface to realize controller startup, fertilization mode selection, and operation parameter setting. The monitoring interface is used to display the fertilizer flow feedback from the fertilizer flow sensor and the operation speed collected by the Hall effect speed measurement module. The system operation interface is used to save data and terminate the system program.

[0065] Example 3: This example is a method for detecting fertilizer discharge flow using the above control system, including the following steps:

[0066] Step 1: After the fertilizer flow sensor is powered on, the LED lamp emits light to illuminate the falling fertilizer particles, and the industrial camera sets the camera field of view and the timing interval T for collection;

[0067] Step 2: The aperture-type fertilizer discharge device is started and reaches the preset fertilizer discharge port opening. Granular fertilizer flows out from the fertilizer outlet at the bottom of the fertilizer box. The granular fertilizer passes through the camera field of view acquisition area of ​​the fertilizer flow sensor. The industrial camera starts to collect granular fertilizer images at regular intervals and transmits the images to the core controller for analysis and processing.

[0068] Step 3: The core controller uses the improved YOLOv5s-seg network segmentation model to identify, detect and segment the fertilizer particles in the image, and outputs the mask contour information of the fertilizer particles in each frame of the image.

[0069] Step 4. The core controller uses rounding fitting to process the mask contour information of the granular fertilizer in each frame to obtain the corresponding granular fertilizer diameter, and treats the particles in the image as approximately spheres to estimate the particle volume. The mass of each particle is further calculated by the density of the granular fertilizer. Combined with the total number of particles detected in the image, the total mass of the granular fertilizer in each image is output; then the granular fertilizer flow rate at the current fertilizer discharge opening within the sampling time is calculated through the flow calculation model.

[0070] In step 4, the flow calculation model of granular fertilizer is as follows:

[0071]

[0072] Where N represents the fertilizer flow rate, N represents the number of granular fertilizer frames captured by the camera, i represents the image frame used to calculate the granular fertilizer flow rate, M represents the number of granular fertilizer detected in each frame, j represents the granular fertilizer in each frame, D represents the diameter of the granular fertilizer in each frame, p represents the fertilizer density, and t represents the total time for collecting granular fertilizer frames.

[0073] Step 5: driving the aperture-type fertilizer discharge device according to the control output to adjust the fertilizer flow rate.

[0074] The present invention adopts a dark box shell, a stable LED light source and a high-contrast black background board to provide an excellent image acquisition environment for industrial cameras, reduce the interference of external lighting changes and complex backgrounds, and ensure the quality of the original image. The improved YOLOv5s-seg model is optimized for the segmentation detection of granular fertilizers, especially small-granular fertilizers, and can accurately extract the mask contour information of each particle. Combined with the volume estimation and density parameters based on the mask contour, the real-time fertilizer discharge flow can be accurately calculated. Compared with the traditional indirect measurement method, this solution based on direct visual measurement has higher accuracy and better real-time performance. The black background board also has a diversion function, which helps the fertilizer particles pass more evenly within the camera's field of view, avoids accumulation and occlusion, and improves the reliability of detection. The calculation formula for the timed acquisition time of the industrial camera takes into account the free-fall motion characteristics of the fertilizer, ensuring the effectiveness of sampling.

[0075] The system compares the real-time monitored fertilizer flow rate with the target flow rate calculated based on operating parameters (target fertilizer rate, travel speed, and operating width). A fuzzy PID controller calculates the control variable, driving the aperture-type fertilizer discharge device to precisely adjust the fertilizer discharge opening. This closed-loop control method can quickly respond to various disturbances (such as changes in material level, fluctuations in fertilizer characteristics, and changes in vehicle speed), ensuring that the actual fertilizer rate always closely tracks the target fertilizer rate, significantly improving the uniformity and accuracy of fertilization, reducing fertilizer waste, and improving fertilizer utilization.

[0076] The improved YOLOv5s-seg model not only achieves high segmentation accuracy but also reduces computational complexity by removing redundant detection heads, making it more suitable for real-time execution on embedded core controllers. The introduction of the AKConv architecture enhances the model's adaptability to fertilizer particles of varying shapes and sizes. The newly added small particle segmentation detection head ensures effective recognition of fine fertilizer particles.

[0077] Each functional module (sensing, control, execution, speed measurement, human-computer interaction) is connected through I 2 C bus and other effective integration methods to form a complete intelligent fertilization control system. Users can easily set parameters, start and stop control and monitor status through the human-computer interaction terminal, which is easy to operate.

[0078] In summary, the granular fertilizer flow sensing device, centrifugal fertilizer spreader fertilizer flow control system and method provided by the present invention significantly improve the fertilization accuracy and operating efficiency of the centrifugal fertilizer spreader through high-precision visual sensing technology, advanced image processing algorithms and intelligent closed-loop control strategies, and have important application value for realizing precision agriculture.

[0079] As described above, although the present invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the present invention itself. Various changes may be made to it in form and detail without departing from the spirit and scope of the present invention as defined in the appended claims.

Claims

1. Granular fertilizer flow sensor device, characterized in that, include: A dark box housing, used to form an at least partially enclosed image acquisition environment; an industrial camera, disposed in the dark box housing, for collecting images of the granular fertilizer; An LED lamp is arranged in the dark box housing and is used to provide light for the granular fertilizer; A black background plate is provided in the dark box housing and serves as a high-contrast background when the granular fertilizer is photographed by the industrial camera; The core controller is connected to the industrial camera data and is configured to: receive images captured by the industrial camera; segment and detect the granular fertilizer in the image using an image processing algorithm and extract mask contour information of the particles; and calculate the fertilizer discharge flow in real time based on the mask contour information to obtain a real-time fertilizer discharge flow signal.

2. The granular fertilizer flow sensor device according to claim 1, characterized in that: The image processing algorithm is an improved YOLOv5s-seg model, including: Input layer, used to input granular fertilizer images; The backbone network (Backbone) adopts the HGNetV2 network structure, which includes a Stem layer for extracting initial features and at least one HG Stage layer. The HG Stage layer includes a downsampling layer for reducing the spatial dimension of the feature map to extract multi-scale features of the granular fertilizer image; The neck network (Neck) adopts the AKConv structure, which includes an offset layer that adjusts the convolution sampling shape by learning an offset, and a resampling layer that resamples the feature map according to the adjusted sampling shape. The resampled feature map is processed by convolution again, normalization, and SiLU activation function before output. A detection head (Detect) is connected to the neck network Neck, and the detection head includes a small particle detection branch for processing a high-resolution feature map with a spatial resolution of 160×160, and detection branches for processing 80×80 and 40×40 feature maps respectively.

3. The granular fertilizer flow sensor device according to claim 1, characterized in that: The core controller calculates the granular fertilizer flow rate through the following steps: The rounding fitting method is used to approximate the granular fertilizer into a spherical model, estimate the mass of a single granular fertilizer, and calculate the total mass of the granules in each frame of the image; The granular fertilizer flow rate per unit time is calculated according to the following formula: Where Q represents the fertilizer flow rate, N represents the total number of granular fertilizer image frames captured by the camera, i represents the i-th frame image used to calculate the fertilizer flow rate, M represents the number of granular fertilizer detected in each frame image, j represents the j-th granular fertilizer in each frame image; D represents the equivalent diameter of each granular fertilizer, p represents the fertilizer density, and t represents the total time for capturing granular fertilizer image frames.

4. The granular fertilizer flow sensor device according to claim 1, characterized in that: The industrial camera is configured to capture images of falling granular fertilizers in a timed interval manner, and the timed interval T of the industrial camera is determined by the following formula: Among them, H1 is the distance from the top of the camera field of view to the fertilizer drop port, H2 is the distance from the bottom of the camera field of view to the fertilizer drop port, and g is the acceleration of gravity.

5. The granular fertilizer flow sensor device according to claim 1, characterized in that: The black background plate is fixed to the inner side of the dark box shell through a card slot, and the black background plate is also used to divert the granular fertilizer falling from the fertilizer box drop port.

6. A centrifugal fertilizer spreader fertilizer flow control system comprising the granular fertilizer flow sensor device according to any one of claims 1 to 5, characterized in that: include: A fertilizer box is provided with a bottom discharge port for storing granular fertilizer; A granular fertilizer flow sensor device includes an industrial camera for capturing images of falling granular fertilizers and a core controller for processing the images to obtain a real-time fertilizer discharge flow signal. Hall effect speed measurement module, used to monitor the speed of the fertilizer spreader; An aperture-type fertilizer discharge device is installed at the bottom discharge port of the fertilizer box and has a blade group with adjustable opening and a microprocessor; the microprocessor is configured to: calculate a control output through a controller algorithm based on a preset target fertilizer flow rate and a real-time fertilizer discharge flow rate signal output by the core controller, and adjust the fertilizer discharge port opening of the blade group according to the control output to achieve closed-loop precision control of the fertilizer flow rate; I 2 C communication bus for implementing I 2 Bus communication between C parsing module, Hall speed measurement module, microprocessor and core controller; I 2 C parsing module, responsible for identifying and parsing I 2 The data frame information transmitted by the C communication bus includes the vehicle speed collected by the Hall effect speed measurement module, the control quantity output by the microprocessor, and the fertilizer discharge flow input by the core controller. The transmitted data is classified and processed to ensure that different data frames are accurately received and parsed.

7. The fertilizer flow control system for a centrifugal fertilizer spreader according to claim 6, characterized in that: The aperture-type fertilizer discharge device (2) further comprises a fixed seat (6), a fixed ring (7), a rotating disk (8), a gear (10) and a stepping motor (11), wherein the fixed seat (6) is fixed to the bottom of the fertilizer box (1), the blade group (9) forms a circular fertilizer discharge opening with adjustable opening by stacking multiple blades, the rotating disk (8) is sleeved on the outside of the blade group (9) and connected to the stepping motor (11) via the gear (10), the stepping motor (11) is controlled by a microprocessor (12) to drive the rotating disk (8) to rotate, and the blade group (9) rotates synchronously with the rotating disk (8) to adjust the relative positions of the multiple blades, thereby forming circular fertilizer discharge openings with different openings.

8. The fertilizer flow control system for a centrifugal fertilizer spreader according to claim 6, characterized in that: It also includes a fertilizer spreading device, which includes a fertilizer spreading disc (5) and fertilizer spreading blades (4), which are arranged below the fertilizer flow sensor device and are used to evenly spread granular fertilizer into the field through centrifugal force.

9. A method for using the centrifugal fertilizer spreader fertilizer flow control system according to claim 6, characterized in that: The following steps are involved: Step 1: Start the fertilizer flow sensor device, the LED light starts to emit a stable light source, and the industrial camera presets the collection field of view and timing interval parameters; Step 2: When the granular fertilizer is discharged from the aperture-type fertilizer discharge device and passes through the detection area of ​​the fertilizer flow sensor device, the industrial camera captures images of the falling granular fertilizer at preset time intervals under the conditions of LED lighting and a black background, and transmits the granular fertilizer images to the core controller in real time; Step 3: The core controller uses the improved YOLOv5s-seg segmentation model to identify and segment the received image, extracting the mask contour information of the granular fertilizer in the image. Based on the extracted mask contour information, the core controller uses a rounding fitting method to approximate the particles into a spherical model, estimate the mass of each particle, and calculate the total mass of all particles in the frame image. The real-time fertilizer discharge flow rate per unit time is calculated based on the sampling time parameter. Step 4: Based on the target fertilization flow rate and the real-time fertilizer discharge flow rate, the microprocessor calculates the control output through the controller algorithm; Step 5: driving the aperture-type fertilizer discharge device according to the control output to adjust the fertilizer flow rate.

10. The method according to claim 9, characterized in that The controller algorithm is a fuzzy PID control algorithm, which compares the fertilizer discharge flow rate detected in real time by the fertilizer flow sensor with the target fertilizer flow rate calculated based on the set target fertilizer amount, machine travel speed and working width, calculates the control output, adjusts the fertilizer discharge port opening of the aperture-type fertilizer discharge device, and realizes closed-loop precise control of the fertilizer flow rate.

Images