Grain flow detection device and plot yield prediction method based on raccoon algorithm

Abstract

The present invention discloses a grain flow detection device and a plot yield prediction method based on the raccoon algorithm. The detection device includes an impeller-type grain collection device and a control box. The impeller-type collection device includes a device housing, a bearing seat, a central rotating shaft, blades, an electromagnet, an electromagnet end block, a capacitive proximity switch and a brush. The control box is composed of a single-chip microcomputer, a control button, an LED display module, and a CAN communication module. The yield prediction method is to optimize the BP neural network based on the raccoon algorithm for prediction. The BP neural network uses the golden section algorithm to optimize the number of nodes in the hidden layer of the BP neural network, and then optimizes the connection weights and hidden layer thresholds of the BP neural network through the raccoon optimization algorithm. Finally, parameters such as grain collection time and grain volume are used as inputs of the trained BP neural network to output the predicted value of grain yield. The present invention can enable a combine harvester to obtain more accurate grain yield information in a complex operating environment and improve the accuracy of yield prediction.

Description

Technical Field

[0001] The present invention relates to the field of agricultural machinery measurement, and in particular to a grain flow detection device and a plot yield prediction method based on a raccoon algorithm. Background Art

[0002] my country is a major agricultural country. In 2021, the cultivated area for grain reached 117.63 million hectares, an increase of 86 hectares over the previous year. Of this total, 29.92 million hectares were planted with rice, and 23.57 million hectares with wheat, both increases of 190,000 hectares over the previous year. With the expansion of large-scale cultivation, the development of precision agriculture is essential.

[0003] Precision agriculture, also known as fine farming or precision agriculture, is a new trend in modern agricultural development worldwide. It refers to a modern agricultural production system based on information and knowledge management. Technologies supporting precision agriculture primarily include intelligent agricultural equipment and system integration, modern 3S technology, yield map generation, and intelligent agricultural management and decision-making. Yield map generation technology is an essential component of precision agriculture.

[0004] The collection of yield information is crucial for yield map generation. This information is obtained through a grain yield measurement system. Grain flow sensors, a core component of these systems, are currently used primarily in the following categories: impulse flow sensors, photoelectric flow sensors, radiation-based grain flow sensors, and gravimetric flow sensors.

[0005] Due to the harsh working environment of combine harvesters, impulse flow sensors are greatly affected by vibration interference, resulting in low measurement accuracy. Photoelectric flow sensors detect the distribution of grain accumulation inside the elevator. Their probes are easily contaminated and require frequent cleaning. Their performance is unstable and they are only suitable for scraper elevators. They are also difficult to install. Radiation grain flow sensors have high measurement accuracy, but the radiation poses certain hazards to operators and the environment. The high cost of use has hindered their widespread application to a certain extent. Weighing flow sensors are mainly used to weigh the grain silos during the harvesting process of combine harvesters. They require the installation of a pressure sensor at the bottom of the silo, which requires the silo to be disassembled for installation, making the installation process cumbersome. Summary of the Invention

[0006] The technical problem addressed by this invention is that existing grain yield detection devices are susceptible to vibration interference during operation, resulting in low detection accuracy. Furthermore, this invention is applicable not only to spiral elevators but also to scraper elevators, resolving the drawback of existing photoelectric grain flow detection devices that cannot be applied to spiral elevators.

[0007] To address these issues, the present invention proposes a novel grain flow detection device based on the Raccoon algorithm. This device installs a grain collection device at the end of a combine harvester's clean grain elevator to collect grain flowing out of the elevator. The flow rate is calculated by measuring the time required to collect a fixed volume of grain.

[0008] The present invention achieves the above technical objectives through the following means.

[0009] A grain flow detection device based on the raccoon algorithm, comprising:

[0010] The device housing, central shaft, blades, and brushes. The central shaft is fixed to the device's central axis, with bearings at each end securing it to the housing. The blades are welded to the central shaft, with the three blades evenly distributed around the central shaft. The brushes are located in the gap between the blades and the housing, fixed to the blades so they rotate with them.

[0011] The device shell is provided with a grain inlet and a grain outlet, which are in roughly opposite directions; the grain inlet is surrounded by a rectangular baffle;

[0012] Grain level detection sensor. The grain level detection sensor is installed on the housing of the grain flow collection device near the grain inlet, and is evenly arranged in parallel along the axis of the central shaft. The straight line formed by the arrangement is parallel to the straight line of the central shaft.

[0013] The blade rotation control device includes an electromagnet and a stopper at the end of the electromagnet. The electromagnet is mounted on the device housing, outside the grain collection area. A hole is drilled in the device housing, and the end of the electromagnet with the stopper is inserted into the hole. The main body of the electromagnet is fixed to the outside of the device housing.

[0014] The grain flow data acquisition and control device mainly consists of a voltage-stabilized power supply, an STM32F407 single-chip microcomputer, an LED display module, control buttons, and a CAN communication module. The single-chip microcomputer is connected to the LED display module, the voltage-stabilized power supply, the control buttons, an electromagnet, and the CAN communication module.

[0015] In the above technical solution, the grain level detection sensor, electromagnet and LED display module are also connected to a voltage-stabilized power supply respectively.

[0016] In the above technical solution, the grain level detection sensors are normally open capacitive proximity switches, and there are three of them.

[0017] In the above technical solution, the electromagnet is a push-pull DC12v, 3A electromagnet with a maximum suction force of 74N.

[0018] The working process of the grain flow detection device is as follows:

[0019] The combine harvester begins harvesting, the electromagnet is in its initial state, and the block blocks the blades from rotating. Grain enters the harvesting area from the device's grain inlet. Brushes are fixed to the edges of the blades, distributed between the blades and the device housing, effectively preventing grain from flowing out of the gap between the blades and the device housing during harvesting. At this time, the control button controls the microcontroller to start timing. When a certain volume of grain is collected, the grain level rises to the grain level detection sensor, triggering the sensor. The sensor sends a signal to the microcontroller, which stops timing and resets the timer. At the same time, it sends a control signal to the electromagnet, controlling it to operate, pull the block, and cause the blades to begin rotating under the weight of the grain, releasing the harvested grain. At the same time, the electromagnet resets, blocking the next blade, and the timer starts counting, starting the next grain harvesting cycle.

[0020] After receiving the signal from the grain level detection sensor, the single chip microcomputer reads the data in the timer as the collection time t, saves it, and transmits it to the host computer through the CAN communication module. The host computer combines the collection time t with the volume v of grain collected each time, the bulk density of grain ρ, the travel speed v of the combine harvester, the cutting width l during the harvester operation, the moisture content w of the grain, and the average loss rate P during the harvesting process. i As the input of BP neural network, the predicted value of grain flow and yield is output.

[0021] The BP neural network is optimized using the raccoon algorithm;

[0022] The raccoon algorithm is an optimization algorithm derived from simulating the hunting behavior of raccoons. It has strong optimization capabilities and fast convergence. It includes initialization, hunting and attacking in Phase 1, and escaping from predators in Phase 2. The details are as follows:

[0023] Step 3.1, population initialization, calculate the threshold and weight number M to be optimized according to the number of neurons in the input layer, hidden layer and output layer of the BP neural network, and calculate according to the following formula

[0024] M=n×n1+2n+1+m×n1+m

[0025] In the search space, N raccoon individuals are randomly generated according to the following formula. The dimension of each raccoon individual is M. The optimization of the threshold and weight is converted into the optimal solution for the position of the raccoon individual.

[0026]

[0027] Where: x i,j for individuals; To find the optimal lower bound; is the upper boundary of the optimization; r is a random number between [0,1];

[0028] Step 3.2: Initialize the optimization algorithm parameters, determine the maximum number of iterations, upper and lower limits of variable values, and fitness function, and use the training error in the BP neural network as the fitness function;

[0029] Step 3.3, exploration phase; assuming the position of the best member of the population is the position of the iguana, half of the raccoons climb the tree and the other half wait for the iguana to fall to the ground. Therefore, the position simulation is:

[0030]

[0031] Where: is the new position of the i-th raccoon in the j-th dimension; r is a random number between [0,1]; G j is the position of the iguana in the jth dimension, which actually refers to the position of the best member; I is a number randomly selected from the set {1,2}; N is the number of raccoons; [N / 2] is the largest integer not exceeding [N / 2];

[0032] In step 3.4, after the iguana falls to the ground, it is placed at a random position in the search space. Based on this random position, the raccoon on the ground moves in the search space, simulated according to the following formula:

[0033]

[0034] Where: is the position of the iguana on the ground in the jth dimension;

[0035]

[0036] Where: is the objective function value after the iguana falls to the ground in the jth dimension; F i,j is the objective function value of the i-th raccoon in the j-th dimension;

[0037] Step 3.5: If the updated individual is better, update the current individual; otherwise, keep it as it is:

[0038]

[0039] Where: is the objective function value of the i-th raccoon in the new position; F i is the objective function value of the i-th raccoon at its previous position;

[0040] Step 3.6, Escape from Predator. The second stage of the process of updating the raccoon's position in the search space is mathematically modeled based on the raccoon's natural behavior when encountering and fleeing from predators. When a predator attacks a raccoon, it flees from its position. The raccoon's moves in this strategy put it in a safe position close to its current position. This behavior is simulated based on the following equation:

[0041]

[0042] Where: is the local lower bound of the j-th decision variable; is the local upper bound of the jth decision variable, t is the number of iterations; T is the maximum number of iterations;

[0043]

[0044] Where: is the new position of the i-th raccoon in the j-th dimension;

[0045] Step 3.7, bring the updated individual into the BP neural network, use the test data set to calculate the error between the predicted value and the true value as the fitness to judge, if the updated individual is better, update the current individual, otherwise keep it as it is

[0046]

[0047] Furthermore, the number of hidden layer nodes of the BP neural network is optimized using the golden section algorithm.

[0048] Step 1: Construct a BP neural network; the number of input layer nodes is 7, the number of output layer nodes is 1, and the number of hidden layer nodes n1 is determined by the following formula:

[0049]

[0050] Where n is the number of input layer nodes, m is the number of output layer nodes, and a is between 1 and 10. It can be determined that the number of hidden layer nodes ranges from [3, 13].

[0051] Step 2: Optimize the number of nodes in the hidden layer of the BP neural network; including the following:

[0052] Step 2.1: The number of hidden layer nodes is within the interval [p, q]. Find the first split point h1=0.382*(qp) within the interval [p, q]. Calculate the mean square error MSE(h1) of the first split point. The calculation formula is as follows:

[0053]

[0054] Where y iis the true value, is the predicted value, and N is the number of samples;

[0055] Step 2.2: Find the second splitting point within [p, q], h2 = 0.618 * (q - p), and obtain the mean square error MSE(h2) of the second splitting point;

[0056] Step 2.3: Compare MSE(h1) and MSE(h2). If MSE(h1) < MSE(h2), the hidden layer node interval becomes [p, h2]; if MSE(h1) > MSE(h2), the hidden layer node interval becomes [h1, q]; if MSE(h1) = MSE(h2), the hidden layer node interval becomes [h1, h2];

[0057] Step​​​​​​​​​​​​​​​​​​​​​​​

[0064] Figure 2 Schematic diagram of the internal structure of the detection device according to the present invention.

[0065] Figure 3 This is a schematic diagram of the installation position of the detection device of the present invention on a combine harvester.

[0066] Figure 4 This is a structural block diagram of the control system of the grain flow detection device described in the present invention.

[0067] Figure 5 This is a working flow chart of the flow detection device of the present invention.

[0068] Figure 6 This is a flow chart of the raccoon optimization algorithm described in the present invention.

[0069] Figure 7 This is a flow chart of the grain yield prediction method of the present invention.

[0070] In the figure: 1-grain inlet, 2-electromagnet, 3-stopper, 4-capacitive proximity switch, 5-blade, 6-device housing, 7-bearing seat, 8-center shaft, 9-grain outlet, 10-grain flow detection device, 11-LED display module, 12-control box, 13-brush, 14-clean grain elevator. DETAILED DESCRIPTION

[0071] The detection device designed in the present invention includes an impeller-type grain collection device and a control box. The impeller-type collection device is mainly composed of a grain flow collection device inlet, a device housing, a bearing seat, a central rotating shaft, blades, an electromagnet, an electromagnet end block, a capacitive proximity switch and a brush. The control box is composed of a single-chip microcomputer, a control button, an LED display module, a storage module, and a CAN communication module. The cell yield prediction method is based on the raccoon algorithm to optimize the BP neural network to predict the yield. It includes: building a BP neural network, using the golden section algorithm to optimize the number of hidden layer nodes of the BP neural network, and then optimizing the connection weights and hidden layer thresholds of the BP neural network through the raccoon optimization algorithm. Finally, the acquired collection time and the collected grain volume are combined with other parameters as the input of the trained BP neural network to obtain the predicted value of grain yield. The present invention can enable the combine harvester to obtain more accurate grain yield information in a complex operating environment and improve the accuracy of yield prediction.

[0072] The present invention will be further described below with reference to the accompanying drawings and embodiments. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are illustrative and intended to explain the present invention, and are not to be construed as limiting the present invention.

[0073] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "axial", "radial", "vertical", "horizontal", "inside", "outside" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0074] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integral connection; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; or internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0075] The specific implementation of the grain flow detection device and plot yield prediction method based on the raccoon algorithm of the present invention is as follows:

[0076] like Figure 1 、 2As shown, a grain flow detection device based on the Raccoon algorithm includes a grain inlet 1, a grain outlet 9, a device housing 6, a bearing block 7, a central shaft 8, a blade 5, an electromagnet 2, a stopper 3, a capacitive proximity switch 4, an LED display module 11, a control box 12 for the detection device, and a brush 13. The brush 13 is connected to the blade 5 by solid glue. When the blade 5 rotates, the brush 13 moves along with it. The blade 5 is fixedly connected to the central shaft 8 by welding. When the central shaft 8 rotates, the blade 5 rotates. The central shaft 8 transverses the central axis of the device housing 6 and is fixed to the device housing 6 on both sides by bearing blocks 7. The bearing blocks 6 ensure that the central shaft 8 does not come into direct contact with the device housing 6 during rotation. The stopper 3 at the end of the electromagnet is fixed to the electromagnet 2. When the electromagnet 2 is energized, it drives the stopper 3 at the end of the electromagnet to move. Electromagnet 2 is fixed to the outside of device housing 6. A stopper 3 at the end of the electromagnet extends into the interior of device housing 6 through a circular hole in the housing, controlling the rotation of central shaft 8 by blocking blade 5. A capacitive proximity switch 4 is fixed to the device through a circular hole in the housing 6 and is used to detect the grain level collected within the device.

[0077] like Figure 4 As shown, the control system of the grain flow detection device is composed of an STM32F407 single chip microcomputer and a grain level detection sensor, a control button, an electromagnet 2, an LED display module 11, a data storage module, and a CAN communication module. The grain level detection sensor is a capacitive proximity switch 4.

[0078] like Figure 3 The figure shows the installation position of the grain flow detection device on the combine harvester. In a specific embodiment, it is installed at the outlet of the clean grain elevator 14.

[0079] like Figure 5 The figure shows the overall working process of the flow detection device. Figure 3As shown, the detection device is installed on the combine harvester at the illustrated location. The system is initialized, the microcontroller timer is set, and the blades are in their initial position. The combine harvester begins harvesting, and the timer is turned on. After being processed by the harvester, grain flows out from the end of the clean grain elevator 14 and enters the grain flow detection device through the grain inlet 1 of the detection device, beginning collection. When the grain reaches the height of the capacitive proximity switch 4, the capacitive proximity switch 4 detects the grain signal and transmits it to the STM32F407 microcontroller. The microcontroller receives the signal, stops the timer, and records the timer timing as the collection time t in the memory module. Simultaneously, a control signal is sent to the electromagnet 2, which pulls up the stopper 3 at the end of the electromagnet, allowing the blades 5 to rotate under the influence of the grain's gravity, releasing the grain for collection. Simultaneously, the electromagnet 2 returns to its original position, blocking the next blade 5. The microcontroller then turns on the timer, beginning the second collection. This cycle repeats.

[0080] like Figure 7 As shown, after obtaining the grain collection time, the grain yield in the harvesting area can be obtained by combining the information such as the travel speed and harvesting amplitude of the combine harvester. The problem of predicting yield information based on measured data is actually a nonlinear function fitting problem, and it is difficult to determine the yield value through a simple linear formula. The present invention uses the raccoon algorithm to perform rapid global optimization of nonlinear extreme values, optimizes the initial value and threshold of the BP neural network, can increase the speed of nonlinear fitting, enhance the nonlinear fitting ability of the BP neural network, and can accurately predict the value of grain yield and reduce prediction errors. The above process is specifically as follows:

[0081] Step (1): Construct BP neural network

[0082] The BP neural network is a multi-layer feedforward network trained by error back propagation (abbreviated as error back propagation). It has a fast processing speed and strong fault tolerance. For any continuous function in a closed interval, the network can be approximated by a single hidden layer feedforward neural network. The BP neural network prediction model used in this embodiment of the present invention has three layers.

[0083] The input of the BP neural network is the grain collection time t, the volume of grain collected each time V, the bulk density of grain ρ, the travel speed v of the combine harvester, the cutting width l during the harvester operation, the grain moisture content w, and the average loss rate P during the harvesting process. i , the activation function selects the Sigmoid function: The output is the predicted value of grain yield within the harvest distance L = v × t; the number of input layer nodes is 7, the number of output layer nodes is 1, and the number of hidden layer nodes n1 is determined by the following formula:

[0084]

[0085] Where n is the number of nodes in the input layer, m is the number of nodes in the output layer, and a is between 1 and 10. The range of the number of hidden layer nodes can be determined as [3, 13].

[0086] Step (2), use the golden section algorithm to optimize the number of hidden layer nodes of the BP neural network.

[0087] Step (2.1), assume that the number of hidden layer nodes is within the range [p, q]. Find the first splitting point h1 = 0.382 * (q - p) within the range [p, q], and calculate the mean square error MSE(h1) of the first splitting point. The calculation formula is as follows.

[0088]

[0089] In the formula, y i is the true output value of the test set data, is the predicted value of the test set data with the number of hidden layer nodes being h1, and N is the number of samples.

[0090] Step (2.2), find the second splitting point within [p, q], h2 = 0.618 * (q - p), and obtain the mean square error MSE(h2) of the second splitting point;

[0091] Step (2.3), compare MSE(h1) and MSE(h2). If MSE(h1) < MSE(h2), the hidden layer node interval becomes [p, h2]; if MSE(h1) > MSE(h2), the hidden layer node interval becomes [h1, q]; if MSE(h1) = MSE(h2), the hidden layer node interval becomes [h1, h2];

[0092] Step (2.4), within the new interval generated in step (2.3), repeat the above steps (2.1)-(2.3). Until the optimal hidden layer node L on [p, q] is found, where the mean square error of point L satisfies that MSE(L) is the minimum root mean square error of all points on the interval [p, q].

[0093] Step (3), use the raccoon algorithm to optimize the BP neural network. The raccoon optimization algorithm is as Figure 6 shown.

[0094] Step (3.1), initialize the population. Calculate the number M of thresholds and weights to be optimized according to the number of neurons in the input layer, hidden layer, and output layer in the BP neural network, and calculate according to the following formula.

[0095] M = n × n1 + 2n + 1 + m × n1 + m

[0096] In the search space, N raccoon individuals are randomly generated according to the following formula. The dimension of each raccoon individual is M. The optimization of the threshold and weight is converted into the optimal solution for the position of the raccoon individual.

[0097]

[0098] Where: x i,j is the initial position of the individual; To find the optimal lower bound; is the upper bound of the optimization; r is a random number between [0,1].

[0099] In step (3.2), the optimization algorithm parameters are initialized, the maximum number of iterations, the upper and lower limits of the variable values, and the fitness function are determined. The training error in the BP neural network is used as the fitness function.

[0100] In step (3.3), the iguana's hunting and attack strategy (exploration phase), the first phase of the raccoon population in the search space is modeled based on a strategy that simulates their attack on iguanas. In this strategy, a group of raccoons climbs a tree to reach an iguana and scare it. Several other raccoons wait at the foot of the tree until the iguana falls to the ground. After the iguana falls to the ground, the raccoons attack and kill it. This strategy causes the raccoons to move to different locations in the search space, demonstrating the exploration ability of COA in solving global searches in the problem space.

[0101] In the COA algorithm, assuming the best position of the population member is the position of the iguana, half of the raccoons climb the tree and the other half wait for the iguana to fall to the ground. Therefore, the mathematical simulation of the position is:

[0102]

[0103] Where: is the new position of the i-th raccoon in the j-th dimension; r is a random number between [0,1]; G j is the position of the iguana in the jth dimension, which actually refers to the position of the best member; I is a number randomly selected from the set {1,2}; N is the number of raccoons; [N / 2] is the largest integer not exceeding [N / 2].

[0104] In step (3.4), the iguana falls to the ground and is placed at a random position in the search space. Based on this random position, the raccoon on the ground moves in the search space. The simulation is performed according to the following formula:

[0105]

[0106] Where: is the new position of the iguana on the ground in the jth dimension.

[0107]

[0108] Where: is the objective function value after the iguana falls to the ground in the jth dimension; F i,j is the objective function value of the i-th raccoon in the j-th dimension.

[0109] In step (3.5), if the updated individual position is better, update the current individual position; otherwise, keep it as it is:

[0110]

[0111] Where: is the objective function value of the i-th raccoon in the new position; F i is the objective function value of the i-th raccoon at its previous position, X i is the raccoon's current location.

[0112] Step (3.6), escape from the predator. The second stage of the process of updating the raccoon's position in the search space is mathematically modeled based on the raccoon's natural behavior when encountering and fleeing from a predator. When a predator attacks a raccoon, it flees from its position. The raccoon's moves in this strategy place it in a safe position close to its current position. This behavior is simulated based on the following equation:

[0113]

[0114] Where: is the local lower bound of the j-th decision variable; is the local upper bound of the j-th decision variable, t is the number of iterations; T is the maximum number of iterations.

[0115]

[0116] Where: is the new position of the i-th raccoon in the j-th dimension.

[0117] In step (3.7), the updated individual position is brought into the BP neural network, and the error between the predicted value and the true value is calculated using the test data set as the fitness evaluation. If the updated individual position is better, the current individual position is updated, otherwise it remains unchanged.

[0118]

[0119] The updated individual positions are the optimized thresholds and weights.

[0120] Step (4), yield prediction. The number of hidden layer nodes of the BP neural network is set according to the golden section algorithm. The optimal weights and thresholds between the input layer, hidden layer, and output layer obtained by the raccoon optimization algorithm are assigned to the BP neural network to complete the training of the BP neural network. The trained BP neural network is the required prediction model, and its output value is the final predicted value of grain yield.

[0121] It should be understood that although this specification is described according to various embodiments, not every embodiment contains only one independent technical solution. This narrative method of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment can also be appropriately combined to form other implementation methods that can be understood by those skilled in the art.

[0122] The series of detailed descriptions listed above are only specific descriptions of feasible implementation methods of the present invention. They are not intended to limit the scope of protection of the present invention. Any equivalent methods or changes that do not deviate from the technology of the present invention should be included in the scope of protection of the present invention.

Claims

1. Grain flow detection device based on raccoon algorithm, characterized in that, include: Device housing (6), central rotating shaft (8), blades (5), brushes (13), grain level detection sensor, blade rotation control device, grain flow data acquisition and control device; The central rotating shaft (8) is fixed to the central axis of the grain flow detection device (10), and the two ends are fixed to the device housing (6) using bearing seats (7); the blades (5) are fixedly connected to the central rotating shaft (8); the brushes (13) are fixed to the blades (5) and distributed in the gap between the blades (5) and the device housing (6), and when the central rotating shaft (8) rotates, the brushes (13) can rotate along with the blades (5); The device housing (6) is provided with a grain inlet (1) and a grain outlet (9), which are in opposite directions; The grain level detection sensor is located near the grain inlet (1) and is used to detect the height of grain accumulation in the detection device; a blade rotation control device for controlling the start and stop of the rotation of the blade (5), for allowing grain to accumulate in the device when the blade (5) stops rotating, and for releasing the accumulated grain when the blade (5) rotates; A blade rotation control device comprises an electromagnet (2) and a stopper (3) located at the end of the electromagnet; the electromagnet (2) is mounted on the device housing (6) and is located outside the grain collection area; a hole is opened on the device housing (6); one end of the electromagnet (2) with the stopper (3) is inserted into the hole; the stopper (3) is located inside the device housing (6); the electromagnet (2) is controlled by a grain flow data acquisition and control device so that the stopper (3) can move telescopically within the device housing; when the stopper (3) is extended, the blade (5) is prevented from rotating; when the stopper (3) is retracted, the blade (5) can rotate; The grain flow data acquisition and control device is used to control the working status of the entire detection device and calculate the grain flow.

2. The grain flow detection device based on the raccoon algorithm according to claim 1, characterized in that: The grain inlet (1) is surrounded by rectangular baffles.

3. The grain flow detection device based on the raccoon algorithm according to claim 1, characterized in that: The blades (5) are evenly distributed along the circumference of the central rotation axis (8), and the number of the blades (5) is three.

4. The grain flow detection device based on the raccoon algorithm according to claim 1, characterized in that: The grain level detection sensors are arranged in parallel and evenly along the axial direction of the central rotating shaft (8), and the straight line formed by the arrangement is parallel to the straight line where the central rotating shaft (8) is located.

5. The grain flow detection device based on the raccoon algorithm according to claim 4, characterized in that: The grain level detection sensor adopts capacitive proximity switches (4), the number of which is 3.

6. The grain flow detection device based on the raccoon algorithm according to claim 1, characterized in that: The grain flow data acquisition and control device includes: a voltage-stabilized power supply, a single-chip microcomputer, an LED display module, control buttons, a CAN communication module and a host computer; the single-chip microcomputer is connected to the LED display module, the voltage-stabilized power supply, the control buttons, the electromagnet and the CAN communication module, and the CAN communication module is connected to the host computer. The single-chip microcomputer transmits the acquired information to the host computer, and the host computer calculates the grain yield.

7. A method for predicting plot yield using the grain flow detection device according to any one of claims 1 to 5, characterized in that: After the grain is processed by the harvester, it flows out from the end of the clean grain elevator (14), enters the shell of the grain flow detection device from the grain inlet (1) of the detection device, and starts the acquisition work. The single-chip microcomputer is controlled by the control button to start timing, and the electromagnet (2) is controlled to make the stopper (3) block the rotation of the blade (5). The brush (13) is fixed at the edge part of the blade (5) and is distributed between the blade (5) and the device shell (6) to prevent the grain from flowing out through the gap between the blade (5) and the device shell (6) when collecting the grain. When a certain volume of grain is collected and the grain level rises to the grain level detection sensor, the sensor is triggered. The sensor sends a signal to the single-chip microcomputer, and the single-chip microcomputer stops timing. At the same time, a control signal is sent to the electromagnet (2) to control the electromagnet (2) to pull the stopper (3) to make the blade (5) break away from the block of the stopper (3). The blade (5) starts to rotate under the gravity of the grain, releasing the grain collected this time. At the same time, the electromagnet (2) is controlled to make the stopper (3) extend to block the rotation of the blade (5), block the next blade (5), and reset the timer. The timer starts timing and starts the next grain collection, working in a cycle. After receiving the signal from the grain level detection sensor, the single chip reads the data in the timer and saves it as the collection time t. The data is then transmitted to the host computer through the CAN communication module. The host computer combines the collection time t, the volume v of each grain collection, the bulk density ρ of the grain, the travel speed v of the combine harvester, the cutting width l of the harvester, the moisture content w of the grain, and the average loss rate P during the harvesting process. i , as the input of the BP neural network, outputs the predicted value of production, and displays the value through the LED display module.

8. The cell yield prediction method according to claim 7, characterized in that: The construction process of the BP neural network is as follows: Step 1, construct a BP neural network; the number of nodes in the input layer is 7, the number of nodes in the output layer is 1, and the number of nodes n1 in the hidden layer is determined by the following formula: where n is the number of nodes in the input layer, m is the number of nodes in the output layer, a is between 1 and 10, and the range of the number of nodes in the hidden layer is [3, 13] when determining the range; Step 2, optimize the number of nodes in the hidden layer of the BP neural network; including the following: Step 2.1, when the number of nodes in the hidden layer is within the range [p, q], find the first segmentation point h1 = 0.382*(q - p) within the range [p, q], and calculate the mean square error MSE(h1) of the first segmentation point. The calculation formula is as follows Where y i is the true value, is the predicted value, N is the number of samples; Step 2.2, find the second segmentation point h2 = 0.618*(q - p) within [p, q], and obtain the mean square error MSE(h2) of the second segmentation point; Step 2.3, compare MSE(h1) and MSE(h2). If MSE(h1) < MSE(h2), the hidden layer node interval becomes [p, h2]; if MSE(h1) > MSE(h2), the hidden layer node interval becomes [h1, q]; if MSE(h1) = MSE(h2), the hidden layer node interval becomes [h1, h2]; Step 2.4, in the new interval generated in Step 2.3, repeat the above Steps 2.1 - 2.3 until the optimal node L in the hidden layer on [p, q] is found, where the mean square error of point L satisfies that MSE(L) is the minimum root mean square error of all points on the interval [p, q].

9. The cell yield prediction method according to claim 8, characterized in that: The optimization of the BP neural network adopts the raccoon algorithm, specifically as follows: Step 3.1, population initialization. Calculate the number M of thresholds and weights to be optimized according to the number of neurons in the input layer, hidden layer and output layer in the BP neural network, and calculate according to the following formula M = n×n1 + 2n + 1 + m×n1 + m In the search space, N raccoon individuals are randomly generated according to the following formula. The dimension of each raccoon individual is M. The optimization of the threshold and weight is converted into the optimal solution for the position of the raccoon individual. Where: x i,j is the initial position of the individual; To find the optimal lower bound; is the upper boundary of the optimization; r is a random number between [0,1]; Step 3.2: Initialize the optimization algorithm parameters, determine the maximum number of iterations, upper and lower limits of variable values, and fitness function, and use the training error in the BP neural network as the fitness function; Step 3.3, exploration phase; assuming the position of the best member of the population is the position of the iguana, half of the raccoons climb the tree and the other half wait for the iguana to fall to the ground. Therefore, the position simulation is: Where: is the new position of the i-th raccoon in the j-th dimension, M is the dimension of each raccoon individual; r is a random number between [0,1]; G j is the position of the iguana in the jth dimension, which actually refers to the position of the best member; I is a number randomly selected from the set {1,2}; N is the number of raccoons; [N / 2] is the largest integer not exceeding [N / 2]; In step 3.4, after the iguana falls to the ground, it is placed at a random position in the search space. Based on this random position, the raccoon on the ground moves in the search space, simulated according to the following formula: Where: is the position of the iguana on the ground in the jth dimension; Where: is the objective function value after the iguana falls to the ground in the jth dimension; F i,j is the objective function value of the i-th raccoon in the j-th dimension; Step 3.5: If the updated individual is better, update the current individual; Otherwise keep it as is: Where: is the objective function value of the i-th raccoon in the new position; F i is the objective function value of the i-th raccoon at its previous position; Step 3.6, Escape from Predator. The second stage of the process of updating the raccoon's position in the search space is mathematically modeled based on the raccoon's natural behavior when encountering and fleeing from predators. When a predator attacks a raccoon, it flees from its position. The raccoon's moves in this strategy put it in a safe position close to its current position. This behavior is simulated based on the following equation: Where: is the local lower bound of the j-th decision variable; is the local upper bound of the jth decision variable, t is the number of iterations; T is the maximum number of iterations; Where: is the new position of the i-th raccoon in the j-th dimension; Step 3.7, bring the updated individual into the BP neural network, use the test data set to calculate the error between the predicted value and the true value as the fitness to judge, if the updated individual is better, update the current individual, otherwise keep it as it is

Images