Yield measurement method and system based on grain accumulation form online identification and measurement

Abstract

The invention discloses a yield measurement method and system based on grain accumulation form online identification and measurement, and relates to the field of agricultural machinery intelligence. Acquiring a first operation position, an operation speed and a header image of the current harvester, and calculating a current actual effective cutting width according to the header image; when the current output voltage of the proximity switch is a high level, collecting a grain image on a scraper of the elevator, constructing a grain three-dimensional accumulation model based on the grain image in combination with the reconstructed three-dimensional point cloud, and calculating the grain volume; obtaining the water content of the current grain, and converting the volume of the grain into the weight of the grain; and acquiring a real-time second operation position of the harvester, calculating an effective harvesting length in unit time and calculating a harvesting area in unit time by combining the first operation position and the operation speed, and further completing calculation of the grain quality and the yield of the land parcel. According to the invention, key technical support is provided for intelligent and precise upgrading of agricultural mechanization.

Description

Technical Field

[0001] This invention relates to the field of intelligent agricultural machinery, and more specifically to a yield measurement method and system based on online identification and measurement of grain stacking morphology. Background Technology

[0002] Current methods for measuring yield during agricultural machinery harvesting largely rely on traditional weighing or volume estimation, which suffer from poor real-time performance, low accuracy, and limited versatility. This is especially true in grain harvesting, where scraper conveyors, as crucial components for transporting grain, operate at high speeds and handle complex grain shapes, making it difficult to achieve real-time and accurate yield measurement using traditional methods.

[0003] With the development of image processing, 3D reconstruction, and multi-source sensor fusion technologies, new intelligent sensing methods have been provided for agricultural equipment. Currently, there is no publicly available technical literature that systematically combines scraper elevator image acquisition, 3D stacking modeling, bulk density and humidity estimation, and BeiDou positioning of cutting width information to achieve a complete technical solution for real-time yield measurement.

[0004] Therefore, how to solve the above-mentioned technical problems still needs to be further studied by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a yield measurement method and system based on online identification and measurement of grain stacking morphology.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A yield measurement method based on online identification and measurement of grain stacking morphology includes the following steps:

[0008] Acquire the current first working position, working speed, and header image of the harvester, and calculate the current actual effective cutting width based on the header image;

[0009] When the output voltage of the proximity switch is high, the image of the grain on the scraper of the elevator is acquired. Based on the grain image and the reconstructed 3D point cloud, a 3D grain stacking model is constructed, and the grain volume is calculated.

[0010] Obtain the current moisture content of the grain and convert the grain volume into the grain weight;

[0011] The system obtains the real-time second operating position of the harvester, combines it with the first operating position and operating speed, calculates the effective harvest length per unit time, calculates the harvest area per unit time, and then completes the calculation of grain quality and plot yield.

[0012] Optionally, the output voltage U of the current header height sensor can be collected. Header Compare it with the set threshold U HeadersThe relationship; if U Header >U Headers This indicates that the header is not lowered and the harvester has not started harvesting operations; if U Header ≤U Headers This indicates that the header is down and the harvester is in the process of harvesting. When the harvester is in the process of harvesting, the output information of the Beidou positioning system is read to obtain the current first working position and working speed of the harvester.

[0013] Optionally, the calculation of the current actual effective cutting width is as follows: the front-end vision sensor captures image information of the front end of the header, and through binarization, bilateral filtering, and edge detection, the number of pixels P between the headers of the harvester is obtained. Header The number of pixels P of grain between the harvester's cutterhead Grain According to the design width W of the harvester header Header Calculate the current actual effective cutting width W Grain :

[0014]

[0015] Optionally, when the output voltage of the current proximity switch is high, an image of the grain on the scraper of the elevator is acquired. The specific process is as follows:

[0016] Collect the current output voltage U of the proximity switch proswitch ;

[0017] If U proswitch A low level indicates that there is no grain on the elevator scraper and the grain has not yet entered the elevator.

[0018] If U proswitch A high level indicates that there is grain on the elevator scraper, and the image is captured to measure the volume of grain accumulation on the scraper.

[0019] The system synchronously acquires images of grain on the scraper of the elevator. Each frame of the image is recorded with a system timestamp. Camera-1 is selected as the "master reference". The image frames with the closest timestamps are selected from the frame buffers of other cameras. The time difference threshold Tthreshold is set to ±10ms. If the time difference of a frame is too large, the frame is discarded and the next acquisition round begins. If the time difference is less than the set threshold, a group of "synchronous frames" is formed for image processing.

[0020] Optionally, a three-dimensional grain stacking model is constructed based on the grain image and the reconstructed three-dimensional points, specifically including the following steps:

[0021] An improved grain depth-texture fusion matching algorithm with adaptive dynamic matching is used for scraper grain point cloud reconstruction. Initialization is guided by depth prior, and the initial candidate depth d0 is estimated using the coarse geometric relationships between multiple cameras to narrow the search space.

[0022]

[0023] d coare (x i ) is the initial depth estimated by the camera pose / geometric model, and δ is the range of random perturbation;

[0024] For grain edges or overlapping areas, the matching window size is adaptively adjusted.

[0025] S i =S base ·(1+α·σ I (x i ))

[0026] S base Based on the block size, ·σ I (x i ) represents the image at pixel x i The ambient brightness variance, where α is an adjustment factor;

[0027] By introducing a weighted matching function that combines cost confidence and edge enhancement, a depth matching cost function is constructed:

[0028] C(x,d)=ω c ·C SAD (x,d)+ω g G(x,d)

[0029] C SAD (x,d) represents the original block matching cost, G(x,d) represents the image gradient difference, and ω represents the image gradient difference. c ω g To dynamically adjust the weights;

[0030] Construct a sparse support point graph, and use the propagation confidence related to the distance between support points in each round of propagation:

[0031]

[0032] By introducing a local motion prediction model and grain texture aggregation constraints, the point cloud reconstruction accuracy is enhanced under complex textures and overlapping occlusion conditions. At the same time, a three-frame sliding window mechanism is adopted to extract stable visual anchor points in consecutive image frames to achieve dynamic point cloud completion.

[0033] After matching is complete, the depth value d(x,y) of each pixel is obtained, and the 3D point coordinates are reconstructed using the camera model:

[0034]

[0035] K is the camera intrinsic parameter matrix, and P(x,y) is the corresponding 3D point coordinate. By combining the reconstructed 3D points through a multi-view system, a point cloud map of grain accumulation on the scraper is formed.

[0036] The acquired point cloud data is discretized and mapped to unit voxel grids to construct a 3D grain stacking model. A classifier is used to identify and remove "grain boundary voxels" and "background noise voxels" to improve the accuracy of volume estimation. The volume of grain within the scraper corresponding to the frame image is calculated by integrating the number of voxels, where each voxel is a cube with a side length of v. VoxelGrid The total effective cereal gluten count is N. grain Then the total volume of grain V Grain for:

[0037]

[0038] Optionally, the moisture content of the current grain can be obtained. Specifically, the moisture content of the current grain can be obtained as follows:

[0039] Read the data from the moisture sensor, collect the digital value output by the current moisture sensor, and obtain the real-time moisture content R of the grain at the current moment. water ;

[0040] If R waterh A reading of 0 indicates that the moisture sensor is not in contact with the grain;

[0041] If R waterh The reading is non-zero, indicating that the moisture sensor is in contact with the grain. This is consistent with the grain bulk density model k(R) calibrated before harvest. water The volume data of the grain on the scraper is converted into the weight m of the grain on that scraper. ScrGrain :

[0042] m ScrGrain =V Grain .k(R Water ).

[0043] Optionally, the real-time second operating position of the harvester is obtained, and combined with the first operating position and operating speed, the effective harvested length per unit time and the harvested area per unit time are calculated, thereby completing the calculation of grain quality and plot yield. The specific calculation process is as follows:

[0044] The system acquires the real-time location information of the harvester, identifies the valid coordinates within the location information per unit time, eliminates duplicate location coordinates that have already been calculated through tag cyclic mapping, and calculates the effective harvest length L per unit time using the continuous coordinate difference method. Grain (t):

[0045]

[0046] Actual harvested area A per unit time Grain (t):

[0047] A Grain (t)=L Grain (t).W Grain (t);

[0048] The cumulative grain mass is M ScrGrain :

[0049]

[0050] Using a sliding window and Kalman filtering, the area coverage and quality estimation are smoothed to obtain the yield of the plot per unit time, Y. Grai :

[0051]

[0052] A yield measurement system based on online identification and measurement of grain stacking morphology, utilizing any one of the yield measurement methods based on online identification and measurement of grain stacking morphology, includes a header height sensor, a smart terminal, a front-end vision sensor, a Beidou positioning antenna, a heterogeneous image acquisition module, a moisture sensor, a proximity switch, a scraper, and a Hall effect speed sensor.

[0053] The header height sensor is used to monitor the header height of the harvester. When the sensor output voltage is lower than the set threshold, it indicates that the harvester is performing harvesting operations.

[0054] The intelligent terminal is used to collect data from all sensors during the harvester's operation and run relevant algorithms to achieve online analysis of yield, display, and storage of data from the operation process;

[0055] A front-end vision sensor is used to acquire image information of the front end of the header, and subsequent image analysis is used to obtain the actual harvest width during the harvesting process;

[0056] The Beidou positioning antenna is used to receive Beidou satellite signals, and after data processing, the real-time location and speed of the harvester are obtained.

[0057] The heterogeneous image acquisition module is used to acquire real-time images of the grain on the scraper in the elevator, and then use 3D reconstruction to obtain the accumulated volume of the grain on the scraper.

[0058] Moisture sensor is used to monitor the real-time moisture content of grains in the elevator;

[0059] A proximity switch is used to detect whether there is grain on the scraper.

[0060] Scrapers, used in elevators to load grains, lifting them from the bottom to the top;

[0061] Hall effect speed sensor is used to detect the real-time speed of the elevator.

[0062] Optionally, the heterogeneous image acquisition module includes four cameras, one structured light generator, and four LEDs. The cameras are used to acquire images from different angles, the structured light generator is used for grain positioning analysis and auxiliary modeling, and the LEDs are used for supplementary lighting.

[0063] As can be seen from the above technical solutions, compared with the prior art, the present invention discloses a yield measurement method and system based on online identification and measurement of grain stacking morphology. Through multimodal perception and intelligent modeling, it solves the core pain points of traditional yield monitoring such as low accuracy, poor environmental adaptability and high maintenance costs, and provides key technical support for the upgrading of agricultural mechanization to intelligence and precision. Attached Figure Description

[0064] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0065] Figure 1 This is a schematic diagram of the harvester yield measurement system of the present invention;

[0066] Figure 2 This is a schematic diagram illustrating the effective cutting width calculation of the present invention;

[0067] Figure 3 This is a flowchart of the production measurement system of the present invention;

[0068] Among them, 1. Cutting table height sensor; 2. Smart terminal; 3. Front-end vision sensor; 4. Beidou positioning antenna; 5. Heterogeneous image acquisition module; 6. Moisture sensor; 7. First LED light; 8. First camera; 9. Proximity switch; 10. Scraper; 11. Hall speed sensor; 12. Second LED light; 13. Second camera; 14. Third LED light; 15. Structured light generator; 16. Third camera; 17. Fourth camera; 18. Fourth LED light. Detailed Implementation

[0069] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0070] Example 1

[0071] This invention discloses a yield measurement method and system based on online identification and measurement of grain stacking morphology, wherein the harvester yield measurement system is as follows: Figure 1 As shown, it mainly consists of a cutting table height sensor 1, a smart terminal 2, a front-end vision sensor 3, a Beidou positioning antenna 4, a heterogeneous image acquisition module 5, a moisture sensor 6, a proximity switch 9, a scraper 10, and a Hall effect speed sensor 11.

[0072] The header height sensor 1 is used to monitor the header height of the harvester. When the sensor output voltage is lower than the set threshold, it indicates that the harvester is performing harvesting operations.

[0073] The intelligent terminal 2 is used to collect data from all sensors during the harvester's operation and run relevant algorithms to achieve online analysis of yield, display, and storage of data from the operation process;

[0074] Front-end vision sensor 3 is used to acquire image information of the front end of the header, and subsequent image analysis is used to obtain the actual harvest width during the harvesting process.

[0075] Beidou positioning antenna 4 is used to receive Beidou satellite signals, and after data processing, the real-time location and speed of the harvester are obtained;

[0076] The heterogeneous image acquisition module is used to acquire real-time images of the grain on the scraper 10 in the elevator, and then to obtain the accumulation volume of the grain on the scraper 10 through 3D reconstruction. The mechanism mainly consists of 4 cameras, 1 structured light generator 15, and 4 LED lights. The LED lights include a first LED light 7, a second LED light 12, a third LED light 14, and a fourth LED light 18. The cameras include a first camera 8, a second camera 13, a third camera 16, and a fourth camera 17. The cameras are used to acquire images from different angles, the structured light generator 15 is used for grain positioning analysis and auxiliary modeling, and the LEDs are used for supplementary lighting.

[0077] Moisture sensor 6 is used to monitor the real-time moisture content of the grain in the elevator;

[0078] Proximity switch 9 is used to detect whether there is grain on scraper 10;

[0079] Scraper 10, used in the elevator to load grain, so that the grain is lifted from the bottom to the top;

[0080] Hall effect speed sensor 11 is used to detect the real-time speed of the elevator.

[0081] like Figure 3 As shown, the specific workflow of the yield measurement method is as follows:

[0082] (1) Determine whether the harvester is in harvesting mode.

[0083] Read the data from the header height sensor and acquire the current output voltage U of the header height sensor. Header Compare it with the set threshold U Headers The relationship.

[0084] ①If U Header >U Headers This indicates that the header was not lowered and the harvester was not performing harvesting operations.

[0085] Continue to collect the output signal from the header height sensor.

[0086] ②If U Header ≤U Headers This indicates that the header has been lowered and the harvester is in the process of harvesting.

[0087] The system reads the output information from the BeiDou positioning system to obtain the current operating position and speed of the harvester.

[0088] like Figure 2 As shown, the front-end vision sensor captures image information of the front end of the header, and obtains the number of pixels P between the headers of the harvester through binarization, bilateral filtering, and edge detection. Header The number of pixels P of grain between the harvester's cutterhead Grain According to the design width W of the harvester header Header Calculate the current actual effective cutting width W (unit: m). Grain (t), (unit, m):

[0089]

[0090] (2) Determine if there is grain on the scraper of the harvester's elevator.

[0091] Read proximity switch data and acquire the current output voltage U of the proximity switch. proswitch

[0092] ①If U proswitch A low level indicates that there is no grain on the elevator scraper and the grain has not yet entered the elevator.

[0093] Continue to acquire the output signal of the proximity switch.

[0094] ②If U proswitch A high level indicates that there is grain on the conveyor scraper, and it is necessary to start taking pictures to measure the volume of grain accumulation on the scraper.

[0095] Four cameras in the heterogeneous image acquisition module synchronously acquire images of grain on the conveyor scraper. Each frame from the four cameras records a system timestamp. Camera -1 in the heterogeneous image acquisition module is selected as the "master reference," and image frames with the closest timestamps are selected from the frame buffers of the other three cameras. A time difference threshold T is set. threshold The time difference is ±10ms. If the time difference of a frame is too large, the frame is discarded and the next frame is collected. If the time difference is less than the set threshold, a set of "synchronization frames" is formed for image processing.

[0096] Using "synchronized frame" images, an adaptive dynamic matching grain depth-texture fusion matching algorithm is employed to reconstruct the point cloud of scraped grain. First, initialization is guided by depth priors, and the initial candidate depth d0 is estimated using the coarse geometric relationships between multiple cameras to narrow the search space.

[0097]

[0098] d coare (x i The initial depth is estimated by the camera pose / geometric model and is the range of random perturbations.

[0099] For grain edges or overlapping areas, the matching window size is adaptively adjusted.

[0100] S i =S base ·(1+α·σ I (x i ))

[0101] S base Based on the block size, σ I (x i ) represents the image at pixel x i The ambient brightness variance, α is an adjustment coefficient that controls the response speed.

[0102] By introducing a weighted matching function that combines cost confidence and edge enhancement, a depth matching cost function is constructed:

[0103] C(x,d)=ω c ·C SAD (x,d)+ω g G(x,d)

[0104] C SAD (x,d) represents the original block matching cost, G(x,d) represents the image gradient difference, which controls the penalty for discontinuous edge regions, and ω c ω g To dynamically adjust the weights, adjustments are made based on factors such as the current pixel gradient magnitude and texture complexity.

[0105] Construct a sparse support point graph, and use the propagation confidence related to the distance between support points in each round of propagation:

[0106]

[0107] Pixels closer to reliable feature points have higher propagation confidence, reducing error propagation.

[0108] By introducing a local motion prediction model and grain texture clustering constraints, the accuracy of point cloud reconstruction under complex textures and overlapping occlusion conditions is enhanced. Simultaneously, a three-frame sliding window mechanism is employed to extract stable visual anchor points in consecutive image frames, achieving dynamic point cloud completion.

[0109] After matching is complete, the depth value d(x,y) of each pixel is obtained, and the 3D point coordinates are reconstructed using the camera model:

[0110]

[0111] K is the camera intrinsic parameter matrix, and P(x,y) is the corresponding 3D point coordinate. By combining the reconstructed 3D points through a multi-view system, a point cloud map of grain accumulation on the scraper is formed.

[0112] The acquired point cloud data is discretized and mapped to unit voxel grids to construct a 3D grain stacking model. A Random Forest Grain classifier is used to identify and remove "grain boundary voxels" and "background noise voxels," improving volume estimation accuracy. Finally, the volume of grain within the scraper corresponding to the frame image is calculated using the number of integrated voxels; each voxel is a cube with a side length of v. VoxelGrid (Unit: m), the total effective cereal voxels are N grain Then the total volume of grain V Grain (Unit: m) 3 )for:

[0113]

[0114] (3) Determine the real-time moisture content of grains

[0115] Read the data from the moisture sensor, collect the digital value output by the current moisture sensor, and obtain the real-time moisture content R of the grain at the current moment. water .

[0116] ①If R waterh A reading of 0 indicates that the moisture sensor is not in contact with the grain; continue collecting data from the moisture sensor.

[0117] ②If R waterh The reading is a non-zero value of 0, indicating that the moisture sensor is in contact with the grain. This is consistent with the grain bulk density model k(R) calibrated before harvest. water (Based on real-time moisture content R)water The calculated dynamic bulk density of grain, kg / m³ 3 The volume data of the grain on the scraper is converted into the weight m of the grain on that scraper. ScrGrain (Unit: kg)

[0118] m ScrGrain =V Grain .k(R Water )

[0119] (4) Grain mass of scraper is mapped to yield per unit area

[0120] The real-time location information of the harvester is obtained using the BeiDou positioning system. The valid coordinates within the location information per unit time are determined. Repeatedly calculated coordinates are eliminated through tag cyclic mapping. The effective harvest length L per unit time is calculated using the continuous coordinate difference method. Grain (t), (unit, m):

[0121]

[0122] Actual harvested area A per unit time Grain (t), (unit: m) 2 ):

[0123] A Grain (t)=L Grain (t).W Grain (t)

[0124] Harvesting grain within this work area requires N Scraper Only with a scraper conveyor can the grain be completely transported into the grain silo, with a cumulative grain mass of M. ScrGrain (Unit: kg)

[0125]

[0126] Then, a sliding window and Kalman filter are used to smooth the area coverage and quality estimation, yielding the plot's yield Y per unit time. Grain (Unit: kg / ha)

[0127]

[0128] Example 2

[0129] The harvester yield measurement system uses an AA-ROT-120 height sensor manufactured by Wenzhou Jinxing Auto Parts Co., Ltd., and the intelligent terminal uses an APOLLO sensor manufactured by Beijing Congping Technology Co., Ltd. The 12PRO all-in-one large-screen rugged vehicle-mounted smart tablet and the GEAC90TAI edge computing controller from Suzhou Tianzhun Technology Co., Ltd. are used. The front-end vision sensor uses the RER-USB4KCAM30H industrial camera from Shenzhen Ruier Vision Technology Co., Ltd., the Beidou positioning antenna uses the XYZ-GNSS mushroom-head high-precision RTK measurement antenna from Shenzhen Yonghao Innovation Technology Co., Ltd., the heterogeneous image acquisition model uses the SHY01 heterogeneous image acquisition module designed by the Nanjing Institute of Agricultural Mechanization of the Ministry of Agriculture and Rural Affairs, the moisture sensor uses the Skc01 grain moisture sensor designed by the Nanjing Institute of Agricultural Mechanization of the Ministry of Agriculture and Rural Affairs, the proximity switch uses the E2E-X5E1 Omron proximity switch from Shanghai Weisite Automation Co., Ltd., the Hall speed sensor uses the M5M8M12M18 Hall proximity switch from Chongqing Shengyixin Electronic Technology Co., Ltd., and the grain combine harvester uses the CM100 Gushen grain harvester from Weichai Lovol Smart Agriculture Technology Co., Ltd.

[0130] For rice harvesting, before the CM100 Grain God harvester begins harvesting, the harvester's yield measurement system is activated. First, the system initializes, establishes communication with each sensor, acquires information from each sensor, and reads the output voltage U from the header sensor. Header Compare its relationship with the set threshold of 3.7.

[0131] If U Header >3.7 indicates the header is not lowered and the harvester is not performing harvesting operations. Continue collecting the output signal from the header height sensor. If U Header ≤3.7 indicates that the header is down and the harvester is in the process of harvesting. The system reads the output information from the Beidou positioning system to obtain the current operating position and speed of the harvester.

[0132] The front-end vision sensor captures image information of the front end of the header, and obtains the number of pixels P between the headers of the harvester through binarization, bilateral filtering, and edge detection. Header The number of pixels P of rice between the harvester's cutter table Grain The CM100 Grain God grain harvester is equipped with a header with a width of W. Header =2.67m, calculate the current actual effective cutting width W Grain (t):

[0133]

[0134] Read proximity switch data and acquire the current output voltage U of the proximity switch. proswitch If Uproswitch =0, indicating there are no rice grains on the elevator scraper, and the rice grains have not yet entered the elevator. Continue to collect the output signal of the proximity switch.

[0135] If U proswitch =3.7V indicates there is grain on the elevator scraper, and image capture needs to begin to measure the volume of rice accumulation on the scraper. Four cameras in the heterogeneous image acquisition module simultaneously capture images of the rice on the elevator scraper. Each frame from the four cameras records a system timestamp. Camera -1 in the heterogeneous image acquisition module is selected as the "master reference," and image frames with the closest timestamps are selected from the frame buffers of the other three cameras. A time difference threshold T is set. threshold The time difference is ±10ms. If the time difference of a frame is too large, the frame is discarded and the next frame is collected. If the time difference is less than the set threshold, a set of "synchronization frames" is formed for image processing.

[0136] Using synchronized frame images, an adaptive dynamic matching grain depth-texture fusion matching algorithm was employed to reconstruct the point cloud of scraped rice. The acquired point cloud data was discretized and mapped to unit voxel grids to construct a 3D stacking model of rice. A Random Forest Grain classifier was used to identify and remove "rice boundary voxels" and "background noise voxels," improving the accuracy of volume estimation. Finally, the volume of rice within the scraper corresponding to the frame image was calculated by integrating the number of voxels; each voxel is a cube with a side length of v. VoxelGrid The total effective rice voxels are N grain Then the total volume of rice, V Grain for:

[0137]

[0138] Read the data from the moisture sensor, collect the digital value output by the current moisture sensor, and obtain the real-time moisture content R of the rice at the current moment. water If R waterh =0, indicating that the moisture sensor is not in contact with the rice grains; continue collecting moisture sensor data. If R waterh A reading of 0 indicates that the moisture sensor has come into contact with the rice grains. Based on the rice grain bulk density model calibrated before harvest, the grain volume data on the scraper is converted into the weight (m) of the rice on that scraper. ScrGrain :

[0139] m ScrGrain =V Grain .(3.0206·R Water +533.75)

[0140] The real-time location information of the harvester is obtained using the BeiDou positioning system. The valid coordinates within the location information per unit time are determined. Repeatedly calculated coordinates are eliminated through tag cyclic mapping. The effective harvest length L per unit time is calculated using the continuous coordinate difference method. Grain (t), (unit, m):

[0141]

[0142] Actual harvested area A per unit time Grain (t), (unit: m) 2 ):

[0143] A Grain (t)=L Grain (t).W Grain (t)

[0144] Harvesting grain within this work area requires N Scraper Only with a scraper conveyor can the grain be completely transported into the grain silo, with a cumulative grain mass of M. ScrGrain (Unit: kg)

[0145]

[0146] Then, a sliding window and Kalman filter are used to smooth the area coverage and quality estimation, yielding the plot's yield Y per unit time. Grain (Unit: kg / ha)

[0147]

[0148] This method is also applicable to the harvesting of crops such as wheat, corn, and soybeans.

[0149] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0150] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A yield measurement method based on online identification and measurement of grain stacking morphology, characterized in that, Includes the following steps: Acquire the current first working position, working speed, and header image of the harvester, and calculate the current actual effective cutting width based on the header image; When the output voltage of the proximity switch is high, the image of the grain on the scraper of the elevator is acquired. Based on the grain image and the reconstructed 3D point cloud, a 3D grain stacking model is constructed, and the grain volume is calculated. Obtain the current moisture content of the grain and convert the grain volume into the grain weight; The system obtains the real-time second operating position of the harvester, combines it with the first operating position and operating speed, calculates the effective harvest length per unit time, calculates the harvest area per unit time, and then completes the calculation of grain quality and plot yield.

2. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, Collect the output voltage U of the current header height sensor Header Compare it with the set threshold U Headers The relationship; if U Header >U Headers This indicates that the header is not lowered and the harvester has not started harvesting operations; if U Header ≤U Headers This indicates that the header is down and the harvester is in the process of harvesting. When the harvester is in the process of harvesting, the output information of the Beidou positioning system is read to obtain the current first working position and working speed of the harvester.

3. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, The calculation of the current actual effective cutting width is as follows: The front-end vision sensor captures image information of the front end of the header, and through binarization, bilateral filtering, and edge detection, the number of pixels P between the headers of the harvester is obtained. Header The number of pixels P of grain between the harvester's cutterhead Grain According to the design width W of the harvester header Header Calculate the current actual effective cutting width W Grain :

4. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, The current proximity switch output voltage is high, and the image of the grain on the elevator scraper is acquired. The specific process is as follows: Collect the current output voltage U of the proximity switch proswitch ; If U proswitch A low level indicates that there is no grain on the elevator scraper and the grain has not yet entered the elevator. If U proswitch A high level indicates that there is grain on the elevator scraper, and the image is captured to measure the volume of grain accumulation on the scraper. The system synchronously acquires images of grain on the scraper of the elevator. Each frame of the image is recorded with a system timestamp. Camera-1 is selected as the "master reference". The image frames with the closest timestamps are selected from the frame buffers of other cameras. The time difference threshold Tthreshold is set to ±10ms. If the time difference of a frame is too large, the frame is discarded and the next acquisition round begins. If the time difference is less than the set threshold, a group of "synchronous frames" is formed for image processing.

5. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, A 3D grain stacking model is constructed based on grain images combined with reconstructed 3D point clouds, specifically including the following steps: An improved grain depth-texture fusion matching algorithm with adaptive dynamic matching is used for scraper grain point cloud reconstruction. Initialization is guided by depth prior, and the initial candidate depth d0 is estimated using the coarse geometric relationships between multiple cameras to narrow the search space. d coare (x i ) is the initial depth estimated by the camera pose / geometric model, and δ is the range of random perturbation; For grain edges or overlapping areas, the matching window size is adaptively adjusted. S i =S base ·(1+a·s I (x i )) S base Based on the block size, ·σ I (x i ) represents the image at pixel x i The ambient brightness variance, where α is an adjustment factor; By introducing a weighted matching function that combines cost confidence and edge enhancement, a depth matching cost function is constructed: C(x,d)=ω c ·C SAD (x,d)+ω g G(x,d) C SAD (x,d) represents the original block matching cost, G(x,d) represents the image gradient difference, and ω represents the image gradient difference. c ω g To dynamically adjust the weights; Construct a sparse support point graph, and use the propagation confidence related to the distance between support points in each round of propagation: By introducing a local motion prediction model and grain texture aggregation constraints, the point cloud reconstruction accuracy is enhanced under complex textures and overlapping occlusion conditions. At the same time, a three-frame sliding window mechanism is adopted to extract stable visual anchor points in consecutive image frames to achieve dynamic point cloud completion. After matching is complete, the depth value d(x,y) of each pixel is obtained, and the 3D point coordinates are reconstructed using the camera model: K is the camera intrinsic parameter matrix, and P(x,y) is the corresponding 3D point coordinate. By combining the reconstructed 3D points through a multi-view system, a point cloud map of grain accumulation on the scraper is formed. The acquired point cloud data is discretized and mapped to unit voxel grids to construct a 3D grain stacking model. A classifier is used to identify and remove "grain boundary voxels" and "background noise voxels" to improve the accuracy of volume estimation. The volume of grain within the scraper corresponding to the frame image is calculated by integrating the number of voxels, where each voxel is a cube with a side length of v. VoxelGrid The total effective cereal creatinine content is N. grain Then the total volume of grain V Grain for:

6. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, To obtain the current moisture content of the grain, the specific steps are as follows: Read the data from the moisture sensor, collect the digital value output by the current moisture sensor, and obtain the real-time moisture content R of the grain at the current moment. water ; If R waterh A reading of 0 indicates that the moisture sensor is not in contact with the grain; If R waterh The reading is non-zero, indicating that the moisture sensor is in contact with the grain. This is consistent with the grain bulk density model k(R) calibrated before harvest. water The volume data of the grain on the scraper is converted into the weight m of the grain on that scraper. ScrGrain : m ScrGrain =V Grain .k(R Water )。 7. The yield measurement method based on online identification and measurement of grain stacking morphology according to claim 1, characterized in that, The real-time second operating position of the harvester is obtained. Combined with the first operating position and operating speed, the effective harvested length per unit time and the harvested area per unit time are calculated. This leads to the calculation of grain quality and plot yield. The specific calculation process is as follows: The system acquires the real-time location information of the harvester, identifies the valid coordinates within the location information per unit time, eliminates duplicate location coordinates that have already been calculated through tag cyclic mapping, and calculates the effective harvest length L per unit time using the continuous coordinate difference method. Grain (t): Actual harvested area A per unit time Grain (t): A Grain (t)=L Grain (t).W Grain (t); The cumulative grain mass is M ScrGrain : Using a sliding window and Kalman filtering, the area coverage and quality estimation are smoothed to obtain the yield of the plot per unit time, Y. Grai :

8. A yield measurement system based on online identification and measurement of grain stacking morphology, utilizing the yield measurement method based on online identification and measurement of grain stacking morphology as described in any one of claims 1-7, characterized in that, This includes a cutting table height sensor, a smart terminal, a front-end vision sensor, a Beidou positioning antenna, a heterogeneous image acquisition module, a moisture sensor, a proximity switch, a scraper, and a Hall effect speed sensor. The header height sensor is used to monitor the header height of the harvester. When the sensor output voltage is lower than the set threshold, it indicates that the harvester is performing harvesting operations. The intelligent terminal is used to collect data from all sensors during the harvester's operation and run relevant algorithms to achieve online analysis of yield, display, and storage of data from the operation process; A front-end vision sensor is used to acquire image information of the front end of the header, and subsequent image analysis is used to obtain the actual harvest width during the harvesting process; The Beidou positioning antenna is used to receive Beidou satellite signals, and after data processing, the real-time location and speed of the harvester are obtained. The heterogeneous image acquisition module is used to acquire real-time images of the grain on the scraper in the elevator, and then use 3D reconstruction to obtain the accumulated volume of the grain on the scraper. Moisture sensor is used to monitor the real-time moisture content of grains in the elevator; A proximity switch is used to detect whether there is grain on the scraper. Scrapers, used in elevators to load grains, lifting them from the bottom to the top; Hall effect speed sensor is used to detect the real-time speed of the elevator.

9. A yield measurement system based on online identification and measurement of grain stacking morphology according to claim 8, characterized in that, The heterogeneous image acquisition module includes four cameras, one structured light generator, and four LEDs. The cameras are used to acquire images from different angles, the structured light generator is used for grain positioning analysis and auxiliary modeling, and the LEDs are used for supplementary lighting.

Images