Unmanned agricultural machine operation data analysis method and system based on three-dimensional scene reconstruction

Abstract

The application discloses an unmanned agricultural machine operation data analysis method and system based on three-dimensional scene reconstruction, relates to the field of unmanned agricultural machines, and constructs a farm three-dimensional scene model through field reconnaissance, grid division, point cloud data collection by laser scanning, data processing, maps unmanned agricultural machine trajectories and poses in the three-dimensional scene model, realizes time sequence correlation through timestamp marking, divides operation intervals according to historical multi-modal data, determines operation types in combination with a weighted algorithm, evaluates operation results by taking operation coverage and spraying uniformity as indexes, adjusts parameters for substandard operation, finally sets an adjustment period, constructs a knowledge graph, adjusts operation parameters according to the knowledge graph, realizes intelligent monitoring of unmanned agricultural machine operation, effectively guarantees the operation quality of the unmanned agricultural machine, and improves the operation efficiency of the unmanned agricultural machine.

Description

Technical Field

[0001] This invention relates to the field of unmanned agricultural machinery, specifically to a method and system for analyzing unmanned agricultural machinery operation data based on three-dimensional scene reconstruction. Background Technology

[0002] In the field of modern unmanned agricultural machinery operations, unmanned agricultural machinery is a type of unmanned aerial vehicle that integrates navigation, intelligent monitoring, intelligent control and precision operation technologies. It can perform sowing, irrigation and fertilization operations in the air without human operation. Precision and intelligent management are the core requirements for improving operational efficiency and ensuring crop growth quality. In traditional unmanned agricultural machinery farm operations, it is mostly necessary to manually survey the farm boundaries and crop distribution and manually plan the operation trajectory. This is not only inefficient, but also prone to errors due to human factors. At the same time, it lacks accurate judgment of different operation types, which can easily lead to misjudgment and affect the evaluation results.

[0003] Existing evaluation indicators lack quantitative standards, making it impossible to scientifically measure operational results. When operations fail to meet standards, parameter adjustments rely heavily on experience, resulting in unstable optimization effects. Furthermore, operational data is not accumulated in a structured manner, requiring repeated parameter adjustments for each operation. The lack of reuse of historical experience leads to low efficiency. These problems make it difficult for unmanned agricultural machinery operations to meet the needs of large-scale, high-precision agricultural production. Therefore, a systematic solution is urgently needed. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for analyzing unmanned agricultural machinery operation data based on three-dimensional scene reconstruction, so as to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for analyzing unmanned agricultural machinery operation data based on three-dimensional scene reconstruction, the method comprising the following steps:

[0006] Step S1: Use terrestrial laser scanning technology to scan the farm and construct a three-dimensional scene model;

[0007] Step S1-1: Conduct an on-site survey of the farm boundaries, crop distribution, and fixed obstacles;

[0008] Step S1-2: Draw a farm site survey sketch based on the field survey data, and mark crops, boundaries and fixed obstacles;

[0009] Steps S1-3: Divide the scanning area according to the grid division method. Divide the farm into multiple sub-grids according to the "a×a" grid and mark the center point of each grid. Set a target at the center point. a is the side length of the sub-grid, which is determined according to the unmanned agricultural machinery operation range.

[0010] Steps S1-4: Collect high-density point cloud data for each sub-region, use statistical filtering algorithm to remove discrete noise points from the point cloud data, then use voxel grid downsampling method to downsample the data, and finally use kd tree to remove duplicate points from the point cloud.

[0011] Step S1-5: Based on the center point coordinates of the target, the preprocessed point clouds of multiple sub-regions are stitched together and their coordinates are unified using the target matching stitching method.

[0012] Steps S1-6: Import the assembled point cloud model into 3D modeling software to construct a 3D scene model of the farm.

[0013] By conducting on-site surveys to clarify farm boundaries, crop distribution, and fixed obstacles, and by combining grid division with the setting of targets to avoid missing key areas, a benchmark was provided for the construction of a 3D scene model. Data processing provided reliable model support for subsequent unmanned agricultural machinery trajectory mapping.

[0014] Step S2: Map the pose and working trajectory of the unmanned agricultural machinery into the 3D scene model;

[0015] Step S2-1: Use the National Geodetic Coordinate System CGCS2000 to record the coordinate points of the 3D scene model, use the BeiDou positioning geographic coordinate system to record the coordinate points of the unmanned agricultural machinery, and construct the unmanned agricultural machinery track data through the coordinate points of the unmanned agricultural machinery.

[0016] Step S2-2: Clean and standardize the coordinate points and attitude angles in the unmanned agricultural machinery track data, and timestamp the coordinate points and attitude angles in the track data.

[0017] Step S2-3: Map the coordinate points of the unmanned agricultural machinery track data to the 3D model, and match the attitude angle data of the unmanned agricultural machinery with the coordinate points of the track data through timestamps.

[0018] The operation trajectory of unmanned agricultural machinery is visualized, which facilitates real-time monitoring of the operation status of unmanned agricultural machinery. The data is marked with timestamps, which provides the basic data with correct time sequence for subsequent steps.

[0019] Step S3: Based on the multimodal operation data of unmanned agricultural machinery, divide the data into different intervals to determine the operation behavior of the unmanned agricultural machinery;

[0020] Step S3-1: Collect multimodal data on the normal operation of historical unmanned agricultural machinery, perform data cleaning and standardization on the multimodal data, use filtering algorithms to eliminate environmental interference, use linear interpolation to process missing values, and use box plots to identify outliers.

[0021] Step S3-2: Use timestamps to unify the data into the same time series;

[0022] Step S3-3: Divide typical intervals according to the characteristics of different types of unmanned agricultural machinery operations, specifically as follows:

[0023] Seeding: The flight speed range is V[v1, v2], the flight altitude range is H[h1, h2], and the spray width range is W[w1, w2].

[0024] Irrigation: The flight speed range is V[v2, v3], the flight altitude range is H[h2, h3], and the spray width range is W[w2, w3].

[0025] Fertilization: The flight speed range is V[v3, v4], the flight altitude range is H[h3, h4], and the spray width range is W[w3, w4].

[0026] Step S3-4: Compare the flight speed, flight altitude, and jet width at each moment with the typical range, and mark the "candidate operation range" for each parameter;

[0027] Step S3-5: Set the weight of spray width to 0.4, flight altitude to 0.3, and flight speed to 0.3. Use a weighted algorithm to calculate the total weight score for each operation type, specifically:

[0028]

[0029]

[0030] In the formula, S core (T) represents the weighted score for a certain job type, where T represents the job type, and ω V For the flight speed weight, ω H For the flight altitude weight, ω W As the weight of the spray width, The function for determining flight speed. For flight altitude determination function, This is a function for determining the spray width.

[0031] Step S3-6: If the total score of a certain type of task is >0.7, it is directly determined to be that type of task. If all scores are <0.7, then proceed to "Boundary and Timing Auxiliary Judgment".

[0032] Step S3-7: When a certain type of operation enters the "Boundary and Timing Auxiliary Judgment", the flight speed, flight altitude, and spray width of the operation type are judged with reference to the interval division in step S3-3. When the flight speed, flight altitude, and spray width of the operation type are exactly at the operation interval division threshold boundary v2 or v3, h2 or h3, w2 or w3 in S3-3, they are exactly in the middle of two operation types and the operation type cannot be determined. Then, the operation types of the previous five time points are referenced. If the previous five operation types are all the same operation type, it is determined that the operation type is consistent with the previous five operation types. If the operation type is different from the other four time points in the middle of the previous five time points, it is judged as data fluctuation, and the operation type is still judged to be consistent with the four same operation types.

[0033] By using historical multimodal data, different types of operations can be distinguished, the results of operation type determination can be clarified, and managers can grasp the specific operation content and progress of unmanned agricultural machinery, providing a basis for subsequent operation result evaluation.

[0034] Step S4: Analyze the results of unmanned agricultural machinery operations, and adjust the position and operation trajectory of the unmanned agricultural machinery based on the analysis results;

[0035] Step S4-1: Set the coverage rate and spray uniformity as the core evaluation indicators, specifically:

[0036] Job coverage:

[0037]

[0038] In the formula, C represents the job coverage rate, and S... act S represents the actual coverage area of ​​unmanned agricultural machinery operations. tar Where W is the target operating area, L is the spray width, k is the trajectory length, a is the trajectory overlap coefficient, b is the spray height coefficient, and H is the base spray width.

[0039] Spray uniformity:

[0040] In the formula, U is the spray uniformity, F is the spray flow rate, V is the flight speed, Q is the spray volume per unit area, μ is the average spray volume, σ is the standard deviation, a is the spray width coefficient, b is the base spray width, i is the summation index variable, and Q... i This represents the actual spray volume of the i-th sample, where n is the total number of sampling points.

[0041] Step S4-2: Obtain the actual flight speed V of the unmanned agricultural aircraft through the flight control system. act Actual flight altitude H act and actual spray width W act ;

[0042] Step S4-3: Set the target operating area before the unmanned agricultural machinery operation is completed, collect the actual coverage area after the unmanned agricultural machinery operation, and collect the spraying volume at the sampling point using the unmanned agricultural machinery flight control system.

[0043] Step S4-4: Calculate the actual operation coverage rate C act Usage uniformity of actual spraying act Set the evaluation threshold C std with U std When the actual operational coverage rate C act ≥C std And the actual spray uniformity U act ≤U std The work was completed on time and met the standards.

[0044] Steps S4-5: When the work fails to meet the standards, adjust the parameters according to different influencing factors, focusing on different dimensions:

[0045] Adjusting the flight altitude, specifically:

[0046]

[0047] In the formula, H tar W represents the target altitude value for the flight altitude of the unmanned agricultural machinery. tar denoted as target spray width, 'a' as spray height coefficient, and 'b' as base spray width;

[0048] Adjust the spacing of the work trajectory as follows:

[0049] If the actual work trajectory spacing D act >W act Keep H act The spacing of the work trajectory remains unchanged, but the following adjustments are made:

[0050] D tar =W act ×(1-k tar );

[0051] In the formula, D tar W represents the target trajectory spacing. act k represents the actual spray width. tar The overlap rate of the target trajectory;

[0052] Adjusting flight speed, specifically:

[0053]

[0054] In the formula, V tar Let F be the target flight speed, F be the spray flow rate, and Q be the spray volume. tar W is the target spraying rate per unit area. act This refers to the actual spray width;

[0055] Maintain a stable flight altitude, controlling fluctuations to ≤0.5m.

[0056] By setting evaluation indicators to assess work results, the evaluation results are quantified, making the work quality assessment more objective. For substandard work, adjustment plans are provided in different dimensions, thereby improving the quality of subsequent work.

[0057] Step S5: Set the pose and operation trajectory adjustment cycle, construct a knowledge graph based on the parameter adjustment changes within the cycle, and adjust the pose and operation trajectory of the unmanned agricultural machinery based on the knowledge graph.

[0058] Step S5-1: Record the target operating area of ​​the unmanned agricultural machinery, and record the parameters adjusted in steps S1-S4 based on the operating area of ​​the unmanned agricultural machinery.

[0059] Step S5-2: Establish a target table, an actual table, and a parameter table. The target table includes the target operating area and target spray uniformity of the unmanned agricultural machine. The actual table includes the actual operating area, actual spray uniformity, and trajectory length of the unmanned agricultural machine. The parameter table includes the spray width, flight altitude, and flight speed of the unmanned agricultural machine.

[0060] Step S5-3: Set a unique identifier ID for each target work area, and link the data in the actual table and the parameter table in the target table through primary and foreign keys;

[0061] Step S5-4: When the unmanned agricultural machinery is working, the target working area is compared with the actual table to find the same value. Then, based on the same value found in the actual table, the parameters of the unmanned agricultural machinery are set by searching the parameter table through the index. Finally, the unmanned agricultural machinery starts working.

[0062] By adjusting the current operating parameters of unmanned agricultural machinery based on historical normal data, the operating efficiency of unmanned agricultural machinery is improved. By adjusting the parameters of the knowledge graph, the operating parameters are updated, which promotes continuous improvement in operating results.

[0063] To achieve the above objectives, the present invention provides an unmanned agricultural machinery operation data analysis system based on three-dimensional scene reconstruction. The system includes a three-dimensional scene model construction module, a three-dimensional mapping module, an operation behavior determination module, an operation analysis and parameter adjustment module, and a trajectory optimization and knowledge graph construction module.

[0064] The three-dimensional scene model construction module obtains farm spatial information through on-site surveys, divides areas and performs point cloud collection operations, and generates a three-dimensional farm model after data preprocessing, providing a spatial reference for mapping the operation trajectory of unmanned agricultural machinery.

[0065] The three-dimensional mapping module uses dual coordinates to record the three-dimensional scene and unmanned agricultural machinery data. After cleaning and standardizing the raw data of the unmanned agricultural machinery, the operation trajectory is mapped to the three-dimensional model and associated through timestamps to realize the three-dimensional visualization of unmanned agricultural machinery operation.

[0066] The operation behavior determination module is based on historical multimodal data. Through preprocessing, typical interval division, weighted score calculation and boundary time series-assisted determination, it accurately identifies the real-time operation type of unmanned agricultural machinery and provides behavioral basis for operation evaluation.

[0067] The parameter adjustment module evaluates the operation results with operation coverage and spraying uniformity as core indicators, compares the threshold to determine the standard, and adjusts the flight altitude, trajectory spacing and flight speed in different dimensions for substandard operations to optimize the position and trajectory of the unmanned agricultural machinery.

[0068] The trajectory optimization and knowledge graph construction module sets an adjustment period, records the operation parameters and adjustment data within the period, constructs a "target-actual-parameter" association table, and dynamically optimizes the pose and trajectory of unmanned agricultural machinery based on historical parameters of the knowledge graph.

[0069] The 3D scene model construction module includes a data acquisition unit and a 3D model construction unit;

[0070] The data acquisition unit is responsible for collecting farm spatial information and drawing sketches;

[0071] The 3D model building unit is responsible for building a 3D model of the farm based on the sketches and point cloud data.

[0072] The three-dimensional mapping module includes a coordinate and trajectory data acquisition unit and an unmanned agricultural machinery pose and trajectory mapping unit.

[0073] The trajectory data acquisition unit records the coordinate points of the farm's three-dimensional model using the National Geodetic Coordinate System and records the real-time coordinates of the unmanned agricultural machinery using the BeiDou positioning geographic coordinate system. The data is then cleaned and timestamped to ensure time consistency.

[0074] The unmanned agricultural machinery pose and trajectory mapping unit maps the time-stamped track coordinate points of the unmanned agricultural machinery to the three-dimensional model of the farm according to the coordinate system correspondence, generates the spatial position trajectory of the unmanned agricultural machinery, and then associates the attitude angle data with the track coordinate points through the timestamp.

[0075] The job behavior determination module includes a multimodal data processing unit and a job determination unit;

[0076] The multimodal data processing unit is responsible for collecting and processing multimodal data from historical unmanned agricultural machinery operations.

[0077] The operation determination unit is responsible for dividing typical operation areas and determining the type of unmanned agricultural machinery operation.

[0078] The task analysis and parameter adjustment module includes a result evaluation unit and a parameter adjustment unit;

[0079] The results evaluation unit is responsible for determining whether unmanned agricultural machinery operations meet the standards.

[0080] The parameter adjustment unit is responsible for adjusting the multimodal data of the unmanned agricultural machinery.

[0081] The trajectory optimization and knowledge graph construction module includes a parameter adjustment recording unit and a trajectory optimization unit;

[0082] The adjustment parameter recording unit establishes a database table by setting an adjustment period and recording the target working area and the adjusted core parameters within the period, thereby constructing a basic data structure for the knowledge graph.

[0083] The trajectory optimization unit is responsible for adjusting the parameters of the current unmanned agricultural machinery's required operation type based on historical parameters.

[0084] Compared with the prior art, the beneficial effects of the present invention are:

[0085] 1. This invention ensures that unmanned agricultural machinery can complete its work accurately and efficiently according to the expected goals by constructing a three-dimensional scene model, model mapping, work behavior determination, and parameter adjustment.

[0086] 2. This invention uses quantitative evaluation indicators and thresholds to objectively determine whether the work meets the standards, thus avoiding potential quality problems.

[0087] 3. This invention enables parameter reuse through knowledge graphs, reducing the cost and time of manual debugging. The overall process reduces manual intervention, shortens operation time, and improves operation efficiency and the intelligence of unmanned agricultural machinery operation through automated data processing. Attached Figure Description

[0088] Figure 1 This is a flowchart illustrating the unmanned agricultural machinery operation data analysis method based on 3D scene reconstruction according to the present invention.

[0089] Figure 2 This is a schematic diagram of the structure of the unmanned agricultural machinery operation data analysis system based on three-dimensional scene reconstruction according to the present invention. Detailed Implementation

[0090] 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.

[0091] Example 1: As Figure 1 As shown, the present invention provides a technical solution: a method for analyzing unmanned agricultural machinery operation data based on three-dimensional scene reconstruction. The method includes the following steps:

[0092] Step S1: Use terrestrial laser scanning technology to scan the farm and construct a three-dimensional scene model;

[0093] Step S1-1: Conduct an on-site survey of the farm boundaries, crop distribution, and fixed obstacles;

[0094] Step S1-2: Draw a farm site survey sketch based on the field survey data, and mark crops, boundaries and fixed obstacles;

[0095] Steps S1-3: Divide the scanning area according to the grid division method. Divide the farm into multiple sub-grids according to the "a×a" grid and mark the center point of each grid. Set a target at the center point. a is the side length of the sub-grid, which is determined according to the unmanned agricultural machinery operation range.

[0096] Steps S1-4: Collect high-density point cloud data for each sub-region, use statistical filtering algorithm to remove discrete noise points from the point cloud data, then use voxel grid downsampling method to downsample the data, and finally use kd tree to remove duplicate points from the point cloud.

[0097] Step S1-5: Based on the center point coordinates of the target, the preprocessed point clouds of multiple sub-regions are stitched together and their coordinates are unified using the target matching stitching method.

[0098] Steps S1-6: Import the assembled point cloud model into 3D modeling software to construct a 3D scene model of the farm.

[0099] By conducting on-site surveys to clarify farm boundaries, crop distribution, and fixed obstacles, and by combining grid division with the setting of targets to avoid missing key areas, a benchmark was provided for the construction of a 3D scene model. Data processing provided reliable model support for subsequent unmanned agricultural machinery trajectory mapping.

[0100] Step S2: Map the pose and working trajectory of the unmanned agricultural machinery into the 3D scene model;

[0101] Step S2-1: Use the National Geodetic Coordinate System CGCS2000 to record the coordinate points of the 3D scene model, use the BeiDou positioning geographic coordinate system to record the coordinate points of the unmanned agricultural machinery, and construct the unmanned agricultural machinery track data through the coordinate points of the unmanned agricultural machinery.

[0102] Step S2-2: Clean and standardize the coordinate points and attitude angles in the unmanned agricultural machinery track data, and timestamp the coordinate points and attitude angles in the track data.

[0103] Step S2-3: Map the coordinate points of the unmanned agricultural machinery track data to the 3D model, and match the attitude angle data of the unmanned agricultural machinery with the coordinate points of the track data through timestamps.

[0104] The operation trajectory of unmanned agricultural machinery is visualized, which facilitates real-time monitoring of the operation status of unmanned agricultural machinery. The data is marked with timestamps, which provides the basic data with correct time sequence for subsequent steps.

[0105] Step S3: Based on the multimodal operation data of unmanned agricultural machinery, divide the data into different intervals to determine the operation behavior of the unmanned agricultural machinery;

[0106] Step S3-1: Collect multimodal data on the normal operation of historical unmanned agricultural machinery, perform data cleaning and standardization on the multimodal data, use filtering algorithms to eliminate environmental interference, use linear interpolation to process missing values, and use box plots to identify outliers.

[0107] Step S3-2: Use timestamps to unify the data into the same time series;

[0108] Step S3-3: Divide typical intervals according to the characteristics of different types of unmanned agricultural machinery operations, specifically as follows:

[0109] Seeding: The flight speed range is V[v1, v2], the flight altitude range is H[h1, h2], and the spray width range is W[w1, w2].

[0110] Irrigation: The flight speed range is V[v2, v3], the flight altitude range is H[h2, h3], and the spray width range is W[w2, w3].

[0111] Fertilization: The flight speed range is V[v3, v4], the flight altitude range is H[h3, h4], and the spray width range is W[w3, w4].

[0112] Step S3-4: Compare the flight speed, flight altitude, and jet width at each moment with the typical range, and mark the "candidate operation range" for each parameter;

[0113] Step S3-5: Set the weight of spray width to 0.4, flight altitude to 0.3, and flight speed to 0.3. Use a weighted algorithm to calculate the total weight score for each operation type, specifically:

[0114]

[0115] In the formula, S core (T) represents the weighted score for a certain job type, where T represents the job type, and ω V For the flight speed weight, ω H For the flight altitude weight, ω W As the weight of the spray width, The function for determining flight speed. For flight altitude determination function, This is a function for determining the spray width.

[0116] Step S3-6: If the total score of a certain type of task is >0.7, it is directly determined to be that type of task. If all scores are <0.7, then proceed to "Boundary and Timing Auxiliary Judgment".

[0117] Step S3-7: When a certain type of operation enters the "Boundary and Timing Auxiliary Judgment", the flight speed, flight altitude, and spray width of the operation type are judged with reference to the interval division in step S3-3. When the flight speed, flight altitude, and spray width of the operation type are exactly at the operation interval division threshold boundary v2 or v3, h2 or h3, w2 or w3 in S3-3, they are exactly in the middle of two operation types and the operation type cannot be determined. Then, the operation types of the previous five time points are referenced. If the previous five operation types are all the same operation type, it is determined that the operation type is consistent with the previous five operation types. If the operation type is different from the other four time points in the middle of the previous five time points, it is judged as data fluctuation, and the operation type is still judged to be consistent with the four same operation types.

[0118] By using historical multimodal data, different types of operations can be distinguished, the results of operation type determination can be clarified, and managers can grasp the specific operation content and progress of unmanned agricultural machinery, providing a basis for subsequent operation result evaluation.

[0119] Step S4: Analyze the results of unmanned agricultural machinery operations, and adjust the position and operation trajectory of the unmanned agricultural machinery based on the analysis results;

[0120] Step S4-1: Set the coverage rate and spray uniformity as the core evaluation indicators, specifically:

[0121] Job coverage:

[0122]

[0123] In the formula, C represents the job coverage rate, and S... act S represents the actual coverage area of ​​unmanned agricultural machinery operations. tar Where W is the target operating area, L is the spray width, k is the trajectory length, a is the trajectory overlap coefficient, b is the spray height coefficient, and H is the base spray width.

[0124] Spray uniformity:

[0125] In the formula, U is the spray uniformity, F is the spray flow rate, V is the flight speed, Q is the spray volume per unit area, μ is the average spray volume, σ is the standard deviation, a is the spray width coefficient, b is the base spray width, i is the summation index variable, and Q... i This represents the actual spray volume of the i-th sample, where n is the total number of sampling points.

[0126] Step S4-2: Obtain the actual flight speed V of the unmanned agricultural aircraft through the flight control system. act Actual flight altitude H act and actual spray width W act ;

[0127] Step S4-3: Set the target operating area before the unmanned agricultural machinery operation is completed, collect the actual coverage area after the unmanned agricultural machinery operation, and collect the spraying volume at the sampling point using the unmanned agricultural machinery flight control system.

[0128] Step S4-4: Calculate the actual operation coverage rate C act Usage uniformity of actual spraying act Set the evaluation threshold C std with U std When the actual operational coverage rate C act ≥C std And the actual spray uniformity U act ≤U std The work was completed on time and met the standards.

[0129] Steps S4-5: When the work fails to meet the standards, adjust the parameters according to different influencing factors, focusing on different dimensions:

[0130] Adjusting the flight altitude, specifically:

[0131]

[0132] In the formula, H tar W represents the target altitude value for the flight altitude of the unmanned agricultural machinery. tar denoted as target spray width, 'a' as spray height coefficient, and 'b' as base spray width;

[0133] Adjust the spacing of the work trajectory as follows:

[0134] If the actual work trajectory spacing D act >W act Keep H act The spacing of the work trajectory remains unchanged, but the following adjustments are made:

[0135] D tar =W act ×(1-k tar );

[0136] In the formula, D tar W represents the target trajectory spacing. act k represents the actual spray width. tar The overlap rate of the target trajectory;

[0137] Adjusting flight speed, specifically:

[0138]

[0139] In the formula, V tar Let F be the target flight speed, F be the spray flow rate, and Q be the spray volume. tar W is the target spraying rate per unit area. act This refers to the actual spray width;

[0140] Maintain a stable flight altitude, controlling fluctuations to ≤0.5m.

[0141] By setting evaluation indicators to assess work results, the evaluation results are quantified, making the work quality assessment more objective. For substandard work, adjustment plans are provided in different dimensions, thereby improving the quality of subsequent work.

[0142] Step S5: Set the pose and operation trajectory adjustment cycle, construct a knowledge graph based on the parameter adjustment changes within the cycle, and adjust the pose and operation trajectory of the unmanned agricultural machinery based on the knowledge graph.

[0143] Step S5-1: Record the target operating area of ​​the unmanned agricultural machinery, and record the parameters adjusted in steps S1-S4 based on the operating area of ​​the unmanned agricultural machinery.

[0144] Step S5-2: Establish a target table, an actual table, and a parameter table. The target table includes the target operating area and target spray uniformity of the unmanned agricultural machine. The actual table includes the actual operating area, actual spray uniformity, and trajectory length of the unmanned agricultural machine. The parameter table includes the spray width, flight altitude, and flight speed of the unmanned agricultural machine.

[0145] Step S5-3: Set a unique identifier ID for each target work area, and link the data in the actual table and the parameter table in the target table through primary and foreign keys;

[0146] Step S5-4: When the unmanned agricultural machinery is working, the target working area is compared with the actual table to find the same value. Then, based on the same value found in the actual table, the parameters of the unmanned agricultural machinery are set by searching the parameter table through the index. Finally, the unmanned agricultural machinery starts working.

[0147] By adjusting the current operating parameters of unmanned agricultural machinery based on historical normal data, the operating efficiency of unmanned agricultural machinery is improved. By adjusting the parameters of the knowledge graph, the operating parameters are updated, which promotes continuous improvement in operating results.

[0148] To achieve the above objectives, the present invention provides an unmanned agricultural machinery operation data analysis system based on three-dimensional scene reconstruction. The system includes a three-dimensional scene model construction module, a three-dimensional mapping module, an operation behavior determination module, an operation analysis and parameter adjustment module, and a trajectory optimization and knowledge graph construction module.

[0149] The three-dimensional scene model construction module obtains farm spatial information through on-site surveys, divides areas and performs point cloud collection operations, and generates a three-dimensional farm model after data preprocessing, providing a spatial reference for mapping the operation trajectory of unmanned agricultural machinery.

[0150] The three-dimensional mapping module uses dual coordinates to record the three-dimensional scene and unmanned agricultural machinery data. After cleaning and standardizing the raw data of the unmanned agricultural machinery, the operation trajectory is mapped to the three-dimensional model and associated through timestamps to realize the three-dimensional visualization of unmanned agricultural machinery operation.

[0151] The operation behavior determination module is based on historical multimodal data. Through preprocessing, typical interval division, weighted score calculation and boundary time series-assisted determination, it accurately identifies the real-time operation type of unmanned agricultural machinery and provides behavioral basis for operation evaluation.

[0152] The parameter adjustment module evaluates the operation results with operation coverage and spraying uniformity as core indicators, compares the threshold to determine the standard, and adjusts the flight altitude, trajectory spacing and flight speed in different dimensions for substandard operations to optimize the position and trajectory of the unmanned agricultural machinery.

[0153] The trajectory optimization and knowledge graph construction module sets an adjustment period, records the operation parameters and adjustment data within the period, constructs a "target-actual-parameter" association table, and dynamically optimizes the pose and trajectory of unmanned agricultural machinery based on historical parameters of the knowledge graph.

[0154] The 3D scene model construction module includes a data acquisition unit and a 3D model construction unit;

[0155] The data acquisition unit is responsible for collecting farm spatial information and drawing sketches;

[0156] The 3D model building unit is responsible for building a 3D model of the farm based on the sketches and point cloud data.

[0157] The three-dimensional mapping module includes a coordinate and trajectory data acquisition unit and an unmanned agricultural machinery pose and trajectory mapping unit.

[0158] The trajectory data acquisition unit records the coordinate points of the farm's three-dimensional model using the National Geodetic Coordinate System and records the real-time coordinates of the unmanned agricultural machinery using the BeiDou positioning geographic coordinate system. The data is then cleaned and timestamped to ensure time consistency.

[0159] The unmanned agricultural machinery pose and trajectory mapping unit maps the time-stamped track coordinate points of the unmanned agricultural machinery to the three-dimensional model of the farm according to the coordinate system correspondence, generates the spatial position trajectory of the unmanned agricultural machinery, and then associates the attitude angle data with the track coordinate points through the timestamp.

[0160] The job behavior determination module includes a multimodal data processing unit and a job determination unit;

[0161] The multimodal data processing unit is responsible for collecting and processing multimodal data from historical unmanned agricultural machinery operations.

[0162] The operation determination unit is responsible for dividing typical operation areas and determining the type of unmanned agricultural machinery operation.

[0163] The task analysis and parameter adjustment module includes a result evaluation unit and a parameter adjustment unit;

[0164] The results evaluation unit is responsible for determining whether unmanned agricultural machinery operations meet the standards.

[0165] The parameter adjustment unit is responsible for adjusting the multimodal data of the unmanned agricultural machinery.

[0166] The trajectory optimization and knowledge graph construction module includes a parameter adjustment recording unit and a trajectory optimization unit;

[0167] The adjustment parameter recording unit establishes a database table by setting an adjustment period and recording the target working area and the adjusted core parameters within the period, thereby constructing a basic data structure for the knowledge graph.

[0168] The trajectory optimization unit is responsible for adjusting the parameters of the current unmanned agricultural machinery's required operation type based on historical parameters.

[0169] Example 2: The surveyed farm measures 100m in length and 50m in width, with a total area of ​​5000m². 2 The central area is a wheat planting area, surrounded by a 1-meter-wide road, with three fixed obstacles: irrigation wells.

[0170] The unmanned agricultural machinery was set to spray at a fixed flow rate F = 1.2 L / min, and the spray width coefficient a = 0.05 m. -1 The basic spray width (b) is 1.5m, completing the fertilization operation in the wheat planting area, with a target area of ​​4000m². 2 The target coverage rate is 95%, the target spray uniformity U = 15%, and the target spray rate per unit area Q = 0.3 L / m². 2 The target trajectory overlap rate k = 10%;

[0171] The farm boundary was surveyed on-site as (0,0)-(100,0)-(100,50)-(0,50). The boundary of the wheat planting area was marked as (5,5)-(95,5)-(5,45) and three irrigation locations were marked. A survey sketch was drawn.

[0172] Based on the unmanned agricultural machinery operating radius of 10m, the subgrid side length is set to a = 10m, and it is divided into 50 subgrids. The center coordinates of each subgrid are (5,5), (15,5)...(95,45), and targets are set at each center point;

[0173] Point cloud data of sub-regions are collected using a laser scanner, discrete noise points are extracted using a statistical filtering algorithm, voxel grid downsampling is used, and then kd-tree deduplication is performed.

[0174] Using the target coordinates as a reference, 50 sub-region point clouds were stitched together using the target matching stitching method, with the coordinates unified to the CGCS2000 coordinate system, and the stitching error ≤0.05m;

[0175] Import the stitched point cloud into 3D modeling software to generate a 3D scene model of the farm.

[0176] Example 3: Collect multimodal data of historical fertilization operations, process the data and mark it with timestamps, and divide typical operation intervals. Fertilization operation V = [1.5, 2.5], W = [2.0, 3.0], H = [2.5, 3.5];

[0177] Taking the data at 10:05:00 as an example, V = 2.0 m / s, H = 3.0 m, W = 2.5 m, these three values ​​are marked as "fertilization candidate intervals". According to the weighted algorithm, the score is 1.0. A score > 0.7 is determined to be a fertilization operation.

[0178] Example 4: After the operation is completed, the actual flight speed V is obtained through the flight control system. act = 2.2 m / s, actual flight altitude H act = 3.2m, actual spray width W act = 2.6m, actual operating coverage area S act =3750m 2Spraying amounts were collected at 20 sampling points: 0.28, 0.32, 0.29, 0.31, 0.30, 0.27, 0.33, 0.28, 0.30, 0.31, 0.29, 0.32, 0.28, 0.30, 0.31, 0.29, 0.32, 0.27, 0.33, 0.30.

[0179] The calculated coverage rate was 93.75%, which is less than 95%, and was therefore deemed substandard.

[0180] The spraying rate per unit area is 3.652 L / min, which translates to Q = 0.21 L / m². 2 The spray uniformity U = 6.67 < 15%, which is considered substandard.

[0181] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for unmanned agricultural machine operation data analysis based on three-dimensional scene reconstruction, characterized in that: The method comprises the following steps: Step S1, scanning the farm using ground laser scanning technology to construct a three-dimensional scene model; Step S2, mapping the unmanned agricultural machine pose and operation trajectory into the three-dimensional scene model; Step S3, according to the multi-modal operation data of the unmanned agricultural machine, dividing each operation data interval to judge the operation behavior of the unmanned agricultural machine; Step S3-1, collect multi-modal data of the normal work of the unmanned agricultural machine, and perform data cleaning and standardization processing on the multi-modal data, use filtering algorithm to eliminate environmental interference, use linear interpolation method to process missing values, and use box plot method to identify outliers; Step S3-2, use timestamp to unify the data to the same time sequence; Step S3-3, according to the characteristics of different operation types of the unmanned agricultural machine, divide the typical interval, specifically: Seeding: flight speed interval V[v1, v2], flight height interval H[h1, h2], and spray width interval W[w1, w2]; Irrigation: flight speed interval V[v2, v3], flight height interval H[h2, h3], and spray width interval W[w2, w3]; Fertilization: flight speed interval V[v3, v4], flight height interval H[h3, h4], and spray width interval W[w3, w4]; Step S3-4, compare the flight speed, flight height and spray width of each time with the typical interval respectively, and mark the "candidate operation interval" of each parameter; Step S3-5, set the spray width weight to 0.4, the flight height weight to 0.3, and the flight speed weight to 0.3, and use a weighted algorithm to calculate the total weight score of each operation type, specifically: ； ； wherein, is a weight score for a certain operation type, T represents an operation type, is a flight speed weight, is a flight height weight, is a swath width weight, is a judgment function for flight speed, is a judgment function for flight height, is a judgment function for swath width; Step S3-6, if the total score of a certain operation type is >0.7, it is directly determined as that operation type, if all scores are <0.7, then enter "boundary and time sequence auxiliary judgment"; Step S3-7, when a certain operation type enters "boundary and time sequence auxiliary judgment", the flight speed, flight height and spray width of the operation type are determined according to the interval division of step S3-3, when the flight speed, flight height and spray width of the operation type are exactly at the threshold boundary v2 or v3, h2 or h3, w2 or w3 of the operation interval division of S3-3, it is exactly between two operation types and cannot be determined, then refer to the operation type of the previous five time points, if the previous five operation types are the same operation type, it is determined that the operation type is consistent with the previous five operation types, if the operation type of a certain time point is different from the other four time points in the previous five time points, it is determined that the data fluctuates, and it is still determined that the operation type is consistent with the four same operation types; Step S4, analyze the operation results of the unmanned agricultural machine, and adjust the unmanned agricultural machine pose and operation trajectory according to the analysis results; Step S4-1, set operation coverage and spray uniformity as core evaluation indexes, specifically: Operation coverage: ； wherein C is the operation coverage rate, is the actual operation coverage area of unmanned agricultural machinery, is the target operation area, W is the swath width, L is the track length, k is the track overlap coefficient, a is the swath height coefficient, b is the base swath width, and H is the flight height. Spray uniformity: ; ; ; ; where U is the spray uniformity, F is the spray flow rate, V is the flight speed, and Q is the spray volume per unit area, is the average spray volume, is the standard deviation, a is the swath height coefficient, b is the base swath width, and i is the summation index variable, represents the ith actual spray volume, and n is the total number of sampling points; Step S4-2, obtaining the actual flight speed V of the unmanned agricultural machine by the flight control system act , the actual flight height H act , and the actual swath width W act ; Step S4-3, set the target operation area before the unmanned agricultural machine operation, collect the actual coverage area after the unmanned agricultural machine operation, and collect the spray amount of the sampling points by the flight control system of the unmanned agricultural machine; Step S4-4, calculating actual job coverage C act with actual spray uniformity U act , setting evaluation threshold C std with U std , when actual job coverage C act ≥ C std and actual spray uniformity U act ≤ U std , job is up to standard; Step S4-5, when the task is not up to standard, adjust the parameters in different dimensions according to different influencing factors: Adjust the flight height, specifically: ； In the formula, is a target height value of the unmanned agricultural machine flight height, is a target swath width, a is a swath height coefficient, and b is a basic swath width. Adjust the flight speed, specifically: If the actual working track interval D act >W act , keep H act unchanged, adjust the working track interval, specifically: ； wherein is the target trajectory spacing, is the actual swath width, is the target trajectory overlap rate; Adjust the flight speed, specifically: ； In the formula, F is the target flight speed, F is the spray flow rate, F is the target unit area spray amount, F is the actual spray width; Stabilize the flight height, control the fluctuation amplitude ≤0.5m; Step S5, set the pose and operation track adjustment period, construct a knowledge graph according to the parameter adjustment changes in the period, and adjust the pose and operation track of the unmanned agricultural machine according to the knowledge graph.

2. The method for unmanned farm machine operation data analysis based on three-dimensional scene reconstruction according to claim 1, characterized in that: The specific steps of step S1 are as follows: Step S1-1, field survey of farm boundaries, crop distribution and fixed obstacles; Step S1-2, draw a farm site survey sketch according to the field survey data, mark crops, boundaries and fixed obstacles; Step S1-3, according to the grid division method, divide the scanning area into multiple subgrids according to the "a x a" grid, mark the center point of each grid, and set a target at the center point, a is the length of the subgrid, and is determined according to the operation range of the unmanned agricultural machine; Step S1-4, collect high-density point cloud data of each subarea, use statistical filtering algorithm to remove discrete noise points, then use voxel grid downsampling method for data downsampling, and finally use k-d tree for point cloud deduplication; Step S1-5, according to the center point coordinate target, use target matching splicing method to splice and unify the coordinates of the pretreated multiple subarea point clouds; Step S1-6, import the spliced whole point cloud model into a three-dimensional modeling software to build a three-dimensional scene model of the farm.

3. The method of claim 2, wherein: The specific steps of step S2 are as follows: Step S2-1, record the three-dimensional scene model coordinate points using the national geodetic coordinate system CGCS2000, record the unmanned agricultural machine coordinate points using the Beidou positioning geographic coordinate system, and build the unmanned agricultural machine track data through the unmanned agricultural machine coordinate points; Step S2-2, clean and standardize the coordinate points and attitude angles in the unmanned agricultural machine track data, and mark the track data coordinate points and attitude angles with time stamps; Step S2-3, map the coordinate points of the unmanned agricultural machine track data to the three-dimensional model, and correspond the attitude angle data of the unmanned agricultural machine to the coordinate points of the track data through time stamp.

4. The method of claim 3, wherein: The specific steps of step S5 are as follows: Step S5-1, record the target operation area of the unmanned agricultural machine, and record the parameters adjusted through steps S1-S4 according to the unmanned agricultural machine operation area; Step S5-2, establish target table, actual table and parameter table, target table includes target operation area and target spraying uniformity of unmanned agricultural machine, actual table includes actual operation area, actual spraying uniformity and track length of unmanned agricultural machine, parameter table includes swath width, flight height and flight speed of unmanned agricultural machine; Step S5-3, set a unique identifier ID for each target operation area, and associate the data in the actual table and the parameter table through the primary key in the target table; Step S5-4, when the unmanned agricultural machine is working, compare the target operation area in the actual table, find the same value, then find the parameters in the parameter table through index according to the same value obtained by comparing in the actual table, and finally set the parameters of the unmanned agricultural machine, and the unmanned agricultural machine works.

5. An unmanned agricultural machine operation data analysis system based on three-dimensional scene reconstruction, used for executing the unmanned agricultural machine operation data analysis method based on three-dimensional scene reconstruction according to any one of claims 1-4, characterized in that: The system comprises a three-dimensional scene model construction module, a three-dimensional mapping module, a work behavior judgment module, a work analysis and parameter adjustment module, and a trajectory optimization and knowledge graph construction module; The three-dimensional scene model construction module obtains farm space information through field investigation, divides regions, and performs point cloud collection operations, and generates a farm three-dimensional model after preprocessing data, thereby providing a spatial reference for unmanned agricultural machine work trajectory mapping; The three-dimensional mapping module records three-dimensional scenes and unmanned agricultural machine data using double coordinates, maps work trajectories to the three-dimensional model after standardizing unmanned agricultural machine original data, correlates through timestamps, and realizes three-dimensional visualization of unmanned agricultural machine work; The work behavior judgment module accurately identifies real-time work types of unmanned agricultural machines based on historical multi-modal data through preprocessing, typical interval division, weighted score calculation, and boundary time sequence auxiliary judgment, thereby providing a behavior basis for work evaluation; The parameter adjustment module takes work coverage and spraying uniformity as core indicators to evaluate work results, compares threshold values to determine whether standards are met, adjusts flight height, trajectory spacing, and flight speed in different dimensions for substandard work, and optimizes unmanned agricultural machine pose and trajectory; The trajectory optimization and knowledge graph construction module records work parameters and adjustment data within a period through setting an adjustment period, constructs a "target-actual-parameter" correlation table, dynamically optimizes unmanned agricultural machine pose and trajectory based on historical parameters of the knowledge graph.

6. The system for unmanned farm machinery operation data analysis based on three-dimensional scene reconstruction according to claim 5, characterized in that: The three-dimensional scene model construction module comprises a data acquisition unit and a three-dimensional model construction unit; The data acquisition unit is responsible for acquiring farm space information and drawing sketches; The three-dimensional model construction unit is responsible for constructing a farm three-dimensional model according to sketches and point cloud data; The three-dimensional mapping module comprises a coordinate and trajectory data acquisition unit and an unmanned agricultural machine pose and trajectory mapping unit; The trajectory data acquisition unit records farm three-dimensional model coordinate points through the national geodetic coordinate system, records real-time coordinates of unmanned agricultural machines using the Beidou positioning geographic coordinate system, then cleanses data and marks timestamps to ensure time sequence consistency; The unmanned agricultural machine pose and trajectory mapping unit maps timestamped unmanned agricultural machine trajectory coordinate points to the farm three-dimensional model according to the corresponding relationship of the coordinate system, generates an unmanned agricultural machine spatial position trajectory, and then correlates attitude angle data with trajectory coordinate points through timestamps.

7. The system for unmanned farm machinery operation data analysis based on three-dimensional scene reconstruction of claim 5, wherein: The work behavior judgment module comprises a multi-modal data processing unit and a work judgment unit; The multi-modal data processing unit is responsible for collecting multi-modal data of normal work of unmanned agricultural machines and processing data; The work judgment unit is responsible for dividing typical work intervals and judging work types of unmanned agricultural machines; The work analysis and parameter adjustment module comprises a result evaluation unit and a parameter adjustment unit; The result evaluation unit is responsible for making compliance judgments on unmanned agricultural machine work; The parameter adjustment unit is responsible for adjusting multi-modal data of unmanned agricultural machines.

8. The three-dimensional scene reconstruction based unmanned farm machine operation data analysis system of claim 5, wherein: The trajectory optimization and knowledge graph construction module comprises an adjustment parameter recording unit and a trajectory optimization unit; The adjustment parameter recording unit records the target working area in a period and the adjusted core parameters to establish a database table and construct a knowledge graph basic data structure by setting an adjustment period; The trajectory optimization unit is responsible for adjusting the parameters of the required working type of the unmanned agricultural machine according to historical parameters.

Images