Multi-sensor fusion pose measurement method and measurement device for tractor and seeder

Abstract

A multi-sensor fusion pose measurement method for a tractor and a seeder comprises: collecting original state data during operation of the tractor and the seeder and pre-processing the original state data, the original state data including satellite positioning data, inertial positioning data and visual observation data; using an error state Kalman filter model to preliminarily fuse the pre-processed original state data to obtain a preliminary estimation result of the pose of the tractor; using a multi-source information deep fusion model to optimally fuse the satellite positioning data, the visual observation data and the preliminary estimation result of the pose of the tractor to obtain accurate pose information of the tractor; and using a pre-constructed combined kinematics model to solve multi-degree-of-freedom pose information of an end effector of the seeder based on the accurate pose information of the tractor. The present application provides a multi-sensor fusion pose measurement method and a measurement device for a tractor and a seeder, which can accurately measure the pose state of the tractor and the seeder.

Description

Technical Field

[0001] This invention relates to the field of agricultural machinery automation and intelligent sensing technology, specifically to a multi-sensor fusion pose measurement method and device for tractors and seeders. Background Technology

[0002] Precision agriculture is an important direction for the development of modern agriculture, and its core is to automate and intelligentize the agricultural production process. In farmland tillage and sowing operations, tracked tractors pulling rotary tillers and seeders is a common combination. When operating in complex terrains such as hilly and mountainous areas, tracked tractor-rotary tiller seeders need to acquire real-time and accurate six-degree-of-freedom positional information of the tractor body, hydraulic suspension system, and rotary tiller end effector to achieve precise operational control.

[0003] However, at present, single sensors are difficult to adapt to the complex operating environment of hilly and mountainous areas, and have obvious measurement limitations: BDS is susceptible to signal loss and positioning accuracy degradation due to vegetation and terrain obstruction; IMU inertial measurement data suffers from time-accumulated drift error; visual sensors are susceptible to interference from changes in lighting and missing field textures, leading to feature matching failure and degradation of observation accuracy. Multi-sensor fusion is the core means to solve the problems of poor reliability and insufficient accuracy of single sensors. By combining the advantages of BDS absolute positioning, IMU high-frequency inertial measurement, and visual local fine observation, and combining ESKF filtering and DS evidence theory to achieve multi-level data fusion, the stability and accuracy of agricultural machinery pose measurement can be effectively improved. However, for the multi-joint serial configuration of tracked tractors-rotary tillers, there is currently no mature and complete multi-sensor system deployment scheme, robust fusion strategy for multi-source heterogeneous data, and precise pose calculation technology system for the end effector, which restricts the accuracy and stability of intelligent seeding operations in complex terrain. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a multi-sensor fusion pose measurement method and device for tractors and seeders, which can accurately measure the pose state of tractors and seeders.

[0005] To achieve the above objectives, the specific solution adopted by the present invention is as follows: a multi-sensor fusion pose measurement method for tractors and seeders, comprising: Raw state data is collected during the operation of tractors and seeders and preprocessed. The raw state data includes satellite positioning data, inertial positioning data and visual observation data. The error state Kalman filter model is used to perform preliminary fusion of the preprocessed raw state data to obtain the preliminary estimation result of the tractor pose. A deep fusion model of multi-source information is used to optimize and fuse satellite positioning data, visual observation data and preliminary estimation results of tractor pose to obtain precise tractor pose information. The deep fusion model of multi-source information is constructed based on DS evidence theory. The multi-degree-of-freedom pose information of the seeder end effector is calculated based on the precise pose information of the tractor using a pre-constructed combined kinematic model.

[0006] As a further optimization of the aforementioned multi-sensor fusion pose measurement method for tractors and seeders, the preprocessing method for the raw state data includes noise reduction filtering, time synchronization, coordinate unification, and standardization. After preprocessing, the raw state data is as follows: ; in, For satellite positioning data, For inertial positioning data, The data represents visual observations, and k represents time.

[0007] As a further optimization of the above-mentioned multi-sensor fusion pose measurement method for tractors and seeders, the noise reduction filtering method is as follows: ; in, The initial observations for the original state data, These are the optimized observations after the original state data has been denoised and filtered. This is the length of the filtering window; The method for unifying the coordinates is as follows: ; in, For the observations in the sensor coordinate system, This is the coordinate transformation matrix from the sensor coordinate system to the global coordinate system. Observations in the global coordinate system; The standardization process is as follows: ; in, For observations before standardization, For all The mean, For all standard deviation These are the observations after standardization.

[0008] As a further optimization of the above-mentioned multi-sensor fusion pose measurement method for tractors and seeders: the system state vector of the error state Kalman filter model is: ; in, Position status, In terms of speed state, For state quaternions, To achieve zero bias in the accelerometer, Zero bias for the gyroscope; The state propagation equation of the error state Kalman filter model is: ; in, Let be the prior state estimate at time k. For the inertial positioning data used as input, Let be the combined kinematic state transition function of the tractor and the seeder. This refers to system process noise. The error state propagation equation for the error state Kalman filter model is: ; in, Here is the state transition matrix. This is the noise-driven matrix.

[0009] As a further optimization of the aforementioned multi-sensor fusion pose measurement method for tractors and seeders, the following methods are used to optimize and fuse satellite positioning data, visual observation data, and preliminary tractor pose estimation results using a multi-source information deep fusion model: Satellite positioning attitude estimation results are generated based on satellite positioning data, and visual observation attitude estimation results are generated based on visual observation data. An evidence set containing multiple evidence sources was constructed based on satellite positioning attitude estimation results, visual observation attitude estimation results, and preliminary tractor pose estimation results. The basic probability allocation results of each evidence source are calculated based on the Gaussian membership function. The degree of evidence conflict in the evidence set is calculated based on the basic probability allocation results, and the DS combination rule is selected according to the degree of evidence conflict. The DS combination rule is used to optimize and fuse all evidence sources to obtain fused evidence, and the basic probability allocation results of the fused evidence are calculated. Based on the basic probability allocation results of the fused evidence, the confidence levels of the satellite positioning attitude estimation results, the visual observation attitude estimation results, and the preliminary tractor pose estimation results are calculated, and the one with the highest confidence level is selected as the precise pose information of the tractor.

[0010] As a further optimization of the multi-sensor fusion pose measurement method for tractors and seeders: when selecting the DS combination rule based on the degree of evidence conflict, if the degree of evidence conflict is less than the preset conflict threshold, the standard DS combination rule is selected; if the degree of evidence conflict reaches the preset conflict threshold, the weighted modified DS combination rule is selected.

[0011] As a further optimization of the above-mentioned multi-sensor fusion pose measurement method for tractors and seeders: the combined kinematic model includes a combined system pose transfer model and a combined system state space model; The pose transfer model of the combined system is as follows: ; in, Used to describe the position and attitude of the tractor relative to the global coordinate system. Used to describe the connection and motion relationship between the hydraulic suspension mechanism and the hydraulic press between the tractor and the seeder. Used to describe the relative motion relationship between the seeder and the hydraulic suspension mechanism. Used to describe the positional relationship between the end effector of the seeder and the seeder; The state-space model of the combined system is as follows: ; in, Let be the system state vector. To control the input amount, Let be the combined kinematic state transition function of the tractor and the seeder. This refers to system process noise.

[0012] As a further optimization of the aforementioned multi-sensor fusion pose measurement method for tractors and seeders, a method for calculating the multi-degree-of-freedom pose information of the seeder's end effector based on the tractor's precise pose information using a pre-built combined kinematics model includes: The preliminary pose information of the seeder end effector in the global coordinate system was calculated based on the precise pose information of the tractor using a combined kinematics model. The initial pose information is corrected using a pre-built hybrid error compensation model; The multi-degree-of-freedom pose information of the seeder end effector is calculated based on the corrected preliminary pose information.

[0013] As a further optimization of the above-mentioned multi-sensor fusion pose measurement method for tractors and seeders: the hybrid error compensation model includes a multiplicative error compensation part and an additive error compensation part.

[0014] A multi-sensor fusion pose measurement device for tractors and seeders includes: A dual-antenna BDS module, mounted on top of the tractor, is used to collect satellite positioning data; Several multi-axis IMU modules are installed on the tractor body, the seeder body, and the hydraulic suspension mechanism between the tractor and the seeder to collect inertial positioning data; A binocular vision sensor is installed on the body of the seeder to collect visual observation data; The data processing module is used to measure the precise pose information of the tractor and the multi-degree-of-freedom pose information of the end effector of the seeder using the multi-sensor fusion pose measurement method for tractors and seeders described above.

[0015] Beneficial Effects: This invention addresses the problems of low accuracy, poor continuity, and insufficient reliability in pose measurement of agricultural machinery combination systems in complex field environments. While balancing measurement performance and system cost, it first establishes a multi-sensor pose perception platform consisting of a dual-antenna BDS, a multi-nine-axis IMU, and a binocular vision sensor. Secondly, based on the connection relationship and motion coupling characteristics between the tracked tractor, hydraulic suspension mechanism, and rotary tiller / seeder, a kinematic model and a multi-sensor measurement model of the tractor-rotary tiller / seeder combination system are established. Then, the multi-source data acquired by the BDS, IMU, and vision sensors are subjected to denoising filtering, time synchronization, coordinate unification, and standardization. Based on this, using IMU inertial information as the state prediction input, an error state Kalman filter model (ESKF) is constructed, fusing BDS position and heading observation information with visual odometry pose information to achieve real-time estimation of the tractor's pose. Abnormal data detection is achieved through observation residual analysis. This invention further introduces DS evidence theory to construct a deep fusion model of multi-source information. Evidence synthesis is adaptively completed based on the consistency and conflict level of multi-source observation information, and the preliminary positioning results are optimized and fused to obtain high-precision and high-stability tractor pose information. Finally, the fused tractor pose information is input into a pre-established kinematic model of the combined system. The pose of the rotary tiller-seeder end effector is solved by combining ZYX Euler angle rotation transformation and DH parameter chain transfer relationship, and corrected using an error compensation model to obtain the six-degree-of-freedom pose information of the end effector. This invention can improve the accuracy, continuity, and stability of pose measurement of the tractor-rotary tiller-seeder combined system in complex field environments, and can provide reliable pose information support for intelligent agricultural machinery autonomous navigation, path planning, and precision seeding operations. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the overall process framework of the method of the present invention. Detailed Implementation

[0017] 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.

[0018] like Figure 1 As shown, the present invention first provides a multi-sensor fusion pose measurement method for tractors and seeders, including S1 to S4.

[0019] S1. During the operation of the tractor and seeder, raw state data is collected and preprocessed. The raw state data includes satellite positioning data, inertial positioning data, and visual observation data. The satellite positioning data is denoted as... Satellite positioning data is used to determine the absolute position and heading information of the tractor, while inertial positioning data is denoted as... Used to determine the motion status information of the tractor, the visual observation data is denoted as This is used to determine the relative pose changes of the tractor. After integrating satellite positioning data, inertial positioning data, and visual observation data, the original state data can be denoted as... , where K is time.

[0020] Specifically, the methods for preprocessing raw state data include noise reduction filtering, time synchronization, coordinate unification, and standardization. The specific formulas for each processing method are as follows.

[0021] The noise reduction filtering method is as follows: .

[0022] in, The initial observations for the original state data, These are the optimized observations after the original state data has been denoised and filtered. This represents the length of the filtering window.

[0023] During time synchronization, the timestamp of the inertial positioning data is used as the reference time, and then the satellite positioning data and visual observation data are mapped to the corresponding time to complete the time synchronization. After time synchronization, the original state data can be recorded as follows: ,in These represent satellite positioning data, inertial positioning data, and visual observation data after time synchronization, respectively.

[0024] The method for unifying the coordinates is as follows: .

[0025] in, For the observations in the sensor coordinate system, This is the coordinate transformation matrix from the sensor coordinate system to the global coordinate system. These are observations in the global coordinate system.

[0026] The standardization process is as follows: .

[0027] in, For observations before standardization, For all The mean, For all standard deviation These are the observations after standardization.

[0028] After preprocessing, the original state data is as follows: .

[0029] in, For satellite positioning data, For inertial positioning data, The data represents visual observations, and k represents time.

[0030] S2. The error state Kalman filter model is used to perform preliminary fusion of the preprocessed raw state data to obtain the preliminary estimation result of the tractor pose. Specifically, the core design of the error state Kalman filter model is as follows.

[0031] The system state vector of the error-state Kalman filter model is: .

[0032] in, Position status, In terms of speed state, For state quaternions, To achieve zero bias in the accelerometer, This is for zero bias of the gyroscope.

[0033] The state propagation equation of the error state Kalman filter model is: .

[0034] in, Let be the prior state estimate at time k. For the inertial positioning data used as input, Let be the combined kinematic state transition function of the tractor and the seeder. This refers to system process noise.

[0035] The error state propagation equation for the error state Kalman filter model is: .

[0036] in, Here is the state transition matrix. This is the noise-driven matrix.

[0037] Based on the error state propagation equation, the prediction error covariance matrix of the error state Kalman filter model is: ; in, For the prediction error covariance matrix, Let be the process noise covariance matrix.

[0038] Furthermore, a unified observation equation can be constructed: ; in, For the observation matrix, To observe noise.

[0039] The observation residual between the predicted state and the actual observed value is: .

[0040] Based on the observation residuals, the Kalman gain of the error-state Kalman filter model can be calculated as follows: ; in, To observe the noise covariance matrix.

[0041] Furthermore, the predicted state and error covariance matrix can be updated based on the Kalman gain, specifically as follows: ; .

[0042] Finally, the preliminary estimation results of the tractor pose generated by the error state Kalman filter model are as follows: ; in, The combination of represents the estimated position of the tractor. These are the estimated values ​​for roll angle, pitch angle, and yaw angle, respectively. S3. A multi-source information deep fusion model is used to optimize and fuse satellite positioning data, visual observation data, and the preliminary tractor pose estimation results to obtain the precise tractor pose information. The multi-source information deep fusion model is constructed based on the DS evidence theory. Specifically, the methods for optimizing and fusing satellite positioning data, visual observation data, and the preliminary tractor pose estimation results using the multi-source information deep fusion model include S31 to S36.

[0043] S31. Generate satellite positioning attitude estimation results based on satellite positioning data, and generate visual observation attitude estimation results based on visual observation data.

[0044] S32. Based on the satellite positioning attitude estimation results, visual observation attitude estimation results, and preliminary tractor pose estimation results, an evidence set containing multiple evidence sources is constructed. This evidence set is denoted as... ,in, Evidence generated based on satellite positioning attitude estimation results, visual observation attitude estimation results, or preliminary tractor pose estimation results.

[0045] S33. Calculate the basic probability allocation results for each evidence source based on the Gaussian membership function. More specifically, the basic probability allocation results are as follows: ; in, The error between other evidence sources and the evidence sources generated based on the preliminary estimation results of the tractor pose is considered. This represents the standard deviation of the error corresponding to the source of evidence.

[0046] Furthermore, the basic probability assignment results need to meet the following conditions. .

[0047] S34. Calculate the degree of evidence conflict in the evidence set based on the basic probability allocation results, and select the DS combination rule according to the degree of evidence conflict. The method for calculating the degree of evidence conflict is as follows: .

[0048] Furthermore, when selecting the DS combination rule based on the degree of evidence conflict, if the degree of evidence conflict is less than the preset conflict threshold, the standard DS combination rule is selected; if the degree of evidence conflict reaches the preset conflict threshold, the weighted modified DS combination rule is selected.

[0049] S35. Optimize and fuse all evidence sources using the DS combination rule to obtain fused evidence, and calculate the basic probability allocation result of the fused evidence. More specifically, the basic probability allocation result of the fused evidence is as follows: ; in, and These represent the basic probability allocation results of the evidence to be fused. Because there are multiple sources of evidence in the evidence set, a recursive approach is needed to complete the fusion, specifically: ; in, For operators of DS combination rules.

[0050] S36. Based on the basic probability allocation results of the fused evidence, calculate the confidence levels of the satellite positioning attitude estimation results, the visual observation attitude estimation results, and the preliminary tractor pose estimation results, and select the one with the highest confidence level as the precise pose information of the tractor. The specific method is as follows: ; in, Let A be the confidence level of evidence source A. Based on this, the precise pose information of the tractor obtained is: ; in, The combination of these coordinates provides the precise position coordinates of the tractor. These represent the precise roll angle, pitch angle, and yaw angle, respectively.

[0051] S4. Using a pre-constructed combined kinematics model, the multi-degree-of-freedom pose information of the seeder's end effector is calculated based on the tractor's precise pose information. The combined kinematics model includes a combined system pose transfer model and a combined system state-space model.

[0052] The pose transfer model of the combined system is as follows: .

[0053] in, Used to describe the position and attitude of the tractor relative to the global coordinate system. Used to describe the connection and motion relationship between the hydraulic suspension mechanism and the hydraulic press between the tractor and the seeder. Used to describe the relative motion relationship between the seeder and the hydraulic suspension mechanism. Used to describe the positional relationship between the end effector of the seeder and the seeder.

[0054] Furthermore, the attitude transformation relationship between the tractor coordinate system and the global coordinate system is as follows: ; ; ; .

[0055] Based on the structure of the hydraulic suspension mechanism and the seeder, a DH parameter chain can also be established, where the coordinate transformation matrix of the i-th link is: ; in, The length of the link. This is the linkage offset. For the link torsion angle, This refers to the joint angle.

[0056] Furthermore, based on the DH parameter chain, the pose transfer relationship between each connecting unit of the hydraulic suspension mechanism can be calculated in one step, thereby obtaining the pose transformation matrix of the seeder end effector relative to the tractor coordinate system, specifically: ; Therefore, the pose transformation matrix of the seeder end effector in the global coordinate system is: ; in, ,and .

[0057] The state-space model of the combined system is as follows: .

[0058] in, Let be the system state vector. To control the input amount, Let be the combined kinematic state transition function of the tractor and the seeder. This represents system process noise. Additionally, the system state vector... ,in, For position coordinates, For velocity components, These are roll angle, pitch angle, and yaw angle, respectively.

[0059] Furthermore, methods for calculating the multi-degree-of-freedom pose information of the seeder end effector based on the precise pose information of the tractor using a pre-constructed combined kinematic model include S41 to S43.

[0060] S41. Calculate the preliminary pose information of the seeder end effector in the global coordinate system based on the precise pose information of the tractor using a combined kinematics model.

[0061] S42. The preliminary pose information is corrected using a pre-constructed hybrid error compensation model. More specifically, the hybrid error compensation model includes a multiplicative error compensation part and an additive error compensation part. The hybrid error compensation model is as follows: ; in, This is the bearing error compensation coefficient matrix. This is the additive error compensation vector. The result is the pose after compensation.

[0062] S43. Calculate the multi-degree-of-freedom pose information of the seeder end effector based on the corrected preliminary pose information. More specifically, the multi-degree-of-freedom pose information of the seeder end effector is as follows: ; in, The combination determines the spatial position of the seeder's end effector. These are the roll angle, pitch angle, and heading angle of the end effector of the seeder.

[0063] The present invention further provides a multi-sensor fusion pose measurement device for tractors and seeders, including a dual-antenna BDS module, several multi-axis IMU modules, a binocular vision sensor and a data processing module.

[0064] A dual-antenna BDS module, mounted on top of the tractor, is used to collect satellite positioning data. In other embodiments of the invention, the dual-antenna BDS module can be replaced with other modules, also known as global satellite navigation modules, all of which are mature existing technologies and will not be described in detail here.

[0065] Several multi-axis IMU modules are installed on the tractor body, the seeder body, and the hydraulic suspension mechanism between the tractor and the seeder to collect inertial positioning data.

[0066] A binocular vision sensor is installed on the seeder body to collect visual observation data.

[0067] The data processing module is used to measure the precise pose information of the tractor and the multi-degree-of-freedom pose information of the end effector of the seeder using the multi-sensor fusion pose measurement method for tractors and seeders described above.

[0068] 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 multi-sensor fusion pose measurement method for tractors and seeders, characterized in that, include: Raw state data is collected during the operation of tractors and seeders and preprocessed. The raw state data includes satellite positioning data, inertial positioning data and visual observation data. The error state Kalman filter model is used to perform preliminary fusion of the preprocessed raw state data to obtain the preliminary estimation result of the tractor pose. A deep fusion model of multi-source information is used to optimize and fuse satellite positioning data, visual observation data and preliminary estimation results of tractor pose to obtain precise tractor pose information. The deep fusion model of multi-source information is constructed based on DS evidence theory. The multi-degree-of-freedom pose information of the seeder end effector is calculated based on the precise pose information of the tractor using a pre-constructed combined kinematic model.

2. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 1, characterized in that, Methods for preprocessing raw state data include noise reduction filtering, time synchronization, coordinate unification, and standardization. After preprocessing, the raw state data becomes: ； in, For satellite positioning data, For inertial positioning data, The data represents visual observations, and k represents time.

3. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 2, characterized in that, The noise reduction filtering method is as follows: ； in, The initial observations for the original state data, These are the optimized observations after the original state data has been denoised and filtered. This is the length of the filtering window; The method for unifying the coordinates is as follows: ； in, For the observations in the sensor coordinate system, This is the coordinate transformation matrix from the sensor coordinate system to the global coordinate system. For observations in the global coordinate system; The standardization process is as follows: ； in, For observations before standardization, For all The mean, For all standard deviation These are the observations after standardization.

4. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 1, characterized in that, The system state vector of the error-state Kalman filter model is: ； in, Position status, In terms of speed state, For state quaternions, To achieve zero bias in the accelerometer, Zero bias for the gyroscope; The state propagation equation of the error state Kalman filter model is: ； in, Let be the prior state estimate at time k. For the inertial positioning data used as input, Let be the combined kinematic state transition function of the tractor and the seeder. This refers to system process noise. The error state propagation equation for the error state Kalman filter model is: ； in, Here is the state transition matrix. This is the noise-driven matrix.

5. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 1, characterized in that, Methods for optimizing and fusing satellite positioning data, visual observation data, and preliminary tractor pose estimation results using a multi-source information deep fusion model include: Satellite positioning attitude estimation results are generated based on satellite positioning data, and visual observation attitude estimation results are generated based on visual observation data. An evidence set containing multiple evidence sources was constructed based on satellite positioning attitude estimation results, visual observation attitude estimation results, and preliminary tractor pose estimation results. The basic probability allocation results of each evidence source are calculated based on the Gaussian membership function. The degree of evidence conflict in the evidence set is calculated based on the basic probability allocation results, and the DS combination rule is selected according to the degree of evidence conflict. The DS combination rule is used to optimize and fuse all evidence sources to obtain fused evidence, and the basic probability allocation results of the fused evidence are calculated. Based on the basic probability allocation results of the fused evidence, the confidence levels of the satellite positioning attitude estimation results, the visual observation attitude estimation results, and the preliminary tractor pose estimation results are calculated, and the one with the highest confidence level is selected as the precise pose information of the tractor.

6. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 5, characterized in that, When selecting the DS combination rule based on the degree of evidence conflict, if the degree of evidence conflict is less than the preset conflict threshold, the standard DS combination rule is selected; if the degree of evidence conflict reaches the preset conflict threshold, the weighted modified DS combination rule is selected.

7. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 1, characterized in that, The combined kinematics model includes a pose transfer model of the combined system and a state space model of the combined system. The pose transfer model of the combined system is as follows: ； in, Used to describe the position and attitude of the tractor relative to the global coordinate system. Used to describe the connection and motion relationship between the hydraulic suspension mechanism and the hydraulic press between the tractor and the seeder. Used to describe the relative motion relationship between the seeder and the hydraulic suspension mechanism. Used to describe the positional relationship between the end effector of the seeder and the seeder; The state-space model of the combined system is as follows: ； in, Let be the system state vector. To control the input amount, Let be the combined kinematic state transition function of the tractor and the seeder. This refers to system process noise.

8. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 7, characterized in that, Methods for calculating the multi-degree-of-freedom pose information of a seeder end effector based on the precise pose information of a tractor using a pre-constructed combined kinematics model include: The preliminary pose information of the seeder end effector in the global coordinate system was calculated based on the precise pose information of the tractor using a combined kinematics model. The initial pose information is corrected using a pre-built hybrid error compensation model; The multi-degree-of-freedom pose information of the seeder end effector is calculated based on the corrected preliminary pose information.

9. The multi-sensor fusion pose measurement method for tractors and seeders as described in claim 8, characterized in that, The hybrid error compensation model includes a multiplicative error compensation component and an additive error compensation component.

10. A multi-sensor fusion pose measurement device for tractors and seeders, characterized in that, include: A dual-antenna BDS module, mounted on top of the tractor, is used to collect satellite positioning data; Several multi-axis IMU modules are installed on the tractor body, the seeder body, and the hydraulic suspension mechanism between the tractor and the seeder to collect inertial positioning data; A binocular vision sensor is installed on the body of the seeder to collect visual observation data; The data processing module is used to measure the precise pose information of the tractor and the multi-degree-of-freedom pose information of the end effector of the seeder using the multi-sensor fusion pose measurement method for tractors and seeders as described in any one of claims 1-9.

Images