Navigation precision self-checking system for intelligent inspection vehicle of farm

Abstract

The invention discloses a navigation precision self-checking system for a farm intelligent inspection vehicle, and the system comprises a data collection module which is used for collecting the environment parameters of a farm and the state parameters of the intelligent inspection vehicle in real time; the data processing module is used for carrying out data integration on the environment parameters and the state parameters and then carrying out data processing and analysis to obtain a target data set; the model construction module is used for constructing an initial multi-modal network model and training the initial multi-modal network model based on the target data set to obtain a target multi-modal network model; and the monitoring prediction module is used for monitoring the running state of the intelligent inspection vehicle in real time based on the target multi-mode network model, and automatically triggering a self-checking algorithm to calibrate the path according to the monitoring result when the detected actual path is deviated from the preset path. The navigation precision of the intelligent inspection vehicle is remarkably improved, manual intervention is reduced through automatic self-inspection and calibration, and the maintenance cost and time are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent patrol inspection, and in particular relates to a navigation accuracy self-inspection system of an intelligent patrol inspection vehicle for a farm. Background Art

[0002] With the development of modern aquaculture, the scale of farms continues to expand, and the traditional manual inspection method can no longer meet the needs of high efficiency and refined management. As an emerging automation equipment, the intelligent inspection vehicle can realize real-time monitoring and automatic management of the farm environment by carrying sensors and actuators. However, the intelligent inspection vehicle itself has the problem of navigation deviation accumulation. After a long period of operation, due to factors such as sensor errors and wheel slippage, the intelligent inspection vehicle may accumulate navigation deviations, causing the inspection path to deviate from the predetermined trajectory. The farm environment is changeable. When facing environmental changes, such as slippery ground and moving obstacles, the environmental adaptability of the existing intelligent inspection vehicle is insufficient, and it is difficult to quickly adjust the navigation strategy. At the same time, there is a lack of self-checking and calibration mechanisms. The existing intelligent inspection vehicles lack the function of automatically detecting navigation accuracy and mainly rely on manual regular calibration, which not only increases maintenance costs, but also makes it difficult to ensure the timeliness of calibration. In addition, due to the problem of navigation accuracy, the environmental monitoring data collected by the intelligent inspection vehicle may be biased, which affects the accuracy of the data, and thus affects the accurate assessment and management decision of the farm environment. It is known that the navigation accuracy of the intelligent inspection vehicle is directly related to the inspection effect and data accuracy. Most of the existing intelligent inspection vehicles lack effective self-detection and calibration mechanisms. Once the navigation system deviates, it will affect the efficiency and quality of the entire inspection process. Therefore, there is an urgent need for a navigation accuracy self-checking system for the intelligent inspection vehicle of the farm to complete the navigation accuracy self-check before the intelligent inspection, automatically detect and calibrate the navigation accuracy, to ensure the accuracy of the inspection data. Summary of the invention

[0003] In order to solve the above technical problems, the present invention provides a navigation accuracy self-checking system for a farm intelligent inspection vehicle, comprising:

[0004] Data acquisition module, used to collect the environmental parameters of the farm and the status parameters of the intelligent inspection vehicle in real time;

[0005] A data processing module, connected to the data acquisition module, is used to first integrate the environmental parameters and state parameters, and then process and analyze the data to obtain a target data set;

[0006] A model building module, connected to the data processing module, for building an initial multimodal network model, and training the initial multimodal network model based on the target data set to obtain a target multimodal network model;

[0007] The monitoring prediction module is connected to the model building module and is used to monitor the operating status of the intelligent inspection vehicle in real time based on the target multimodal network model. According to the monitoring results, when the actual path detected deviates from the preset path, the self-checking algorithm is automatically triggered to perform path calibration.

[0008] Preferably, the data acquisition module includes a first acquisition unit and a second acquisition unit;

[0009] The first acquisition unit is used to collect environmental parameters of the farm based on distributed sensors to obtain an environmental image;

[0010] The second collection unit is used to collect status parameters of the intelligent inspection vehicle;

[0011] The state parameters of the intelligent inspection vehicle include but are not limited to the real-time position, speed, acceleration, and direction of the intelligent inspection vehicle.

[0012] Preferably, the distributed sensors include but are not limited to laser radar, camera, industrial panoramic camera;

[0013] The industrial panoramic camera is arranged on the roof of the intelligent inspection vehicle to obtain a panoramic image of the road of the farm;

[0014] The industrial panoramic camera includes an automatic control unit, a photographing level control circuit unit, and a tilt photogrammetry unit;

[0015] The photographing level control circuit unit is provided with a flexible flat cable, and the tilt photogrammetry unit is flexibly connected to the automatic control unit through the flexible flat cable; the flexible flat cable is used to control the power on / off, photographing, and focusing of the camera, and to perform feedback detection on the state of the camera, detect whether the camera is turned on / off, detect whether the camera has completed the photographing action, and the real-time state of the camera;

[0016] The tilt photogrammetry unit is connected to the photographing level control circuit unit;

[0017] The oblique photogrammetry unit includes five high-precision cameras, which are tilted and placed at a 90-degree vertical angle to each other.

[0018] The second acquisition unit includes at least a GPS unit and an IMU laser radar;

[0019] The GPS unit is used to obtain the real-time position of the intelligent inspection vehicle;

[0020] The IMU laser radar is used to obtain speed and direction information of the intelligent inspection vehicle.

[0021] Preferably, the data processing module includes a data processing unit, an image processing unit, and a feature extraction unit;

[0022] The data processing unit is used to de-noise the original data, convert it into text data, and then merge the texts of different sources and formats;

[0023] The image processing unit is used to perform pixel preprocessing on the environment image and then perform image correction to obtain a target image;

[0024] The feature extraction unit is used to extract features from the processed images and data.

[0025] Preferably, the data processing unit includes a data cleaning unit, a text processing unit, and a data integration unit;

[0026] The data cleaning unit is used to remove noise and outliers in the original data and fill in missing data;

[0027] The text processing unit is used to perform word segmentation, dictionary building, text vectorization, and text normalization on the text data in sequence;

[0028] The data integration unit is used to unify and merge data from different sources and formats, establish associations between data, and obtain data views.

[0029] Preferably, the image processing unit includes a pixel processing unit and an image correction unit;

[0030] The pixel processing unit is used to perform pixel normalization, pixel centering, data enhancement, noise reduction, size adjustment and color space conversion on the environment image in sequence to obtain a target image for subsequent use;

[0031] The image correction unit is used to perform radiation correction, geometric correction, image registration, image fusion, light and color uniformity, image stitching and image cropping on the image processed by the pixel processing unit in sequence to obtain a target optical remote sensing image.

[0032] Preferably, the feature extraction unit comprises:

[0033] The time domain feature extraction unit is used to describe the overall characteristics of the spectral signal through mean, variance and standard deviation;

[0034] A frequency domain feature extraction unit is used to convert the spectral signal into the frequency domain through Fourier transform, so as to analyze different frequency components; or to describe the distribution of the spectral signal in the frequency domain through power spectrum density;

[0035] A wavelet capability feature extraction unit, used to extract energy of different frequency bands using wavelet transform;

[0036] The adaptive feature extraction unit is used to adjust the parameters of feature extraction according to real-time data using an adaptive filter or an adaptive signal processing algorithm.

[0037] Preferably, the model construction module includes a model construction unit, a model training unit, and a model fusion unit;

[0038] The model building unit is used to build a convolutional neural network model and a feedforward neural network model respectively;

[0039] The model training unit is used to train the convolutional neural network model based on the target data set, and then use the output of the trained convolutional neural network model as the input of the feedforward neural network model to train the feedforward neural network model to obtain the target multimodal network model.

[0040] Preferably, the model building unit includes a first building unit and a second building unit;

[0041] The first construction unit is used to construct two identical convolutional neural network models, and add a BatchNorm2d layer, a nonlinear ReLU layer and an attention mechanism module after the two-dimensional convolution layer in the convolutional neural network model to generate a twin adversarial network model;

[0042] The second construction unit is used to generate a convolutional neural network group based on the twin adversarial network model;

[0043] The model training unit also includes a layer selection unit and a loss function introduction unit;

[0044] The layer selection unit is used to delete the input layer and the output layer of the multimodal network model;

[0045] The loss function introduction unit is used to introduce the cross entropy loss function and the attention mechanism module into the multimodal network model to reconstruct the model and obtain the target multimodal neural network.

[0046] Preferably, the monitoring prediction module includes a path planning and generation unit, a navigation accuracy self-checking algorithm unit, a real-time calibration and adjustment unit, an environmental feature recognition and matching unit, and a control execution unit;

[0047] The path planning and generation unit is used to generate the shortest path or the optimal path for the vehicle to travel according to the task requirements and environmental information;

[0048] The navigation accuracy self-checking algorithm unit is used to automatically check the navigation accuracy by comparing the deviation between the actual driving path and the preset path;

[0049] The real-time calibration and adjustment unit is used to automatically start the calibration procedure when the navigation deviation exceeds a preset threshold value, adjust the vehicle's driving state or sensor parameters, and ensure that the vehicle returns to the correct path;

[0050] The environmental feature recognition and matching unit is used to recognize and match specific features in the environment based on image recognition technology to provide auxiliary positioning information; the specific features include but are not limited to road signs, fences, and buildings;

[0051] The control execution unit is used to execute specific control commands according to the self-test results and calibration instructions to ensure that the vehicle travels along a predetermined path; the control commands include but are not limited to adjusting the vehicle speed and steering angle.

[0052] Compared with the prior art, the present invention has the following advantages and technical effects:

[0053] The present invention significantly improves the navigation accuracy of the intelligent inspection vehicle through multi-sensor fusion and environmental feature recognition. The real-time self-checking and calibration mechanism enables the system to quickly adapt to changes in the farm environment, enhancing the system's adaptability. Automated self-checking and calibration reduce manual intervention, maintenance costs and time. It also further improves the accuracy of inspection data and provides reliable decision support for farm management. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0055] Figure 1 Schematic diagram of the system structure of an embodiment of the present invention. DETAILED DESCRIPTION

[0056] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0057] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0058] like Figure 1 As shown, in this embodiment, a navigation accuracy self-checking system for an intelligent inspection vehicle of a farm is provided, comprising:

[0059] Data acquisition module, used to collect the environmental parameters of the farm and the status parameters of the intelligent inspection vehicle in real time;

[0060] The data processing module is connected to the data acquisition module and is used to first integrate the environmental parameters and state parameters, and then process and analyze the data to obtain the target data set;

[0061] A model building module, connected to the data processing module, is used to build an initial multimodal network model and train the initial multimodal network model based on a target data set to obtain a target multimodal network model;

[0062] The monitoring and prediction module is connected to the model building module and is used to monitor the operating status of the intelligent inspection vehicle in real time based on the target multimodal network model. According to the monitoring results, when the actual path detected deviates from the preset path, the self-check algorithm is automatically triggered to perform path calibration.

[0063] Further, the data acquisition module includes a first acquisition unit and a second acquisition unit;

[0064] The first acquisition unit is used to collect environmental parameters of the farm based on distributed sensors to obtain environmental images;

[0065] The second collection unit is used to collect the status parameters of the intelligent inspection vehicle;

[0066] The status parameters of the intelligent patrol vehicle include but are not limited to the real-time position, speed, acceleration, and direction of the intelligent patrol vehicle.

[0067] Furthermore, the distributed sensors include but are not limited to LiDAR, cameras, and industrial panoramic cameras;

[0068] The industrial panoramic camera is installed on the roof of the intelligent inspection vehicle to obtain panoramic images of the roads in the farm;

[0069] The industrial panoramic camera includes an automatic control unit, a photo level control circuit unit, and a tilt photogrammetry unit;

[0070] The photographing level control circuit unit is provided with a flexible cable, and the tilt photogrammetry unit is flexibly connected to the automatic control unit through the flexible cable; the flexible cable is used to control the camera's power on / off, photographing, focusing, and feedback detection of the camera's status, detect whether the camera is powered on / off, detect whether the camera has completed the photographing action, and the camera's real-time status;

[0071] The tilt photogrammetry unit is connected to the photo level control circuit unit;

[0072] The oblique photogrammetry unit consists of five high-precision cameras that are tilted and placed at 90 degrees to each other.

[0073] Furthermore, the automatic control unit is an STM32F103 microcontroller; the crystal oscillator of the STM32F103 microcontroller is 8MHZ, and the instruction execution interval is 1us. The STM32F103 chip contains an AD conversion module, and the AD conversion module includes a plurality of AD conversion units, and the AD conversion unit is used for AD conversion to measure the voltage state of the camera feedback information. When using AD conversion to measure the voltage state of the feedback information of the five cameras, there is no need to carry a complex external AD conversion circuit, and only the STM32F103 chip itself is needed to complete it; the serial port of the STM32F103 chip will connect the NB-IoT remote data control module, the Bluetooth control module, the GNSS (Beidou / GPS) positioning module and the TF card recording module; the GPIO port of the STM32F103 chip will connect the camera's photo control unit and the power on / off control unit. Each camera requires two GPIO ports to control the power on / off and photo taking. The five cameras of this system require a total of 10 GPIO ports to control the power on / off and photo taking of the five cameras; in addition, each camera requires two AD conversions, and a total of 10 AD conversions are required to perform feedback detection on whether the camera is turned on and whether it is taking a photo, but the STM32F103 chip has fewer AD conversion ports, so a rotating line selection method of AD conversion is required here; in addition, this system includes five cameras, a central control system, and various functional units, all of which use centralized power supply to maintain system consistency. The camera at least includes a photographing control unit and a power on / off control unit; the photographing control unit and the power on / off control unit of the camera are connected to the GPIO port of the STM32F103 microcontroller.

[0074] The second acquisition unit includes at least a GPS unit and an IMU laser radar;

[0075] The GPS unit is used to obtain the real-time location of the intelligent inspection vehicle;

[0076] IMU lidar is used to obtain speed and direction information of the intelligent inspection vehicle.

[0077] Further, the data processing module includes a data processing unit, an image processing unit, and a feature extraction unit;

[0078] The data processing unit is used to denoise the original data, convert it into text data, and then merge the texts of different sources and formats;

[0079] The image processing unit is used to perform pixel preprocessing on the environment image and then perform image correction to obtain the target image;

[0080] The feature extraction unit is used to extract features from the processed images and data.

[0081] Further, the data processing unit includes a data cleaning unit, a text processing unit, and a data integration unit;

[0082] The data cleaning unit is used to remove noise and outliers from the original data and fill in missing data;

[0083] The text processing unit is used to perform word segmentation, dictionary building, text vectorization, and text normalization on the text data in sequence;

[0084] The data integration unit is used to unify and merge data from different sources and formats, establish associations between data, and obtain data views.

[0085] Furthermore, the data cleaning unit includes:

[0086] A filtering unit is used to reduce instantaneous fluctuations by calculating the average value of a stable signal within a preset time through a moving average filter, and then remove burst noise by replacing each data point with the median value within the signal window based on a median filter;

[0087] A wavelet transform unit, used for decomposing signal data into components of different frequencies by wavelet transform, and then reconstructing the signal by discarding or adjusting high frequency noise;

[0088] A Kalman filter unit for minimizing the estimation error by estimating the current state and the measurement value through a recursive filtering method;

[0089] The singular value decomposition unit is used to decompose the matrix into singular vectors and singular values, and realize noise reduction by truncating high-frequency components.

[0090] Further, the image processing unit includes a pixel processing unit and an image correction unit;

[0091] The pixel processing unit is used to perform pixel normalization, pixel centering, data enhancement, noise reduction, size adjustment and color space conversion on the environment image in sequence to obtain the target image for subsequent use;

[0092] The image correction unit is used to perform radiation correction, geometric correction, image registration, image fusion, light and color uniformity, image stitching and image cropping on the image processed by the pixel processing unit in sequence to obtain the target optical remote sensing image.

[0093] Furthermore, the feature extraction unit includes:

[0094] The time domain feature extraction unit is used to describe the overall characteristics of the spectral signal through mean, variance and standard deviation;

[0095] A frequency domain feature extraction unit is used to convert the spectral signal into the frequency domain through Fourier transform, so as to analyze different frequency components; or to describe the distribution of the spectral signal in the frequency domain through power spectrum density;

[0096] A wavelet capability feature extraction unit, used to extract energy of different frequency bands using wavelet transform;

[0097] The adaptive feature extraction unit is used to adjust the parameters of feature extraction according to real-time data using an adaptive filter or an adaptive signal processing algorithm.

[0098] Furthermore, the model construction module includes a model construction unit, a model training unit, and a model fusion unit;

[0099] The model building unit is used to build a convolutional neural network model and a feedforward neural network model respectively;

[0100] The model training unit is used to train the convolutional neural network model based on the target data set, and then use the output of the trained convolutional neural network model as the input of the feedforward neural network model to train the feedforward neural network model to obtain the target multimodal network model.

[0101] Further, the model building unit includes a first building unit and a second building unit;

[0102] The first construction unit is used to construct two identical convolutional neural network models, adding a BatchNorm2d layer, a nonlinear ReLU layer and an attention mechanism module after the two-dimensional convolution layer in the convolutional neural network model to generate a twin adversarial network model;

[0103] The second building unit is used to generate a convolutional neural network group based on the twin adversarial network model;

[0104] The model training unit also includes a layer selection unit and a loss function introduction unit;

[0105] The layer selection unit is used to delete the input layer and output layer of the multimodal network model;

[0106] The loss function introduction unit is used to introduce the cross entropy loss function and the attention mechanism module into the multimodal network model to rebuild the model and obtain the target multimodal neural network.

[0107] Furthermore, the convolutional neural network includes a first convolutional layer module, a second convolutional layer module, a dense block module, a third convolutional layer module, a fully connected layer unit module, and a probability conversion unit module;

[0108] The first convolutional layer module includes: a first convolutional layer module convolutional layer unit, a first convolutional layer module batch normalization layer unit, and a first convolutional layer module activation function unit;

[0109] The second convolutional layer module includes: a second convolutional layer module convolutional layer unit, a second convolutional layer module batch normalization layer unit, a second convolutional layer module activation function unit, and a second convolutional layer module maximum pooling layer unit;

[0110] The dense block module includes several dense connection layer modules, wherein the dense connection layer module includes: a first batch of normalized layer units of the dense connection layer module, a first activation function unit of the dense connection layer module, a first convolutional layer unit of the dense connection layer module, a second batch of normalized layer units of the dense connection layer module, a second activation function unit of the dense connection layer module, and a second convolutional layer unit of the dense connection layer module;

[0111] The third convolutional layer module includes: a third convolutional layer module convolutional layer unit, a third convolutional layer module batch normalization layer unit, a third convolutional layer module activation function unit, and a third convolutional layer module average pooling layer unit.

[0112] Furthermore, the monitoring prediction module includes a path planning and generation unit, a navigation accuracy self-checking algorithm unit, a real-time calibration and adjustment unit, an environmental feature recognition and matching unit, and a control execution unit;

[0113] The path planning and generation unit is used to generate the shortest path or optimal path for the vehicle according to the task requirements and environmental information;

[0114] The navigation accuracy self-checking algorithm unit is used to automatically check the navigation accuracy by comparing the deviation between the actual driving path and the preset path;

[0115] The real-time calibration and adjustment unit is used to automatically start the calibration procedure when the navigation deviation exceeds the preset threshold, adjust the vehicle's driving state or sensor parameters, and ensure that the vehicle returns to the correct path;

[0116] The environmental feature recognition and matching unit is used to recognize and match specific features in the environment based on image recognition technology to provide auxiliary positioning information; specific features include but are not limited to road signs, fences, and buildings;

[0117] The control execution unit is used to execute specific control commands according to the self-test results and calibration instructions to ensure that the vehicle travels along the predetermined path; the control commands include but are not limited to adjusting the vehicle speed and steering angle.

[0118] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A navigation accuracy self-checking system for a farm intelligent inspection vehicle, characterized in that: include: Data acquisition module, used to collect the environmental parameters of the farm and the status parameters of the intelligent inspection vehicle in real time; A data processing module, connected to the data acquisition module, is used to first integrate the environmental parameters and state parameters, and then process and analyze the data to obtain a target data set; A model building module, connected to the data processing module, for building an initial multimodal network model, and training the initial multimodal network model based on the target data set to obtain a target multimodal network model; The monitoring prediction module is connected to the model building module and is used to monitor the operating status of the intelligent inspection vehicle in real time based on the target multimodal network model. According to the monitoring results, when the actual path detected deviates from the preset path, the self-checking algorithm is automatically triggered to perform path calibration.

2. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 1 is characterized in that: The data acquisition module includes a first acquisition unit and a second acquisition unit; The first acquisition unit is used to collect environmental parameters of the farm based on distributed sensors to obtain an environmental image; The second collection unit is used to collect status parameters of the intelligent inspection vehicle; The state parameters of the intelligent inspection vehicle include but are not limited to the real-time position, speed, acceleration, and direction of the intelligent inspection vehicle.

3. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 2 is characterized in that: The distributed sensors include but are not limited to laser radar, cameras, and industrial panoramic cameras; The industrial panoramic camera is arranged on the roof of the intelligent inspection vehicle to obtain a panoramic image of the road of the farm; The industrial panoramic camera includes an automatic control unit, a photographing level control circuit unit, and a tilt photogrammetry unit; The photographing level control circuit unit is provided with a flexible flat cable, and the tilt photogrammetry unit is flexibly connected to the automatic control unit through the flexible flat cable; the flexible flat cable is used to control the power on / off, photographing, and focusing of the camera, and to perform feedback detection on the state of the camera, detect whether the camera is turned on / off, detect whether the camera has completed the photographing action, and the real-time state of the camera; The tilt photogrammetry unit is connected to the photographing level control circuit unit; The oblique photogrammetry unit includes five high-precision cameras, which are tilted and placed at a 90-degree vertical angle to each other. The second acquisition unit includes at least a GPS unit and an IMU laser radar; The GPS unit is used to obtain the real-time position of the intelligent inspection vehicle; The IMU laser radar is used to obtain speed and direction information of the intelligent inspection vehicle.

4. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 1 is characterized in that: The data processing module includes a data processing unit, an image processing unit, and a feature extraction unit; The data processing unit is used to de-noise the original data, convert it into text data, and then merge the texts of different sources and formats; The image processing unit is used to perform pixel preprocessing on the environment image and then perform image correction to obtain a target image; The feature extraction unit is used to extract features from the processed images and data.

5. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 4 is characterized in that: The data processing unit includes a data cleaning unit, a text processing unit, and a data integration unit; The data cleaning unit is used to remove noise and outliers in the original data and fill in missing data; The text processing unit is used to perform word segmentation, dictionary building, text vectorization, and text normalization on the text data in sequence; The data integration unit is used to unify and merge data from different sources and formats, establish associations between data, and obtain data views.

6. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 4 is characterized in that: The image processing unit includes a pixel processing unit and an image correction unit; The pixel processing unit is used to perform pixel normalization, pixel centering, data enhancement, noise reduction, size adjustment and color space conversion on the environment image in sequence to obtain a target image for subsequent use; The image correction unit is used to perform radiation correction, geometric correction, image registration, image fusion, light and color uniformity, image stitching and image cropping on the image processed by the pixel processing unit in sequence to obtain a target optical remote sensing image.

7. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 4 is characterized in that: The feature extraction unit comprises: The time domain feature extraction unit is used to describe the overall characteristics of the spectral signal through mean, variance and standard deviation; A frequency domain feature extraction unit is used to convert the spectral signal into the frequency domain through Fourier transform, so as to analyze different frequency components; or to describe the distribution of the spectral signal in the frequency domain through power spectrum density; A wavelet capability feature extraction unit, used to extract energy of different frequency bands using wavelet transform; The adaptive feature extraction unit is used to adjust the parameters of feature extraction according to real-time data using an adaptive filter or an adaptive signal processing algorithm.

8. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 1 is characterized in that: The model construction module includes a model construction unit, a model training unit, and a model fusion unit; The model building unit is used to build a convolutional neural network model and a feedforward neural network model respectively; The model training unit is used to train the convolutional neural network model based on the target data set, and then use the output of the trained convolutional neural network model as the input of the feedforward neural network model to train the feedforward neural network model to obtain the target multimodal network model.

9. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 8 is characterized in that: The model building unit includes a first building unit and a second building unit; The first construction unit is used to construct two identical convolutional neural network models, and add a BatchNorm2d layer, a nonlinear ReLU layer and an attention mechanism module after the two-dimensional convolution layer in the convolutional neural network model to generate a twin adversarial network model; The second construction unit is used to generate a convolutional neural network group based on the twin adversarial network model; The model training unit also includes a layer selection unit and a loss function introduction unit; The layer selection unit is used to delete the input layer and the output layer of the multimodal network model; The loss function introduction unit is used to introduce the cross entropy loss function and the attention mechanism module into the multimodal network model to reconstruct the model and obtain the target multimodal neural network.

10. The navigation accuracy self-checking system of the farm intelligent inspection vehicle according to claim 1 is characterized in that: The monitoring prediction module includes a path planning and generation unit, a navigation accuracy self-checking algorithm unit, a real-time calibration and adjustment unit, an environmental feature recognition and matching unit, and a control execution unit; The path planning and generation unit is used to generate the shortest path or the optimal path for the vehicle to travel according to the task requirements and environmental information; The navigation accuracy self-checking algorithm unit is used to automatically check the navigation accuracy by comparing the deviation between the actual driving path and the preset path; The real-time calibration and adjustment unit is used to automatically start the calibration procedure when the navigation deviation exceeds a preset threshold value, adjust the vehicle's driving state or sensor parameters, and ensure that the vehicle returns to the correct path; The environmental feature recognition and matching unit is used to recognize and match specific features in the environment based on image recognition technology to provide auxiliary positioning information; the specific features include but are not limited to road signs, fences, and buildings; The control execution unit is used to execute specific control commands according to the self-test results and calibration instructions to ensure that the vehicle travels along a predetermined path; the control commands include but are not limited to adjusting the vehicle speed and steering angle.

Images