Locomotive inertial measurement unit calibration method and device based on rotating target detection model

Abstract

The invention belongs to the technical field of locomotive state detection, and particularly relates to a locomotive inertial measurement unit calibration method and device based on a rotating target detection model. The method comprises the following steps: acquiring an image of a locomotive cab console; inputting the image into a locomotive handle detection model, identifying each handle in the image, and determining the relative position of the handle and the corresponding handle table; wherein the locomotive handle detection model is obtained by training a rotating target detection model by using a handle detection data set, and the handle detection data set is a locomotive cab operation table image set marked with a handle and a corresponding handle table; determining the state of the handle according to the relative position of the handle and the corresponding handle table; determining the running state of the locomotive according to the state of the handle; and calibrating the inertial measurement unit of the locomotive according to the determined running state of the locomotive. Therefore, the speed measurement precision and the positioning precision of the locomotive are improved, and the normal operation of the locomotive is ensured.

Description

Technical Field

[0001] The invention belongs to the technical field of locomotive state detection, and in particular relates to a locomotive inertial measurement unit calibration method and device based on a rotating target detection model. Background Art

[0002] Since Beidou satellite signals may be blocked in complex terrains such as urban canyons (high-rise building areas) or mountainous areas, the positioning accuracy will be reduced. At this time, some locomotive inertial measurement units are usually used in the prior art to determine the locomotive running status, so as to estimate and correct the locomotive position and speed according to the locomotive running status. However, locomotive inertial measurement units such as gyroscopes, accelerometers and magnetometers are affected by circuit noise, electromagnetic interference, mechanical vibration and temperature, resulting in unstable locomotive status detection, which causes problems in the calculation of the locomotive speed and position. Not only can the locomotive position not be correctly estimated and corrected, but it may even increase the error of locomotive positioning and affect the normal operation of the locomotive. Summary of the invention

[0003] The purpose of the present invention is to provide a locomotive inertial measurement unit calibration method and device based on a rotating target detection model, so as to solve the technical problem that the locomotive inertial measurement unit is affected by circuit noise, electromagnetic interference, mechanical vibration and temperature, resulting in errors in the calculation of the locomotive speed and position.

[0004] In order to solve the above technical problems, the present invention provides a locomotive inertial measurement unit calibration method based on a rotating target detection model, comprising the following steps: Acquire an image of a locomotive cab operating console; input the image into a locomotive handle detection model, identify each handle in the image, and determine the relative position of the handle and the corresponding handle platform; wherein the locomotive handle detection model is obtained by training a rotating target detection model using a handle detection data set, and the handle detection data set is a locomotive cab operating console image set with handles and corresponding handle platforms marked; Determine the state of the handle according to the relative position of the handle and the corresponding handle platform; determine the running state of the locomotive according to the state of the handle; The locomotive inertial measurement unit is calibrated according to the determined locomotive operating status.

[0005] Further, the process of determining the running state of the locomotive according to the state of the handle includes: determining the running state of the locomotive according to the state of the handle and the locomotive model; The method for obtaining the locomotive model is: inputting the image of the vehicle cab control panel into the locomotive model recognition model to obtain the locomotive model; the locomotive model recognition model is obtained by training the image classification model using the locomotive model data set, and the locomotive model data set is a locomotive cab control panel image set containing locomotive model labels.

[0006] Furthermore, the process of calibrating the locomotive inertial measurement unit according to the locomotive running state includes: when the locomotive running state is stopped, calibrating the initial value of the locomotive inertial measurement unit.

[0007] Further, the locomotive inertial measurement unit includes at least one of a gyroscope, an accelerometer and a magnetometer.

[0008] Furthermore, the rotating target detection model is a deep learning model.

[0009] Furthermore, the image classification model trained using the locomotive model dataset is a deep learning model.

[0010] Further, the locomotive speed and position are determined using the locomotive inertial measurement unit after the locomotive inertial measurement unit has been calibrated.

[0011] Furthermore, the image of the locomotive cab control panel is a panoramic image of the locomotive cab control panel.

[0012] The present invention is a pioneering invention, and its beneficial effects are as follows: the locomotive inertial measurement unit calibration method based on the rotating target detection model of the present invention can use a new method to accurately determine the running state of the locomotive during the operation of the locomotive, so as to calibrate the locomotive inertial measurement unit according to the running state of the locomotive, wherein the method for accurately determining the running state of the locomotive is: obtaining an image of the locomotive cab control panel, identifying the locomotive handle and the corresponding handle platform, and then determining the state of the handle according to the relative position of the handle and the handle platform, and determining the current running state of the locomotive according to the handle state. Since the position and state of the locomotive handle will not be affected by circuit noise, electromagnetic interference, mechanical vibration and temperature, the method of the present invention can accurately determine the running state of the locomotive under any circumstances, thereby calibrating the locomotive inertial measurement unit according to the running state of the locomotive, ensuring the detection accuracy of the locomotive inertial measurement unit, thereby improving the locomotive speed measurement accuracy and positioning accuracy, and ensuring the normal operation of the locomotive.

[0013] To solve the above technical problems, the present invention also provides a locomotive inertial measurement unit calibration device based on a rotating target detection model, comprising a processor, which implements the steps of the locomotive inertial measurement unit calibration method of the present invention when executing a computer program.

[0014] The present invention is a pioneering invention, and its beneficial effects are the same as those of the locomotive inertial measurement unit calibration method based on the rotating target detection model of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1is a flow chart of a method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to a method embodiment of the present invention; Figure 2a , Figure 2b , Figure 2c and Figure 2d is a panoramic view of the cab control panel of different models of locomotives according to the method embodiment of the present invention; Figure 3a , Figure 3b , Figure 3c and Figure 3d It is a panoramic view of a cab control panel of a certain model of locomotive with handles located at different positions according to the method embodiment of the present invention; Figure 4a , Figure 4b , Figure 4c and Figure 4d It is a panoramic view of the cab control panel of another model of locomotive according to the method embodiment of the present invention, with handles located in different positions. DETAILED DESCRIPTION

[0016] The present invention provides a method and device for calibrating a locomotive inertial measurement unit based on a rotating target detection model. During the operation of the locomotive, a new method can be used to accurately determine the operating state of the locomotive, so as to calibrate the locomotive inertial measurement unit according to the locomotive operating state. The method for accurately determining the locomotive operating state is as follows: acquiring an image of the locomotive cab control panel, identifying the locomotive handle, and then determining the current operating state of the locomotive according to the state and position of the handle. Since the position and state of the locomotive handle will not be affected by circuit noise, electromagnetic interference, mechanical vibration and temperature, the method of the present invention can accurately determine the operating state of the locomotive under any circumstances, thereby calibrating the locomotive inertial measurement unit according to the locomotive operating state, ensuring the detection accuracy of the locomotive inertial measurement unit, thereby improving the locomotive speed measurement accuracy and positioning accuracy, and ensuring the normal operation of the locomotive.

[0017] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail below with reference to the accompanying drawings and embodiments.

[0018] Method Example: The locomotive inertial measurement unit calibration method based on the rotating target detection model of the present invention uses the control panel image to judge the running state of the locomotive, and then calibrates the locomotive inertial measurement unit according to the running state of the locomotive. After calibration, the state and speed of the locomotive are updated according to the detection results of the inertial measurement unit, so that the locomotive speed and position can be determined more accurately.

[0019] like Figure 1 As shown, the specific steps are as follows: Step 1: Acquire the image of the locomotive cab control panel.

[0020] In this embodiment, a panoramic video camera installed on the control panel of the locomotive cab is used to capture images of the locomotive control panel. Therefore, the panoramic video camera captures the picture to detect the state of the driver's control handle, and the state of the handle can determine the operating state of the locomotive (stationary, forward running, reverse running, acceleration and deceleration, etc.).

[0021] In this embodiment, the image of the locomotive driving console is a panoramic image, which can include all images of the console without any omissions, and can improve the accuracy of locomotive model recognition and handle status recognition.

[0022] Step 2: Determine the handle status.

[0023] The image of the locomotive cab control panel obtained in step 1 is input into the locomotive handle detection model, from which the images of the handle and the corresponding handle platform are identified, and the relative positions of the handle and the corresponding handle platform are determined; the state of the handle is determined according to the relative positions of the handle and the corresponding handle platform. The state of the handle includes the handle position in front of, behind, or in the middle of the handle platform. When the handle platform has a scale line, the state of the handle includes the handle being above, below, or in the middle of the scale line.

[0024] In this embodiment, the locomotive handle detection model is obtained by training the rotating target detection model using the handle detection data set, and the locomotive handle state is identified based on the relative position of the handle and the handle. Among them, the handle detection data set is a locomotive cab operating console image set with the handle and the corresponding handle console marked, such as Figure 2a~Figure 2d The figure shows the schematic diagram of the control console handles of different models of cabs. The handles and the corresponding handles are marked with boxes in the figure. The marked boxes include not only the position information but also the rotation angle information. Figure 2a There is only one handle on the console. Figure 2b and Figure 2c There are two handles on the console. Figure 2d The control console includes three handles. In this embodiment, the rotating target detection model is a deep learning model, and YOLO, R-CNN, SSD, etc. can be selected.

[0025] Step 3: Determine the locomotive operating status.

[0026] In this embodiment, after determining the state and position of the handle, the running state of the locomotive (including stationary, forward running and reverse running) can be determined in combination with the locomotive model.

[0027] Figure 3a~Figure 3dThis is a panoramic picture of the control panel of a certain type of locomotive. The control panel includes two handles. When handle 2 is below the scale line, the locomotive is in a stationary state; when handle 2 leaves the scale line to the middle position or the front, and handle 1 is above the scale line, the locomotive is in a forward running state; when handle 2 is in the middle position of the scale line and handle 1 is in the middle of the scale line, the locomotive is in a reverse running or standby state.

[0028] Figure 4a~Figure 4d This is a panoramic picture of another type of locomotive control panel. This model has only one handle and no scale lines. When the handle is at the rear, the locomotive is stationary; when the handle is pushed to the front of the handle panel, the locomotive is in a forward running state; when the locomotive is running, the driver's hand will hold the handle. When the handle holding feature is obtained through target detection technology analysis, it can be judged that the locomotive is in a running state. Since the driver's hand will be placed on the handle when the locomotive is running, it may block the handle and cause the handle to fail to be detected. Therefore, this embodiment also analyzes the forward, middle, and backward features of the handle holding hand to determine the position of the handle.

[0029] Due to different locomotive models, the position, number and corresponding locomotive status of the handles may also be different. Therefore, in this embodiment, when determining the running status of the locomotive based on the handles, it is also necessary to determine the running status of the locomotive in combination with the locomotive model.

[0030] In this embodiment, the method for obtaining the locomotive model includes: inputting the acquired image of the vehicle cab control panel into the locomotive model recognition model to obtain the locomotive model; the locomotive model recognition model is obtained by training the image classification model using the locomotive model data set, and the locomotive model data set is a locomotive cab control panel image set containing the locomotive model label.

[0031] In this embodiment, the image classification model trained with the locomotive model data set is a deep learning model, specifically a deep convolutional neural network; as other implementations, the image recognition model based on deep learning can also be other image recognition models, such as recurrent neural networks, long short-term memory networks, and Transformer architectures.

[0032] The image recognition model is used to identify the locomotive model by recognizing the panoramic image of the locomotive cab control panel, and the locomotive model can be accurately determined according to the different structures and layouts of the control panel, thereby providing a basis for subsequently determining the locomotive state according to the handle position. As another implementation, the locomotive model can also be obtained by other means, for example, the driver manually inputs the locomotive model.

[0033] Step 4: Calibrate the locomotive inertial measurement unit according to the locomotive's operating status.

[0034] After determining the running status of the locomotive according to the position of the handle, the locomotive running inertial measurement unit can be calibrated to improve the accuracy of locomotive positioning.

[0035] In this embodiment, in order to make the calibration more accurate, the method for calibrating the detection device according to the running state of the locomotive includes: when it is determined that the locomotive stops running, calibrating the locomotive inertial measurement unit and setting the initial value of the state detection device, such as the initial position, speed and posture of the device. As another embodiment, the inertial measurement unit can also be calibrated according to the running state of the locomotive when the locomotive is not stopped.

[0036] In this embodiment, the calibrated locomotive inertial measurement unit includes an accelerometer, a magnetometer, and a gyroscope. As another embodiment, the calibrated locomotive inertial measurement unit may also include any one or two of them.

[0037] Step 5: Use the detection data of the calibrated locomotive inertial measurement unit to estimate the locomotive position and speed, and correct the locomotive positioning in combination with the positioning signal.

[0038] By analyzing the handle through panoramic video, the position of the locomotive handle can be accurately determined, thereby accurately determining the start and stop status of the locomotive. When the locomotive stops, the multimodal data is integrated for automatic calibration, which improves the accuracy of the predicted speed and position.

[0039] Device Example: The locomotive inertial measurement unit calibration device based on the rotating target detection model of the present invention includes a processor, and the processor implements the steps of the locomotive inertial measurement unit calibration method based on the rotating target detection model as described in the method embodiment of the present invention when executing a computer program. The specific process, principle and beneficial effects of the method have been described in detail in the method embodiment, and will not be repeated in this embodiment.

[0040] The processor may be a microprocessor MCU, a programmable logic device FPGA or other processing device.

[0041] The present invention provides a method and device for calibrating a locomotive inertial measurement unit based on a rotating target detection model. By using an image recognition handle, the state of the locomotive handle can be accurately determined, thereby accurately determining the start and stop state of the locomotive. Since the image recognition method is not affected by circuit interference, magnetic field interference, and temperature, the inertial measurement unit can be calibrated when the accuracy is affected, thereby ensuring the measurement accuracy of the inertial measurement unit and improving the accuracy of the locomotive position and speed calculation. Furthermore, when determining the running state of the locomotive, the present invention, in addition to using the handle state, also combines the model of the locomotive, which can more accurately determine the running state of the locomotive and ensure the calibration effect of the inertial measurement element.

Claims

1. A locomotive inertial measurement unit calibration method based on a rotating target detection model, characterized in that: The following steps are involved: Acquire an image of a locomotive cab operating console; input the image into a locomotive handle detection model, identify each handle in the image, and determine the relative position of the handle and the corresponding handle platform; wherein the locomotive handle detection model is obtained by training a rotating target detection model using a handle detection data set, and the handle detection data set is a locomotive cab operating console image set with handles and corresponding handle platforms marked; Determine the state of the handle according to the relative position of the handle and the corresponding handle platform; determine the running state of the locomotive according to the state of the handle; The locomotive inertial measurement unit is calibrated according to the determined locomotive operating status.

2. The method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to claim 1, characterized in that: The process of determining the running state of the locomotive according to the state of the handle includes: determining the running state of the locomotive according to the state of the handle and the locomotive model; The method for obtaining the locomotive model is: inputting the image of the vehicle cab control panel into the locomotive model recognition model to obtain the locomotive model; the locomotive model recognition model is obtained by training the image classification model using the locomotive model data set, and the locomotive model data set is a locomotive cab control panel image set containing locomotive model labels.

3. The method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to claim 1, characterized in that: The process of calibrating the locomotive inertial measurement unit according to the locomotive running state includes: when the locomotive running state is stopped, calibrating the initial value of the locomotive inertial measurement unit.

4. The method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to claim 1, characterized in that: The locomotive inertial measurement unit includes at least one of a gyroscope, an accelerometer, and a magnetometer.

5. The method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to claim 1, characterized in that: The rotating target detection model is a deep learning model.

6. The method for calibrating a locomotive inertial measurement unit based on a rotating target detection model according to claim 2, characterized in that: The image classification model trained using the locomotive model dataset is a deep learning model.

7. The locomotive inertial measurement unit calibration method based on a rotating target detection model according to any one of claims 1 to 6, characterized in that: The locomotive speed and position are determined using the locomotive inertial measurement unit after the locomotive inertial measurement unit has been calibrated.

8. The locomotive inertial measurement unit calibration method based on a rotating target detection model according to any one of claims 1 to 6, characterized in that: The image of the locomotive cab control panel is a panoramic image of the locomotive cab control panel.

9. A locomotive inertial measurement unit calibration device based on a rotating target detection model, comprising a processor, characterized in that: The processor implements the steps of the locomotive inertial measurement unit calibration method based on the rotating target detection model as described in any one of claims 1 to 8 when executing the computer program.

Images