Vehicle-mounted radar height adjustment method and device based on composite ranging and machine vision

Abstract

The application relates to a vehicle-mounted radar height adjustment method and device based on composite ranging and machine vision, and the method comprises the following steps: in the process of vehicle driving, composite off-ground distance information of a vehicle-mounted ground penetrating radar is collected through a composite ranging module based on a millimeter wave radar range finder and a single-point laser range finder, and driving condition information is collected through a machine vision acquisition module; the composite off-ground distance information is fused and calculated based on a weighted off-ground height algorithm to obtain a composite average off-ground distance and a composite driving pitch angle; a road surface dangerous target is detected and classified according to the driving condition information based on a machine vision model; and adaptive off-ground height adjustment of the vehicle-mounted ground penetrating radar is carried out based on a composite ranging-dangerous target double adjustment mechanism according to the composite average off-ground distance, the composite driving pitch angle and the detection result of the road surface dangerous target. Compared with the prior art, the application realizes unmanned, automation and precision while guaranteeing the safety and stability of the equipment.

Description

Technical Field

[0001] This invention relates to the field of vehicle-mounted ground-penetrating radar adjustment technology, and in particular to a method and device for adjusting the altitude of vehicle-mounted radar based on composite ranging and machine vision. Background Technology

[0002] Under the combined effects of heavy loads, extreme temperature differences, heavy rainfall, and strong ultraviolet radiation, road structures are prone to developing hidden defects such as voids and cavities. The defects around pipelines are particularly severe, affecting road structural stability and posing a significant danger to traffic safety. Currently, various non-destructive testing methods are widely used in the detection and screening of hidden road defects. Among them, ground-penetrating radar (GPR) is a highly efficient, high-precision, and low-labor-cost solution widely used in the inspection of road structures at all levels. Its working principle involves emitting and receiving electromagnetic waves. By analyzing the propagation patterns of electromagnetic waves in different media, it extracts the characteristics and specific locations of hidden defects in the road's concealed structure and presents the raw data as a two-dimensional image, facilitating rapid identification, classification, and maintenance plan development by inspectors.

[0003] In recent years, with the rapid development of deep learning, various machine vision models, represented by YOLOv8, have made rapid progress and gradually realized the automated analysis function of vehicle-mounted ground-penetrating radar data. They have initially achieved real-time target detection and positioning of hidden road defects.

[0004] However, due to the lack of miniaturization of air-coupled ground-penetrating radar (GPR) equipment, its connection to the inspection vehicle typically relies on rigid rods and pins, among other connectors. Although inspection personnel can manually adjust the height of the vehicle-mounted GPR using an installed lift, the entire process is not yet automated, unmanned, or precise, and is ill-suited for handling unexpected situations. Specifically, existing vehicle-mounted GPR inspections face two main problems:

[0005] 1. Due to defects, repairs, and other reasons, the road surface is uneven, making it difficult for inspectors to accurately observe the vehicle's driving conditions from inside the vehicle. Furthermore, the reaction speed is limited when operating the lift, and the onboard ground-penetrating radar equipment is at risk of touching the ground, which can easily cause equipment damage or even affect driving safety.

[0006] 2. Similarly, the ground-penetrating radar altitude cannot be stabilized within a reasonable range in real time, resulting in low quality of collected data, which affects the subsequent process of disease identification, classification, and road maintenance plan formulation. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the existing technology, which requires manual adjustment of the ground-penetrating radar's altitude, and the entire process has not yet achieved automation, unmanned operation, and precision. This invention provides a method and device for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] A method for adjusting the altitude of an onboard radar based on composite ranging and machine vision includes the following steps:

[0010] During vehicle operation, composite ground-penetrating radar distance information is collected by a composite ranging module based on millimeter-wave radar rangefinder and single-point laser rangefinder, and driving status information is collected by a machine vision acquisition module.

[0011] The composite ground clearance information is fused and calculated based on a weighted ground clearance algorithm to obtain the composite average ground clearance and composite pitch angle.

[0012] Based on the machine vision model, dangerous targets on the road surface are detected and classified according to the driving condition information;

[0013] Based on the composite average ground clearance, composite travel pitch angle, and detection results of dangerous targets on the road surface, the vehicle-mounted ground-penetrating radar adaptively adjusts its ground clearance using a composite ranging-dangerous target dual-restriction mechanism.

[0014] Furthermore, the composite ranging module includes a first millimeter-wave radar rangefinder, a second millimeter-wave radar rangefinder, a third millimeter-wave radar rangefinder, and a fourth millimeter-wave radar rangefinder; as well as a first laser rangefinder, a second laser rangefinder, a third laser rangefinder, and a fourth laser rangefinder;

[0015] The first millimeter-wave radar rangefinder, the second millimeter-wave radar rangefinder, the third millimeter-wave radar rangefinder, and the fourth millimeter-wave radar rangefinder are fixed to the four corners of the vehicle-mounted ground-penetrating radar by means of adhesive or embedding; the first laser rangefinder, the second laser rangefinder, the third laser rangefinder, and the fourth laser rangefinder are fixed to the four corners of the vehicle-mounted ground-penetrating radar by means of adhesive or embedding.

[0016] Furthermore, the calculation process of the weighted ground clearance algorithm includes:

[0017] Based on the accuracy characteristics of millimeter-wave radar rangefinders and single-point laser rangefinders in different environments, different weighted ground clearance parameters are assigned; the first average ground clearance and the first travel pitch angle are calculated based on the data acquisition results of each millimeter-wave radar rangefinder; the second average ground clearance and the second travel pitch angle are calculated based on the data acquisition results of each laser rangefinder.

[0018] Based on the first average ground clearance, the first travel pitch angle, the second average ground clearance, and the second travel pitch angle, the composite average ground clearance and the composite travel pitch angle are calculated using the weighted ground clearance parameters.

[0019] Furthermore, the calculation expression for the weighted ground clearance parameter is as follows:

[0020]

[0021] In the formula, This is a weighted height-off-ground parameter;

[0022] Furthermore, the machine vision model can detect and classify road surface hazards including ruts, repairs, and speed bumps.

[0023] Furthermore, the machine vision model is a lightweight YOLOv8 model. The lightweighting method includes replacing the C2f module in the backbone of the YOLOv8 model network with a C2fGhost convolutional module; and adding a deep network detection layer P6 to the YOLOv8 model for the detection and classification of large and dangerous targets.

[0024] Furthermore, the C2fGhost convolution module divides the input feature map into two branches. One branch performs feature extraction operations through two GhostConv modules to obtain a new feature map. The new feature map is then concatenated with the feature map of the other branch, and after one convolution, the final feature map is generated and output.

[0025] Furthermore, the process of adaptively adjusting the ground clearance of the vehicle-mounted ground-penetrating radar using the composite ranging-dangerous target dual-contract mechanism includes:

[0026] When the composite average ground clearance does not conform to the preset reasonable ground clearance range or the composite travel pitch angle does not conform to the preset reasonable pitch angle range, the winch rotation angle is calculated according to the reasonable ground clearance range and the reasonable pitch angle range to drive the motor hinge to lift or lower in the slide rail, thereby realizing the adjustment of the ground clearance or pitch angle of the vehicle-mounted ground-penetrating radar.

[0027] When adjusting the ground clearance, the calculation expression for the winch rotation angle is as follows:

[0028]

[0029] In the formula, The angle of rotation of the winch. The inner diameter of the hinge. Let be the radius of the inner ring of the hinge. It is the difference between the median of the preset reasonable ground clearance range and the current composite average ground clearance value.

[0030] Furthermore, the composite ranging-dangerous target dual-contract mechanism, in its process of adaptively adjusting the ground clearance of the vehicle-mounted ground-penetrating radar, also includes:

[0031] When the machine vision model detects a dangerous target on the road surface in front of the vehicle, it calculates the winch rotation angle based on the classification of the dangerous target, and drives the motor hinge to raise the ground-penetrating radar's height above the ground.

[0032] During the execution of the composite ranging-dangerous target dual-restriction mechanism, the priority of adjusting the ground clearance of the vehicle-mounted ground-penetrating radar based on the identification of dangerous targets on the road surface is higher than the priority of adjusting the ground clearance of the vehicle-mounted ground-penetrating radar based on the composite average ground clearance distance and composite travel elevation angle.

[0033] The present invention also provides a vehicle-mounted radar height adjustment device based on composite ranging and machine vision, comprising: a detection vehicle, a vehicle-mounted ground-penetrating radar, a GPS positioning system, a main control computer, a connection device, a support frame, an automatic lift, a composite ranging module and a machine vision model. The vehicle-mounted ground-penetrating radar includes a ground-penetrating radar array and a ground-penetrating radar protective shell. The ground-penetrating radar array, the GPS positioning system and the main control computer are all encapsulated in the ground-penetrating radar protective shell.

[0034] The connecting devices are respectively connected to the detection vehicle, the ground-penetrating radar protective shell, and the automatic lift;

[0035] The automatic lift is powered by a built-in motor and is connected to the ground-penetrating radar protective shell via a support frame, hinges, and slide rails. It is also connected to the connecting equipment via rigid rods and pins. The height and pitch angle of the ground-penetrating radar protective shell can be adjusted by the extension and retraction of the hinges in the slide rails.

[0036] The composite ranging module is installed at the four corners of the bottom of the ground penetrating radar protective shell;

[0037] The machine vision model is mounted on top of the inspection vehicle;

[0038] The main control computer is used to execute the steps of the vehicle-mounted radar height adjustment method based on composite ranging and machine vision as described above.

[0039] Compared with the prior art, the present invention has the following advantages:

[0040] (1) Traditional methods rely on manual adjustment of the height of vehicle-mounted ground-penetrating radar, which is not only cumbersome and inefficient, but also difficult to achieve precise control. When faced with complex road surface conditions, the testing personnel cannot adjust the height in a timely and accurate manner, which poses a risk of the equipment touching the ground and threatens the safety of the equipment and driving safety.

[0041] This invention proposes an easily detachable composite ranging module and a machine vision acquisition module suitable for vehicle-mounted ground-penetrating radar. The easily detachable composite ranging module consists of a millimeter-wave radar rangefinder and a single-point laser rangefinder. It considers the accuracy characteristics of the two types of rangefinders under different weather conditions and assigns different weight values ​​to achieve fusion calculation based on a weighted ground clearance algorithm, so as to ensure continuous and accurate measurement of ground clearance data and overcome the problem of poor radar data quality caused by imprecise control of ground clearance.

[0042] A lightweight machine vision model is used to detect and classify dangerous targets on the road surface in video data in real time. Then, an adaptive real-time adjustment system for the ground-penetrating radar of the vehicle based on the composite ranging-dangerous target dual-restriction mechanism is established to realize the parameterized and automated control of the elevator.

[0043] (2) This invention utilizes an easily detachable composite ranging module and a machine vision acquisition module to construct an adaptive height adjustment system with a composite ranging-dangerous target dual-response mechanism. This system can automatically and accurately adjust the ground-penetrating radar's altitude and elevation angle according to road conditions and dangerous target conditions, achieving unmanned, automated, and precise operation, effectively avoiding equipment contact with the ground, and ensuring equipment safety and driving safety. Attached Figure Description

[0044] Figure 1 This is a flowchart illustrating the steps of a vehicle-mounted radar height adjustment method based on composite ranging and machine vision provided in an embodiment of the present invention.

[0045] Figure 2 This is a schematic diagram of the program execution flow of a vehicle radar height adjustment method based on composite ranging and machine vision provided in an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the installation position of a composite ranging module provided in an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of a C2fGhost module provided in an embodiment of the present invention;

[0048] Figure 5 This is a side view schematic diagram of a vehicle-mounted radar height adjustment device based on composite ranging and machine vision provided in an embodiment of the present invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0050] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0051] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0052] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product of this invention is usually placed during use. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] It should be noted that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0054] Example 1

[0055] like Figure 1 and Figure 2 As shown, this embodiment provides a method for adjusting the altitude of an onboard radar based on composite ranging and machine vision, including the following steps:

[0056] S1: During vehicle operation, composite ground-penetrating radar distance information is collected by a composite ranging module based on millimeter-wave radar rangefinder and single-point laser rangefinder, and driving status information is collected by a machine vision acquisition module.

[0057] S2: A real-time fusion and calculation program for composite ranging data based on a weighted ground clearance algorithm is used to fuse and calculate composite ground clearance information to obtain composite average ground clearance and composite travel pitch angle;

[0058] S3: Based on machine vision models, detect and classify hazardous targets on the road surface according to driving condition information;

[0059] S4: Based on the composite average ground clearance, composite travel pitch angle, and detection results of dangerous targets on the road surface, the vehicle-mounted ground-penetrating radar adaptively adjusts its ground clearance according to the composite ranging-dangerous target dual-restriction mechanism.

[0060] In step S1, preferably, the composite ranging module includes a first millimeter-wave radar rangefinder, a second millimeter-wave radar rangefinder, a third millimeter-wave radar rangefinder, and a fourth millimeter-wave radar rangefinder; as well as a first laser rangefinder, a second laser rangefinder, a third laser rangefinder, and a fourth laser rangefinder;

[0061] The first, second, third, and fourth millimeter-wave radar rangefinders are fixed to the four corners of the vehicle-mounted ground-penetrating radar by adhesive or embedding. Similarly, the first, second, third, and fourth laser rangefinders are fixed to the four corners of the vehicle-mounted ground-penetrating radar by adhesive or embedding. (See also...) Figure 3 As shown in the diagram. In adverse weather conditions such as rain, snow, and fog, the millimeter-wave radar rangefinder and the single-point laser rangefinder jointly provide composite ground-elevation data, while the single-point laser rangefinder primarily provides high-precision data under favorable weather conditions.

[0062] It should be understood that after installation, both types of instruments should be a certain distance (15-20cm) from the edge of the protective shell to ensure comprehensive measurement of the protective shell's height above the ground. Simultaneously, the edges of the two types of instruments should be joined, placed side-by-side, and perpendicular to the direction of travel to ensure synchronous measurement of single-point height data. Under good weather conditions, the rangefinder should collect ground-free height data at a frequency of no less than 40 data points per second, with a ranging accuracy error not exceeding 5%. Under adverse weather conditions, the ranging accuracy error of the millimeter-wave radar should not exceed 5%, and the ranging accuracy error of the single-point lidar should not exceed 10%, to ensure continuous and accurate measurement of ground-free height data.

[0063] In some embodiments, the machine vision acquisition module may include multiple high-definition cameras that can automatically adjust their orientation, but they must be symmetrically distributed at the front and rear ends of the vehicle roof.

[0064] Specifically, the machine vision acquisition module should be able to acquire video within a rectangular area of ​​about 10-15 meters in front of and behind the vehicle, with a video frame rate of no less than 40 images per second and a pixel count of no less than 8 million.

[0065] In step S2, the calculation process of the weighted ground clearance algorithm includes:

[0066] Based on the accuracy characteristics of millimeter-wave radar rangefinders and single-point laser rangefinders in different environments, different weighted ground clearance parameters are assigned; the first average ground clearance and the first travel pitch angle are calculated based on the data acquisition results of each millimeter-wave radar rangefinder; the second average ground clearance and the second travel pitch angle are calculated based on the data acquisition results of each laser rangefinder.

[0067] Based on the first average ground clearance, the first pitch angle, the second average ground clearance, and the second pitch angle, the composite average ground clearance and the composite pitch angle are calculated using weighted ground clearance parameters.

[0068] Specifically, the real-time fusion and calculation program for composite ranging data based on a weighted ground clearance algorithm should include the following functions:

[0069] The program is stored in the distance-to-ground calculation and control terminal and communicates with the composite ranging module in real time.

[0070] The program calculates weighted ground clearance parameters based on road weather conditions and estimated ground clearance, achieving data fusion. Specifically:

[0071] First, determine the weighted ground clearance parameter, as shown in the formula below:

[0072]

[0073] In the formula, For weighted ground clearance parameters, It uses segmented values ​​and is dimensionless. Different weight values ​​are assigned based on the accuracy characteristics of the two types of rangefinders, which are then used for subsequent calculations of the composite average ground clearance and composite pitch angle.

[0074] The real-time data collected by the first, second, third, and fourth millimeter-wave radar rangefinders were then fused, and the average ground clearance was calculated. and pitch angle .

[0075] Similarly, the real-time data collected by the first, second, third, and fourth single-point laser rangefinders are fused, and the average ground clearance is calculated. and pitch angle .

[0076] Considering data collection quality and driving safety, a preset reasonable ground clearance range is specified as follows: The program will determine in real time whether the calculated value falls within this range; similarly, a preset reasonable range for travel pitch angles is defined as follows: The program then determines in real time whether the calculated value falls within this range.

[0077] In step S3, the road surface hazards are detected and classified based on the driving condition information using a machine vision model. The detectable and classifiable road surface hazards include ruts, repairs, and speed bumps.

[0078] The machine vision model is a lightweight version of the machine vision model. Lightweight improvement methods include replacing / optimizing the backbone network, optimizing the convolution process, and optimizing the detection head.

[0079] Specifically, the following steps are included:

[0080] A lightweight improvement to the high-performance open-source machine vision model YOLOv8 based on the ghost convolution method is performed. The specific steps are as follows:

[0081] The C2f module in the network backbone is improved into a C2fGhost convolutional module with a lower number of model parameters. See [link / reference] Figure 4 As shown.

[0082] The Ghost convolution module works as follows: the output features of the regular convolutional module (C2f module) one layer above the C2fGhost module are input into the C2fGhost module. Specifically, the input feature map is divided into two branches. One branch performs feature extraction operations consecutively through two GhostConv modules.

[0083] The new feature map obtained from this branch is then concatenated with the feature map that has not been processed by these two modules, and then subjected to a regular convolution to generate and output the final feature map. The C2fGhost module can significantly reduce the total number of model parameters, effectively reduce computational complexity, and is suitable for deployment in vehicle-mounted distance calculation and control terminals.

[0084] Furthermore, this embodiment improves upon the original detection layers of YOLOv8 by adding a P6 detection layer with a lower output feature map resolution. The improved model includes detection layers P3, P4, P5, and P6. This new structural design can more effectively identify multi-scale features in images. Specifically, the shallow network detection layers (P3, P4, P5) are mainly responsible for identifying detailed features of the image, which is beneficial for detecting and classifying small, dangerous targets; while the deep network detection layer (P6) is more conducive to the detection and classification of large, dangerous targets. At the same time, since the P6 output feature map has a lower resolution, it can save computing power, thereby improving the training and detection speed of the model.

[0085] Next, the lightweight YOLOv8 will be trained, validated, and tested on multiple datasets (no fewer than 5, with no fewer than 800 training images, no fewer than 100 validation images, and no fewer than 100 test images in each dataset) containing images of dangerous objects (ruts, repairs, speed bumps), and the corresponding performance metrics will be met.

[0086] It should be understood that, among a series of performance metrics, the lightweight YOLOv8 achieves a total training time of no more than 0.5 hours on the training set, a real-time target detection frame rate (FPS) of no less than 40 frames per second on the test set, and an average accuracy of greater than 85% in identifying dangerous targets. Finally, the model is deployed in a ground-distance calculation and control terminal, communicating in real-time with the video acquisition module to achieve the detection and classification of dangerous targets during vehicle movement.

[0087] Specifically, in order to verify the performance of the lightweight YOLOv8 in this embodiment, the calculation method of average accuracy (map) is given in some embodiments.

[0088] The relevant indicators are defined and calculated as follows:

[0089] "Positive" and "Negative" are the predicted sample labels, where "Positive" represents the disease in the image to be detected (positive sample), and "Negative" represents the background in the image to be detected (negative sample). "True" and "False" represent the prediction results, where "True" indicates a correct prediction and "False" indicates an incorrect prediction. Therefore, FP indicates that a negative sample was incorrectly predicted as a positive sample, FN indicates that a positive sample was incorrectly predicted as a negative sample, TN indicates that a negative sample was correctly predicted as a negative sample, and TP indicates that the classification was correctly predicted.

[0090] Precision, also known as accuracy, assesses the accuracy of hazard detection; that is, the proportion of data that are predicted as positive but are actually positive.

[0091]

[0092] Recall rate, also known as completeness, assesses whether disease detection is comprehensive, that is, the proportion of data that are actually positive samples that are correctly predicted as positive samples.

[0093]

[0094] Average accuracy Precision and recall are typically not optimized simultaneously, leading to a neglect of the overall outcome. Therefore, introducing... and To balance the calculation results of the two, for a certain disease type, the Precision-Recall (PR) curve is obtained by plotting Precision on the vertical axis and Recall on the horizontal axis. PR Area under the curve The value of. Furthermore, This represents the average of all classes in the entire test set. .

[0095]

[0096]

[0097] in, Representative categories (ruts, repairs, or speed bumps); This represents the total number of a certain type of disease in the test set; It is a subscript symbol used to distinguish a certain type of disease.

[0098] In step S4, the vehicle-mounted ground-penetrating radar adaptive ground-lift altitude real-time adjustment system based on the composite ranging-dangerous target dual-contract mechanism includes the following functions:

[0099] When the composite average ground clearance or composite travel pitch angle output by the real-time fusion and calculation program of composite ranging data does not conform to the preset reasonable range, the system issues a command to the automatic elevator to calculate the winch rotation angle (control parameters) according to the preset reasonable ground clearance range and the preset reasonable pitch angle range, and then drives the motor hinge to lift or lower in the slide rail to achieve adjustment of the ground clearance or pitch angle of the ground penetrating radar protective shell.

[0100] Taking the adjustment of the height off the ground as an example, the formula for calculating the winch rotation angle when adjusting the height is as follows:

[0101]

[0102] in, It is the winch rotation angle, in degrees; Indicates the inner diameter of the hinge, in cm; Indicates the radius of the inner ring of the hinge, in cm; It represents the difference between the median of the preset reasonable ground clearance range and the current composite average ground clearance value (also refers to the arc length traveled by the winch rotation), in cm.

[0103] Similarly, the pitch angle of a vehicle-mounted ground-penetrating radar device can be adjusted using a similar method based on the angle formed between the two hinges.

[0104] Furthermore, when the machine vision model detects different types of dangerous targets in front of the vehicle, the system issues a command to the automatic lift, calculates the winch rotation angle according to the classification of dangerous targets, and drives the motor to lift the height of the ground-penetrating radar protective shell to avoid collision risks.

[0105] The formula for calculating the winch rotation angle when adjusting the height is the same as the formula above, but... The definitions and calculation methods differ, as shown below:

[0106]

[0107] in, This indicates the distance the ground-penetrating radar protective shell needs to be lifted to avoid dangerous targets (also refers to the arc length traveled by the winch rotation), in cm.

[0108] Then, combining the current vehicle speed information collected by GPS, after a certain interval (when the vehicle-mounted radar passes a dangerous target), the system sends a command to the automatic lift, and the motor drives the hinge to lower the height of the ground-penetrating radar protective shell back to its original position. The calculation formula is as follows:

[0109]

[0110] in, This is the current vehicle speed collected by GPS, in m / s. It is the sum of the farthest actual distance in front of the vehicle and the vehicle length in the video obtained by the machine vision acquisition module, in meters.

[0111] Alternatively, when the machine vision model detects the same dangerous target behind the vehicle, the system issues a command to the automatic lift to lower the height off the ground to its original position;

[0112] Furthermore, the real-time detection and classification program for hazardous targets on the road surface based on lightweight machine vision models takes precedence over the real-time fusion and calculation program for composite ranging data based on a weighted ground clearance algorithm; that is, the latter does not operate once a hazardous target is detected. Otherwise, both operate synchronously during the detection process.

[0113] Furthermore, the zero-point correction procedure is used to eliminate the error caused by the air layer between the antenna and the ground during the lifting, lowering, and shaking of the vehicle-mounted ground-penetrating radar equipment, thereby improving the quality of the collected data.

[0114] Example 2

[0115] like Figure 5As shown, this embodiment provides a vehicle-mounted radar height adjustment device based on composite ranging and machine vision, including: a detection vehicle, a vehicle-mounted ground-penetrating radar, a GPS positioning system, a main control computer, a connection device, a support frame, an automatic lift, a composite ranging module, and a machine vision model. The vehicle-mounted ground-penetrating radar includes a ground-penetrating radar array and a ground-penetrating radar protective shell. The ground-penetrating radar array, the GPS positioning system, and the main control computer are all encapsulated in the ground-penetrating radar protective shell.

[0116] The connection devices are respectively connected to the detection vehicle, the protective shell of the ground-penetrating radar, and the automatic lift;

[0117] The automatic lift is powered by a built-in motor and is connected to the ground-penetrating radar protective shell via a support frame, hinges, and slide rails. It is also connected to the connecting equipment via rigid rods and pins. The height and pitch angle of the ground-penetrating radar protective shell (containing a ground-penetrating radar array, GPS positioning system, and main control computer) can be adjusted by the extension and retraction of the hinges in the slide rails.

[0118] The composite ranging module is installed at the four corners of the bottom of the ground penetrating radar protective shell;

[0119] The machine vision model is mounted on top of the inspection vehicle;

[0120] The main control computer is used to execute the steps of a vehicle radar height adjustment method based on composite ranging and machine vision, as described in Example 1.

[0121] Furthermore, the support frame is connected to the connecting equipment via rigid rods and pins, and welded to the instrument with slide rails to further ensure the stability of the vehicle-mounted ground-penetrating radar during operation;

[0122] Furthermore, the connecting device is connected to the vehicle's rear bumper by welding.

[0123] The preferred embodiments of the present invention have been described in detail above. It should be understood that those skilled in the art can make numerous modifications and variations based on the concept of the present invention without creative effort. Therefore, all technical solutions that can be obtained by those skilled in the art based on the concept of the present invention through logical analysis, reasoning, or limited experimentation on the basis of existing technology should be within the scope of protection defined by the claims.

Claims

1. A method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision, characterized in that, Includes the following steps: During vehicle operation, composite ground-penetrating radar distance information is collected by a composite ranging module based on millimeter-wave radar rangefinder and single-point laser rangefinder, and driving status information is collected by a machine vision acquisition module. The composite ground clearance information is fused and calculated based on a weighted ground clearance algorithm to obtain the composite average ground clearance and composite pitch angle. Based on the machine vision model, dangerous targets on the road surface are detected and classified according to the driving condition information; Based on the composite average ground clearance, composite travel pitch angle, and detection results of dangerous targets on the road surface, an adaptive ground clearance adjustment of the vehicle-mounted ground-penetrating radar is performed based on the composite ranging-dangerous target dual-restriction mechanism. The calculation process of the weighted ground clearance algorithm includes: Based on the accuracy characteristics of millimeter-wave radar rangefinders and single-point laser rangefinders in different environments, different weighted ground clearance parameters are assigned; the first average ground clearance and the first travel pitch angle are calculated based on the data acquisition results of each millimeter-wave radar rangefinder; the second average ground clearance and the second travel pitch angle are calculated based on the data acquisition results of each laser rangefinder. Based on the first average ground clearance, the first travel pitch angle, the second average ground clearance, and the second travel pitch angle, the composite average ground clearance and the composite travel pitch angle are calculated using the weighted ground clearance parameters. The calculation expression for the weighted ground clearance parameter is as follows: In the formula, This is a weighted height-off-ground parameter.

2. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 1, characterized in that, The composite ranging module includes a first millimeter-wave radar rangefinder, a second millimeter-wave radar rangefinder, a third millimeter-wave radar rangefinder, and a fourth millimeter-wave radar rangefinder; as well as a first laser rangefinder, a second laser rangefinder, a third laser rangefinder, and a fourth laser rangefinder; The first millimeter-wave radar rangefinder, the second millimeter-wave radar rangefinder, the third millimeter-wave radar rangefinder, and the fourth millimeter-wave radar rangefinder are fixed to the four corners of the vehicle-mounted ground-penetrating radar by means of adhesive or embedding; the first laser rangefinder, the second laser rangefinder, the third laser rangefinder, and the fourth laser rangefinder are fixed to the four corners of the vehicle-mounted ground-penetrating radar by means of adhesive or embedding.

3. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 1, characterized in that, The machine vision model can detect and classify road surface hazards including ruts, repairs, and speed bumps.

4. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 1, characterized in that, The machine vision model is a lightweight version of the YOLOv8 model. The lightweighting method includes replacing the C2f module in the backbone of the YOLOv8 model network with a C2fGhost convolutional module; and adding a deep network detection layer to the YOLOv8 model for the detection and classification of large and dangerous targets.

5. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 4, characterized in that, The C2fGhost convolution module divides the input feature map into two branches. One branch performs feature extraction operations through two GhostConv modules to obtain a new feature map. The new feature map is then concatenated with the feature map of the other branch, and after one convolution, the final feature map is generated and output.

6. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 1, characterized in that, The composite ranging-dangerous target dual-contract mechanism includes the following process for adaptive ground-penetrating radar altitude adjustment: When the composite average ground clearance does not conform to the preset reasonable ground clearance range or the composite travel pitch angle does not conform to the preset reasonable pitch angle range, the winch rotation angle is calculated according to the reasonable ground clearance range and the reasonable pitch angle range to drive the motor hinge to lift or lower in the slide rail, thereby realizing the adjustment of the ground clearance or pitch angle of the vehicle-mounted ground-penetrating radar. When adjusting the ground clearance, the calculation expression for the winch rotation angle is as follows: In the formula, The angle of rotation of the winch. The inner diameter of the hinge. Let be the radius of the inner ring of the hinge. It is the difference between the median of the preset reasonable ground clearance range and the current composite average ground clearance value.

7. The method for adjusting the altitude of a vehicle-mounted radar based on composite ranging and machine vision according to claim 6, characterized in that, The composite ranging-dangerous target dual-contract mechanism, in its process of adaptively adjusting the ground clearance of the vehicle-mounted ground-penetrating radar, also includes: When the machine vision model detects a dangerous target on the road surface in front of the vehicle, it calculates the winch rotation angle based on the classification of the dangerous target, and drives the motor hinge to raise the ground-penetrating radar's height above the ground. During the execution of the composite ranging-dangerous target dual-restriction mechanism, the priority of adjusting the ground clearance of the vehicle-mounted ground-penetrating radar based on the identification of dangerous targets on the road surface is higher than the priority of adjusting the ground clearance of the vehicle-mounted ground-penetrating radar based on the composite average ground clearance distance and composite travel elevation angle.

8. A vehicle-mounted radar height adjustment device based on composite ranging and machine vision, characterized in that, include: The system includes a vehicle, a vehicle-mounted ground-penetrating radar, a GPS positioning system, a main control computer, connecting equipment, a support frame, an automatic lift, a composite ranging module, and a machine vision model. The vehicle-mounted ground-penetrating radar includes a ground-penetrating radar array and a ground-penetrating radar protective shell. The ground-penetrating radar array, the GPS positioning system, and the main control computer are all encapsulated within the ground-penetrating radar protective shell. The connecting devices are respectively connected to the detection vehicle, the ground-penetrating radar protective shell, and the automatic lift; The automatic lift is powered by a built-in motor and is connected to the ground-penetrating radar protective shell via a support frame, hinges, and slide rails. It is also connected to the connecting equipment via rigid rods and pins. The height and pitch angle of the ground-penetrating radar protective shell can be adjusted by the extension and retraction of the hinges in the slide rails. The composite ranging module is installed at the four corners of the bottom of the ground penetrating radar protective shell; The machine vision model is mounted on top of the inspection vehicle; The main control computer is used to execute the steps of the vehicle radar height adjustment method based on composite ranging and machine vision as described in any one of claims 1-7.

Images