A Vehicle Detection Method and System for Millimeter-Wave Radar

Abstract

The present invention discloses a vehicle detection method and system for millimeter-wave radar, which relates to the technical field of radar signal processing, and includes the following steps: S1. Install a road-end and vehicle-end millimeter-wave radar integrated machine on the top of the underground roadway and the trackless rubber-tyred vehicle respectively. The millimeter-wave radar integrated machine emits millimeter waves and receives echoes to obtain original radar echo data, and the original radar echo data includes the position and movement information of the trackless rubber-tyred vehicle in the detection roadway; The present invention optimizes and adjusts the performance parameter signals of the millimeter-wave radar integrated machines arranged in the mine and on the trackless rubber-tyred vehicle precisely by establishing a mine environment electromagnetic model and a radar simulation model, according to the actual structure and material of the mine roadway, and further combining with the absorption of radar signals, avoiding the interference of reflection, diffraction and occlusion in the complex mine environment, and ensuring the accuracy and precision of the feature data extracted from the original radar echo data.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle control devices, and particularly to a method and system for detecting vehicles using millimeter-wave radar. Background Art

[0002] Currently, most mines use positioning based on UWB technology. Since UWB technology cannot accurately identify the needs of vehicle speed, vehicle direction, and vehicle distance, it often cannot solve the congestion problem. Some mines use lidar, but due to the influence of factors such as the changing underground environment over time, signal detection and signal coverage have blind spots, resulting in high false alarm rates and large distance errors in positioning target objects. It is difficult to capture effective and accurate signal sources, and from a systematic perspective, it is impossible to effectively count the traffic flow and vehicle detection statistics of vehicle speeds in a specific roadway during a specific time period. Therefore, under multiple factors such as equipment hardware failures, complex environmental interference in underground mines, electromagnetic waves generated by wireless signal sources, and unstable positions of base station signal coverage, neither UWB technology nor lidar technology can effectively and accurately detect vehicles.

[0003] The current transportation environment of underground trackless rubber-tired vehicles relies on the self-discipline of drivers to abide by the traffic regulations of underground trackless rubber-tired vehicles. Due to the small space in the underground environment, especially in the roadways for round trips to the working face, drivers often do not know whether there are vehicles driving in the roadway, their driving positions, and how far they are from the chamber, resulting in roadway traffic jams. In this case, one trackless rubber-tired vehicle must reverse to give way to the other, which will lead to the defect of reduced utilization rate of trackless rubber-tired vehicles. Even more, if multiple trackless rubber-tired vehicles enter the same roadway during the same time period, it will inevitably lead to vehicle congestion and may also cause unnecessary accidents.

[0004] The prior art has the following deficiencies: However, in a complex mining environment, the transportation environment of trackless rubber-tired vehicles has unique terrain and irregular tunnel walls, which may cause reflection and diffraction phenomena of millimeter-wave radar in the complex mining environment, resulting in misjudgment of the vehicle detection system. And in a narrow mine roadway, when there are multiple trackless rubber-tired vehicle targets at the same time, the millimeter-wave radar signal may be affected by reflection and occlusion, making it difficult for the vehicle detection system to accurately identify and locate multiple targets. In addition, electromagnetic interference generated by radio transmission equipment and electric mechanical equipment will affect the stability of radar signals and the robustness of the vehicle detection system.

[0005] The above information disclosed in the background art section is only used to enhance the understanding of the background of the present disclosure, and therefore it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0006] The objective of the present invention is to provide a vehicle detection method and system for millimeter-wave radar. The present invention obtains the position and movement information of vehicles in the roadway, the radar information and movement information of trackless rubber-tired vehicles through a millimeter-wave radar integrated machine, extracts feature data through Kalman filtering and performs radar imaging processing, establishes a mine environment electromagnetic model and a radar simulation model, improves the accuracy of radar performance parameters, and uses a convolutional neural network model to deeply analyze and optimize the detection data to generate a risk controllable coefficient, and performs visual warning processing by comparing the warning decision threshold to solve the problems in the above-mentioned background technology.

[0007] To achieve the above objective, the present invention provides the following technical solution: A vehicle detection method for millimeter-wave radar, including the following steps:

[0008] S1. Install a roadside and vehicle-mounted millimeter-wave radar integrated machine on the top of the underground roadway and the trackless rubber-tired vehicle respectively. Transmit millimeter waves through the millimeter-wave radar integrated machine and receive the echo to obtain the original radar echo data. The original radar echo data includes the position and movement information of the trackless rubber-tired vehicle in the detected roadway and the self-movement speed information of the detected trackless rubber-tired vehicle.

[0009] S2. Apply Kalman filtering technology to perform preprocessing operations of noise removal and data fusion on the original radar echo data for vehicle detection, generate preprocessed data, and perform feature extraction and interference suppression on the preprocessed data to generate feature data for subsequent processing.

[0010] S3. Use the signal processing algorithms of synthetic aperture radar (SAR) and inverse synthetic aperture radar (ISAR) for radar imaging recognition to obtain radar imaging data, and combine the UWB technology to locate the position information of the target vehicle, and track the movement state of the target vehicle through the multi-target tracking (MHT) algorithm. According to the preset traffic rules and strategies in the mine, combine the traffic facilities such as the linked traffic signal control box, chamber LED lamp device, and gate device to perform necessary traffic management on multiple vehicles in the mine passage.

[0011] S4. Establish a mine environment electromagnetic model, accurately model the electromagnetic characteristics in the roadway, use Maxwell's equations to simulate the propagation and reflection characteristics of millimeter-wave signals under mine tunnel conditions, and construct a radar simulation model. Use computer simulation software to simulate and test the performance of the millimeter-wave radar under different working conditions. The performance includes signal-to-noise ratio (SNR), range-velocity resolution, detection probability, and false alarm rate. Use the genetic algorithm to find the optimal parameter set of the radar and adaptively optimize and adjust the parameters and update the model.

[0012] S5. Use the convolutional neural network algorithm to conduct in-depth intelligent analysis on the feature data, and use the automatic learning technology that has been trained, tested, verified, analyzed, and optimized to identify the features of different targets, judge the detected target attributes and whether they are potential hazard sources, and generate a risk controllability coefficient to improve the accuracy and robustness of the detection system;

[0013] S6. Use WIFI, 4G or MESH technology to conduct data communication and exchange for each module in the detection system, convert the risk controllability coefficient into visual information and synchronize it to the drivers in the trackless rubber-tired vehicle. The visual information includes the traffic flow status of the full-scene passage, and compare the determination result of the visual information with the warning decision threshold, and then trigger an alarm.

[0014] Preferably, the acquisition logic of the original radar echo data is as follows:

[0015] A1. Install a road-end millimeter-wave radar integrated machine on the top of the underground roadway. By continuously transmitting pulse signals and recording the echo signals, obtain the radar echo signals in the detection roadway, and calibrate them as ;

[0016] A2. Analyze and process the radar echo signals in the roadway By calculating the delay and frequency change of to determine the position and movement information of the trackless rubber-tired vehicle in the roadway, and calibrate them as and ;

[0017] A3. The calculation steps for the position and movement information of the trackless rubber-tired vehicle in the roadway are as follows:

[0018] By measuring the frequency difference between the millimeter-wave radar transmitted signal and the received signal in the roadway , calculate the distance, and calibrate it as , then , where in the formula, represents the speed of light, that is , represents the bandwidth of the radar echo signal in the roadway;

[0019] Use the Doppler effect to measure the speed of the detection target relative to the radar in the roadway according to the Doppler frequency shift , and calibrate it as , then , where in the formula, represents the speed of light, that is , represents the carrier frequency, and , where in the formula, Denoted as the wavelength of radar transmission;

[0020] Using a beamforming algorithm, measure the angle of the detection target relative to the phased array radar in the roadway and calibrate it as , then , Denoted as the wavelength of radar transmission, Denoted as the spacing between the th radar array elements, Denoted as the phase difference of the

[0021] A4. By measuring the distances calculated separately through radar stations in the roadway, speed and angle , use triangulation technology to accurately locate the vehicle position, deduce the vehicle's movement trajectory, and use the Kalman filtering algorithm to integrate multiple measurement results, filter out noise and more accurately predict the vehicle position, and provide a smooth and continuous movement trajectory of the vehicle in the roadway;

[0022] A5. The vehicle-mounted millimeter-wave radar installed on the trackless rubber-tyred vehicle continuously emits pulse signals to detect the surrounding environment and records the echo signals to obtain the vehicle radar echo signal, calibrated as ;

[0023] A6. Analyze and process the vehicle radar echo signal , obtain the vehicle's own movement speed information based on the radar echo delay and Doppler frequency change, calibrated as , and install a UWB positioning module on the trackless rubber-tyred vehicle, and combine the UWB positioning technology to determine the position information of the trackless rubber-tyred vehicle in the roadway, calibrated as ;

[0024] A7. The calculation steps for the movement speed information of the trackless rubber-tyred vehicle and the position information in the roadway are as follows:

[0025] The movement speed information of the trackless rubber-tyred vehicle includes the radar echo delay time and Doppler frequency change, calibrated as and respectively, and are respectively used to calculate the distance between the detection target in the roadway and the vehicle radar, calibrated as , and calculate the speed of the detection target in the roadway relative to the vehicle radar, calibrated as , then the distance calculation formula is , where is expressed as the time delay between the radar emission pulse and the received echo, and the velocity calculation formula is , where is expressed as the speed of the trackless rubber-tyred vehicle relative to the radar, is expressed as the emission frequency of the radar;

[0026] A8. The UWB positioning module provides high security, high-speed data transmission and high positioning accuracy technologies, and absolutely locates and calibrates the radar information by analyzing the time difference and track information of the vehicle radar echo signal . Then, the Kalman filtering algorithm is used again to integrate multiple measurement results and reduce errors, so as to provide the position information of the trackless rubber-tyred vehicle under the actual motion state in the roadway .

[0027] Preferably, the acquisition logic of the preprocessed data is as follows:

[0028] For noise filtering and data fusion in the preprocessing operation, the Kalman filtering technology is combined with the original radar echo data, calibrated as , and , , , for prediction and update;

[0029] Define the original radar echo data as the state variable, that is . Noise removal and data fusion adopt the Kalman filtering algorithm. By predicting and updating the state variable, the preprocessed data is obtained, calibrated as , that is , where is expressed as the position of the vehicle in the roadway and the motion information , the motion speed information on the trackless rubber-tyred vehicle and the position information in the roadway at the time step to generate the optimal state estimation value

[0030] Preferably, the acquisition logic of the characteristic data is as follows:

[0031] For the preprocessed data , the least mean square error algorithm is used for adaptive filtering to eliminate the interference signal, and the characteristic information that can distinguish the target is extracted from the data after interference suppression for subsequent detection and tracking tasks. Among them, the calculation formula of the least mean square error algorithm for adaptive interference cancellation is , where is expressed as the error signal of the preprocessed data at the current time step , Represented as preprocessed data The updated weight coefficient vector, Represented as preprocessed data Of the weight coefficient vector, Represented as the step size parameter, Represented as preprocessed data At the current time step Of the complex conjugate of the input signal;

[0032] Then, according to the preprocessed data after interference suppression Extract feature data and calibrate it as Then Wherein, Represents the set of feature data changing with time Of, Represents the feature data At the moment The state value at which it is located, and the Represents the feature data The position features of the vehicle and the trackless rubber-tyred vehicle in the roadway, including distance, speed, and angle, contained therein.

[0033] Preferably, the recognition and acquisition logic of the radar imaging data is as follows:

[0034] Transmit and receive the original radar echo data reflected by the target by the millimeter wave radar integrated machine , convert the time series of the echo into an electrical signal, convert it into a digital signal through an AD converter, and extract feature data through a series of Kalman filtering processes to obtain feature data ;

[0035] Use the synthetic aperture radar SAR signal processing algorithm for range compression, R-D domain conversion, and azimuth compression calculation to improve the range resolution and azimuth resolution of the millimeter wave radar. Among them, the calculation formula for range compression is Wherein, Represents the compressed signal, Represents the characteristic signal received in the millimeter wave radar echo signal, Represents the millimeter wave radar transmission signal;

[0036] The calculation formula for R-D domain conversion is Wherein, Represents the Fourier transform, Represents the RD domain signal, Respectively represent the range frequency and Doppler frequency describing the target motion parameters;

[0037] The calculation formula for azimuth compression is Wherein, Expressed as the azimuth difference resolution, Expressed as the signal sampling rate, Expressed as the linear minimum interval of the target relative motion;

[0038] The inverse synthetic aperture radar (ISAR) signal processing algorithm is used for Doppler center frequency estimation and motion compensation calculation to obtain the distance and speed information of the detected target and obtain a clearer target image. Among them, the calculation formula for Doppler center frequency estimation is , where Expressed as the frequency shift caused by the motion of the detected target in the Doppler spectrum of the processed radar echo signal, Expressed as the frequency at which the maximum value is taken, Expressed as the complex convolution function, Expressed as the time-domain signal received by the radar, that is, the time series of the echo signal at moment;

[0039] The calculation formula for motion compensation is , where Expressed as the detected target at After motion compensation at the moment of the signal, Expressed as the imaginary part exponential function of the complex number, Expressed as the sampling time;

[0040] After SAR and ISAR imaging processing, radar imaging data is generated. Combining with the UWB base stations arranged in the mine, using the UWB tags equipped on the trackless rubber-tired vehicle to transmit UWB signals, the UWB base stations receive the signals from the UWB tags of the trackless rubber-tired vehicle, and according to the received signal strength and time difference, calculate the accurate position of the vehicle UWB tag relative to the base station through trilateration, that is, combining SAR, ISAR imaging data and UWB positioning data to construct a panoramic view of the detailed situation of the trackless rubber-tired vehicle inside the mine.

[0041] Preferably, the steps of the multi-target tracking (MHT) algorithm for tracking the motion state of the target vehicle are as follows:

[0042] From the feature data Measure the observed value of the initial motion state of the detected target vehicle at the time step , and transfer the state of the observed value into matrix representation, marked as ;

[0043] Use the motion model to perform state prediction on the initial motion state of each known detected target vehicle to predict the position and speed of the detected target vehicle at the next time step , that is, the prediction formula is , where represents the predicted state of the detected target vehicle at time , represents the state estimation of the detected target vehicle at time , represents the control input matrix of the detected target vehicle at time , represents the control vector of the detected target vehicle at time , represents the process noise of the detected target vehicle at time ;

[0044] If the Mahalanobis distance is used to calculate the data association degree, then for each newly received measurement value of the detected target vehicle, calculate the matching degree with the predicted motion state of the target vehicle. The calculation formula of the Mahalanobis distance is , where represents the Mahalanobis distance between the observation and the target , respectively represent the predicted value of the predicted target and the measured value of the observed target, represents the actual observation, represents the observation matrix, represents the covariance matrix associated with the observation and the target ;

[0045] According to the result of the data association degree, perform data update, and use the observed data to perform Kalman filter calculation to update the target state estimation. The calculation formula of the Kalman filter is , where represents the state estimation of the detected target vehicle at time after update, represents the Kalman gain at time , represents the actual observation of the detected target vehicle at time , represents the observation matrix of the detected target vehicle at time .

[0046] Preferably, the steps for establishing the electromagnetic model of the mine environment are as follows:

[0047] Since the mine tunnel structure and the electromagnetic properties of the wall materials underground will affect the propagation process of radar signals, that is, affect the reflection and absorption of millimeter waves, the Maxwell equation is introduced to calculate the losses during the signal propagation process, including the free space loss and the additional loss of material attenuation, which are respectively calibrated as , ;

[0048] Among them, the free space loss has the following calculation formula , where represents the distance, represents the radar signal frequency;

[0049] The additional loss of material attenuation has the following calculation formula , where represents the magnetic permeability of the material, represents the conductivity of the material.

[0050] Preferably, the logical steps for parameter optimization and model update of the radar simulation model are as follows:

[0051] The radar simulation model is established and simulated using computer radar simulation software, which simulates the transmitter, receiver, and signal processing part of the radar, and accurately simulates according to multiple factors such as the size, shape, and reflection characteristics of the detected target vehicle, the mine tunnel structure, and the wall material properties. The genetic algorithm is used to optimize the parameter indicators that quantify the radar performance. The parameter indicators include signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate, which are respectively calibrated as ;

[0052] The genetic algorithm is a cycle of encoding parameters, performing selection, crossover, mutation, and fitness evaluation according to the radar performance indicators to find the optimal solution, and continuously adjusting the parameters in the simulation model according to the results of the genetic algorithm. Among them, the steps of genetic algorithm optimization are as follows: Generate an initial population composed of radar parameters, calibrated as , and the initial population ;

[0053] Calculate the fitness of the initial population , calibrated as , and score according to the performance indicators of the predetermined signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate. Then the function formula for fitness evaluation is , where represents the fitness for evaluating the th radar performance parameter , represents the weight factor of signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate, represents the signal-to-noise ratio parameter set, represents the range resolution parameter set, represents the velocity resolution parameter set, represents the false alarm rate parameter set, represents the An individual;

[0054] Select the parent individuals based on the fitness values and use single-point crossover calculation to generate the next generation, and randomly change the gene values in the offspring genes with a relatively low probability of 0.1% - 1.0% to obtain gene mutation individuals to form a new population , which is used to maintain the population diversity and find a new solution space;

[0055] After several gene genetic calculations, until a predetermined number of generations is reached or an individual with the best fitness appears, the individual with the highest fitness is extracted as the optimal parameter solution for the radar performance.

[0056] Preferably, the acquisition logic steps of the risk controllable coefficient are as follows:

[0057] Establish a convolutional neural network algorithm, use the data obtained from the mine environment electromagnetic model and the radar simulation model as the validation set, and combine the feature data as the training data. During training, the prediction result is obtained through forward propagation, and the error between the actual output and the expected output is calculated by the loss function, and the weights and biases are updated through the backpropagation algorithm. After training is completed, the convolutional neural network model is evaluated on the validation set to predict and determine the probability of the detected target vehicle being dangerous, the probability of abnormal behavior, and the contribution value of environmental factors to the degree of danger, which are respectively calibrated as , , ;

[0058] The risk controllable coefficient, calibrated as is calculated according to the probability of danger directly output by a neuron in the fully connected layer of the convolutional neural network model , the probability of abnormal behavior and the contribution value of environmental factors to the degree of danger of the risk score, which is used to evaluate and control the degree of danger of the detected target vehicle. Among them, the risk controllable coefficient has a calculation formula of , in the formula, respectively represent the probability of danger , the probability of abnormal behavior and the contribution value of environmental factors to the degree of danger of the relative weight factors, and , represents the deviation constant of the risk coefficient.

[0059] Preferably, it includes a millimeter-wave radar integrated machine, which is responsible for transmitting millimeter-wave signals and receiving the radar signals reflected from the vehicle to provide the original radar echo data for vehicle detection;

[0060] The data acquisition preprocessing module directly receives the original radar echo data provided by the millimeter-wave radar integrated device, performs preliminary noise filtering and data fusion preprocessing, and performs feature extraction and interference suppression on the preprocessed data again;

[0061] The environment modeling and radar simulation module creates a model of the trackless rubber-tyred vehicle and the underground passage environment in the mine based on the preprocessed data in the data acquisition preprocessing module, and simulates the performance of the millimeter-wave radar under different conditions by setting environmental characteristic data;

[0062] The vehicle detection and tracking module receives the radar data from the environment modeling and radar simulation modules, detects vehicles from the radar data in real time, and tracks the detected vehicles in real time;

[0063] The signal processing AI analysis and optimization module uses the deep learning algorithm in machine learning technology to intelligently analyze and optimize the radar signal and vehicle trajectory data of the vehicle detection and tracking module, determine the dangerous state of the detection target, and output the analysis results to the visual warning module;

[0064] The visual warning module receives the target hazard analysis results of the signal processing AI analysis and optimization module, and displays them to the user in an intuitive manner, providing real-time alarm function.

[0065] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0066] The present invention establishes a mine environment electromagnetic model and a radar simulation model, and according to the actual structure and material of the mine tunnel, further combines the absorption of radar signals, and accurately optimizes and adjusts the performance parameter signals of the millimeter wave radar integrated machine deployed in the mine and on the trackless rubber-wheeled vehicle, thereby avoiding the interference of reflection, diffraction and occlusion in the complex mining environment, ensuring the accuracy and precision of the feature data extracted from the original radar echo data, and streamlining the number of detection targets, so as to improve the accurate recognition efficiency and accurate positioning effect of target vehicle detection;

[0067] The present invention adopts a convolutional neural network model to deeply analyze and optimize the detection data to generate a controllable coefficient of danger, compares the warning decision threshold to perform visual warning processing, and further accurately predicts and analyzes the danger source of the detection target vehicle, thereby ensuring the safety of driving underground in the mine. In addition, the traffic signal control box in the lane is linked according to the warning strategy to implement necessary traffic management to improve the operating efficiency of the trackless rubber-tyred vehicle, and UWB technology is used to locate the position of the vehicle in the lane, thereby ensuring high-speed, reliable and low-latency transmission of communication, while improving the positioning effect of the detection target vehicle. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments described in the present invention. For those of ordinary skill in the art, other accompanying drawings can also be obtained based on these drawings.

[0069] Figure 1 It is a flowchart of the vehicle detection method for the millimeter-wave radar of the present invention.

[0070] Figure 2 It is a schematic structural diagram of the vehicle detection system for the millimeter-wave radar of the present invention. Detailed implementation manners

[0071] Now, the exemplary embodiments will be described more comprehensively with reference to the accompanying drawings. However, the exemplary embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these exemplary embodiments are provided so that the present disclosure will be more comprehensive and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. Embodiment 1

[0072] As Figure 1 shown, the present invention provides a vehicle detection method for a millimeter-wave radar, including the following steps:

[0073] S1. Install a road-end and vehicle-end millimeter-wave radar integrated machine on the top of the underground roadway and the trackless rubber-tired vehicle respectively. The millimeter-wave radar integrated machine emits millimeter waves and receives echoes to obtain the original radar echo data, which includes the position and movement information of the trackless rubber-tired vehicle in the detected roadway and the self-movement speed information of the detected trackless rubber-tired vehicle.

[0074] It should be noted that the acquisition logic of the original radar echo data is as follows:

[0075] A1. Install a road-end millimeter-wave radar integrated machine on the top of the underground roadway. By continuously emitting pulse signals and recording the echo signals, the radar echo signals in the detected roadway are obtained and calibrated as ;

[0076] A2. Analyze and process the radar echo signals in the roadway. By calculating the delay and frequency change of, the position and movement information of the trackless rubber-tired vehicle in the roadway are determined and calibrated as and ;

[0077] A3. The calculation steps for the position and movement information of the trackless rubber-tired vehicle in the roadway are as follows:

[0078] By measuring the frequency difference between the transmitted signal and the received signal of the millimeter-wave radar in the roadway , calculate the distance, calibrated as , then , where represents the speed of light, i.e., , represents the bandwidth of the radar echo signal in the roadway;

[0079] Using the Doppler effect, according to the Doppler frequency shift measure the speed of the detection target relative to the radar in the roadway, calibrated as , then , where represents the speed of light, i.e., , represents the carrier frequency, and , where represents the wavelength transmitted by the radar;

[0080] Using the beamforming algorithm, measure the angle of the detection target relative to the phased array radar in the roadway, calibrated as , then , represents the wavelength transmitted by the radar, represents the spacing between the th radar array elements, represents the phase difference of the

[0081] A4. The distances , speeds , and angles calculated separately by measuring

[0082] radar stations in the roadway are used to accurately locate the vehicle position using triangulation technology, and the vehicle's motion trajectory is deduced. The Kalman filtering algorithm is used to integrate multiple measurement results, filter out noise, and more accurately predict the vehicle position, and provide a smooth and continuous motion trajectory of the vehicle in the roadway;A5. The vehicle-mounted millimeter-wave radar installed on the trackless rubber-tyred vehicle continuously emits pulse signals to detect the surrounding environment and records the echo signals to obtain the vehicle radar echo signal, calibrated as ;

[0083] A6. Analyze and process the vehicle radar echo signal to obtain the vehicle's own motion speed information based on the radar echo delay and Doppler frequency change, calibrated as , and an Ultra-Wideband (UWB) positioning module is installed on the trackless rubber-tyred vehicle, and the UWB positioning technology is combined to determine the position information of the trackless rubber-tyred vehicle in the roadway and calibrate it as ;

[0084] A7. For the moving speed information of the trackless rubber-tyred vehicle and the position information in the roadway , the calculation steps are as follows:

[0085] The moving speed information of the trackless rubber-tyred vehicle includes the radar echo delay time and the Doppler frequency change, which are respectively calibrated as and , and are respectively used to calculate the distance between the detection target in the roadway and the vehicle radar, calibrated as , and calculate the speed of the detection target in the roadway relative to the vehicle radar, calibrated as , then the distance calculation formula is , where represents the time delay between the radar transmitted pulse and the received echo, and the speed calculation formula is , where represents the speed of the trackless rubber-tyred vehicle relative to the radar, represents the transmission frequency of the radar;

[0086] A8. The UWB positioning module provides high security, high-speed data transmission and high positioning accuracy technology, and absolutely locates and calibrates the radar information by analyzing the time difference and track information of the vehicle radar echo signal , and then uses the Kalman filtering algorithm again to integrate multiple measurement results and reduce errors to provide the position information of the trackless rubber-tyred vehicle under the actual motion state in the roadway .

[0087] S2. Apply the Kalman filtering technology to perform preprocessing operations of noise removal and data fusion on the original radar echo data detected by the vehicle, generate preprocessed data, and perform feature extraction and interference suppression on the preprocessed data to generate feature data for subsequent processing;

[0088] It should be noted that the acquisition logic of the preprocessed data is as follows:

[0089] The preprocessing operations include noise filtering, data fusion, feature extraction and interference suppression. For noise filtering and data fusion, the Kalman filtering technology is combined with the original radar echo data, calibrated as , and , , , to perform prediction and update to evaluate the position and speed of the vehicle; for feature extraction, the Fourier transform algorithm is used to extract from the original radar echo data Extract frequency-domain features; for interference suppression, a spatial filtering algorithm is used to remove the signals of non-target vehicles to suppress the interference generated by the environment;

[0090] Define the original radar echo data as the state variable, that is , noise removal and data fusion are performed using the Kalman filtering algorithm. By predicting and updating the state variable, preprocessed data is obtained and calibrated as , that is , where represents the position and motion information of the vehicle in the roadway, the motion speed information on the trackless rubber-tyred vehicle, and the position information in the roadway at time step to produce the optimal state estimate.

[0091] It should be noted that the acquisition logic of the feature data is as follows:

[0092] For the preprocessed data the least mean square error algorithm is used for adaptive filtering to eliminate the interference signal, and the feature information that can distinguish the target is extracted from the data after interference suppression for subsequent detection and tracking tasks. Among them, the calculation formula of the least mean square error algorithm for adaptive interference cancellation is , where represents the preprocessed data at the current time step of the error signal, represents the updated weight coefficient vector of the preprocessed data represents the weight coefficient vector of the preprocessed data represents the step size parameter, represents the complex conjugate of the input signal of the preprocessed data at the current time step

[0093] Then, according to the preprocessed data after interference suppression the feature data is extracted and calibrated as , then , where represents the set of feature data changing with time , represents the feature data at the moment in the state value, and the represents the feature data The position characteristics of the distance, speed, and angle between the vehicles in the roadway and the trackless rubber-tired vehicle contained therein.

[0094] S3. Use the signal processing algorithms of synthetic aperture radar (SAR) and inverse synthetic aperture radar (ISAR) for radar imaging recognition to obtain radar imaging data. Combine with UWB technology to locate the position information of the target vehicle, and track the motion state of the target vehicle through the multi-target tracking (MHT) algorithm. According to the preset traffic rules and strategies in the mine, combine the traffic facilities such as the linkage traffic signal control box, chamber LED lamp device, and gate device to conduct necessary traffic management on multiple vehicles in the mine passage.

[0095] It should be noted that the recognition and acquisition logic of the radar imaging data is as follows:

[0096] The millimeter-wave radar integrated machine emits and receives the original radar echo data reflected by the target object , convert the time series of the echo into an electrical signal, convert it into a digital signal through an AD converter, and extract the characteristic data through a series of processes such as Kalman filtering to obtain the characteristic data ;

[0097] Use the synthetic aperture radar (SAR) signal processing algorithm for range compression, R-D domain conversion, and azimuth compression calculation to improve the range resolution and azimuth resolution of the millimeter-wave radar. Among them, the calculation formula for range compression is , where represents the compressed signal, represents the characteristic signal received in the millimeter-wave radar echo signal, represents the millimeter-wave radar transmitted signal;

[0098] The calculation formula for R-D domain conversion is , where represents the Fourier transform, represents the RD domain signal, respectively represent the range frequency and Doppler frequency describing the target motion parameters;

[0099] The calculation formula for azimuth compression is , where represents the azimuth difference resolution, represents the signal sampling rate, represents the linear minimum interval of the target relative motion.

[0100] Use the inverse synthetic aperture radar (ISAR) signal processing algorithm for Doppler center frequency estimation and motion compensation calculation to obtain the distance and speed information of the detected target and get a clearer target image. Among them, the calculation formula for Doppler center frequency estimation is , where represents the frequency shift caused by the detection of target motion in the Doppler spectrum of the processed radar echo signal, represents the frequency at which the maximum value is taken, represents the complex convolution function, represents the time-domain signal received by the radar, that is, the echo signal at the time series at the moment;

[0101] The calculation formula for motion compensation is , where represents the signal after motion compensation of the detected target at the moment, ; represents the imaginary part exponential function of the complex number, represents the sampling time;

[0102] After SAR and ISAR imaging processing, radar imaging data is generated. Combining with the UWB base stations arranged in the mine, using the UWB tags equipped on the trackless rubber-tired vehicle to transmit UWB signals, the UWB base stations receive the signals from the UWB tags of the trackless rubber-tired vehicle, and according to the received signal strength and time difference, calculate the precise position of the vehicle UWB tag relative to the base station through trilateration, that is, combining SAR, ISAR imaging data and UWB positioning data to construct a panoramic view of the detailed situation of the trackless rubber-tired vehicle inside the mine.

[0103] It should be noted that the steps of the multi-target tracking MHT algorithm for tracking the motion state of the target vehicle are as follows: From the feature data Measure the observed value of the initial motion state of the detected target vehicle at the time step , and transform the state of the observed value into matrix representation, marked as ;

[0104] Use the motion model to perform state prediction on the initial motion state of each known detected target vehicle to predict the position and speed of the detected target vehicle at the next time step , that is, the prediction formula is , where represents the predicted state of the detected target vehicle at the moment , represents the state estimation of the detected target vehicle at the moment , represents the control input matrix of the detected target vehicle at the moment , represents the control vector of the detected target vehicle at the moment , represents the detected target vehicle at the moment Process noise;

[0105] Use the Mahalanobis distance to calculate the data association degree. For each newly received measurement value of the detected target vehicle, calculate the matching degree with the predicted motion state of the target vehicle. The calculation formula of the Mahalanobis distance is , where Denote as the observation And the target The Mahalanobis distance between them, Respectively denote as the predicted value of the predicted target and the measured value of the observed target, Denote as the actual observation, Denote as the observation matrix, Denote as the observation And the target The covariance matrix associated between them;

[0106] Update the data according to the result of the data association degree, and use the observed data to perform Kalman filter calculation to update the target state estimation. The calculation formula of the Kalman filter is , where Denote as the state estimation of the detected target vehicle at time , Denote as at time The Kalman gain, Denote as the detected target vehicle at time The actual observation, Denote as the detected target vehicle at time The observation matrix.

[0107] S4. Establish an electromagnetic model of the mine environment. By accurately modeling the electromagnetic characteristics in the roadway, use Maxwell's equations to simulate the propagation and reflection characteristics of millimeter-wave signals under mine tunnel conditions, and construct a radar simulation model. Use computer simulation software to simulate and test the performance of millimeter-wave radar under different working conditions. The performance includes signal-to-noise ratio SNR, range-velocity resolution, detection probability, and false alarm rate. Use the genetic algorithm to find the optimal parameter set of the radar and adaptively optimize and adjust the parameters and update the model;

[0108] It should be noted that the steps for establishing the electromagnetic model of the mine environment are as follows:

[0109] Since the mine tunnel structure and the electromagnetic characteristics of the wall materials underground will both affect the propagation process of radar signals, that is, affect the reflection and absorption of millimeter waves, introduce Maxwell's equations to calculate the losses in the signal propagation process, including free-space loss and additional loss of material attenuation, which are respectively calibrated as , ;

[0110] Among them, the free-space loss The calculation formula for is as follows. In the formula, represents the distance, represents the radar signal frequency;

[0111] The additional loss of material attenuation The calculation formula for is as follows. In the formula, represents the magnetic permeability of the material, represents the conductivity of the material.

[0112] It should be noted that the logical steps for parameter optimization and model update of the radar simulation model are as follows:

[0113] The radar simulation model is built and simulated using computer radar simulation software, which simulates the transmitter, receiver, and signal processing parts of the radar, and accurately simulates based on multiple factors such as the size, shape, and reflection characteristics of the detected target vehicle, the mine tunnel structure, and the wall material properties. The genetic algorithm is used to optimize the parameter indicators that quantify the radar performance. The parameter indicators include signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate, which are respectively calibrated as ;

[0114] The genetic algorithm is a loop of encoding parameters, performing selection, crossover, mutation, and fitness evaluation according to the radar performance indicators to find the optimal solution, and continuously adjusting the parameters in the simulation model according to the results of the genetic algorithm to maximize the detection radar performance and minimize the error. Among them, the steps of genetic algorithm optimization are as follows: Generate an initial population composed of radar parameters, calibrated as , and the initial population ;

[0115] Calculate the fitness of the initial population , calibrated as , and score according to the performance indicators of the predetermined signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate. Then the function formula for fitness evaluation is In the formula, represents the fitness for evaluating the th radar performance parameter , represents the weight factor of signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate, represents the signal-to-noise ratio parameter set, represents the range resolution parameter set, represents the velocity resolution parameter set, represents the false alarm rate parameter set, represents the Individuals;

[0116] Select parent individuals based on fitness values and use single-point crossover calculation to generate the next generation, and randomly change the gene values in the offspring genes with a relatively low probability of 0.1% - 1.0% to obtain gene mutation individuals to form a new population , which is used to maintain population diversity and find new solution spaces. Among them, single-point crossover is that the genes of the next generation are exchanged from the gene sequences of the parent individuals to generate new individuals with different characteristics, and gene mutation means that each gene in the gene sequence of each individual has a mutation probability of 0.1% - 1.0%. When any gene is selected to mutate, the gene sequence value of the individual itself will change;

[0117] After several gene inheritance calculations, until a predetermined number of generations is reached or an individual with the best fitness appears, the individual with the highest fitness is extracted as the optimal parameter solution for radar performance.

[0118] S5. Use the convolutional neural network algorithm to conduct in-depth intelligent analysis on the feature data, and use the automatic learning technology of training, testing, verification, analysis and optimization to identify the features of different targets, judge the detected target attributes and whether they are potential hazard sources, and generate a risk control coefficient to improve the accuracy and robustness of the detection system;

[0119] It should be noted that the acquisition logic steps of the risk control coefficient are as follows:

[0120] Establish a convolutional neural network algorithm, use the data obtained from the mine environment electromagnetic model and the radar simulation model as the validation set, and combine the feature data as the training data. During training, the prediction result is obtained through forward propagation, and the loss function calculates the error between the actual output and the expected output, and updates the weights and biases through the backpropagation algorithm. After training, the convolutional neural network model is evaluated on the validation set to predict and determine the probability of the detected target vehicle being dangerous, the probability of abnormal behavior, and the contribution value of environmental factors to the degree of danger, which are respectively calibrated as , , ;

[0121] The risk control coefficient, calibrated as, is the probability of directly outputting danger by a neuron in the fully connected layer of the convolutional neural network model , the probability of abnormal behavior and the contribution value of environmental factors to the degree of danger to calculate the risk score, which is used to evaluate and control the degree of danger of the detected target vehicle. Among them, the risk control coefficient The calculation formula of is , where, The probability of dangerousness , the probability of abnormal behavior and environmental factors' contribution to the risk level The relative weight factor of , Expressed as a deviation constant for the hazard factor.

[0122] S6. Use WIFI, 4G or MESH technology to exchange data among various modules in the detection system, and convert the controllable coefficient of danger into visual information to synchronize with the driver in the rubber-tyred trackless vehicle. The visual information includes the traffic flow status of the entire channel, and compares the warning decision threshold with the judgment result of the visual information, thereby triggering an alarm.

[0123] It should be noted that the setting logic of the warning decision threshold is as follows:

[0124] Based on the accident records of rubber-tyred trackless vehicles in the past, traffic data, and the environment and structure of the nodes where accidents frequently occur in the mine tunnel, different scenarios are simulated to test different risk controllability coefficients. The effect of triggering alarms can be adjusted to ensure that dangers can be effectively prevented and sufficient response time is provided in the actual mine tunnel situation;

[0125] Set the warning threshold to ,and , where It is expressed as the minimum value of the controllable range of danger. It is expressed as the maximum value of the controllable range of danger. When the alarm is triggered continuously in real time, the location of vehicles in the tunnel and the visualization of the traffic flow status of the whole scene channel are synchronized, and the traffic signal control box in the tunnel is linked to implement necessary traffic management for multiple vehicles in the mine channel according to the preset traffic rules and strategies in the mine. When the vehicle position in the tunnel is synchronized, the alarm is triggered at intervals, and the visualization information of the vehicle position in the tunnel and the traffic flow status of the whole scene channel is synchronized, so that the trackless rubber-tyred vehicle driver can actively control the traffic operation status according to the traffic rules of the mine tunnel;

[0126] At the same time, combined with the alarm frequency, alarm sensitivity and false alarm rate, data is transmitted using WIFI, 4G or MESH technology, and the driver experience data is fed back to the convolutional neural network model for optimization training, and the risk control coefficient is accurately generated. , and further carry out early warning processing, respond quickly when potential target vehicle risks are detected in time, ensure the safety of driving underground in the mine while ensuring high-speed, reliable and low-latency transmission of communications.

[0127] Example 2

[0128] As Figure 2 shown, the present invention provides a vehicle detection system for a millimeter-wave radar, including a millimeter-wave radar integrated machine, which is the core component used for vehicle detection in a mining machine and for use in a trackless rubber-tired vehicle. It is responsible for transmitting millimeter-wave signals and receiving radar signals reflected from vehicles to provide original radar echo data for vehicle detection;

[0129] A data acquisition and preprocessing module directly receives the original radar echo data provided by the millimeter-wave radar integrated machine, and performs preliminary noise filtering and data fusion preprocessing, and further performs feature extraction and interference suppression on the preprocessed data to improve the quality of the radar signal data and provide a cleaner signal for the subsequent vehicle detection and tracking module and signal processing AI analysis and optimization module processing stages;

[0130] An environment modeling and radar simulation module creates a model of the trackless rubber-tired vehicle and the underground passage environment where it is located according to the preprocessed data in the data acquisition and preprocessing module. By setting environmental characteristic data, it simulates the performance of the millimeter-wave radar under different conditions to optimize the parameter settings in actual applications and improve the accuracy of vehicle detection;

[0131] A vehicle detection and tracking module receives the radar data that has passed through the environment modeling and radar simulation module, and detects vehicles in real time from the radar data and tracks the detected vehicles in real time;

[0132] A signal processing AI analysis and optimization module uses deep learning algorithms in machine learning technology to perform intelligent analysis and optimization on the radar signals and vehicle trajectory data of the vehicle detection and tracking module, judges the dangerous state of the detection target, and outputs the analysis results to the visualization warning module to improve the accuracy and robustness of vehicle detection and tracking;

[0133] A visualization warning module receives the target danger analysis results of the signal processing AI analysis and optimization module and displays them to the user in an intuitive manner, providing a real-time alarm function to assist vehicle driving and improve the monitoring task.

[0134] A vehicle detection method for a millimeter-wave radar provided by an embodiment of the present invention is implemented through the above vehicle detection system for a millimeter-wave radar. The specific methods and processes of a vehicle detection method and system for a millimeter-wave radar are as described in the embodiments of the above vehicle detection method and system for a millimeter-wave radar, and will not be elaborated here.

[0135] The above formulas are all dimensionless and take their numerical values for calculation. The formulas are obtained by collecting a large amount of data for software simulation to obtain a formula that is closest to the actual situation. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0136] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wired or wireless (such as infrared, wireless, microwave, etc.) manner. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more sets of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0137] It should be understood that in various embodiments of the present application, the order numbers of the above processes do not indicate the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0138] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0139] As described above, the above are only specific embodiments of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present application, and all such changes or substitutions should be covered by the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A vehicle detection method using millimeter wave radar, characterized in that: The steps include: S1. Install a road-side integrated millimeter-wave radar on the top of the underground tunnel and on the trackless rubber-tyred vehicle, respectively. The millimeter-wave radar integrated device transmits millimeter waves and receives echoes to obtain original radar echo data. The original radar echo data includes the position and movement information of the trackless rubber-tyred vehicle in the tunnel and the movement speed information of the trackless rubber-tyred vehicle itself. S2. Apply Kalman filtering technology to perform preprocessing operations of noise removal and data fusion on the original radar echo data of vehicle detection to generate preprocessed data, and perform feature extraction and interference suppression on the preprocessed data to generate feature data for subsequent processing; S3. Use the signal processing algorithms of synthetic aperture radar (SAR) and inverse synthetic aperture radar (ISAR) to perform radar imaging recognition, obtain radar imaging data, and locate the target vehicle position information in combination with UWB technology, and track the movement status of the target vehicle through the multi-target tracking (MHT) algorithm. According to the preset traffic rules and strategies in the mine, combined with the traffic facilities of the linkage traffic signal control box, chamber LED light device, and gate device, necessary traffic management is carried out for multiple vehicles in the mine passage; S4. Establish an electromagnetic model of the mine environment. By accurately modeling the electromagnetic characteristics in the tunnel, use Maxwell's equations to simulate the propagation and reflection characteristics of millimeter-wave signals under mine tunnel conditions, and build a radar simulation model. Use computer simulation software to simulate and test the performance of millimeter-wave radar under different working conditions, including signal-to-noise ratio (SNR), distance and velocity resolution, detection probability and false alarm rate. Use genetic algorithms to find the best radar parameter set and adaptively optimize and adjust the parameters and update the model. S5. Use convolutional neural network algorithm to conduct in-depth intelligent analysis of feature data, and use automatic learning technology that has been trained, tested, verified, analyzed and optimized to identify the characteristics of different targets, determine the attributes of the detected targets and whether they are potential sources of danger, and generate risk controllability coefficients to improve the accuracy and robustness of the detection system; S6. Use WIFI, 4G or MESH technology to exchange data among various modules in the detection system, and convert the controllable coefficient of danger into visual information to synchronize with the driver in the rubber-tyred trackless vehicle. The visual information includes the traffic flow status of the entire channel, and compares the warning decision threshold with the judgment result of the visual information, thereby triggering an alarm.

2. The vehicle detection method of millimeter wave radar according to claim 1, characterized in that: The acquisition logic of the original radar echo data is as follows: A1. Install a road-end millimeter-wave radar integrated device at the top of the underground tunnel, continuously transmit pulse signals and record echo signals to obtain radar echo signals in the detection tunnel, calibrated as ; A2. Radar echo signal in the tunnel Analyze and process The delay and frequency change of the trackless rubber-tyred vehicle in the lane are determined to determine the position and motion information of the trackless rubber-tyred vehicle in the lane, which are calibrated as and ; A3. Position of rubber-tyred trackless vehicles in the lane and sports information The calculation steps are as follows: By measuring the frequency difference between the transmitted and received signals of the millimeter-wave radar in the lane , calculate the distance and calibrate it as ,but , where Expressed as the speed of light, , Represented as radar echo signal in the tunnel bandwidth; Using the Doppler effect, according to the Doppler frequency shift Measure the speed of the detected target relative to the radar in the lane and calibrate it as ,but , where Expressed as the speed of light, , is the carrier frequency, and , where Expressed as the wavelength of radar transmission; The beamforming algorithm is used to measure the angle of the detected target relative to the phased array radar in the tunnel and calibrate it as ,but , Expressed as the wavelength of radar transmission, Expressed as The spacing between radar array elements is Expressed as The phase difference of the receiving array elements; A4. By measuring the inside of the tunnel The distance calculated for each radar station ,speed and angle , using triangulation technology to accurately locate the vehicle position and infer the vehicle's trajectory, and using the Kalman filter algorithm to integrate multiple measurement results, filter out noise and more accurately predict the vehicle's position, and provide a smooth and continuous trajectory of the vehicle in the lane; A5. The vehicle-side millimeter-wave radar installed on the rubber-tyred trackless vehicle detects the surrounding environment by continuously emitting pulse signals and recording echo signals to obtain the vehicle radar echo signal, which is calibrated as ; A6. Vehicle radar echo signal Analyze and process the vehicle's own speed information based on the radar echo delay and Doppler frequency change, and calibrate it as , and installed a UWB positioning module on the trackless rubber-tyred vehicle, and combined with UWB positioning technology to determine the position information of the trackless rubber-tyred vehicle in the lane and calibrate it ; A7. Speed ​​information on rubber-tyred trackless vehicles and location information in the lane The calculation steps are as follows: Movement speed information on rubber-tyred trackless vehicles Including radar echo delay time and Doppler frequency change, calibrated as and , and are used to calculate the distance between the detected target and the vehicle radar in the lane and calibrate , and calculate the speed of the detected target in the lane relative to the vehicle radar, calibrated as , then the distance calculation formula is , where It is expressed as the time delay between the radar transmitting pulse and the received echo, and the speed calculation formula is , where It is expressed as the speed of the rubber-tyred trackless vehicle relative to the radar, Expressed as the transmitting frequency of the radar; A8, UWB positioning module provides high security, high-speed data transmission and high positioning accuracy technology, and analyzes the vehicle radar echo signal The time difference and track information are used to absolutely locate and calibrate the radar information, and the Kalman filter algorithm is used again to integrate multiple measurement results and reduce errors to provide the position information of the trackless rubber-tyred vehicle in the actual motion state in the lane. .

3. The vehicle detection method of millimeter wave radar according to claim 2, characterized in that: The acquisition logic of the preprocessed data is as follows: The preprocessing operation uses Kalman filtering technology to combine the original radar echo data and calibrate it for noise filtering and data fusion. ,and , , , make predictions and updates; Defining raw radar echo data is the state variable, that is ,Noise removal and data fusion are carried out by using Kalman filtering algorithm, which obtains preprocessed data and calibrates them by predicting and updating state variables. ,Right now , where Represented as the position of the vehicle in the lane and sports information , Movement speed information on rubber-tyred trackless vehicles and location information in the lane At time step The best state estimate produced.

4. The vehicle detection method of millimeter wave radar according to claim 3, characterized in that: The acquisition logic of the feature data is as follows: Preprocessing data The minimum mean square error algorithm is used for adaptive filtering to eliminate interference signals, and feature information that can distinguish targets is extracted from the interference suppressed data for subsequent detection and tracking tasks. The calculation formula of the minimum mean square error algorithm used in adaptive interference elimination is: , where Represented as preprocessed data At the current time step The error signal, Represented as preprocessed data The updated weight coefficient vector, Represented as preprocessed data The weight coefficient vector of is represented as the step size parameter, Represented as preprocessed data At the current time step The complex conjugate of the input signal; According to the preprocessed data after interference suppression Extract feature data and mark it as ,but , where Represented as feature data over time A collection of changes, Represented as feature data At the moment The state value, Represented as feature data The position characteristics of distance, speed and angle of vehicles and rubber-tyred trackless vehicles in the lane are included.

5. The vehicle detection method of millimeter wave radar according to claim 4, characterized in that: The recognition and acquisition logic of the radar imaging data is as follows: The millimeter wave radar integrated device transmits and receives the original radar echo data reflected by the target object , so that the time series of echoes is converted into electrical signals, and then converted into digital signals through AD converters, and feature data is extracted after a series of processing through Kalman filtering to obtain feature data ; The synthetic aperture radar SAR signal processing algorithm is used to perform range compression, RD domain conversion and azimuth compression calculations to improve the range resolution and azimuth resolution of the millimeter wave radar. The calculation formula for range compression is: , where Represented as the compressed signal, It is represented as the characteristic signal received in the millimeter-wave radar echo signal. Represents the millimeter wave radar transmission signal; The calculation formula for RD domain conversion is: , where Expressed as Fourier transform, Represented as an RD domain signal, They are respectively represented as the range frequency and Doppler frequency describing the target motion parameters; The calculation formula for azimuth compression is: , where Expressed as the azimuth difference resolution, Expressed as the signal sampling rate, It is expressed as the linear minimum interval of the target's relative motion; The inverse synthetic aperture radar (ISAR) signal processing algorithm is used to perform Doppler center frequency estimation and motion compensation calculation to obtain the distance and speed information of the detected target and obtain a clearer target image. The calculation formula for Doppler center frequency estimation is: , where It is expressed as the frequency shift caused by detecting target motion in the Doppler spectrum of the radar echo signal. It is expressed as the frequency of taking the maximum value, is represented as a complex convolution function, It is represented as the time domain signal received by the radar, that is, the echo signal is Time series of moments; The calculation formula for motion compensation is: , where Represents the detection target in After motion compensation signal, Expressed as the exponential function of the imaginary part of a complex number, It is represented as sampling time; Radar imaging data is generated after SAR and ISAR imaging processing. Combined with the UWB base station arranged in the mine, the UWB tag equipped with the trackless rubber-tyred vehicle is used to transmit the UWB signal. The UWB base station receives the signal from the UWB tag of the trackless rubber-tyred vehicle and calculates the precise position of the vehicle's UWB tag relative to the base station through trilateral measurement based on the received signal strength and time difference. That is, the SAR, ISAR imaging data and UWB positioning data are combined to construct a panoramic picture of the detailed situation of the trackless rubber-tyred vehicle inside the mine.

6. The vehicle detection method of millimeter wave radar according to claim 5, characterized in that: The steps of the multi-target tracking MHT algorithm to track the motion state of the target vehicle are as follows: From the feature data The target vehicle is detected at the time step The observation value of the initial motion state is converted into a matrix representation, marked as ; Use the motion model to predict the initial motion state of each known detection target vehicle to predict the detection target vehicle at the next time step. The position and speed of , where Represents the detection of the target vehicle at time The predicted state of Represents the detection of the target vehicle at time The state estimate of Represents the detection of the target vehicle at time The control input matrix, Represents the detection of the target vehicle at time The control vector of Represents the detection of the target vehicle at time process noise; The Mahalanobis distance is used to calculate the data association degree. For each newly received measurement value of the detected target vehicle, the matching degree with the predicted target vehicle motion state is calculated. The calculation formula of the Mahalanobis distance is: , where Represented as observation With the goal The Mahalanobis distance between They are respectively represented as the predicted value of the predicted target and the measured value of the observed target, Represents actual observations, It is represented as the observation matrix, Represented as observation With the goal The covariance matrix of the associations between ; The data is updated according to the result of data association, and the target state estimation is updated by Kalman filter calculation using the observed data. The calculation formula of Kalman filter is: , where Represents the updated detection target vehicle at time The state estimate of Expressed as at time The Kalman gain, Represents the detection of the target vehicle at time The actual observation of Represents the detection of the target vehicle at time The observation matrix.

7. The vehicle detection method of millimeter wave radar according to claim 6, characterized in that: The steps for establishing the mine environment electromagnetic model are as follows: Since the electromagnetic properties of the tunnel structure and wall materials in the mine will affect the propagation process of the radar signal, that is, affect the reflection and absorption of the millimeter wave, the Maxwell equations are introduced to calculate the loss in the signal propagation process, including free space loss and additional loss due to material attenuation, which are calibrated as , ; Among them, free space loss The calculation formula is , where Expressed as distance, Expressed as radar signal frequency; Additional loss due to material attenuation The calculation formula is , where Expressed as the magnetic permeability of the material, Expressed as the electrical conductivity of the material.

8. The vehicle detection method using millimeter wave radar according to claim 7, characterized in that: The logical steps of parameter optimization and model updating of the radar simulation model are as follows: The radar simulation model is modeled and simulated using computer radar simulation software to simulate the transmitter, receiver and signal processing parts of the radar. It is accurately simulated based on the size, shape and reflection characteristics of the target vehicle, the tunnel structure and the wall material properties. The genetic algorithm is used to optimize the parameter indicators of the quantitative radar performance. The parameter indicators include signal-to-noise ratio, distance resolution, velocity resolution, detection probability and false alarm rate, which are calibrated as ; The genetic algorithm is based on radar performance indicators, encoding parameters, and a cycle of selection, crossover, mutation, and fitness evaluation to find the optimal solution, and continuously adjust the parameters in the simulation model according to the results of the genetic algorithm. The steps of the genetic algorithm optimization process are as follows: Generate an initial population consisting of radar parameters, calibrated as , and the initial population ; Calculate the initial population The fitness of , and score according to the predetermined performance indicators of signal-to-noise ratio, distance resolution, velocity resolution, detection probability and false alarm rate, then the function formula for fitness evaluation is: , where Represents the evaluation Radar performance parameters The fitness of Expressed as weighting factors for signal-to-noise ratio, range resolution, velocity resolution, detection probability, and false alarm rate, Expressed as a signal-to-noise ratio parameter set, Expressed as a range resolution parameter set, Expressed as a velocity resolution parameter set, Expressed as a false alarm rate parameter set, Represented as the radar performance parameter set individual; The parent individuals are selected based on the fitness value and the next generation is generated using single-point crossover calculation. The gene values ​​in the offspring genes are randomly changed with a relatively low probability of 0.1%-1.0%, and the gene mutation individuals are obtained to form a new population. , used to maintain population diversity and find new solution spaces; After several genetic calculations, until a predetermined number of generations or an individual with the best fitness appears, the individual with the highest fitness is extracted as the optimal parameter solution for radar performance.

9. The vehicle detection method using millimeter wave radar according to claim 8, characterized in that: The logical steps for obtaining the risk controllable coefficient are as follows: A convolutional neural network algorithm was established, and the data obtained from the mine environment electromagnetic model and radar simulation model were used as the validation set, combined with the feature data As training data, the prediction results are obtained through forward propagation during training, and the loss function calculates the error between the actual output and the expected output, and the weights and biases are updated through the back propagation algorithm. After the training is completed, the convolutional neural network model is evaluated on the validation set to predict the probability of detecting the target vehicle being dangerous, the probability of abnormal behavior, and the contribution of environmental factors to the degree of danger, which are calibrated as , , ; Danger controllable coefficient, calibrated as It is the probability of directly outputting danger based on a neuron in the fully connected layer of the convolutional neural network model. , the probability of abnormal behavior and environmental factors' contribution to the risk level The risk score is calculated to evaluate and control the danger level of the detected target vehicle, among which the controllable risk coefficient The calculation formula is , where The probability of dangerousness , the probability of abnormal behavior and environmental factors' contribution to the risk level The relative weight factor of , Expressed as a deviation constant for the hazard factor.

10. A vehicle detection system using a millimeter wave radar, according to a vehicle detection method using a millimeter wave radar according to any one of claims 1 to 9, characterized in that: It includes a millimeter wave radar integrated device, which is responsible for transmitting millimeter wave signals and receiving radar signals reflected from vehicles to provide raw radar echo data for vehicle detection; The data acquisition preprocessing module directly receives the original radar echo data provided by the millimeter-wave radar integrated device, performs preliminary noise filtering and data fusion preprocessing, and performs feature extraction and interference suppression on the preprocessed data again; The environment modeling and radar simulation module creates a model of the trackless rubber-tyred vehicle and the underground passage environment in the mine based on the preprocessed data in the data acquisition preprocessing module, and simulates the performance of the millimeter-wave radar under different conditions by setting environmental characteristic data; The vehicle detection and tracking module receives the radar data from the environment modeling and radar simulation modules, detects vehicles from the radar data in real time, and tracks the detected vehicles in real time; The signal processing AI analysis and optimization module uses the deep learning algorithm in machine learning technology to intelligently analyze and optimize the radar signal and vehicle trajectory data of the vehicle detection and tracking module, determine the dangerous state of the detection target, and output the analysis results to the visual warning module; The visual warning module receives the target hazard analysis results of the signal processing AI analysis and optimization module, and displays them to the user in an intuitive manner, providing real-time alarm function.

Images