Low-speed vehicle detection system for long uphill road section of expressway

Abstract

The invention relates to the technical field of highway low-speed vehicle detection, and discloses a highway long uphill road section low-speed vehicle detection system, a sensor module is used for installing a millimeter wave radar on the roadside of a highway long uphill road section, and a vehicle is detected by using millimeter wave band electromagnetic wave emission and reception reflection signals. The method comprises the following steps: shooting a road in real time through video monitoring equipment, and monitoring a live scene of a long uphill road section of an expressway; the data processing module is used for receiving the data sent by the sensor module and analyzing and processing the data so as to determine the speed, the position and the driving state of the vehicle, and when a low-speed vehicle is detected, the data processing module can send the information of the vehicle to the warning module; the warning module is used for sending warning information to a driver according to the warning mode and the warning duration threshold value and warning vehicles behind in time; the communication module is used for sending the detected low-speed vehicle information to a traffic management department; the human cost is reduced, and the management efficiency is improved.

Description

Technical Field

[0001] The invention relates to the technical field of highway low-speed vehicle detection, and in particular to a low-speed vehicle detection system on a long uphill section of a highway. Background Art

[0002] With the rapid growth of highway construction and traffic volume, its traffic efficiency and driving safety have attracted much attention. Highways have clear speed limits. On long uphill sections of highways, some vehicles may travel at low speeds due to vehicle load, insufficient power and other reasons. Such low-speed vehicles will not only affect the traffic efficiency of highways, but also easily cause traffic accidents such as rear-end collisions, posing serious hidden dangers to traffic safety. At present, the detection of low-speed vehicles on long uphill sections of highways mainly relies on manual patrols and fixed speed measuring equipment, but these methods have problems such as low detection efficiency and limited coverage. Summary of the invention

[0003] The purpose of the present invention is to solve the above problems and to design a low-speed vehicle detection system for a long uphill section of a highway.

[0004] The present invention provides a low-speed vehicle detection system for a long uphill section of a highway, wherein the low-speed vehicle detection system for a long uphill section of a highway comprises a sensor module, a data processing module, a warning module and a communication module, wherein:

[0005] The sensor module is used to install the millimeter-wave radar on the roadside of the long uphill section of the highway, use the millimeter-wave electromagnetic wave to transmit and receive reflected signals to detect vehicles, and use video monitoring equipment to take real-time photos of the road to monitor the real scene of the long uphill section of the highway;

[0006] A data processing module is used to receive the data sent by the sensor module, and analyze and process the data to determine the speed, position and driving status of the vehicle. When a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module;

[0007] The warning module is used to send warning information to the driver according to the alarm mode and alarm duration threshold, and timely warn the rear vehicle;

[0008] The communication module is used to send the detected low-speed vehicle information to the traffic management department so that the traffic management department can grasp the on-site situation in time and take measures to deal with it.

[0009] Optionally, in the first implementation of the present invention, when the millimeter wave radar transmits electromagnetic waves, the transmitting time t1 is recorded. When the electromagnetic waves are reflected by the vehicle, the reflected signals are received and the receiving time t2 is recorded. According to the electromagnetic wave propagation speed c, the distance d between the radar and the vehicle is calculated:

[0010] d=c×(t2-t1) / 2

[0011] The distance is measured multiple times within a short time interval Δt, and the radial velocity v of the vehicle is calculated by the ratio of the distance change Δd to the time interval Δt:

[0012] v=Δd / Δt

[0013] The beamforming algorithm and signal processing algorithm are used to analyze the reflected signals in multiple directions. Assume that there are N antenna units for the reflected signal, and the received reflected signal is represented by x n (t)(n=1,2,…,N), the signal after beamforming is y(t):

[0014]

[0015] Among them, w n is the weighting coefficient of the nth antenna unit, and the scanning and focusing of the beam in different directions are determined by adjusting the weighting coefficient;

[0016] The beamformed signal is used for target detection, and the distance between the target and the radar is calculated based on the propagation time of the electromagnetic wave. The position information of the vehicle in the two-dimensional plane is determined by target detection and distance measurement in multiple beam directions.

[0017] Optionally, in a second implementation of the present invention, the data processing module includes a denoising processing submodule, a model recognition submodule, a speed calculation submodule and a low-speed judgment submodule, wherein:

[0018] The denoising processing submodule is used to obtain the image collected by the video monitoring device and perform denoising processing on the image using a Gaussian filtering algorithm to obtain a denoised image;

[0019] A model recognition submodule is used to input the denoised image into a vehicle recognition model for recognition to obtain a vehicle recognition result, wherein the vehicle recognition model is constructed based on a YOLOv7 network structure, and the YOLOv7 network structure consists of four parts: an input end, a backbone network, a neck network, and a prediction end. The vehicle recognition model introduces a CA coordinate attention mechanism into the neck network of the YOLOv7 network structure, and combines the efficient aggregation network module of the backbone network with the lightweight network FasterNet into an efficient aggregation lightweight network;

[0020] A speed calculation submodule is used to analyze each frame of the image collected by the video monitoring device during real-time monitoring, detect the vehicle in the image according to the vehicle recognition result output by the vehicle recognition model, determine the vehicle position in combination with the millimeter-wave radar data, compare the vehicle positions in adjacent frames through an inter-frame difference algorithm, calculate the movement speed of the vehicle in the image, and then convert the speed in the image into the speed on the actual road according to the installation parameters of the video monitoring device to obtain the vehicle speed, wherein the installation parameters at least include focal length and shooting angle;

[0021] The low-speed judgment submodule is used to make judgments based on the vehicle's position, speed and driving status. When the vehicle speed is lower than the preset low-speed threshold, it is judged as a low-speed vehicle. The data processing module encapsulates the vehicle's information into a data frame and sends it to the warning module. The preset low-speed threshold is 60 kilometers per hour.

[0022] Optionally, in a third implementation of the present invention, the vehicle identification model includes the following process when performing identification:

[0023] The input end uniformly scales the input original image to a standard size. The backbone network extracts features from the original image passing through the input end through the convolution normalization activation module, extracts target features through the efficient aggregation lightweight network module, expands the receptive field of the feature layer through the maximum pooling operation of the downsampling structure, and connects it in parallel with the convolution normalization activation module to fuse feature information.

[0024] The neck network uses a spatial pyramid pooling network combined with channel amplitude propagation convolution, an efficient aggregation network, a convolution normalization activation module, upsampling operations, and downsampling operations to fuse feature layers of different sizes. In the CA coordinate attention mechanism, the input tensor is split into a tensor in the width direction and a tensor in the height direction. The two tensors are concatenated after average pooling. After passing through two-dimensional convolution, batch normalization, and nonlinear activation functions in turn, the output retains the channel information correlation and fully utilizes the intermediate tensor of position information. The intermediate tensor is split into an intermediate tensor in the height direction and an intermediate tensor in the width direction, and two-dimensional convolution and Sigmoid function activation are performed respectively. Finally, the two outputs are multiplied by the original input tensor to obtain the final output tensor.

[0025] The prediction end trains the feature layer through re-parameterized convolution, and then generates detection boxes for small, medium and large targets through the convolutional normalization activation module to obtain the vehicle recognition result.

[0026] Optionally, in a fourth implementation of the present invention, the vehicle driving state is judged by the following process: calculating the speed change rate Δv / Δt of the vehicle at consecutive time points, and if Δv / Δt is greater than a set acceleration threshold a th1, the vehicle is judged to be in an accelerating state; if Δv / Δt is less than the set deceleration threshold a th2 , the vehicle is judged to be in a deceleration state; if Δv / Δt is close to zero within a certain range, the vehicle is judged to be in a constant speed state; if the vehicle speed continues to be close to zero for a certain period of time, the vehicle is judged to be stationary, where Δv is the speed difference of the vehicle at two consecutive time points, Δt is the time interval between two consecutive speed measurement moments, and the acceleration threshold a is th1 is 0.8m / s2, a th2 <0.

[0027] Optionally, in a fifth implementation of the present invention, the warning module includes a warning light linkage submodule, a laser light and voice broadcast submodule and a road network terminal submodule, wherein:

[0028] The warning light linkage submodule is used to trigger the linkage alarm mechanism when a low-speed vehicle is detected, so that the warning lights on the roadside work. The warning lights are normally yellow, and the red light is immediately lit when a low-speed event is triggered;

[0029] The laser light and voice broadcast submodule is used to turn on the laser light and flash mode when a low-speed vehicle is detected, and the tweeter makes on-site voice broadcasts;

[0030] The road network terminal submodule is used to display relevant prompt information through variable information signs in the road section when a low-speed vehicle is detected, reminding the driver to pay attention to the driving conditions.

[0031] Optionally, in a sixth implementation of the present invention, a method for implementing a low-speed vehicle detection system on a long uphill section of a highway comprises the following steps:

[0032] Millimeter-wave radars are installed on the roadside of long uphill sections of highways. Vehicles are detected by transmitting and receiving reflected signals of electromagnetic waves in the millimeter-wave band. The road is photographed in real time through video monitoring equipment to monitor the actual scene of the long uphill sections of highways.

[0033] Receive the data sent by the sensor module, analyze and process the data to determine the speed, position and driving status of the vehicle. When a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module;

[0034] Send warning information to the driver according to the alarm mode and alarm duration threshold, and warn the vehicles behind in time;

[0035] The detected low-speed vehicle information is sent to the traffic management department so that the traffic management department can grasp the on-site situation in time and take measures to deal with it.

[0036] Optionally, in a seventh implementation of the present invention, the receiving of data sent by the sensor module, and analyzing and processing the data to determine the speed, position and driving state of the vehicle, when a low-speed vehicle is detected, the data processing module sends the vehicle information to the warning module, including:

[0037] Acquire the image collected by the video monitoring device, and use the Gaussian filter algorithm to denoise the image to obtain a denoised image;

[0038] Input the denoised image into a vehicle recognition model for recognition, and obtain a vehicle recognition result, wherein the vehicle recognition model is constructed based on a YOLOv7 network structure, and the YOLOv7 network structure consists of four parts: an input end, a backbone network, a neck network, and a prediction end. The vehicle recognition model introduces a CA coordinate attention mechanism into the neck network of the YOLOv7 network structure, and combines the efficient aggregation network module of the backbone network with the lightweight network FasterNet into an efficient aggregation lightweight network;

[0039] In the real-time monitoring process, each frame of the image collected by the video monitoring device is analyzed, the vehicle in the image is detected according to the vehicle recognition result output by the vehicle recognition model, and the vehicle position is determined in combination with the millimeter wave radar data, the vehicle positions in adjacent frames are compared by an inter-frame difference algorithm, the movement speed of the vehicle in the image is calculated, and then the speed in the image is converted into the speed on the actual road according to the installation parameters of the video monitoring device to obtain the vehicle speed, wherein the installation parameters at least include the focal length and the shooting angle;

[0040] The vehicle position, speed and driving status are combined for judgment. When the vehicle speed is lower than the preset low-speed threshold, it is determined to be a low-speed vehicle. The data processing module encapsulates the vehicle information into a data frame and sends it to the warning module. The preset low-speed threshold is 60 kilometers per hour.

[0041] In the technical solution provided by the present invention, the front-end radar vision equipment, with the help of advanced algorithms and detection technologies, can accurately detect the vehicle's travel trajectory and speed in real time, and quickly and accurately identify low-speed vehicles; the system links to warn vehicles coming from behind, effectively reminding the driver to pay attention to the driving conditions, and significantly reducing safety incidents such as rear-end collisions caused by vehicles traveling at low speeds; the system has the advantages of rapid detection, accurate identification, timely alarm and efficient management, and has high work efficiency, providing strong guarantees for road traffic safety and smoothness, and can also provide data support for traffic management, assist in formulating scientific management strategies, reduce labor costs, and improve management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0043] Figure 1 A schematic diagram of the structure of a low-speed vehicle detection system on a long uphill section of a highway provided by an embodiment of the present invention;

[0044] Figure 2 A schematic diagram of the data processing module structure of a low-speed vehicle detection system for a long uphill section of a highway provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0046] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 A schematic diagram of the structure of a low-speed vehicle detection system for a long uphill section of a highway provided by an embodiment of the present invention, the system includes a sensor module, a data processing module, a warning module and a communication module, wherein:

[0047] The sensor module is used to install the millimeter-wave radar on the roadside of the long uphill section of the highway, use the millimeter-wave electromagnetic wave to transmit and receive reflected signals to detect vehicles, and use video monitoring equipment to take real-time photos of the road to monitor the real scene of the long uphill section of the highway;

[0048] The data processing module is used to receive the data sent by the sensor module, analyze and process the data to determine the speed, position and driving status of the vehicle. When a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module;

[0049] The warning module is used to send warning information to the driver according to the alarm mode and alarm duration threshold, and timely warn the rear vehicle;

[0050] The communication module is used to send the detected low-speed vehicle information to the traffic management department so that the traffic management department can grasp the on-site situation in time and take measures to deal with it.

[0051] In this embodiment, when the millimeter wave radar transmits electromagnetic waves, the transmitting time t1 is recorded. When the electromagnetic waves are reflected by the vehicle, the reflected signals are received and the receiving time t2 is recorded. According to the electromagnetic wave propagation speed c, the distance d between the radar and the vehicle is calculated:

[0052] d=c×(t2-t1) / 2

[0053] The distance is measured multiple times within a short time interval Δt, and the radial velocity v of the vehicle is calculated by the ratio of the distance change Δd to the time interval Δt:

[0054] v=Δd / Δt

[0055] The beamforming algorithm and signal processing algorithm are used to analyze the reflected signals in multiple directions. Assume that there are N antenna units for the reflected signal, and the received reflected signal is represented by x n (t)(n=1,2,…,N), the signal after beamforming is y(t):

[0056]

[0057] Among them, w n is the weighting coefficient of the nth antenna unit, and the scanning and focusing of the beam in different directions are determined by adjusting the weighting coefficient;

[0058] The beamformed signal is used for target detection, and the distance between the target and the radar is calculated based on the propagation time of the electromagnetic wave. The position information of the vehicle in the two-dimensional plane is determined by target detection and distance measurement in multiple beam directions.

[0059] In this embodiment, Figure 2 As shown, the data processing module includes a denoising processing submodule, a model recognition submodule, a speed calculation submodule and a low-speed judgment submodule, wherein:

[0060] The denoising processing submodule is used to obtain the image collected by the video monitoring device and perform denoising processing on the image using a Gaussian filtering algorithm to obtain a denoised image;

[0061] The model recognition submodule is used to input the denoised image into the vehicle recognition model for recognition to obtain the vehicle recognition result. The vehicle recognition model is built based on the YOLOv7 network structure. The YOLOv7 network structure consists of four parts: input end, backbone network, neck network and prediction end. The vehicle recognition model introduces the CA coordinate attention mechanism in the neck network of the YOLOv7 network structure, and combines the efficient aggregation network module of the backbone network with the lightweight network FasterNet into an efficient aggregation lightweight network.

[0062] The speed calculation submodule is used to analyze each frame of the image collected by the video monitoring device during the real-time monitoring process, detect the vehicle in the image according to the vehicle recognition result output by the vehicle recognition model, determine the vehicle position in combination with the millimeter-wave radar data, compare the vehicle positions in adjacent frames through the inter-frame difference algorithm, calculate the movement speed of the vehicle in the image, and then convert the speed in the image into the speed on the actual road according to the installation parameters of the video monitoring device to obtain the vehicle speed, wherein the installation parameters at least include the focal length and the shooting angle;

[0063] The low-speed judgment submodule is used to make judgments based on the vehicle's position, speed and driving status. When the vehicle speed is lower than the preset low-speed threshold, it is judged as a low-speed vehicle. The data processing module encapsulates the vehicle's information into a data frame and sends it to the warning module. The preset low-speed threshold is 60 kilometers per hour.

[0064] In this embodiment, the vehicle identification model includes the following process when performing identification:

[0065] The input end uniformly scales the input original image to a standard size. The backbone network extracts features from the original image passing through the input end through the convolution normalization activation module, extracts target features through the efficient aggregation lightweight network module, expands the receptive field of the feature layer through the maximum pooling operation of the downsampling structure, and connects it in parallel with the convolution normalization activation module to fuse feature information.

[0066] The neck network uses a spatial pyramid pooling network combined with channel amplitude propagation convolution, an efficient aggregation network, a convolution normalization activation module, upsampling operations, and downsampling operations to fuse feature layers of different sizes. In the CA coordinate attention mechanism, the input tensor is split into a tensor in the width direction and a tensor in the height direction. The two tensors are concatenated after average pooling. After passing through two-dimensional convolution, batch normalization, and nonlinear activation functions in turn, the output retains the channel information correlation and fully utilizes the intermediate tensor of position information. The intermediate tensor is split into an intermediate tensor in the height direction and an intermediate tensor in the width direction, and two-dimensional convolution and Sigmoid function activation are performed respectively. Finally, the two outputs are multiplied by the original input tensor to obtain the final output tensor.

[0067] The prediction end trains the feature layer through re-parameterized convolution, and then generates detection boxes for small, medium and large targets through the convolutional normalization activation module to obtain the vehicle recognition result.

[0068] In this embodiment, the vehicle driving state is judged by the following process: calculating the speed change rate of the vehicle at consecutive time points Δv / Δt, and if Δv / Δt is greater than the set acceleration threshold a th1 , the vehicle is judged to be in an accelerating state; if Δv / Δt is less than the set deceleration threshold a th2 , the vehicle is judged to be in a deceleration state; if Δv / Δt is close to zero within a certain range, the vehicle is judged to be in a constant speed state; if the vehicle speed continues to be close to zero for a certain period of time, the vehicle is judged to be stationary, where Δv is the speed difference of the vehicle at two consecutive time points, Δt is the time interval between two consecutive speed measurement moments, and the acceleration threshold a is th1 is 0.8m / s2, a th2 <0.

[0069] In this embodiment, the warning module includes a warning light linkage submodule, a laser light and voice broadcast submodule and a road network terminal submodule, wherein:

[0070] The warning light linkage submodule is used to trigger the linkage alarm mechanism when a low-speed vehicle is detected, so that the warning lights on the roadside work. The warning lights are normally yellow, and the red light is immediately lit when a low-speed event is triggered;

[0071] The laser light and voice broadcast submodule is used to turn on the laser light and flash mode when a low-speed vehicle is detected, and the tweeter makes on-site voice broadcasts;

[0072] The road network terminal submodule is used to display relevant prompt information through variable information signs in the road section when a low-speed vehicle is detected, reminding the driver to pay attention to the driving conditions.

[0073] In this embodiment, the inter-frame difference algorithm can be used to preliminarily detect the presence and position change of the vehicle. By processing the continuous video frames collected by the video monitoring device, the area where the motion change occurs in the picture can be quickly located, that is, the possible position of the vehicle. Since the vehicle moves relative to the background during driving, it will produce obvious gray value changes between adjacent frames, which will appear as a highlight area in the difference image.

[0074] In this embodiment, the front-end radar vision equipment can detect the vehicle's trajectory and speed in real time. Once a low-speed vehicle is detected, the system will immediately link to warn the vehicle behind and issue a prompt to the driver to pay attention to the driving situation. In this way, it can effectively reduce safety incidents such as rear-end collisions caused by low-speed vehicles. The long uphill vehicle low-speed detection device has the advantages of rapid detection, accurate identification, timely alarm and efficient management, with high work efficiency, which can provide strong guarantee for road traffic safety and smoothness.

[0075] In this embodiment, real-time detection: through the front-end sensing detection terminal, the millimeter wave radar speed measurement system uses electromagnetic waves in the millimeter wave band for detection, emits electromagnetic waves and receives reflected signals; millimeter waves have good penetration and anti-interference capabilities, making the speed measurement performance of millimeter wave radar relatively stable in complex environments; when the vehicle is traveling on a long uphill section, the radar can accurately measure the vehicle changes and relative speed; the detection distance is adjustable from 100 to 300 meters; video monitoring equipment, through the camera to take real-time photos of the road, and monitor the real scene in real time; custom settings: users can customize the detection area and the minimum speed limit according to actual needs; accurate alarm: when the vehicle speed is detected to be lower than the set value, the system immediately triggers an alarm prompt; data linkage module: the collected radar data and video data are transmitted to the processing unit for comprehensive analysis; if the speed of the vehicle is lower than the preset threshold (the threshold is usually set according to the road design standards and safety requirements), it is determined to be a low-speed vehicle; users can customize the alarm mode and alarm duration threshold according to actual needs; warning light linkage: when a low-speed vehicle is detected, the system will trigger the linkage The system can also be used to generate an automatic alarm mechanism; the warning light set beside the road can flash to send a visual warning to the rear vehicles; the warning light is in the yellow light state under normal circumstances, and once a low-speed event is triggered, the warning light in the front warning unit will immediately turn red to attract the driver's attention; users can also select the warning light according to their needs (e.g., the light is off under normal circumstances, and the warning light will work only when a low-speed event is triggered); laser light and voice broadcast: according to the on-site situation, you can choose to add a highway-specific laser light and a tweeter to perform on-site voice broadcast, such as "There is a low-speed vehicle ahead, please drive carefully", for a more intuitive warning; road network terminal: according to the needs, add terminal equipment to further process the front-end detection data, and display relevant prompt information through the variable information signs in the road section, such as "There is a low-speed vehicle ahead, please drive carefully", etc., to remind the driver to pay attention to the driving situation; it can be expanded to provide data to map application manufacturers for early warning of road conditions ahead in car and mobile phone navigation; remote monitoring and command: the system supports access to the existing monitoring platform of the operating company and remote monitoring of the site, so that the management department can understand the on-site situation in a timely manner and make corresponding treatment.

[0076] Through the implementation of the above scheme, the current highway network lacks a low-speed detection device for long uphill vehicles. In view of this problem, a low-speed vehicle detection device for long uphill sections of highways is now developed, and supporting methods such as linkage alarm are developed; the long uphill vehicle low-speed detection device can not only detect low-speed vehicles, but also provide valuable data for traffic management. Through the analysis of the detection data, the traffic management department can understand the traffic flow, vehicle speed distribution and other conditions of the long uphill section, so as to formulate a more scientific and reasonable traffic management strategy; the use of this device can reduce traffic accidents caused by low-speed vehicles and improve road traffic efficiency; at the same time, it can also reduce the manpower cost of the traffic management department and improve management efficiency.

[0077] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A low-speed vehicle detection system for long uphill sections of highways, characterized in that: The low-speed vehicle detection system on a long uphill section of a highway includes a sensor module, a data processing module, a warning module and a communication module, wherein: The sensor module is used to install the millimeter-wave radar on the roadside of the long uphill section of the highway, use the millimeter-wave electromagnetic wave to transmit and receive reflected signals to detect vehicles, and use video monitoring equipment to take real-time photos of the road to monitor the real scene of the long uphill section of the highway; A data processing module is used to receive the data sent by the sensor module, and analyze and process the data to determine the speed, position and driving status of the vehicle. When a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module; The warning module is used to send warning information to the driver according to the alarm mode and alarm duration threshold, and timely warn the rear vehicle; The communication module is used to send the detected low-speed vehicle information to the traffic management department so that the traffic management department can grasp the on-site situation in time and take measures to deal with it.

2. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 1, characterized in that: When the millimeter wave radar transmits electromagnetic waves, the transmitting time t1 is recorded. When the electromagnetic waves are reflected by the vehicle, the reflected signals are received and the receiving time t2 is recorded. According to the electromagnetic wave propagation speed c, the distance d between the radar and the vehicle is calculated: d=c×(t2-t1) / 2 The distance is measured multiple times within a short time interval Δt, and the radial velocity v of the vehicle is calculated by the ratio of the distance change Δd to the time interval Δt: v=Δd / Δt The beamforming algorithm and signal processing algorithm are used to analyze the reflected signals in multiple directions. Assume that there are N antenna units for the reflected signal, and the received reflected signal is represented by x n (t)(n=1,2,…,N), the signal after beamforming is y(t): Among them, w n is the weighting coefficient of the nth antenna unit, and the scanning and focusing of the beam in different directions are determined by adjusting the weighting coefficient; The beamformed signal is used for target detection, and the distance between the target and the radar is calculated based on the propagation time of the electromagnetic wave. The position information of the vehicle in the two-dimensional plane is determined by target detection and distance measurement in multiple beam directions.

3. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 1, characterized in that: The data processing module includes a denoising processing submodule, a model recognition submodule, a speed calculation submodule and a low-speed judgment submodule, wherein: The denoising processing submodule is used to obtain the image collected by the video monitoring device and perform denoising processing on the image using a Gaussian filtering algorithm to obtain a denoised image; A model recognition submodule is used to input the denoised image into a vehicle recognition model for recognition to obtain a vehicle recognition result, wherein the vehicle recognition model is constructed based on a YOLOv7 network structure, and the YOLOv7 network structure consists of four parts: an input end, a backbone network, a neck network, and a prediction end. The vehicle recognition model introduces a CA coordinate attention mechanism into the neck network of the YOLOv7 network structure, and combines the efficient aggregation network module of the backbone network with the lightweight network FasterNet into an efficient aggregation lightweight network; A speed calculation submodule is used to analyze each frame of the image collected by the video monitoring device during real-time monitoring, detect the vehicle in the image according to the vehicle recognition result output by the vehicle recognition model, determine the vehicle position in combination with the millimeter-wave radar data, compare the vehicle positions in adjacent frames through an inter-frame difference algorithm, calculate the movement speed of the vehicle in the image, and then convert the speed in the image into the speed on the actual road according to the installation parameters of the video monitoring device to obtain the vehicle speed, wherein the installation parameters at least include focal length and shooting angle; The low-speed judgment submodule is used to make judgments based on the vehicle's position, speed and driving status. When the vehicle speed is lower than the preset low-speed threshold, it is judged as a low-speed vehicle. The data processing module encapsulates the vehicle's information into a data frame and sends it to the warning module. The preset low-speed threshold is 60 kilometers per hour.

4. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 3, characterized in that: The vehicle identification model includes the following process when performing identification: The input end uniformly scales the input original image to a standard size. The backbone network extracts features from the original image passing through the input end through the convolution normalization activation module, extracts target features through the efficient aggregation lightweight network module, expands the receptive field of the feature layer through the maximum pooling operation of the downsampling structure, and connects it in parallel with the convolution normalization activation module to fuse feature information. The neck network uses a spatial pyramid pooling network combined with channel amplitude propagation convolution, an efficient aggregation network, a convolution normalization activation module, upsampling operations, and downsampling operations to fuse feature layers of different sizes. In the CA coordinate attention mechanism, the input tensor is split into a tensor in the width direction and a tensor in the height direction. The two tensors are concatenated after average pooling. After passing through two-dimensional convolution, batch normalization, and nonlinear activation functions in turn, the output retains the channel information correlation and fully utilizes the intermediate tensor of position information. The intermediate tensor is split into an intermediate tensor in the height direction and an intermediate tensor in the width direction, and two-dimensional convolution and Sigmoid function activation are performed respectively. Finally, the two outputs are multiplied by the original input tensor to obtain the final output tensor. The prediction end trains the feature layer through re-parameterized convolution, and then generates detection boxes for small, medium and large targets through the convolutional normalization activation module to obtain the vehicle recognition result.

5. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 3, characterized in that: The vehicle's driving state is judged by the following process: Calculate the speed change rate of the vehicle at consecutive time points Δv / Δt, if Δv / Δt is greater than the set acceleration threshold a th1 , the vehicle is judged to be in an accelerating state; if Δv / Δt is less than the set deceleration threshold a th2 , the vehicle is judged to be in a deceleration state; if Δv / Δt is close to zero within a certain range, the vehicle is judged to be in a constant speed state; if the vehicle speed continues to be close to zero for a certain period of time, the vehicle is judged to be stationary, where Δv is the speed difference of the vehicle at two consecutive time points, Δt is the time interval between two consecutive speed measurement moments, and the acceleration threshold a is th1 is 0.8m / s2, a th2 <0.

6. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 1, characterized in that: The warning module includes a warning light linkage submodule, a laser light and voice broadcast submodule and a road network terminal submodule, wherein: The warning light linkage submodule is used to trigger the linkage alarm mechanism when a low-speed vehicle is detected, so that the warning lights on the roadside work. The warning lights are normally yellow, and the red light is immediately lit when a low-speed event is triggered; The laser light and voice broadcast submodule is used to turn on the laser light and flash mode when a low-speed vehicle is detected, and the tweeter makes on-site voice broadcasts; The road network terminal submodule is used to display relevant prompt information through variable information signs in the road section when a low-speed vehicle is detected, reminding the driver to pay attention to the driving conditions.

7. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 1, characterized in that: The method for implementing a low-speed vehicle detection system on a long uphill section of a highway comprises the following steps: Millimeter-wave radars are installed on the roadside of long uphill sections of highways. Vehicles are detected by transmitting and receiving reflected signals of electromagnetic waves in the millimeter-wave band. The road is photographed in real time through video monitoring equipment to monitor the actual scene of the long uphill sections of highways. Receive the data sent by the sensor module, analyze and process the data to determine the speed, position and driving status of the vehicle. When a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module; Send warning information to the driver according to the alarm mode and alarm duration threshold, and warn the vehicles behind in time; The detected low-speed vehicle information is sent to the traffic management department so that the traffic management department can grasp the on-site situation in time and take measures to deal with it.

8. The low-speed vehicle detection system for a long uphill section of a highway as claimed in claim 7, characterized in that: The receiving of data sent by the sensor module, and analyzing and processing the data to determine the speed, position and driving status of the vehicle, when a low-speed vehicle is detected, the data processing module will send the vehicle information to the warning module, including: Acquire the image collected by the video monitoring device, and use the Gaussian filter algorithm to denoise the image to obtain a denoised image; Input the denoised image into a vehicle recognition model for recognition, and obtain a vehicle recognition result, wherein the vehicle recognition model is constructed based on a YOLOv7 network structure, and the YOLOv7 network structure consists of four parts: an input end, a backbone network, a neck network, and a prediction end. The vehicle recognition model introduces a CA coordinate attention mechanism into the neck network of the YOLOv7 network structure, and combines the efficient aggregation network module of the backbone network with the lightweight network FasterNet into an efficient aggregation lightweight network; In the real-time monitoring process, each frame of the image collected by the video monitoring device is analyzed, the vehicle in the image is detected according to the vehicle recognition result output by the vehicle recognition model, and the vehicle position is determined in combination with the millimeter wave radar data, the vehicle positions in adjacent frames are compared by an inter-frame difference algorithm, the movement speed of the vehicle in the image is calculated, and then the speed in the image is converted into the speed on the actual road according to the installation parameters of the video monitoring device to obtain the vehicle speed, wherein the installation parameters at least include the focal length and the shooting angle; The vehicle position, speed and driving status are combined for judgment. When the vehicle speed is lower than the preset low-speed threshold, it is determined to be a low-speed vehicle. The data processing module encapsulates the vehicle information into a data frame and sends it to the warning module. The preset low-speed threshold is 60 kilometers per hour.

Images