CN110228413B - Safety warning system to avoid pedestrians getting caught under the car when a large vehicle turns - Google Patents

Abstract

The invention relates to a safety early warning system for preventing pedestrians from being involved under a large vehicle during turning, which comprises a distance measuring radar sensor, a camera, a corner sensor, an information integration unit, a road dangerous area dividing unit and an alarm device. The distance measuring radar sensor collects distance information of objects around the vehicle body, the camera collects images of two sides of the vehicle, and the images are identified by using a convolutional neural network algorithm. The turning angle sensor collects vehicle turning angle information, and a K-means clustering analysis algorithm is adopted to divide the vehicle turning angle information into a dangerous area, an early warning area or a safe area. The information integration unit and the road dangerous area division unit perform information fusion, processing and analysis on the acquired data information, judge whether pedestrians exist in the dangerous area, and output the judgment result to the alarm device for early warning. The invention detects the environment of the large vehicle, judges whether pedestrians exist in the dangerous area, and transmits the judgment result to the alarm device for early warning, thereby improving the safety of road traffic.

Description

Safety warning system to avoid pedestrians getting caught under the car when a large vehicle turns

technical field

The invention belongs to the technical field of road traffic safety, and relates to a pedestrian safety early warning device, in particular to a safety early warning system for preventing pedestrians from being involved under the vehicle when a large vehicle turns.

Background technique

At present, accidents involving pedestrians on both sides of the vehicle frequently occur due to the large difference between the inner wheels and the driver's blind spot when turning a large vehicle. During the turning process, there are pedestrians or cyclists beside the vehicle. If the driver or pedestrian is not careful, the pedestrian or cyclist will easily be involved in the vehicle. May be life-threatening.

The invention patent application with publication number CN 108791166 A discloses "a semi-trailer capable of protecting pedestrians on a turning road, including a trailer plate and a rectangular block, including a trailer plate and a rectangular block, and the left and right sides of the trailer plate are provided with storage slots, The left and right inner walls of the receiving groove are provided with moving grooves, the left and right ends of the rectangular block are fixedly connected with moving plates, the moving plates are slidably connected with the moving grooves, and the rear end of the rectangular block is connected with the rear sidewall of the receiving groove through a locking mechanism. The lower end is provided with a first storage slot, and the inside of the first storage slot is connected with the protective mechanism." This patent application isolates pedestrians in the safe area of the semi-trailer by setting protective mechanisms on both sides of the semi-trailer to avoid the risk of being caught under the vehicle. However, the patent application does not give double warning to pedestrians and drivers, and it is not suitable for accidents caused by large vehicle turning inner wheel difference and driver's visual blind spot.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a safety warning system that avoids pedestrians from being involved in the vehicle when a large vehicle turns, detects the environment on the left and right sides of the vehicle when turning, identifies the surrounding dangerous environment, and gives safety warnings to the driver and pedestrians on both sides. Ensure the safe turning of the vehicle and the safety of pedestrians around the vehicle, and improve the safety of road traffic.

The technical scheme of the present invention is: a safety early warning system for avoiding pedestrians from being involved in the vehicle when a large vehicle turns, including a ranging radar sensor, a camera, a turning angle sensor, an information integration unit, a road danger zone division unit and an alarm device. The ranging radar sensor collects the distance information of objects around the vehicle body, and the collected target distance information is input into the information integration unit for analysis and processing. The camera collects the images on both sides of the vehicle, uses the convolutional neural network algorithm to identify the image, determines whether the target identified by the ranging radar sensor is a pedestrian, and obtains the position data information of the pedestrian in the identified image, and transmits it to the information integration unit. The above-mentioned corner sensor collects vehicle steering angle information, and the road dangerous area division unit uses K-means cluster analysis algorithm to divide the areas on both sides of the large vehicle into dangerous areas, early warning areas or safety areas according to the information collected by the corner sensor and vehicle performance parameters. . The information integration unit and the road dangerous area division unit perform information fusion, processing and analysis on the collected data information, determine whether there are pedestrians in the dangerous area, and output the judgment result to the alarm device. If there is danger in the judgment result, the alarm device will alarm to remind the driver. If there is no danger, the alarm device will not act.

The ranging radar sensor is installed on the vehicle body within the blind area of the vehicle. The ranging radar sensor transmits radio waves through the areas on both sides of the vehicle, and uses the time delay between the received target reflected signal and the transmitted signal to measure the position information of the front target. The camera is installed above the vehicle body in the blind spot area of the vehicle. The angle sensor is installed in the steering column under the steering wheel. The information integration unit includes a data acquisition module, a data processing module, an early warning information analysis module and an early warning information output module. The data acquisition module collects the output information of the ranging radar sensor, the angle sensor and the camera. The information includes the distance between the surrounding target and the vehicle body in the dangerous area, the steering angle of the vehicle and the surrounding environment information. The data processing module matches the position information of the surrounding targets collected by the ranging radar sensor with the target position information collected by the camera, and imports the steering angle information into the road area division system for area division. The early warning information analysis module analyzes the judgment result of the dangerous area where the target around the vehicle is located and the judgment result of whether the target is a pedestrian, and judges the response level of the alarm device in the current state. The early warning information output module outputs the response level result to the alarm device. Vehicle performance parameters include tire section width, wheel outer width, wheelbase, front distance and front outer wheel turning angle.

The K-means clustering analysis algorithm is used to classify the areas on both sides of the vehicle into three parts: danger area, early warning area and safety area. Three offline cluster centroids are obtained through offline clustering. The K-means cluster analysis process is as follows:

1) Determine the number of clusters K=3 as needed;

2) Initialize the cluster centroids, and take points at equal intervals to determine the initial cluster centroids;

3) Set the maximum number of iteration steps and centroid offset: △d=0.0001, J=1000;

Among them: △d is the center of mass offset; J is the maximum number of iteration steps;

4) Calculate the Euclidean distance between each object and the cluster centroid, and classify these objects into the most similar clusters according to the value of the Euclidean distance;

5) According to the clustering results, recalculate the respective centroids of the 3 clusters, and the calculation method is the respective arithmetic squares of each cluster;

6) Repeat steps 3 and 4 until the clustering result does not change;

7) Output the result.

The formula for calculating Euclidean distance is:

Assuming that _{two n} -dimensional _vectors are _A =( _{a 1} , a ₂ , a ₃ . The distance is:

In the formula: ρ(A, B) is the Euclidean distance;

a ₁ , a ₂ , a ₃ ...... a _n is the n dimensions of A;

The above method is used to determine the range of each grade area of the road environment around the vehicle.

The process that the camera uses the convolutional neural network (CNN) algorithm to process the collected images is composed of six layers:

The first layer is the input layer, the images on both sides of the car collected by the input camera;

The second layer includes two independent convolution layers C1 and C2, and the number of convolution kernels is 40; the convolution kernel size of the convolution layer C1 is 3 × 3, and the size is 12 × 12 × 40; the convolution layer C2 volume The size of the product kernel is 7×7 and the size is 10×10×40; the convolution layer extracts features from the input image through the convolution operation. The convolution kernel is equivalent to a filter. The convolution formula of the first layer is:

In the formula: l is the number of layers, x _m ^l-1 is the input of the l-1th hidden layer, n, m are the two-dimensional matrix values, _{wn, m} ^l is the mapping weight matrix of the lth hidden layer , b _i ^l is the bias matrix of the lth hidden layer, y _n ^l is the input image of the input layer, f is the activation function, the ReLU function is used for nonlinear mapping, and the expression is:

y=max(0,x)

In the formula: y is the output, x is the input;

The next layer of each convolutional layer is a pooling layer, which aggregates the features of different positions in the image to reduce the feature dimension and extract the image features twice. The image is divided into non-overlapping matrices, and each subregion outputs the maximum value;

The third layer contains two independent convolution layers C3 and C4, and the number of convolution kernels is 65; the convolution kernel size of the convolution layer C1 is 3×3 and the size is 5×5×65; The size of the product kernel is 3×3 and the size is 4×4×65. The step size of the four convolutional layers is 2; the convolution formula, mapping method and pooling formula are the same as the second layer;

The fourth layer is the fully connected layer F1, which is designed with 300 neurons. The input of F1 comes from C1, C2, C3, and C4; the fully connected layer can enhance the nonlinear mapping capability of the network and limit the size of the network; the F1 fully connected layer The calculation formula is:

where l is the number of layers, x _i ^l-1 is the input of the l-1th hidden layer, i, j are the two-dimensional matrix values, w _{j, i} ^l is the mapping weight matrix of the lth hidden layer, b _j ^l is the bias matrix of the lth hidden layer;

The fifth layer is the fully connected layer F2, and the calculation formula is the same as that of F1; F1 and F2 are fully connected;

The sixth layer is the output layer, and the classification is judged by the softmax function. The calculation formula of the softmax function is:

X _i = _wi x+b

In the formula: θ is the parameter vector, i and j are the two-dimensional matrix values, and T is the temperature parameter. When T is large, that is, it tends to positive infinity, the activation probabilities corresponding to all activation values tend to be the same; When it is low, that is, when it tends to 0, the activation probability corresponding to different activation values is very different;

Complete the process of identifying pedestrians with a convolutional neural network.

The alarm device includes an in-vehicle alarm device and an out-of-vehicle alarm device. The in-vehicle alarm device is composed of a warning light and an in-vehicle alarm, and is used to remind the driver to pay attention to the danger outside the vehicle. The exterior alarm device is an exterior alarm, which is used to remind pedestrians outside the vehicle to pay attention to avoid vehicles. According to the level of the danger area where pedestrians are located, the alarm device sends out different levels of vehicle alarm signals, different levels of alarm signals, and the brightness of the alarm lights and the sound intensity of the alarm are different.

The safety early warning system of the present invention, which avoids pedestrians from being involved in the vehicle when a large vehicle turns, detects the environment on the left and right sides of the vehicle when turning through the ranging radar sensor, the camera and the angle sensor, identifies the surrounding dangerous environment, and inputs the information to the information integration unit and the road. In the dangerous area division system, information fusion, processing and analysis are carried out to determine whether there are pedestrians in the dangerous area, and the judgment result is input to the alarm device for early warning, which provides safety warnings to the driver and pedestrians on both sides, ensuring the safe turning of the vehicle and the surrounding area of the vehicle. Pedestrian safety improves road traffic safety. The invention effectively predicts in real time whether the surrounding environment is dangerous when the large vehicle turns, and provides a new idea for solving the safety problem of the large vehicle turning.

Description of drawings

1 is a schematic diagram of a safety alarm device for preventing pedestrians from being involved in the vehicle when a large vehicle turns;

Figure 2 is a schematic diagram of the alarm level;

Among them: 1—angle sensor, 2—ranging radar sensor, 3—camera, 4—road dangerous area division unit, 5—integrated control unit, 6—alarm device, 7—data acquisition module, 8—data processing module, 9 - Early warning information analysis module, 10 - early warning information output module, 11 - car alarm, 12 - warning light, 13 - car alarm.

Detailed ways

The present invention will be described in detail below with reference to the embodiments and accompanying drawings. The protection scope of the present invention is not limited to the embodiments, and any changes made by those skilled in the art within the scope defined by the claims also belong to the protection scope of the present invention.

The safety warning system of the present invention for preventing pedestrians from getting involved in the vehicle when a large vehicle turns is shown in FIG. 1 , including a ranging radar sensor 2, a camera 3, a turning angle sensor 1, an information integration unit 5, a road dangerous area division unit 4 and an alarm device 6. The information integration unit 5 includes a data acquisition module 7 , a data processing module 8 , an early warning information analysis module 9 and an early warning information output module 10 . The alarm device 6 includes an in-vehicle alarm device and an out-of-vehicle alarm device. The in-vehicle alarm device is composed of a warning light 12 and an in-vehicle alarm 11, and is used to remind the driver to pay attention to the danger outside the vehicle. The exterior alarm device is an exterior alarm 13, which is used to remind pedestrians outside the vehicle to pay attention to avoiding the vehicle. The ranging radar sensor and the camera are communicated with the data acquisition module, and the rotation angle sensor is communicated with the data acquisition module through the road dangerous area division unit. The road hazard area division unit is provided with an input end of vehicle performance parameters, and the vehicle performance parameters include tire section width, wheel outer width, wheelbase, front distance and front outer wheel turning angle. The data acquisition module communicates with the data processing module, the early warning information analysis module and the early warning information output module in sequence, and the early warning information output module communicates with the alarm device. The ranging radar sensor collects the distance information of objects around the vehicle body, and the collected target distance information is input into the information integration unit for analysis and processing. The camera collects the images on both sides of the vehicle, uses the convolutional neural network algorithm to identify the image, determines whether the target identified by the ranging radar sensor is a pedestrian, and obtains the position data information of the pedestrian in the identified image, and transmits it to the information integration unit. The corner sensor collects the vehicle steering angle information, and the road dangerous area division unit uses the K-means clustering analysis algorithm to divide the areas on both sides of the large vehicle into dangerous areas, early warning areas or safety areas according to the information collected by the corner sensors and vehicle performance parameters. The information integration unit and the road dangerous area division unit perform information fusion, processing and analysis on the collected data information, determine whether there are pedestrians in the dangerous area, and output the judgment result to the alarm device. If there is danger in the judgment result, the alarm device will alarm to remind the driver. If there is no danger, the alarm device will not act.

The ranging radar sensor 2 is installed on the vehicle body within the blind area of the vehicle. The ranging radar sensor transmits radio waves through the areas on both sides of the vehicle, and uses the time delay between the received target reflected signal and the transmitted signal to measure the position information of the front target. The camera 3 is installed above the vehicle body in the blind area of the vehicle, and the corner sensor 1 is installed in the steering column below the steering wheel.

The data acquisition module 7 collects the output information of the ranging radar sensor, the angle sensor and the camera, and the information includes the distance between the surrounding target and the vehicle body in the dangerous area, the steering angle of the vehicle and the surrounding environment information. The data processing module 8 matches the surrounding target position information collected by the ranging radar sensor with the target position information collected by the camera, and imports the steering angle information into the road area division system for area division. The early warning information analysis module 9 analyzes the judgment result of the dangerous area where the target around the vehicle is located and the judgment result of whether the target is a pedestrian, and judges the response level of the alarm device in the current state. The early warning information output module 10 outputs the response level result to the alarm device 6 .

The purpose of offline clustering is to classify the areas on both sides of the vehicle, and obtain offline cluster centroids through offline clustering. The information collected by the corner sensor and the vehicle performance parameters are input into the road dangerous area division unit 4 using the K-means clustering analysis algorithm, and the areas on both sides of the large vehicle are divided into dangerous areas, early warning areas and safety areas. part, three offline cluster centroids are obtained by offline clustering. The K-means cluster analysis process is as follows:

1) Determine the number of clusters K=3 as needed;

2) Initialize the cluster centroids, and take points at equal intervals to determine the initial cluster centroids;

3) Set the maximum number of iteration steps and centroid offset: △d=0.0001, J=1000;

Among them: △d is the center of mass offset; J is the maximum number of iteration steps;

4) Calculate the Euclidean distance between each object and the cluster centroid, and classify these objects into the most similar clusters according to the value of the Euclidean distance;

5) According to the clustering results, recalculate the respective centroids of the 3 clusters, and the calculation method is the respective arithmetic squares of each cluster;

6) Repeat steps 3 and 4 until the clustering result does not change;

7) Output the result.

The formula for calculating Euclidean distance is:

Assuming that _{two n} -dimensional _vectors are _A =( _{a 1} , a ₂ , a ₃ . The distance is:

where ρ(A, B) is the Euclidean distance;

a ₁ , a ₂ , a ₃ ...... a _n is the n dimensions of A;

The above method is used to determine the range of each grade area of the road environment around the vehicle.

The process that the camera uses the convolutional neural network (CNN) algorithm to process the collected images is composed of six layers:

The first layer is the input layer, the images on both sides of the car collected by the input camera;

The second layer includes two independent convolution layers C1 and C2, and the number of convolution kernels is 40; the convolution kernel size of the convolution layer C1 is 3 × 3, and the size is 12 × 12 × 40; the convolution layer C2 volume The size of the product kernel is 7×7 and the size is 10×10×40; the convolution layer extracts features from the input image through the convolution operation. The convolution kernel is equivalent to a filter. The convolution formula of the first layer is:

In the formula: l is the number of layers, x _m ^l-1 is the input of the l-1th hidden layer, n, m are the two-dimensional matrix values, _{wn, m} ^l is the mapping weight matrix of the lth hidden layer, b _i ^l is the bias matrix of the lth hidden layer, y _n ^l is the input image of the input layer, f is the activation function, the ReLU function is used for nonlinear mapping, and the expression is:

y=max(0,x)

In the formula: y is the output, x is the input;

The next layer of each convolutional layer is a pooling layer, which aggregates the features of different positions in the image to reduce the feature dimension and extract the image features twice. The image is divided into non-overlapping matrices, and each subregion outputs the maximum value;

The third layer contains two independent convolution layers C3 and C4, and the number of convolution kernels is 65; the convolution kernel size of the convolution layer C1 is 3×3 and the size is 5×5×65; The size of the product kernel is 3×3 and the size is 4×4×65. The step size of the four convolutional layers is 2; the convolution formula, mapping method and pooling formula are the same as the second layer;

The fourth layer is the fully connected layer F1, which is designed with 300 neurons. The input of F1 comes from C1, C2, C3, and C4; the fully connected layer can enhance the nonlinear mapping capability of the network and limit the size of the network; the F1 fully connected layer The calculation formula is:

where l is the number of layers, x _i ^l-1 is the input of the l-1th hidden layer, i, j are the two-dimensional matrix values, w _{j, i} ^l is the mapping weight matrix of the lth hidden layer, b _j ^l is the bias matrix of the lth hidden layer;

The fifth layer is the fully connected layer F2, and the calculation formula is the same as that of F1; F1 and F2 are fully connected;

The sixth layer is the output layer, and the classification is judged by the softmax function. The calculation formula of the softmax function is:

X _i = _wi x+b

In the formula: θ is the parameter vector, i and j are the two-dimensional matrix values, and T is the temperature parameter. When T is large, that is, it tends to positive infinity, the activation probabilities corresponding to all activation values tend to be the same; When it is low, that is, when it tends to 0, the activation probability corresponding to different activation values is very different;

Complete the process of identifying pedestrians with a convolutional neural network.

The working process of the safety warning system for avoiding pedestrians from being involved under the vehicle when the large vehicle turns in the present invention, the steps are as follows:

(1) First, the steering angle sensor 1 collects the vehicle steering angle information, and inputs the vehicle steering angle information into the road danger area division unit 4, and divides the areas on both sides of the vehicle into three parts: danger area, early warning area and safety area;

(2) The ranging radar sensor 2 transmits radio waves to the front area, uses the time delay between the received target reflected signal and the transmitted signal to measure the position information of the front target, inputs the target distance information into the information integration unit 5, and determines that the current target is in the in which area;

(3) The camera 3 installed on the vehicle body collects image information around the large vehicle, and uses the convolutional neural network algorithm to structurally process the image information to identify whether the target of the ranging radar is a pedestrian;

(4) The information integration unit 5 and the road dangerous area division unit 4 perform information fusion, processing and analysis on the collected data information, and determine whether there are pedestrians in the dangerous area;

⑸ Input the judgment result to the early warning information processing module 10 for analysis, and judge whether the target is a pedestrian; if no pedestrian is identified in the judgment result, input the alarm level information to the alarm device according to the target area; The upper-level alarm level information alarm in the area where the pedestrian is located;

⑹ Send the alarm information to the alarm device 6. If there is danger, the alarm device will alarm to remind the driver. If there is no danger, the alarm device will not act.

The alarm device of the present invention adopts graded alarm. As shown in FIG. 2 , when the ranging radar and the camera determine the current target and the area where the target is located, the graded alarm device responds. There are four levels of early warning devices, which are first-level alarm, second-level alarm, third-level alarm and no response. If the target is a pedestrian, when the pedestrian is in the dangerous area, the first-level alarm will be used; if the pedestrian is in the early warning area, the second-level alarm will be used; if the pedestrian is in the safe area, the alarm device will not respond. If the target is not a pedestrian, the alarm level is one level lower than that of pedestrians. If the target is in the dangerous area, the second-level alarm will be adopted; if the target is in the early warning area, the third-level alarm will be adopted. If the target is in the safe area, the alarm device will not respond.

Claims (6)

1. a safety warning system for avoiding pedestrians getting involved under the vehicle when a large vehicle turns, it is characterized in that: described system comprises ranging radar sensor (2), camera (3), angle sensor (1), information integration unit ( 5), a road dangerous area division unit (4) and an alarm device (6); the ranging radar sensor collects the distance information of objects around the vehicle body, and the collected target distance information is input into the information integration unit for analysis and processing; the The camera collects the images on both sides of the vehicle, uses the convolutional neural network algorithm to identify the image, determines whether the target identified by the ranging radar sensor is a pedestrian, and obtains the position data information of the pedestrian in the identified image, and transmits it to the information integration unit; The turning angle sensor collects vehicle steering angle information, and the road dangerous area dividing unit adopts the K-means cluster analysis algorithm to divide the areas on both sides of the large vehicle into dangerous areas, early warning areas or safe area; the information integration unit and the road danger area division unit perform information fusion, processing and analysis on the collected data information, determine whether there are pedestrians in the danger area, and output the judgment result to the alarm device; if the judgment result is dangerous, The alarm device alarms and reminds the driver that if there is no danger, the alarm device does not act; the K-means cluster analysis algorithm is used to classify the areas on both sides of the vehicle, which are divided into three parts: danger area, early warning area and safety area. Clustering obtains three offline cluster centroids; the process of the K-means cluster analysis algorithm is as follows: 1) Determine the number of clusters K=3 as needed; 2) Initialize the cluster centroids, and take points at equal intervals to determine the initial cluster centroids; 3) Set the maximum number of iteration steps and centroid offset: △d=0.0001, J=1000; Among them: △d is the center of mass offset; J is the maximum number of iteration steps; 4) Calculate the Euclidean distance between each object and the cluster centroid, and classify these objects into the most similar clusters according to the value of the Euclidean distance; 5) According to the clustering results, recalculate the respective centroids of the 3 clusters, and the calculation method is the respective arithmetic squares of each cluster; 6) Repeat steps 3 and 4 until the clustering result does not change; 7) Output the result. 2. The safety warning system for avoiding pedestrians getting involved under the vehicle when turning a large vehicle according to claim 1, is characterized in that: the ranging radar sensor (2) is installed on the vehicle body within the blind area of the vehicle, and measures The distance radar sensor transmits radio waves through the areas on both sides of the vehicle, and uses the time delay between the received target reflected signal and the transmitted signal to measure the position information of the front target; the camera (3) is installed above the vehicle body in the blind area of the vehicle; the The angle sensor (1) is installed in the steering column under the steering wheel. 3. The safety warning system for avoiding pedestrians getting involved under the vehicle when turning a large vehicle according to claim 1, is characterized in that: the information integration unit (5) comprises a data acquisition module (7), a data processing module (8) , an early warning information analysis module (9) and an early warning information output module (10); the data acquisition module collects the output information of the ranging radar sensor, the angle sensor and the camera, and the information includes the distance between the surrounding target and the vehicle body in the dangerous area , vehicle steering angle and surrounding environment information; the data processing module matches the surrounding target position information collected by the ranging radar sensor with the target position information collected by the camera, and imports the steering angle information into the road area division system for area division; The early warning information analysis module analyzes the judgment result of the dangerous area where the target around the vehicle is located and the judgment result of whether the target is a pedestrian, and judges the response level of the alarm device in the current state; the early warning information output module outputs the response level result to the alarm device. . 4. The safety warning system for preventing pedestrians from getting involved in the vehicle when a large vehicle turns according to claim 1, wherein the vehicle performance parameters include tire section width, wheel outer width, wheelbase, front distance and front Outer wheel corner. 5. the safety warning system that avoids pedestrians getting involved under the vehicle when the large vehicle turns according to claim 1, is characterized in that: described Euclidean distance calculation formula is: Assuming that _{two n} -dimensional _vectors are _A =( _{a 1} , a ₂ , a ₃ . The distance is:

In the formula: ρ(A, B) is the Euclidean distance; a ₁ , a ₂ , a ₃ ...... a _n is the n dimensions of A; The above method is used to determine the range of each grade area of the road environment around the vehicle. 6. The safety warning system for preventing pedestrians from getting involved under the vehicle when turning a large vehicle according to claim 1, wherein the alarm device (6) comprises an in-vehicle alarm device and an out-of-vehicle alarm device. The alarm device is composed of a warning light (12) and an in-vehicle alarm (11), and is used to remind the driver to pay attention to the danger outside the vehicle; the outer-vehicle alarm device is an out-of-vehicle alarm (13), which is used to remind pedestrians outside the vehicle to pay attention Avoid vehicles; according to the level of the dangerous area where pedestrians are located, the alarm device sends out different levels of vehicle alarm signals, different levels of alarm signals, and the brightness of the alarm lights and the sound intensity of the alarm are different.

Images