Tunnel monitoring method based on machine vision

Abstract

The invention relates to the technical field of tunnel monitoring, and particularly discloses a tunnel monitoring method based on machine vision, which comprises the following steps: S1, dividing sub-periods, acquiring a monitoring image at the starting point of the sub-periods, identifying a target vehicle in the monitoring image, and determining a minimum enclosing rectangle in the monitoring image; calculating a vehicle type score based on the minimum enclosing rectangle, calculating an exhaust emission index, determining a total emission amount in the sub-time period, and if the total emission amount is smaller than a total emission amount threshold value, obtaining a new total emission amount in the next sub-time period; if the total discharge amount is greater than or equal to the total discharge amount threshold value, taking the starting point of the corresponding sub-time period as an adjustment time point; s3, calculating a theoretical distance, determining a target position, and determining a target line; and when the sub-tunnel belongs to any tunnel part, adding 1 to the selection number of the sub-tunnel, determining the total selection number of the single sub-tunnel, and controlling the ventilation system based on the total selection number. The energy consumption of the ventilation system in the tunnel can be reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision, and particularly relates to a tunnel monitoring method based on machine vision. Background Art

[0002] Machine vision is a technology that uses computers, image sensors, and algorithms to simulate the human visual system. By preprocessing the acquired images (such as filtering, enhancement, segmentation, etc.), key features are extracted, and then identification is performed through algorithms (including traditional image processing algorithms and deep learning-based algorithms), and it is widely used in fields such as security monitoring.

[0003] The exhaust gas emitted during vehicle driving contains various harmful substances such as carbon monoxide, nitrogen oxides, and hydrocarbons. Its high-concentration emission not only seriously affects air quality but also poses a threat to human health and is prone to causing respiratory diseases. As a relatively enclosed environment, tunnels are more likely to accumulate exhaust gas concentration, further exacerbating the environmental pollution risk. Although ventilation systems are installed in tunnels, long-term operation of the ventilation systems will result in high energy consumption and additional operating costs. Summary of the Invention

[0004] The purpose of the present invention is to provide a tunnel monitoring method based on machine vision to solve the following technical problems:

[0005] The exhaust gas emitted during vehicle driving contains various harmful substances such as carbon monoxide, nitrogen oxides, and hydrocarbons. Its high-concentration emission not only seriously affects air quality but also poses a threat to human health and is prone to causing respiratory diseases. As a relatively enclosed environment, tunnels are more likely to accumulate exhaust gas concentration, further exacerbating the environmental pollution risk. Although ventilation systems are installed in tunnels, long-term operation of the ventilation systems will result in high energy consumption and additional operating costs.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A tunnel monitoring method based on machine vision includes the following steps:

[0008] S1: Divide a preset monitoring period into n sub-periods with a length of t, where n is a preset first quantity, and t represents the length of a single sub-period. At the starting point of the sub-period, collect monitoring images in the tunnel, identify the vehicles in the monitoring images, denoted as target vehicles, mark the minimum bounding rectangle of the target vehicles in the monitoring images, and obtain the length c and width k of the minimum bounding rectangle.

[0009] S2: Obtain the width d1 of the lane in the monitored image, and obtain the actual width d2 of the lane; calculate the actual length C of the target vehicle as C = c * (d2 / d1), and calculate the actual width K of the target vehicle as K = k * (d2 / d1). Then, the vehicle type score P = η1 * K * D, where η1 is a preset first correction coefficient. Obtain the average speed V of the target vehicle, and calculate the exhaust emission index Y = η2 * V * P, where η2 is a preset second correction coefficient;

[0010] Calculate the emission f of a single target vehicle within a sub - time period as f = η3 * Y * t, where η3 is a preset third correction coefficient, calculate the total emission F, set a total emission threshold Fys. If the total emission F < Fys, then obtain a new total emission in the next sub - time period; if the total emission F ≥ Fys, then use the starting point of the corresponding sub - time period as the adjustment time point;

[0011] S3: Calculate the theoretical distance L = V * t, determine the position on the lane where the target vehicle is located and the position with a distance of the theoretical distance L from the target vehicle's position. Take the position in front of the target vehicle as the target position, and connect the position where the target vehicle is located and the target position to obtain the target line;

[0012] Divide the tunnel into m sub - tunnels with the same length, where m is a preset second quantity. Obtain the tunnel section where the target line is located. When a sub - tunnel belongs to any of the tunnel sections, increment the selection count of the sub - tunnel by 1. Determine the total selection count of a single sub - tunnel. When the total selection count is greater than or equal to a preset total selection count threshold, then turn on the ventilation system in the sub - tunnel.

[0013] As a further solution of the present invention: In step S3, there is one ventilation system in the sub - tunnel.

[0014] As a further solution of the present invention: In step S3, after turning on the ventilation system, the ventilation power G of the ventilation system is G = ((N - Nys) / Nys) * Gys, where Gys represents the preset standard ventilation power, and N and Nys represent the total selection count and the total selection count threshold respectively.

[0015] As a further solution of the present invention: In step S3, the following steps are further included:

[0016] When the number of sub - tunnels with the ventilation system turned on is less than 0.3m, except for the total selection counts corresponding to the sub - tunnels with the ventilation system already turned on, sort the remaining total selection counts in descending order. Starting from the sub - tunnel corresponding to the first total selection count in the sorting, turn on the ventilation system in sequence, and keep the ventilation power of the newly turned - on ventilation system as Gys until the number of sub - tunnels with the ventilation system turned on is greater than or equal to 0.3m.

[0017] As a further solution of the present invention: in the step S3, the following steps are further included:

[0018] When the ventilation power G < Gmin, let the ventilation power G = Gmin;

[0019] When the ventilation power G > Gmax, let the ventilation power G = Gmax, and perform the following steps:

[0020] Except for the total number of selections corresponding to the sub-tunnels with the ventilation system already turned on, sort the remaining total number of selections in descending order. Starting from the sub-tunnel corresponding to the first total number of selections in the sorting, turn on the ventilation system in sequence, and keep the ventilation power of the newly turned-on ventilation system as Gys until the total ventilation power of the newly turned-on ventilation system is greater than 3(G - Gmax).

[0021] As a further solution of the present invention: in the step S1, vehicles are identified from the monitoring images based on a pre-trained vehicle recognition model.

[0022] As a further solution of the present invention: in the step S3, the continuous duration for which the ventilation system in the sub-tunnel is turned on is t.

[0023] As a further solution of the present invention: in the step S1, the following steps are further included:

[0024] Obtain H monitoring periods, where H is the preset number of monitoring periods. Calculate the average value of the total number of vehicles entering within H sub-periods h. The entering vehicles are the vehicles entering the tunnel. Mark the sub-periods with an average value greater than or equal to the preset average threshold as working periods. During the sub-periods other than the working periods, the ventilation system is not turned on.

[0025] As a further solution of the present invention: during the process of calculating the average value, when the difference between the total number of vehicles entering corresponding to the sub-period h within a certain monitoring period and the average value is greater than the preset difference threshold, remove the total number of vehicles entering corresponding to the sub-period h within this monitoring period, and calculate the average value again.

[0026] As a further solution of the present invention: in the step S2 and the step S3, if the total emissions F ≥ 3Fys, then turn on all the ventilation systems in the tunnel.

[0027] Advantages of the present invention: By dividing the monitoring period into multiple sub-periods and collecting monitoring images at the start of each sub-period, accurate time and space reference points can be provided for subsequent analysis. By identifying the target vehicles in the images and calculating the minimum circumscribed rectangle size (length and width) of the target vehicles, the basic information of the vehicles can be efficiently extracted, providing the necessary data support for the estimation of the actual size of the vehicles and the calculation of the emissions. This process helps to accurately identify the spatial characteristics of each vehicle in practical applications, thus providing an accurate basis for subsequent steps and ultimately ensuring the high accuracy of the calculation results of emissions and speed. In the second step (S2), further by obtaining the width of the lane and the actual lane width, and combining with the image information of the target vehicle, the actual size (length and width) of the vehicle is accurately calculated using the proportional relationship. This accurate calculation of the vehicle size is crucial for the subsequent calculation of the exhaust emission index (Y) because the size of the vehicle is highly correlated with the emissions. By further introducing the calculation of vehicle speed (V) and emission index, the exhaust emission level of each vehicle can be scientifically evaluated. Correction factors (η1 and η2) are also introduced in this step, which can accurately adjust the emissions according to different conditions, thus better fitting the actual environment and ensuring that the calculated exhaust emissions are more real and reliable. Then, the solution calculates the emissions (f) of a single target vehicle within the sub-period and accumulates the total emissions (F). Combining with the response mechanism of the threshold judgment system, the opening time of the tunnel ventilation system can be dynamically adjusted to avoid over-enabling the ventilation system, thus saving energy and operating costs. The advantage of this step is that through real-time monitoring and adjustment, the ventilation intensity in the tunnel can be accurately controlled according to the actual situation, avoiding unnecessary energy waste, ensuring that the air quality inside the tunnel is always within the safe range, and at the same time reducing the operating expenses. In the third step (S3), by calculating the theoretical driving distance (L) of the target vehicle and determining the target position in front of it based on the vehicle position to form a target line, the tunnel can be segmented and managed. According to the section of the tunnel where the target vehicle is located, the selection number of the sub-tunnel is accurately calculated, thereby determining whether to turn on the ventilation system to ensure that the activation of the ventilation system is based on actual needs rather than a preset time. This step can more precisely control the air flow in the tunnel, ensuring that the ventilation system is only activated when the vehicle emissions reach a certain level, thus reducing energy waste. Combining with the goal of the overall solution, by dynamically adjusting the opening strategies of the emission monitoring and ventilation systems, the management efficiency of the air quality in the tunnel can be greatly improved, while reducing the energy consumption of the tunnel ventilation system, ensuring that the tunnel environment maintains good air quality without incurring excessive operating costs. The overall solution ultimately achieves the dual goals of energy conservation, emission reduction, and environmental protection through accurate vehicle emission monitoring, scientific emission calculation, and dynamic adjustment of the ventilation system. Description of the Drawings

[0028] The present invention will be further described below with reference to the accompanying drawings.

[0029] Figure 1 It is a schematic flow chart of a tunnel monitoring method based on machine vision of the present invention. Specific embodiments

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0031] Please refer to Figure 1 As shown, the present invention is a tunnel monitoring method based on machine vision, including the following steps:

[0032] S1: Divide a preset monitoring period into n sub-periods with a length of t, where n is a preset first quantity, and t represents the length of a single said sub-period. Collect monitoring images inside the tunnel at the starting point of the sub-period, identify the vehicles in the monitoring images, denoted as target vehicles, mark the minimum bounding rectangle of the target vehicles in the monitoring images, and obtain the length c and width k of the minimum bounding rectangle;

[0033] S2: Obtain the width d1 of the lane in the monitoring image, and obtain the actual width d2 of the lane; calculate the actual length C = c * (d2 / d1) of the target vehicle, and calculate the actual width K = k * (d2 / d1) of the target vehicle. Then the vehicle type score P = η1 * K * D, where η1 is a preset first correction coefficient. Obtain the average vehicle speed V of the target vehicle, and calculate the exhaust emission index Y = η2 * V * P, where η2 is a preset second correction coefficient;

[0034] Calculate the emission f = η3 * Y * t of a single said target vehicle within the sub-period, where η3 is a preset third correction coefficient, calculate the total emission F, set a total emission threshold Fys. If the total emission F < Fys, obtain a new total emission in the next sub-period; if the total emission F ≥ Fys, use the starting point of the corresponding sub-period as the adjustment time point;

[0035] S3: Calculate the theoretical distance L = V * t, determine the position on the lane where the target vehicle is located and the distance between the position of the target vehicle is the theoretical distance L, and use the position in front of the target vehicle as the target position. Connect the position of the target vehicle and the target position to obtain the target line; divide the tunnel into m sub-tunnels with the same length, where m is a preset second quantity, obtain the tunnel section where the target line is located. When a sub-tunnel belongs to any of the tunnel sections, add 1 to the selection number of the sub-tunnel, determine the total selection number of a single sub-tunnel. When the total selection number is greater than or equal to the preset total selection number threshold, turn on the ventilation system in the sub-tunnel.

[0036] It should be noted that by dividing the monitoring period into multiple sub-periods and collecting monitoring images at the starting point of each sub-period, accurate time and space reference points can be provided for subsequent analysis. By identifying the target vehicle in the image and calculating the minimum bounding rectangle size (length and width) of the target vehicle, the basic information of the vehicle can be efficiently extracted, providing the necessary data support for the estimation of the actual size of the vehicle and the calculation of emissions. This process helps to accurately identify the spatial characteristics of each vehicle in practical applications, providing a precise basis for subsequent steps and ultimately ensuring the high accuracy of the calculated results of emissions and speed. The second step (S2) further accurately calculates the actual size (length and width) of the vehicle by obtaining the width of the lane and the actual lane width, and combining the image information of the target vehicle using the proportional relationship. This accurate calculation of the vehicle size is crucial for the subsequent calculation of the exhaust emission index (Y) because the size of the vehicle is highly correlated with the emissions. By further introducing the vehicle speed (V) and the calculation of the emission index, the exhaust emission level of each vehicle can be scientifically evaluated. Correction factors (η1 and η2) are also introduced in this step, which can precisely adjust the emissions according to different conditions, making it more suitable for the actual environment and ensuring that the calculated exhaust emissions are more real and reliable. Then, the solution calculates the emissions (f) of a single target vehicle within a sub-period and accumulates the total emissions (F). Combining the response mechanism of the threshold judgment system, the opening time of the tunnel ventilation system can be dynamically adjusted to avoid over-enabling the ventilation system, thus saving energy and operating costs. The advantage of this step is that through real-time monitoring and adjustment, the ventilation intensity in the tunnel can be accurately controlled according to the actual situation, avoiding unnecessary energy waste, ensuring that the air quality inside the tunnel is always within the safe range, and at the same time reducing the operating expenses. The third step (S3) forms a target line by calculating the theoretical driving distance (L) of the target vehicle and determining the target position in front of it based on the vehicle position, so as to manage the tunnel in sections. According to the part of the tunnel where the target vehicle is located, the number of selected sub-tunnels is accurately calculated, and then it is judged whether to turn on the ventilation system to ensure that the ventilation system is started based on actual needs rather than a preset time. This step can more precisely control the air flow in the tunnel, ensuring that the ventilation system is only started when the vehicle emissions reach a certain level, thus reducing energy waste. Combining the goals of the overall solution, by dynamically adjusting the opening strategies of the emission monitoring and ventilation systems, the management efficiency of the air quality in the tunnel can be greatly improved, while reducing the energy consumption of the tunnel ventilation system, ensuring that the tunnel environment maintains good air quality without increasing excessive operating costs. The overall solution ultimately achieves the dual goals of energy conservation, emission reduction, and environmental protection through precise vehicle emission monitoring, scientific emission calculation, and dynamic adjustment of the ventilation system;

[0037] It should be noted that the target line is along the road traffic direction. If the road traffic direction is curved, the target line is also curved.

[0038] The total emissions F are for all target vehicles. If the tunnel is too long and a single monitoring device cannot capture the entire picture, then all target vehicles are determined based on multiple monitoring devices. The emissions of the same vehicle are only counted once within a sub-period. The specific number of monitoring devices can be determined according to the actual situation.

[0039] When determining the total selection number of a single sub-tunnel, for example: if sub-tunnel x1 belongs to the tunnel section corresponding to target line X1, the selection number of the sub-tunnel is increased by 1. At the same time, if sub-tunnel x1 belongs to the tunnel section corresponding to target line X2, the selection number of the sub-tunnel is increased by 1 again. By this step, the total selection number of sub-tunnel x1 is determined.

[0040] In another preferred embodiment of the present invention, in step S3, the sub-tunnel includes a ventilation system.

[0041] In another preferred embodiment of the present invention, in step S3, after the ventilation system is turned on, the ventilation power G of the ventilation system is G = ((N - Nys) / Nys)*Gys, where Gys represents the preset standard ventilation power, and N and Nys respectively represent the total selection number and the total selection number threshold.

[0042] In another preferred embodiment of the present invention, in step S3, the following steps are further included:

[0043] When the number of sub-tunnels with the ventilation system turned on is less than 0.3m, except for the total selection number corresponding to the sub-tunnels with the ventilation system already turned on, the remaining total selection numbers are sorted in descending order. Starting from the sub-tunnel corresponding to the first total selection number in the sorting, the ventilation system is turned on in sequence, and the ventilation power of the newly turned-on ventilation system is maintained at Gys until the number of sub-tunnels with the ventilation system turned on is greater than or equal to 0.3m.

[0044] It should be noted that according to the real-time emission situation, sub-tunnels with larger emissions are preferentially selected to ensure that the concentration of harmful gases in the tunnel is maximally reduced without wasting energy. At the same time, by maintaining the ventilation power of the newly turned-on ventilation system at Gys, energy waste caused by over-turning on the ventilation system can be avoided, ensuring that the entire ventilation operation achieves the best air flow effect on the premise of energy conservation.

[0045] In another preferred embodiment of the present invention, in step S3, the following steps are further included:

[0046] When the ventilation power G < Gmin, let the ventilation power G = Gmin;

[0047] When the ventilation power G > Gmax, set the ventilation power G = Gmax, and perform the following steps:

[0048] Except for the total selection numbers corresponding to the sub-tunnels with the ventilation system already turned on, sort the remaining total selection numbers in descending order. Starting from the sub-tunnel corresponding to the first total selection number in the sorting, turn on the ventilation system in sequence, and keep the ventilation power of the newly turned-on ventilation system as Gys until the total ventilation power of the newly turned-on ventilation system is greater than 3(G - Gmax).

[0049] In another preferred embodiment of the present invention, in step S1, a vehicle is identified from the monitoring image based on a pre-trained vehicle recognition model.

[0050] It can be understood that using a pre-trained vehicle recognition model to identify vehicles from monitoring images in step S1 can not only greatly improve the accuracy and real-time performance of vehicle detection, effectively utilize deep learning technology to reduce the risk of false detection and missed detection in various complex environments, but also provide accurate and reliable data support for subsequent vehicle size measurement, emission calculation, and dynamic ventilation control; the vehicle recognition model is established based on a deep learning model, and the specific establishment and training processes are already relatively mature in the prior art and will not be elaborated here.

[0051] In another preferred embodiment of the present invention, in step S3, the continuous duration for which the ventilation system in the sub-tunnel is turned on is t.

[0052] In another preferred embodiment of the present invention, in step S1, the following steps are further included:

[0053] Obtain H monitoring periods, where H is the preset number of monitoring periods. Calculate the mean value of the total number of vehicles entering within H sub-periods h. The entering vehicles are the vehicles entering the tunnel. Mark the sub-periods with a mean value greater than or equal to the preset mean threshold as working periods. During sub-periods other than the working periods, the ventilation system is not turned on.

[0054] It can be understood that by calculating the mean value of the number of vehicles entering the tunnel within multiple monitoring periods and using this as a basis to determine which periods have more entering vehicles, it is possible to accurately select when to start the ventilation system. The ventilation system is only started when the number of vehicles entering the tunnel reaches a certain level (i.e., the mean value is greater than or equal to the preset mean threshold), which helps to reduce unnecessary energy consumption; during periods when the number of vehicles is small, the exhaust emissions are relatively low, and the ventilation system does not need to be turned on. Through this measure, it is possible to avoid unnecessarily turning on the ventilation system when the number of vehicles is small and reduce energy consumption.

[0055] In another preferred embodiment of the present invention, during the process of calculating the mean value, when the difference between the total number of incoming vehicles corresponding to sub-period h within a certain monitoring period and the mean value is greater than a preset difference threshold, the total number of incoming vehicles corresponding to sub-period h within this monitoring period is removed, and the mean value is recalculated.

[0056] It should be noted that by removing the abnormal data with too large a difference from the mean value, the influence of extreme situations on the mean value can be avoided, ensuring that the calculated mean value is more reliable and accurate. For example, in a certain period, due to special reasons (such as emergencies or instant traffic peaks), the number of vehicles in a certain sub-period may be abnormally high or low. After removing these abnormal values, the calculated mean value can more truly reflect the normal traffic flow situation, thus providing a more accurate data basis for the regulation of the ventilation system.

[0057] In another preferred embodiment of the present invention, in step S2 and step S3, if the total emissions F≥3Fys, then all the ventilation systems in the tunnel are turned on.

[0058] It should be noted that setting the threshold of the total emissions as 3 times the threshold (3Fys) as the trigger condition can quickly respond to sudden severe air pollution situations. When the tail gas emissions in the tunnel reach a relatively high level, immediately enabling all the ventilation systems helps to quickly reduce the concentration of harmful substances in the air and avoid the continuous accumulation of pollutants causing greater harm to the personnel in the tunnel and the surrounding environment. The rapid response can effectively reduce the time window of air quality deterioration and ensure the safety of the tunnel environment.

[0059] The above has described a detailed description of an embodiment of the present invention, but the content described is only the preferred embodiment of the present invention and cannot be considered as used to limit the scope of implementation of the present invention. All equal changes and improvements made according to the scope of the present invention application should still fall within the scope covered by the patent of the present invention.

Claims

1. A tunnel monitoring method based on machine vision, characterized in that: The following steps are involved: S1: Divide a preset monitoring period into n sub-periods of length t, where n is a preset first number and t represents the length of a single preset sub-period, collect a monitoring image in the tunnel at the starting point of the sub-period, identify a vehicle in the monitoring image, record it as a target vehicle, mark the minimum circumscribed rectangle of the target vehicle in the monitoring image, and obtain the length c and width k of the minimum circumscribed rectangle; S2: Obtain the width d1 of the lane in the monitoring image, and obtain the actual width d2 of the lane; calculate the actual length C=c*(d2 / d1) of the target vehicle, and calculate the actual width K=k*(d2 / d1) of the target vehicle, then the vehicle model score P=η1*K*D, η1 is the preset first correction coefficient, obtain the average speed V of the target vehicle, calculate the exhaust emission index Y=η2*V*P, η2 is the preset second correction coefficient; Calculate the emission of a single target vehicle in a sub-period f=η3*Y*t, η3 is a preset third correction coefficient, calculate the total emission F, set the total emission threshold Fys, if the total emission F<Fys, then obtain a new total emission in the next sub-period; if the total emission F≥Fys, then take the starting point of the corresponding sub-period as the adjustment time point; S3: Calculate the theoretical distance L=V*t, determine the distance between the target vehicle's lane and the target vehicle's position as the position of the theoretical distance L, take the position in front of the target vehicle as the target position, and connect the target vehicle's position and the target position to obtain a target line; The tunnel is divided into m sub-tunnels of the same length, where m is a preset second number, and the tunnel portion where the target line is located is obtained. When a sub-tunnel belongs to any of the tunnel portions, the selection number of the sub-tunnel is incremented by 1, and the total selection number of a single sub-tunnel is determined. When the total selection number is greater than or equal to a preset total selection number threshold, the ventilation system in the sub-tunnel is turned on.

2. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: In the step S3, the sub-tunnel includes a ventilation system.

3. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: In the step S3, after the ventilation system is turned on, the ventilation power G of the ventilation system is G=((N-Nys) / Nys)*Gys, Gys represents the preset standard ventilation power, and N and Nys represent the total selection number and the total selection number threshold, respectively.

4. The tunnel monitoring method based on machine vision according to claim 3 is characterized in that: The step S3 further includes the following steps: When the number of sub-tunnels with ventilation systems turned on is less than 0.3m, except for the total selection numbers corresponding to the sub-tunnels with ventilation systems turned on, the remaining total selection numbers are sorted in descending order, starting from the sub-tunnel corresponding to the first total selection number in the sorting, and the ventilation systems are turned on one by one, and the ventilation power of the newly turned on ventilation systems is maintained at Gys until the number of sub-tunnels with ventilation systems turned on is greater than or equal to 0.3m.

5. The tunnel monitoring method based on machine vision according to claim 4 is characterized in that: The step S3 further includes the following steps: When the ventilation power G<Gmin, the ventilation power G=Gmin; When the ventilation power G>Gmax, the ventilation power G=Gmax, and the following steps are performed: Except for the total number of selections corresponding to the sub-tunnels with the ventilation system turned on, the remaining total number of selections shall be sorted in descending order. Starting from the sub-tunnel corresponding to the first total number of selections in the sorting, the ventilation systems shall be turned on one by one, and the ventilation power of the newly turned on ventilation system shall be maintained at Gys until the total ventilation power of the newly turned on ventilation system is greater than 3 (G-Gmax).

6. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: In the step S1, a vehicle is identified from the monitoring image based on a pre-trained vehicle recognition model.

7. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: In the step S3, the duration of opening of the ventilation system in the sub-tunnel is t.

8. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: The step S1 further includes the following steps: Obtain H monitoring periods, where H is the preset number of monitoring periods, calculate the mean of the total number of vehicles entering the H sub-periods h, where the entering vehicles are vehicles entering the tunnel, and mark the sub-periods with a mean greater than or equal to a preset mean threshold as working periods. The ventilation system is not turned on in the sub-periods other than the working periods.

9. The tunnel monitoring method based on machine vision according to claim 8, characterized in that: In the process of calculating the mean, when the difference between the total number of entering vehicles corresponding to the sub-period h in a certain monitoring period and the mean is greater than the preset difference threshold, the total number of entering vehicles corresponding to the sub-period h in the monitoring period is removed and the mean is calculated again.

10. The tunnel monitoring method based on machine vision according to claim 1, characterized in that: In the step S2 and the step S3, if the total emission F≥3Fys, all ventilation systems in the tunnel are turned on.

Images