Road-traffic carbon emission prediction method and apparatus based on deep learning

Abstract

The present application relates to the field of traffic environment monitoring. Disclosed are a road-traffic carbon emission prediction method and apparatus based on deep learning. The method in the present application comprises: using contour features and traveling speeds of vehicles as an input layer and using the types of the vehicles and emission rates of a greenhouse gas equivalent as an output layer, and training a preliminary prediction model until same is converged; using measured emission concentrations of various types of exhausts as an input layer and using the emission rates of a greenhouse gas equivalent of the vehicles as an output layer, training a corrected prediction model until same is converged; setting a monitoring point on a road; at the monitoring point, acquiring the contour feature and traveling speed of a vehicle passing by, and the emission concentrations of various types of exhausts; inputting, into the preliminary prediction model and the corrected prediction model, the contour feature and traveling speed of the vehicle passing by, and the emission concentrations of the various types of exhausts, so as to obtain the emission rate of a greenhouse gas equivalent of the vehicle passing by. The present application improves the accuracy of carbon emission prediction.

Description

A method and device for predicting carbon emissions from highway traffic based on deep learning

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on December 14, 2023, with application number 202311719331.6 and invention name “A method and device for predicting carbon emissions from highway traffic based on deep learning”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application belongs to the technical field of traffic environment monitoring, and in particular relates to a method and device for predicting carbon emissions from highway traffic based on deep learning. Background Art

[0003] Road transportation is one of the main sources of carbon emissions worldwide, with significant impacts on the environment. To reduce carbon emissions, it is necessary to accurately predict carbon emissions from road transportation for effective management and control.

[0004] Related carbon emission prediction methods are mainly based on statistical models, such as linear regression, multivariate linear regression, time series analysis, etc. However, these methods have limited ability to handle complex and nonlinear data relationships, so the prediction results may have large errors. Summary of the Invention

[0005] The purpose of this application is to provide a method and device for predicting carbon emissions from road traffic based on deep learning, which uses historical data training to obtain a preliminary prediction model and a correction prediction model to detect carbon emissions from vehicles passing through the road, thereby improving the accuracy of carbon emission predictions.

[0006] To solve the above technical problems, this application is implemented through the following technical solutions:

[0007] This application provides a road traffic carbon emissions prediction method based on deep learning, including:

[0008] Obtaining contour features of multiple types of vehicles;

[0009] The test obtains the detected emission concentrations of multiple types of exhaust gases and the detected emission rates of equivalent greenhouse gases of different types of vehicles at different driving speeds;

[0010] Obtaining contour features of different types of vehicles;

[0011] The vehicle's profile features and driving speed are used as the input layer, and the vehicle type and equivalent greenhouse gas emission rate are used as the output layer. The preliminary prediction model is trained until convergence;

[0012] The detected emission concentrations of various types of exhaust gas are used as the input layer, and the equivalent greenhouse gas emission rate of the vehicle is used as the output layer. The correction prediction model is trained until convergence;

[0013] Setting up monitoring points on roads;

[0014] Obtaining the profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles at the monitoring point;

[0015] The profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles are input into the preliminary prediction model and the correction prediction model respectively to obtain the emission rate of equivalent greenhouse gases of the passing vehicles.

[0016] This application also discloses a method for predicting carbon emissions from highway transportation based on deep learning, including:

[0017] Set up multiple monitoring points on the road;

[0018] Continuously obtaining the equivalent greenhouse gas emission rate of vehicles passing through each of the monitoring points;

[0019] The equivalent greenhouse gas emission rate of vehicles passing on the road is obtained according to the average value of the equivalent greenhouse gas emission rates of vehicles passing at different times at multiple monitoring points.

[0020] This application also discloses a method for predicting carbon emissions from highway transportation based on deep learning, including:

[0021] Obtaining a road space model;

[0022] Marking the locations of monitoring points in the road space model;

[0023] Obtain the equivalent greenhouse gas emission rate of vehicles passing through the monitoring point;

[0024] The equivalent greenhouse gas emission rate of vehicles passing through the monitoring point is displayed in the road space model.

[0025] This application also discloses a road traffic carbon emission prediction device based on deep learning, comprising:

[0026] Model training module, used to obtain the contour features of multiple types of vehicles;

[0027] The test obtains the detected emission concentrations of multiple types of exhaust gases at different driving speeds, as well as the emission rates of equivalent greenhouse gases, from different types of vehicles;

[0028] Obtaining contour features of different types of vehicles;

[0029] The vehicle's profile features and driving speed are used as the input layer, and the vehicle type and equivalent greenhouse gas emission rate are used as the output layer. The preliminary prediction model is trained until convergence;

[0030] The detected emission concentrations of various types of exhaust gas are used as the input layer, and the equivalent greenhouse gas emission rate of the vehicle is used as the output layer. The correction prediction model is trained until convergence;

[0031] Global prediction module, used to set monitoring points on the road;

[0032] The recognition and prediction module is used to obtain the profile characteristics, driving speed and emission concentration of various types of exhaust gases of passing vehicles at preset monitoring points;

[0033] Inputting the profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles into the preliminary prediction model and the correction prediction model respectively to obtain the equivalent greenhouse gas emission rate of the passing vehicles;

[0034] The global prediction module is also used to set up multiple monitoring points on the road;

[0035] Obtaining the equivalent greenhouse gas emission rate of vehicles passing on the road according to the average value of the equivalent greenhouse gas emission rate of vehicles passing at different times at each of the monitoring points;

[0036] Visualization module, used to obtain road space model;

[0037] Marking the locations of monitoring points in the road space model;

[0038] Obtain the equivalent greenhouse gas emission rate of passing vehicles;

[0039] The equivalent greenhouse gas emission rate of vehicles passing through the monitoring point is displayed in the road space model.

[0040] This application uses historical test data to train a preliminary prediction model and a corrected prediction model for predicting carbon emission rates. During the prediction process, the equivalent greenhouse gas emission rates of passing vehicles are first obtained based on the preliminary prediction model and then verified. If the verification passes, the model is retained; if not, a more accurate equivalent greenhouse gas emission rate is obtained using the corrected prediction model. This dual approach improves the accuracy of carbon emission predictions.

[0041] Of course, any product implementing the present application does not necessarily need to achieve all of the advantages described above at the same time.

[0042] Figures in the specification

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0044] FIG1 is a schematic diagram of functional modules and information flow of a device for predicting carbon emissions from highway traffic based on deep learning according to an embodiment of the present application;

[0045] FIG2 is a schematic diagram of a flow chart of steps of a model training module and an identification and prediction module according to an embodiment of the present application;

[0046] FIG3 is a schematic diagram of a process flow of a global prediction module according to an embodiment of the present application;

[0047] FIG4 is a schematic diagram of a process flow of a visualization module according to an embodiment of the present application;

[0048] FIG5 is a schematic diagram of a flow chart of step S8 in one embodiment of the present application;

[0049] FIG6 is a schematic diagram of a flow chart of step S81 in one embodiment of the present application;

[0050] FIG7 is a schematic diagram of a flowchart of step S812 in one embodiment of the present application;

[0051] FIG8 is a schematic diagram of a flow chart of step S8122 in one embodiment of the present application;

[0052] FIG9 is a schematic diagram of a flow chart of step S84 in one embodiment of the present application;

[0053] FIG. 10 is a schematic diagram of a flow chart of step S845 in one embodiment of the present application.

[0054] In the accompanying drawings, the components represented by each number are listed as follows: 1-model training module, 2-global prediction module, 3-identification prediction module, 4-visualization module. DETAILED DESCRIPTION

[0055] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0056] It should be noted that the terms "first," "second," and the like in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Instead, they are merely examples of devices and methods consistent with certain aspects of the present application as detailed in the appended claims.

[0057] Carbon emissions refer to the amount of carbon dioxide (CO2) gas released during energy consumption and production. It is one of the main greenhouse gases and has a significant impact on climate change. Road transport is a significant source of carbon emissions. The combustion process in motor vehicles, such as cars, trucks, and motorcycles, produces large amounts of carbon dioxide (CO2) gas, which is released into the atmosphere. This contributes to carbon emissions from road transport.

[0058] In addition to carbon dioxide, automobile exhaust may also contain other greenhouse gases and pollutants, such as carbon monoxide (CO), nitrogen oxides (NO x ) and non-methane volatile organic compounds (NMVOCs). These substances also have a certain impact on air quality and environmental health.

[0059] In order to predict road traffic carbon emissions, please refer to Figures 1 to 4. The present application provides a road traffic carbon emissions prediction device based on deep learning. This device is a device for the interaction of multiple application modules, specifically including a model training module 1, a global prediction module 2, an identification prediction module 3, and a visualization module 4. Of course, these divisions are not strictly based on the form of the equipment, but only divide the units that implement and have certain functions. For example, the model training module 1 is used to train and obtain a usable preliminary prediction model and a correction prediction model. In practice, it may be a series of test devices and computing units and combinations. The global prediction module 2 can be a computing unit for global planning, the identification prediction module 3 is a combination of a speed sensor, an image acquisition camera, an exhaust gas detection sensor, and a computing unit, and the visualization module 4 can be a display screen.

[0060] Before each functional module runs, model training module 1 is required to train the model. First, step S1 can be executed to obtain the profile features of multiple types of vehicles. Next, step S2 can be executed to test and obtain the detected emission concentrations of multiple types of exhaust gases and the equivalent greenhouse gas emission rates for different types of vehicles at different driving speeds. Next, step S3 can be executed to obtain the profile features of different types of vehicles. This is because different types of vehicles have different carbon emission rates and their body profile characteristics also vary. Generally speaking, sedans typically have two to four doors and a closed body. They are low and streamlined, with minimal front and rear overhangs, smooth contours, a generally rounded roof, and moderately sized tires that are in proportion to the body. SUVs (Sport Utility Vehicles) have high ground clearance and a large body size to adapt to different terrain conditions. They have relatively square contours, a long and tall body, typically four doors and a large interior space, and relatively large tires, which provide strong off-road capabilities. Trucks have longer bodies, taller cargo boxes or loading areas, longer fronts, and typically have one or two cabs. They also have larger tires, higher load capacities, and taller chassis with greater ground clearance. Contour recognition can effectively distinguish vehicles with different carbon emission rates. It's important to note that the license plates of pure electric new energy vehicles have distinct characteristics and can be identified without contour recognition.

[0061] During the model training process, step S4 can be executed to train the preliminary prediction model until convergence, using the vehicle's profile features (including the vehicle's profile features obtained in step S1 and the vehicle's profile features obtained in step S3) and driving speed as the input layer, and the vehicle type and equivalent greenhouse gas emission rate as the output layer. Next, step S5 can be executed to train the corrected prediction model until convergence, using the detected emission concentrations of various types of exhaust gas as the input layer and the vehicle's equivalent greenhouse gas emission rate as the output layer. This results in two models with different recognition accuracy rates, which balance the efficiency and accuracy of recognition prediction.

[0062] Before predicting road carbon emissions, the global prediction module 2 must execute step S6 to set up monitoring points along the road. The identification prediction module 3 then executes step S7 to obtain the vehicle profile, speed, and exhaust emission concentrations of various types at the pre-set monitoring points. Next, step S8 is executed to input these vehicle profiles, speed, and exhaust emission concentrations into the preliminary prediction model and the revised prediction model, respectively, to determine the equivalent greenhouse gas emission rate for each vehicle.

[0063] Because traffic conditions vary at different road locations, multiple monitoring points can be set to achieve a more accurate prediction of road carbon emissions. This requires the global prediction module 2 to execute step S011 to set up multiple monitoring points on the road. Next, steps S012 to S013 can be executed to receive and calculate the equivalent greenhouse gas emission rate of vehicles passing through the road based on the average of the equivalent greenhouse gas emission rates of vehicles passing through at different times at each monitoring point.

[0064] To visually display the status of road carbon emissions, the visualization module 4 can execute step S021 to obtain a road spatial model. Next, step S022 can be executed to mark the location of the monitoring point within the road spatial model. Then, step S023 can be executed to obtain the equivalent greenhouse gas emission rate of passing vehicles. Finally, step S024 can be executed to display the equivalent greenhouse gas emission rate of passing vehicles at the monitoring point within the road spatial model. This information can be transmitted via the internet and displayed on various network terminals.

[0065] Referring to Figure 5 , since this solution utilizes a preliminary prediction model and a corrected prediction model with varying degrees of accuracy, to balance prediction accuracy and speed, the preliminary prediction model can be prioritized. If the output is unsatisfactory, the corrected prediction model can be invoked to calibrate the results. Specifically, during the implementation of step S8, step S81 can be performed to determine the equivalent greenhouse gas emission ranges for different types of vehicles at different speeds based on the detected equivalent greenhouse gas emission rates of different types of vehicles at different speeds. Next, step S82 can be performed to input the profile characteristics and driving speed of the passing vehicle into the preliminary prediction model to determine the type of passing vehicle and the equivalent greenhouse gas emission rate. Next, step S83 can be performed to determine whether the equivalent greenhouse gas emission rate of the passing vehicle is within the equivalent greenhouse gas emission range for the corresponding driving speed. If so, step S84 can be performed to output the equivalent greenhouse gas emission rate of the passing vehicle. If not, step S85 can be performed to input the emission concentrations of each type of exhaust gas from the passing vehicle into the corrected prediction model to determine the corrected equivalent greenhouse gas emission rate of the passing vehicle. The final output is the equivalent greenhouse gas emission rate of passing vehicles. This process balances accuracy and efficiency in identification and prediction. Accurately and efficiently identifying the equivalent greenhouse gas emission rate of passing vehicles can be used to evaluate the environmental performance of vehicles, thereby encouraging vehicle manufacturers and owners to take appropriate measures to reduce equivalent greenhouse gas emissions, lower air pollution, improve environmental quality, and achieve green travel.

[0066] To supplement the implementation of steps S81 to S85, we provide the source code for some functional modules, with cross-references and explanations provided in the comments. To prevent the leakage of data involving commercial secrets, data that does not affect the implementation of the solution is desensitized. The same applies below.

[0067] The basic flow of the above code is:

[0068] Create a TensorFlow session and load the trained preliminary prediction model and the corrected prediction model. Then, run the preliminary prediction model to obtain preliminary prediction results. If the preliminary prediction results are within the preset valid range, the prediction results are directly output. If the preliminary prediction results are not within the preset valid range, run the corrected prediction model to obtain the corrected prediction results, which are then output. Finally, close the session.

[0069] Referring to Figure 6 , since the dataset used to train the model often contains a large amount of abnormal data, to avoid inaccurate outputs from the preliminary and corrected prediction models, the accuracy of the output results needs to be analyzed. Specifically, for each type of vehicle, during implementation, step S81 can first be performed to repeatedly obtain the vehicle's equivalent greenhouse gas emission rates at multiple driving speeds. Next, step S812 can be performed to obtain the vehicle's equivalent greenhouse gas emission rate range at multiple driving speeds based on the multiple equivalent greenhouse gas emission rates at multiple driving speeds. Next, step S813 can be performed to fit the vehicle's equivalent greenhouse gas emission rate range at multiple driving speeds to obtain the vehicle's equivalent greenhouse gas emission range at different driving speeds. Finally, step S814 can be performed to aggregate the data for different vehicle types to obtain the equivalent greenhouse gas emission ranges for different vehicle types at different driving speeds. Of course, in practice, experienced automotive engineers can also directly set the range.

[0070] Please refer to Figures 7 and 8. Due to China's fuel monopoly system, the quality of vehicle fuel is relatively consistent. At the same time, China's road conditions are relatively good. This allows the equivalent carbon dioxide emissions in exhaust gas to be simplified to be affected by vehicle type and driving speed. In other words, vehicles of the same model have similar equivalent carbon dioxide emission rates at the same speed. Based on this principle, the accuracy of the results output by the preliminary prediction model can be evaluated. To obtain the equivalent greenhouse gas emission rate ranges for different types of vehicles at multiple driving speeds, the above-mentioned step S812 can first be implemented in step S8121. Based on the multiple equivalent greenhouse gas emission rates detected by the vehicle at multiple driving speeds, each detected emission rate of the vehicle and the corresponding driving speed are combined into a two-dimensional vector to obtain multiple detection feature vectors for the vehicle during the multiple detection processes. Next, steps S81221 to S81226 in step S8122 can be executed. That is, step S81221 can first be executed to select several of the multiple detection feature vectors as baseline detection feature vectors. Next, step S81222 can be executed to calculate and obtain the vector difference modulus length between other detection feature vectors and the benchmark detection feature vector. Next, step S81223 can be executed to form a vector set with the other detection feature vectors and the benchmark detection feature vector with the smallest vector difference modulus length. Next, step S81224 can be executed to calculate and obtain the detection feature vector with the smallest vector difference modulus length with the mean vector of all detection feature vectors in each vector set as the updated benchmark detection feature vector. Next, step S81225 can be executed to determine whether the benchmark detection feature vector in the vector set has changed before and after the update. If so, then steps S81223 to S81225 can be executed to continuously update the generated vector set and the benchmark detection feature vector. Otherwise, step S81226 can be executed to use the vector set containing the largest number of detection feature vectors as the target vector set. Next, step S8123 can be executed to obtain multiple detection emission rates corresponding to each driving speed based on the two-dimensional vector of the detection emission rate and driving speed corresponding to the detection feature vector in the target vector set. Finally, step S8124 may be executed to use the range of the minimum and maximum values ​​of the multiple detected emission rates corresponding to each driving speed as the equivalent greenhouse gas emission rate range of the vehicle at the multiple driving speeds.

[0071] In order to supplement the implementation process of the above-mentioned steps S8121 to S8124, the source code of some functional modules is provided, and a comparative explanation is given in the comment section.

[0072] The program first obtains all detection feature vectors and normalizes them. It then divides the feature vectors into several sets by calculating their distance from all other feature vectors. Next, it calculates the baseline detection feature vector for each set—the mean of all feature vectors in that set—and compares these new baseline detection feature vectors to the original baseline detection feature vectors. If there are any differences, the process continues recursively. Otherwise, the set containing the most feature vectors is selected as the target set, and all feature vectors in this set are returned, representing the range of vehicle emissions at different speeds.

[0073] Referring to Figure 9 , to ensure more accurate output from the calibrated prediction model, outliers in the input data can be removed. Specifically, during the implementation of step S84, step S841 can be performed to determine the emission concentration ratio ranges for the multiple exhaust gas types at different speeds for the different types of vehicles based on the detected emission concentrations of the multiple exhaust gas types for the different types of vehicles. Next, step S842 can be performed to determine whether the ratio ranges of the detected emission concentrations of the multiple exhaust gas types for the multiple types of vehicles at different speeds fall within the emission concentration ratio ranges for the multiple exhaust gas types for the different types of vehicles at different speeds. If so, step S843 can be performed to retain the data; otherwise, the data can be removed. Next, step S844 can be performed to input the detected emission concentrations of the multiple groups of exhaust gas types for the retained groups of vehicles into the calibrated prediction model, thereby determining the emission rates of the multiple equivalent greenhouse gases for the vehicles. Finally, step S845 can be performed to determine the calibrated equivalent greenhouse gas emission rates for the vehicles based on the multiple equivalent greenhouse gas emission rates for the vehicles.

[0074] However, multiple input data will result in multiple output results. In order to obtain an equivalent greenhouse gas emission rate that is more in line with the actual situation, please refer to Figure 10. First, step S8451 can be executed to obtain the average driving speed of passing vehicles. This can be done by the vehicle actively uploading data or reading monitoring data from the transportation management department. Next, step S8452 can be executed to obtain the driving speed corresponding to each equivalent greenhouse gas emission rate of the passing vehicles. Finally, step S8453 can be executed to use the equivalent greenhouse gas emission rate corresponding to the driving speed that is consistent with the average driving speed or has the smallest difference as the equivalent greenhouse gas emission rate after correction of the passing vehicles. If the difference is too large, the driving speed corresponding to each equivalent greenhouse gas emission rate can also be fitted to obtain the equivalent greenhouse gas emission rate at the average driving speed of the vehicle.

[0075] The flow charts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, systems, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flow chart or block diagram can represent a part for a module, program segment or instruction, and the part for the module, program segment or instruction comprises one or more executable instructions for realizing the logical function of the specification. In some alternative implementations, the functions marked in the box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous boxes can actually be performed substantially in parallel, and they can sometimes also be performed in the opposite order, depending on the function involved.

[0076] It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart, can be implemented by hardware that performs the corresponding function or action, such as a circuit or ASIC (Application Specific Integrated Circuit), or can be implemented by a combination of hardware and software, such as firmware.

[0077] Although the present application is described herein in conjunction with various embodiments, in the process of implementing the claimed application, those skilled in the art can understand and implement other variations of the disclosed embodiments by reviewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple situations. A single processor or other unit can implement several functions listed in the claims. Certain measures are recorded in different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0078] The embodiments of the present application have been described above. The above description is illustrative and not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope of the described embodiments. The terminology used herein is selected to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons skilled in the art to understand the embodiments disclosed herein.

Claims

1. A highway traffic carbon emission prediction method based on deep learning, characterized in that: include: Obtaining contour features of multiple types of vehicles; The test obtains the detected emission concentrations of multiple types of exhaust gas and the detected emission rates of equivalent greenhouse gases of different types of vehicles at different driving speeds; Obtaining contour features of different types of vehicles; The vehicle profile and driving speed are used as the input layer, and the vehicle type and equivalent greenhouse gas emission rate are used as the output layer, and the preliminary prediction model is trained until convergence; The detected emission concentration of each type of exhaust gas is used as the input layer, and the emission rate of equivalent greenhouse gases of the vehicle is used as the output layer, and the correction prediction model is trained until convergence; Setting up monitoring points on roads; Obtaining the profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles at the monitoring point; The profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles are respectively input into the preliminary prediction model and the correction prediction model to obtain the emission rate of equivalent greenhouse gases of the passing vehicles.

2. The method according to claim 1, characterized in that The step of inputting the profile characteristics, driving speed and emission concentration of each type of exhaust gas of the passing vehicles into the preliminary prediction model and the correction prediction model to obtain the emission rate of equivalent greenhouse gases of the passing vehicles comprises: According to the detection emission rates of equivalent greenhouse gases of different types of vehicles at different driving speeds, the equivalent greenhouse gas emission ranges of different types of vehicles at different driving speeds are obtained; Inputting the profile characteristics and driving speed of the passing vehicles into the preliminary prediction model to obtain the types of passing vehicles and the emission rates of equivalent greenhouse gases; Determine whether the equivalent greenhouse gas emission rate of the passing vehicles is within the equivalent greenhouse gas emission range at the corresponding driving speed; If so, output the equivalent greenhouse gas emission rate of passing vehicles; If not, the emission concentration of each type of exhaust gas of the passing vehicles is input into the correction prediction model to obtain the emission rate of equivalent greenhouse gases of the passing vehicles after correction; Output the equivalent greenhouse gas emission rate of passing vehicles.

3. The method according to claim 2, characterized in that The step of obtaining the equivalent greenhouse gas emission ranges of different types of vehicles at different driving speeds according to the detected emission rates of equivalent greenhouse gases of different types of vehicles at different driving speeds comprises: For each type of vehicle, The detection emission rate of equivalent greenhouse gases of the vehicle at multiple driving speeds is obtained multiple times. According to the multiple equivalent greenhouse gas emission rates detected by the vehicle at multiple driving speeds, the equivalent greenhouse gas emission rate range of the vehicle at multiple driving speeds is obtained. Fitting the range of equivalent greenhouse gas emission rates of the vehicle at multiple driving speeds to obtain the equivalent greenhouse gas emission range of the vehicle at different driving speeds; The equivalent greenhouse gas emission ranges of different types of vehicles at different driving speeds are summarized.

4. The method according to claim 3, characterized in that The step of obtaining the emission rate range of equivalent greenhouse gases of the vehicle at multiple driving speeds according to multiple detected emission rates of equivalent greenhouse gases of the vehicle at multiple driving speeds comprises: According to multiple detection emission rates of equivalent greenhouse gases of the vehicle at multiple driving speeds, each detection emission rate of the vehicle and the corresponding driving speed are combined into a two-dimensional vector to obtain multiple detection feature vectors of the vehicle in the multiple detection processes; Eliminate outliers from multiple detection feature vectors to obtain a target vector set; Obtaining a plurality of detected emission rates corresponding to each driving speed according to the two-dimensional vector of the detected emission rate and the driving speed corresponding to the detected feature vector in the target vector set; The range of the minimum and maximum values ​​of the multiple detected emission rates corresponding to each driving speed is used as the emission rate range of the equivalent greenhouse gas of the vehicle at the multiple driving speeds.

5. The method according to claim 4, characterized in that The steps of eliminating abnormal values ​​from multiple detection feature vectors to obtain a target vector set include: Selecting a plurality of the detection feature vectors as reference detection feature vectors; Calculate and obtain the vector difference modulus length between the other detection feature vectors and the reference detection feature vector; The other detection feature vectors and the reference detection feature vector with the smallest vector difference modulus length form a vector set; Calculate and obtain the detection feature vector with the smallest vector difference modulus between each of the vector sets and the mean vector of all the detection feature vectors as the updated reference detection feature vector; Determining whether the reference detection feature vectors in the vector set have changed before and after the update; If yes, then continuously updating and generating the vector set and the benchmark detection feature vector; If not, the vector set containing the largest number of detected feature vectors is used as the target vector set.

6. The method according to claim 2, characterized in that The step of inputting the emission concentration of each type of exhaust gas of the passing vehicles into the correction prediction model to obtain the emission rate of equivalent greenhouse gases of the passing vehicles after correction includes: Obtaining emission concentration ratio ranges of multiple types of exhaust gases of different types of vehicles at different driving speeds according to detected emission concentrations of multiple types of exhaust gases of different types of vehicles at different driving speeds; Determine whether the ratio range of the detected emission concentrations of various types of exhaust gas of the passing vehicles obtained multiple times is within the emission concentration ratio range of multiple types of exhaust gas of different types of vehicles at different driving speeds; If yes, keep it; If not, remove it; Inputting the detected emission concentrations of various types of tail gas of the reserved groups of passing vehicles into the correction prediction model respectively to obtain emission rates of multiple equivalent greenhouse gases of the passing vehicles; The emission rate of equivalent greenhouse gases of passing vehicles after correction is obtained according to the emission rates of multiple equivalent greenhouse gases of passing vehicles.

7. The method according to claim 6, characterized in that The step of obtaining the emission rate of equivalent greenhouse gases after correction of the passing vehicles according to the emission rates of multiple equivalent greenhouse gases of the passing vehicles comprises: Get the average speed of passing vehicles; Obtain the driving speed corresponding to each equivalent greenhouse gas emission rate of passing vehicles; The equivalent greenhouse gas emission rate corresponding to the driving speed that is consistent with the average driving speed or has the smallest difference is taken as the equivalent greenhouse gas emission rate after correction of the passing vehicles.

8. A method for predicting carbon emissions from highway traffic based on deep learning, characterized in that: include, Set up multiple monitoring points on the road; A method for predicting carbon emissions from highway traffic based on deep learning according to any one of claims 1 to 7 continuously obtains the emission rate of equivalent greenhouse gases of vehicles passing through each of the monitoring points; The equivalent greenhouse gas emission rate of vehicles passing on the road is obtained according to the average value of the equivalent greenhouse gas emission rates of vehicles passing on the road at different times at multiple monitoring points.

9. A method for predicting carbon emissions from highway traffic based on deep learning, characterized in that: include, Obtaining a road space model; Marking the locations of monitoring points in the road space model; Obtaining the emission rate of equivalent greenhouse gases of vehicles passing through the monitoring point in the highway traffic carbon emission prediction method based on deep learning as described in any one of claims 1 to 7; The equivalent greenhouse gas emission rate of the passing vehicles at the monitoring point is displayed in the road space model.

10. A road traffic carbon emission prediction device based on deep learning, characterized in that: include, A model training module is used to obtain the contour features of multiple types of vehicles; The test obtains the detected emission concentrations of multiple types of exhaust gas at different driving speeds and the emission rates of equivalent greenhouse gases of different types of vehicles; Obtaining contour features of different types of vehicles; The vehicle profile and driving speed are used as the input layer, and the vehicle type and equivalent greenhouse gas emission rate are used as the output layer, and the preliminary prediction model is trained until convergence; The detected emission concentration of each type of exhaust gas is used as the input layer, and the emission rate of equivalent greenhouse gases of the vehicle is used as the output layer, and the correction prediction model is trained until convergence; A global prediction module, used to set monitoring points on roads; The recognition and prediction module is used to obtain the profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles at the preset monitoring points; Inputting the profile characteristics, driving speed and emission concentration of various types of exhaust gas of the passing vehicles into the preliminary prediction model and the correction prediction model respectively to obtain the emission rate of equivalent greenhouse gases of the passing vehicles; The global prediction module is also used to set up multiple monitoring points on the road; Obtaining the equivalent greenhouse gas emission rate of vehicles passing on the road according to the average value of the equivalent greenhouse gas emission rate of vehicles passing at different times at each of the monitoring points; Visualization module, used to obtain road space model; Marking the locations of monitoring points in the road space model; Obtain the equivalent greenhouse gas emission rate of passing vehicles; The equivalent greenhouse gas emission rate of the passing vehicles at the monitoring point is displayed in the road space model.

Images