Park access control system and method based on artificial intelligence

Abstract

The invention discloses a park access control system and method based on artificial intelligence, and the method comprises the steps: carrying out the license plate recognition and color classification of a visited vehicle through a license plate recognition model based on an artificial intelligence algorithm, obtaining a license plate number and a vehicle type, and judging the passing authority type of the visited vehicle according to the license plate number; constructing a carbon emission measuring and calculating model according to the vehicle type, and analyzing the carbon emission of the vehicle entering the park by using the carbon emission measuring and calculating model to obtain carbon emission data; according to the carbon emission data, counting the total carbon emission amount of each vehicle entering the park in a preset period, marking the vehicle of which the difference value between the total carbon emission amount of each vehicle and a preset carbon emission limit is smaller than a preset threshold value as a limited vehicle, and sending an early warning signal; and making a passing decision for an external vehicle according to the vehicle type and the carbon emission capacity of the current day. The invention relates to the technical field of deep learning, and solves the technical problem that a traditional access control system cannot accurately identify vehicle types and cannot make intelligent traffic decisions according to vehicle carbon emission.

Description

Technical Field

[0001] The present invention belongs to the field of intelligent access control, involves deep learning technology, and specifically relates to a park access control and management system and method based on artificial intelligence. Background Art

[0002] Traditional access control systems only achieve access control through license plate character recognition, and cannot distinguish vehicle types (such as fuel vehicles, electric vehicles) and carbon emission characteristics, resulting in the disorderly entry and exit of high-emission vehicles and making it difficult to meet the low-carbon management requirements.

[0003] For example, the patent application with the publication number "CN119445708A" discloses "An intelligent access control system for vehicle entry and exit based on image recognition". This application realizes license plate recognition and whitelist management through a high-definition camera and a CNN model; the patent application with the publication number "CN114283284A" discloses "A method and device for controlling access to vehicles in a community". This application proposes a method for segmenting multiple license plates adhered together and improves the recognition rate of license plates through color pixel analysis.

[0004] Therefore, the prior art usually matches the access permission of a vehicle by recognizing the license plate number, lacking a monitoring and management mechanism for vehicle carbon emissions. For example, the prior art cannot accurately calculate and count the carbon emissions of different types of vehicles, making it difficult to comprehensively grasp the total amount and distribution of vehicle carbon emissions in the park; at the same time, since the carbon emission factor is not incorporated into the vehicle access decision-making system, when deciding whether a vehicle can enter the park, it is easy to cause high-emission vehicles to enter and exit the park randomly, increasing the carbon emission burden of the park. Summary of the Invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art; for this purpose, the present invention proposes a park access control and management system and method based on artificial intelligence, which is used to solve the technical problems that the traditional access control and management system cannot accurately identify vehicle types, effectively manage vehicle carbon emissions, and make intelligent access decisions based on real-time carbon emission situations.

[0006] To achieve the above object, the first aspect of the present invention provides a park access control and management system based on artificial intelligence, including:

[0007] A license plate recognition module: used to perform license plate recognition and color classification on the visiting vehicle by using a license plate recognition model based on an artificial intelligence algorithm, obtain the license plate number and vehicle type, and judge the access permission category of the visiting vehicle according to the license plate number; wherein, the vehicle type includes electric vehicles, fuel vehicles, and hybrid vehicles, and the access permission category includes in-park vehicles, reservation-registered vehicles, and external vehicles;

[0008] Carbon emission analysis module: used to build a carbon emission calculation model according to vehicle type, use the carbon emission calculation model to analyze the carbon emissions of vehicles in the park and registered vehicles after entering the park, and obtain some carbon emission data;

[0009] Vehicle warning module: used to count the total carbon emissions of each vehicle entering the park within a preset period based on a number of carbon emission data, mark the vehicle whose total carbon emissions and the preset carbon emission limit are less than the preset threshold as a restricted vehicle, and send a warning signal to the restricted vehicle;

[0010] Traffic decision module: used to make traffic decisions for external vehicles based on vehicle type and carbon emission capacity of the day.

[0011] It should be noted that the license plate recognition module, the carbon emission analysis module and the vehicle warning module of the present invention are in communication connection, and the carbon emission analysis module is in communication connection with the vehicle warning module and the traffic decision module respectively.

[0012] Furthermore, the license plate recognition model based on artificial intelligence algorithm is used to perform license plate recognition and color classification on visiting vehicles, including:

[0013] A1, using a camera to capture an image of an incoming vehicle, and using a target detection model to locate the license plate of the image to obtain a license plate position frame; wherein the target detection model is built based on a deep learning algorithm and deployed on the camera;

[0014] A2, cropping the image of the visiting vehicle according to the license plate position frame to obtain the license plate image;

[0015] A3, using a dual-branch license plate recognition model built based on a deep learning algorithm to perform license plate recognition and color classification on license plate images, and obtain the proportion of license plate numbers and several preset color types;

[0016] A4, obtaining the vehicle type of the visiting vehicle according to the proportion of several preset color types.

[0017] Furthermore, the dual-branch license plate recognition model includes: a shared feature network, a character recognition branch and a color analysis branch; wherein,

[0018] The shared feature network is used to extract mixed features containing spatial semantic information and color distribution information from the input image to obtain character recognition-oriented features and color analysis-oriented features;

[0019] The character recognition branch is used to recognize license plate characters using character recognition guidance features to obtain the license plate number;

[0020] The color analysis branch is used to identify the background color of the license plate using the color analysis guidance feature and quantify the color ratio to obtain the proportion of several preset color types.

[0021] The license plate recognition model extracts mixed features through a shared feature extraction network, and uses a dual-branch network architecture to perform character recognition and color classification simultaneously, achieving automatic classification of vehicle types. Among them, the dual-branch network can complete the license plate character recognition and color classification tasks simultaneously, reducing the time delay of traditional step-by-step processing and significantly improving the recognition efficiency. The shared feature network splits the mixed features into character recognition-oriented features and color parsing-oriented features through a feature decoupling module, which not only preserves the spatial semantic information but also separates the color distribution information, avoiding feature interference and improving the recognition accuracy.

[0022] Furthermore, the shared feature network includes a multi-level feature extraction module and a feature decoupling module. Among them,

[0023] The multi-level feature extraction module adopts an improved ConvNeXt-Tiny architecture, including:

[0024] Stem layer: Using a 4×4 convolutional kernel, layer normalization, and GELU function to downsample the resolution of the input image to 1 / 4;

[0025] A cascaded module composed of several Stage modules, and each Stage module contains several dual-path feature retention units composed of a main path and a color retention path. Among them:

[0026] The main path includes depthwise separable convolution, channel attention mechanism, and GELU function, which are used to extract spatial semantic features;

[0027] The color retention path includes 3×3 average pooling and 1×1 convolution, which are used to compress and retain the original color information;

[0028] And the outputs of the main path and the color retention path are fused by element-wise addition to obtain the mixed features of each Stage module, and then input to the next Stage module;

[0029] The feature decoupling module is connected to the last Stage module of the multi-level feature extraction module, including:

[0030] Feature channel splitting layer: Splitting the mixed features of the last Stage module into character recognition features and color parsing features;

[0031] Channel rearrangement layer: Performing a cross-channel random permutation operation on the character recognition features and color parsing features to obtain the permuted character recognition-oriented features and color parsing-oriented features.

[0032] The multi-level feature extraction module is based on the improved ConvNeXt-Tiny architecture. Through depthwise separable convolution and channel attention mechanism, it can reduce the computational complexity while ensuring the feature extraction ability. It is suitable for deployment on edge devices such as cameras and can achieve real-time processing.

[0033] Furthermore, the character recognition branch includes a multi-scale feature fusion unit, a spatial attention enhancement module, a dynamic temporal modeling unit, and a multi-head classification output layer; among them,

[0034] The input of the multi-scale feature fusion unit comes from the output features of several Stage modules and character recognition guiding features. Through 3×3 convolution kernels and bilinear upsampling operations, the output features of several Stage modules are unified to the same resolution, and feature fusion is performed using the concatenation of channel splicing and 1×1 convolution to obtain fused features;

[0035] The spatial attention enhancement module uses the SE Block to generate the channel weight C of the fused features through global average pooling, and uses a two-dimensional sine-cosine combination function to generate the initial parameter matrix P init : P init [i,j] = sin(i / H·π) + cos(j / W·π), and the initial parameter matrix is input into a 1×1 convolutional layer and a Sigmoid function in sequence to obtain the spatial position weight matrix S. The channel weight C is multiplied element-wise with the spatial position weight matrix S to obtain the spatially attention-enhanced features; where i and j represent the row index and column index in the initial parameter matrix respectively;

[0036] The dynamic temporal modeling unit processes the spatially attention-enhanced features using a bidirectional LSTM architecture to obtain temporal features;

[0037] The multi-head classification output layer uses a fully connected network and a Softmax function to construct a digital recognition head, a letter recognition head, and a province abbreviation recognition head, and classifies and predicts the temporal features to obtain the license plate number.

[0038] The character recognition branch enhances the adaptability of the model to the size and position changes of license plate characters by fusing the output features of different Stage modules, especially having stronger recognition robustness for blurred or tilted license plates; generating a spatial position weight matrix through the SE Block to focus on the license plate character area, suppressing background noise, and reducing the misrecognition rate; while the bidirectional LSTM captures the context relationship of the character sequence, which can effectively solve the problems of character adhesion or occlusion and improve the recognition accuracy of license plate numbers in complex scenarios.

[0039] Furthermore, the color analysis branch includes a shallow feature reuse unit, a mixed color space conversion layer, a multi-receptive field classification unit, and a color statistics output layer; among them,

[0040] The shallow feature reuse unit performs channel splicing on the output feature of the first Stage module of the multi-stage feature extraction module and the color parsing guiding feature of the feature decoupling module through skip connection to obtain a spliced feature;

[0041] The hybrid color space conversion layer performs color conversion on the spliced feature by using HSV transformation subnetworks and LAB transformation subnetworks that include 3×3 convolutional kernels, batch normalization, and ReLU activation functions in parallel, and splices the output of the HSV transformation subnetworks and the output of the LAB transformation subnetworks with the RGB three channels of the input image to obtain a hybrid color feature; among them, the HSV transformation subnetworks are optimized by using a loss function that minimizes the hue channel classification error, and the LAB transformation subnetworks are optimized by using a loss function that minimizes the lightness channel regression error;

[0042] The multi-receptive field classification unit processes the hybrid color feature by using a dilated convolutional pyramid structure, and inputs the processed feature into a 1×1 convolutional layer for dimensionality reduction processing after channel splicing to obtain a color probability map; where the shape of the color probability map is [B, N, H, W], B represents the batch size, N represents the preset number of color categories, and H and W are the length and width of the input image respectively;

[0043] The color statistics output layer includes:

[0044] Calculating along the category dimension N of the color probability map by using the Softmax function to obtain the color category probability distribution M of each pixel point c,i,j ; where c represents the color category, and i and j represent the row index and column index of the pixel point respectively;

[0045] Taking the mean value of the output feature of the first Stage module of the multi-stage feature extraction module along the channel dimension to obtain a single-channel feature map F;

[0046] Inputting the single-channel feature map F into a 1×1 convolution and a Sigmoid function in sequence to generate a spatial weight matrix S’, and c,i,j Performing element-wise multiplication and summing along the spatial dimension of the color category probability distribution M and the spatial weight matrix S’ in sequence to obtain the weighted probability sum P of each category c ;

[0047] Normalizing the weighted probability sum P c to obtain the proportions of several preset color types.

[0048] The color analysis branch uses HSV and LAB color space conversions in parallel, optimizing hue classification and lightness regression respectively, which can enhance the accuracy of color classification and reduce the influence of light changes. At the same time, the dilated convolutional pyramid structure can capture color distribution features at different scales, helping the model adapt to scenarios of gradual change or local occlusion of the license plate background color.

[0049] Further, obtaining the vehicle type of the visiting vehicle according to the proportion of several preset color types includes:

[0050] Marking a vehicle with a green proportion greater than a first preset threshold, or a vehicle with a green proportion greater than a second preset threshold and a yellow proportion greater than a third preset threshold as an electric vehicle;

[0051] Marking a vehicle with a green proportion greater than a fourth preset threshold and a non-zero black proportion or a non-zero yellow proportion as a hybrid vehicle;

[0052] Marking a vehicle with a blue proportion greater than a fifth preset threshold as a small fuel vehicle;

[0053] Marking a vehicle with a yellow proportion greater than a sixth preset threshold as a large fuel vehicle.

[0054] Judging the vehicle type by setting thresholds for different color proportions is simple, direct and scientific. Without additional sensors or complex detection means, the vehicle type can be initially judged only relying on the license plate color information, reducing the system cost and complexity.

[0055] Further, constructing a carbon emission measurement model according to the vehicle type includes:

[0056] Constructing a carbon emission measurement model for fuel vehicles as: C fuel = D × E base × (1 + αt emp ) + β × t idle ; where D represents the driving distance in the park, E base represents the baseline carbon emission coefficient, αt emp represents the temperature correction factor, β represents the idle emission rate, and t idle represents the idle time;

[0057] Constructing a carbon emission measurement model for electric vehicles as: C elec = E charge / η charge × EF grid × (1 + α loss ) ; where E charge represents the charging power, η charge represents the charging efficiency of the charging pile, EF grid represents the grid emission factor, obtained through the national grid regional emission factor database, αloss It represents the transmission loss compensation coefficient, which is obtained from the line loss rate report of the State Grid;

[0058] The carbon emission measurement model of the hybrid vehicle is constructed as: C hybrid = γ × C fuel + × C elec ; where γ represents the fuel drive weight, which is obtained according to the formula m represents the calibration coefficient, and E0 represents the preset power threshold.

[0059] Furthermore, the obtaining methods of the reference carbon emission coefficient and the idle emission rate include:

[0060] When the visiting vehicle is a vehicle in the park or a reserved registration vehicle, it is obtained through the built-in vehicle information database; where the built-in vehicle information database includes license plate number, vehicle type, carbon emission standard, reference carbon emission coefficient, idle carbon emission rate, preset carbon emission limit, and vehicle photos;

[0061] When the visiting vehicle is an external vehicle, according to the vehicle type obtained by the license plate recognition model, several vehicle photos of the same vehicle type are screened out from the built-in information database, and the similarity between the several vehicle photos and the image of the visiting vehicle is calculated, and the reference carbon emission coefficient and the idle carbon emission rate of the vehicle photo with the largest similarity are selected as the reference carbon emission coefficient and the idle carbon emission rate of the external vehicle.

[0062] Furthermore, the calculation of the similarity between several vehicle photos and the image of the visiting vehicle includes:

[0063] Using the pre-trained vehicle recognition model to extract the features of several vehicle photos and the image of the visiting vehicle respectively, obtaining several stored vehicle features and the features of the visiting vehicle; where the pre-trained vehicle recognition model is constructed based on the deep learning algorithm, and the vehicle photos and the image of the visiting vehicle are obtained by the cameras at the same location;

[0064] Using the cosine similarity function to calculate the similarity between several stored vehicle features and the features of the visiting vehicle.

[0065] Separate carbon emission measurement models are constructed for fuel vehicles, electric vehicles, and hybrid vehicles, considering the carbon emission characteristics of different types of vehicles, making the measurement results more in line with the actual situation; and the carbon emission parameters of external vehicles are matched through vehicle image similarity, which can solve the problem of missing data of external vehicles, so as to make more accurate communication decisions.

[0066] Furthermore, the making of the traffic decision for external vehicles according to the vehicle type and the carbon emission capacity on the day includes:

[0067] D1, Determine whether the vehicle type is an electric vehicle; if so, send an allow - passage control signal; if not, jump to D2;

[0068] D2, According to a number of carbon emission data, count the total carbon emissions in the park within a preset period to obtain the used carbon emissions;

[0069] D3, According to the carbon emission measurement model, count the predicted carbon emissions of the vehicles that did not leave the park the previous day in the park from the parking position to the section where they leave the park to obtain the reserved carbon emissions;

[0070] D4, Subtract the used carbon emissions and the reserved carbon emissions from the carbon emission limit in the park within the preset period in turn, and then divide by the remaining days in the preset period to obtain the estimated daily available carbon emissions;

[0071] D5, According to a number of carbon emission data, count the total carbon emissions of the day, and compare whether the difference between the estimated daily available carbon emissions and the total carbon emissions of the day is greater than the preset reserved threshold; if so, send an allow - passage control signal; if not, send a prohibit - passage control signal.

[0072] The decision - making process comprehensively considers factors such as vehicle type and the daily carbon emission capacity, and formulates reasonable access rules. Electric vehicles are directly allowed to pass, which reflects the support for environmental - friendly vehicles; for fuel vehicles, by calculating the used carbon emissions, reserved carbon emissions and the estimated daily available carbon emissions, and comparing with the total carbon emissions of the day, it decides whether to allow passage, which can effectively control the total carbon emissions in the park and improve the intelligent and green level of access control management.

[0073] The second aspect of the present invention provides a park access control method based on artificial intelligence, including:

[0074] S1, Use a license - plate recognition model based on artificial intelligence algorithms to perform license - plate recognition and color classification on the visiting vehicle, obtain the license - plate number and vehicle type, and judge the access - permission category of the visiting vehicle according to the license - plate number; among them, the vehicle type includes electric vehicles, fuel vehicles and hybrid vehicles, and the access - permission category includes in - park vehicles, reservation - registered vehicles and external vehicles;

[0075] S2, Construct a carbon emission measurement model according to the vehicle type, and use the carbon emission measurement model to analyze the carbon emissions after the in - park vehicles and reservation - registered vehicles enter the park to obtain a number of carbon emission data;

[0076] S3, According to a number of carbon emission data, count the total single - vehicle carbon emissions of each vehicle entering the park within a preset period, mark the vehicles whose difference between the total single - vehicle carbon emissions and the preset carbon emission limit is less than the preset threshold as restricted vehicles, and send a warning signal to the restricted vehicles;

[0077] S4. Make a traffic decision for foreign vehicles based on the vehicle type and the daily carbon emission capacity.

[0078] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0079] Traditional access control systems rely on single license plate character recognition, have weak adaptability to complex environments, and cannot distinguish vehicle types. The present invention realizes parallel processing of license plate character and color classification through a dual-branch network architecture. Among them, the shared feature network fuses spatial semantics and color information. The character recognition branch adopts multi-scale feature fusion and temporal modeling techniques to solve the problems of character adhesion and occlusion. The color analysis branch accurately quantifies the proportion of license plate colors through hybrid color space conversion and dilated convolutional pyramid. And automatically classify vehicle types according to the proportion of license plate colors, providing a reliable data basis for carbon emission management.

[0080] Different from traditional systems that only control the entry and exit of vehicles, the present invention constructs a differential carbon emission measurement model. Among them, the fuel vehicle model introduces a temperature correction factor and an idle emission rate, and the electric vehicle model combines the grid emission factor and transmission loss to achieve full-life cycle carbon accounting. Match the parameters of foreign vehicles through vehicle image similarity to solve the problem of data loss. Dynamically mark and warn over-limit vehicles by real-time statistical total carbon emissions of single vehicles. At the same time, combine the prediction of the vehicles that did not leave the park the previous day from the parking position to the exit, reserve carbon emission quotas, and avoid waste of resources. On this basis, the traffic decision-making module gives priority to releasing electric vehicles and intelligently adjusts the traffic permissions of foreign fuel vehicles according to the remaining daily carbon emission capacity to achieve effective control of carbon emissions. Description of the Drawings

[0081] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0082] Figure 1 It is a flow chart of a park access control system based on artificial intelligence provided by the present invention;

[0083] Figure 2 It is a framework diagram of a park access control system based on artificial intelligence provided by the present invention;

[0084] Figure 3 It is an architecture diagram of a license plate recognition model based on artificial intelligence algorithm provided by the present invention. Detailed Embodiments

[0085] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. 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.

[0086] Please refer to Figures 1 - 3 , an embodiment of the first aspect of the present invention provides an artificial intelligence-based park access control system, including:

[0087] License plate recognition module: used to perform license plate recognition and color classification on the visiting vehicle using a license plate recognition model based on artificial intelligence algorithms, obtain the license plate number and vehicle type, and judge the access permission category of the visiting vehicle according to the license plate number; wherein, the vehicle type includes electric vehicles, fuel vehicles, and hybrid vehicles, and the access permission category includes in-park vehicles, reservation-registered vehicles, and external vehicles;

[0088] Carbon emission analysis module: used to construct a carbon emission measurement model according to the vehicle type, and use the carbon emission measurement model to analyze the carbon emissions of in-park vehicles and reservation-registered vehicles after entering the park, and obtain a number of carbon emission data;

[0089] Vehicle warning module: used to count the total carbon emissions of each vehicle entering the park within a preset period according to a number of carbon emission data, mark the vehicle with the difference between the total carbon emissions of the vehicle and the preset carbon emission limit less than the preset threshold as a restricted vehicle, and send a warning signal to the restricted vehicle;

[0090] Access decision-making module: used to make access decisions for external vehicles according to the vehicle type and the current day's carbon emission capacity.

[0091] In the existing access control system, a target detection model is mostly used for license plate position localization, then a convolutional neural network is used for character recognition, and then the access permission of the vehicle is judged according to the recognized license plate number. When encountering an external vehicle, on-site registration is usually adopted and then it is allowed to enter. However, the carbon emissions in the park not only come from industrial production and power consumption, but also from the exhaust gas of vehicles entering and leaving at will. The traditional system cannot automatically distinguish vehicle types (such as fuel vehicles and electric vehicles), and it is difficult to implement differential environmental protection control strategies. In order to achieve refined carbon emission management, the license plate recognition module of the present invention uses a dual-branch network architecture to add a vehicle type classification function on the basis of traditional license plate recognition, so as to provide data support for subsequent carbon emission measurement and access decision-making. Specifically,

[0092] In this embodiment, a dual-branch deep learning network is adopted to synchronously realize license plate character recognition and license plate background color classification, and its core architecture is as follows:

[0093] (1) Shared Feature Network: It includes a multi-level feature extraction module and a feature decoupling module; among them,

[0094] The multi-level feature extraction module is based on the improved ConvNeXt-Tiny architecture, and includes a Stem layer composed of a 4×4 convolutional kernel, layer normalization, and GELU function (downsampling the resolution of the input image to 1 / 4) and a feature retention cascade module composed of several Stage modules; and each Stage module consists of two parallel paths:

[0095] Main path: Use depthwise separable convolution to extract spatial semantic features (such as license plate character contours);

[0096] Color retention path: Compress the original color information (such as license plate background color) through 3×3 average pooling and 1×1 convolution;

[0097] And the outputs of the main path and the color retention path are feature fused by element-wise addition to obtain the mixed features of each Stage module, and then input to the next Stage module;

[0098] The feature decoupling module splits the output mixed features of the last Stage module of the multi-level feature extraction module into character recognition features and color parsing features through a channel splitting layer, and randomly permutes the channels of the character recognition features and color parsing features using a channel rearrangement layer to eliminate the coupling interference between tasks, obtaining the permuted character recognition-oriented features and color parsing-oriented features, which are used as the input features of the two branches respectively;

[0099] The shared feature network adopts a two-stage architecture of shared feature extraction and task decoupling, and realizes the multi-path retention of underlying features through the improved DualPathBlock of ConvNeXt. While extracting general semantic features in the multi-level feature extraction module, the original color distribution information is retained through an independent color path. The feature decoupling layer uses channel splitting and rearrangement techniques to separate the mixed features into character recognition-oriented features and color parsing features, which can reduce the amount of repeated calculations compared with traditional single-task models, and enable high-level semantic features to focus on character structure recognition and shallow detail features to support color analysis, realizing implicit feature interaction while ensuring the independence of the two tasks;

[0100] (2) Character Recognition Branch: It includes a multi-scale feature fusion unit, a spatial attention enhancement module, a dynamic temporal modeling unit, and a multi-head classification output layer; among them,

[0101] The input of the multi-scale feature fusion unit comes from the output features of several Stage modules and the character recognition guiding features. Through the 3×3 convolutional kernel and bilinear upsampling operation, the output features of several Stage modules are unified to the same resolution, and feature fusion is performed by the cascading method of channel concatenation and 1×1 convolution to obtain the fused features;

[0102] The spatial attention enhancement module uses the SEBlock to generate the channel weight C of the fused features through global average pooling, and uses the two-dimensional sine-cosine combination function to generate the initial parameter matrix P init : P init [i,j] = sin(i / H·π)+cos(j / W·π), and then the initial parameter matrix is input into the 1×1 convolutional layer and the Sigmoid function in sequence to obtain the spatial position weight matrix S. The channel weight C is multiplied element by element with the spatial position weight matrix S to obtain the spatially attention-enhanced features; where i and j represent the row index and column index of the initial parameter matrix respectively;

[0103] The dynamic time series modeling unit uses the bidirectional LSTM architecture to process the spatially attention-enhanced features to obtain the time series features;

[0104] The multi-head classification output layer uses a fully connected network and the Softmax function to construct a digital recognition head (including a 10-class fully connected layer, corresponding to the numbers 0-9), a letter recognition head (including a 24-class fully connected layer, excluding the letters O / I that are easily confused with numbers), and a provincial abbreviation recognition head (including a 31-class fully connected layer, corresponding to the abbreviation characters of each province), and classifies and predicts the time series features to obtain the license plate number;

[0105] The character recognition branch uses the spatial attention enhancement module and dynamic LSTM sequence modeling, and embeds the license plate format rules using the classification prediction head (province / letter / number), which can solve the recognition problem of easily confused characters such as "0 / O";

[0106] (3) Color analysis branch: including a shallow feature reuse unit, a hybrid color space conversion layer, a multi-receptive field classification unit, and a color statistics output layer; where,

[0107] The shallow feature reuse unit reuses the high-resolution output features of the multi-level feature extraction module, that is, the output features of the first Stage module, and performs channel concatenation on this feature and the color analysis guiding features of the feature decoupling module through a skip connection to obtain the concatenated features;

[0108] The hybrid color space conversion layer parallelly sets up an HSV transformation sub-network and an LAB transformation sub-network. Each sub-network contains a cascaded structure of a 3×3 convolutional kernel, batch normalization, and a ReLU activation function. The HSV transformation sub-network and the LAB transformation sub-network are optimized respectively by minimizing the loss function of the hue channel classification error and the loss function of the lightness channel regression error. Then, the output features of the two sub-networks are concatenated with the RGB 3 channels of the input image to obtain hybrid color features: original RGB (3 channels) + HSV (3 channels) + LAB (3 channels) = 9-channel hybrid features;

[0109] The multi-receptive field classification unit uses an atrous convolutional pyramid structure and processes the hybrid color features with 3×3 convolutions with different dilation rates to expand the receptive field to cover the color gradient region. Then, the processed features are input into a 1×1 convolutional layer for dimensionality reduction after channel concatenation to obtain a color probability map. The shape of the color probability map is [B, N, H, W], where B represents the batch size, N represents the preset number of color categories, and H and W represent the length and width of the input image respectively;

[0110] The color statistics output layer includes:

[0111] Calculate along the category dimension N of the color probability map using the Softmax function to obtain the color category probability distribution M of each pixel point c,i,j ;

[0112] Then, take the mean of the output features of the first Stage module of the multi-level feature extraction module along the channel dimension to obtain a single-channel feature map F;

[0113] Input the single-channel feature map F into a 1×1 convolution and a Sigmoid function in sequence to generate a spatial weight matrix S'; i,j and multiply the color category probability distribution M c,i,j and the spatial weight matrix S' i,j element-wise and sum along the spatial dimension in sequence to obtain the weighted probability sum P for each category c , and its calculation formula is: where c represents the color category, i and j represent the row index and column index in the color probability map respectively, and h and w represent the height and width of the color probability map;

[0114] Normalize the weighted probability sum P c to obtain the proportions of several preset color types; where N represents the preset number of color categories;

[0115] ​The color analysis branch utilizes the HSV-LAB hybrid space conversion layer to learn the non-linear color mapping through a two-stream convolutional network, and combines the dilated convolutional pyramid to capture the gradient features, which can reduce the color recognition error rate in rainy and cloudy weather compared with the traditional RGB space method; moreover, the introduction of the spatial weighted statistics layer can quantitatively analyze the area ratio of the gradient colors of new energy vehicle license plates, which is more convenient for vehicle type classification.

[0116] In this embodiment, a two-branch deep learning network is built using the pytorch framework, and the CCPD license plate detection dataset (including license plate position annotations), the CRPD character recognition dataset (including character sequence labels), and several artificially annotated license plate color classification datasets are used as experimental data. The robustness of the model is improved through enhancement strategies such as random rotation, illumination perturbation, and occlusion simulation.

[0117] When the two-branch tasks are jointly trained, the AdamW optimizer and the cosine annealing scheduling strategy are adopted. The CTC loss function (Connectionist Temporal Classification) is used for the character recognition branch, and the FocalLoss loss function is used for the color analysis branch to solve the problem of color class imbalance.

[0118] After training is completed, the obtained license plate recognition model will compress the model volume through channel pruning and INT8 quantization, and finally be deployed to a high-definition camera equipped with an NVIDIA Jetson Xavier NX embedded module to achieve real-time single-frame processing.

[0119] In the license plate recognition module, the intelligent recognition of vehicle identity and type is realized through a multi-stage deep learning algorithm. The specific steps are as follows:

[0120] A1, Use a camera to capture an image of the visiting vehicle, and use the object detection model to locate the license plate in the image to obtain the license plate position box; among them, the object detection model is built based on the deep learning algorithm and deployed on the camera.

[0121] A2, Crop the image of the visiting vehicle according to the license plate position box to obtain the license plate image.

[0122] A3, Use the two-branch license plate recognition model built based on the deep learning algorithm to perform license plate recognition and color classification on the license plate image to obtain the license plate number and the proportion of several preset color types.

[0123] A4, Automatically classify the vehicle type based on the color proportion threshold rule:

[0124] Vehicles with a green proportion greater than the first preset threshold, or vehicles with a green proportion greater than the second preset threshold and a yellow proportion greater than the third preset threshold are marked as electric vehicles.

[0125] Vehicles with a green proportion greater than the fourth preset threshold and a non - zero black proportion or a non - zero yellow proportion are marked as hybrid vehicles;

[0126] Vehicles with a blue proportion greater than the fifth preset threshold are marked as small fuel - powered vehicles;

[0127] Vehicles with a yellow proportion greater than the sixth preset threshold are marked as large fuel - powered vehicles.

[0128] In this embodiment, the color proportion threshold rules can be as follows:

[0129] Electric vehicles: green proportion > 90% (pure electric), or green > 60% and yellow > 20% (large new - energy vehicles);

[0130] Hybrid vehicles: green > 70% and there is black / yellow (hybrid or plug - in hybrid);

[0131] Small fuel - powered vehicles: blue > 90% (ordinary blue license plate);

[0132] Large fuel - powered vehicles: yellow > 80% (yellow license plate trucks / buses).

[0133] In the carbon emission analysis module, it is used to construct a carbon emission measurement model according to the vehicle type, which is used as a statistical tool for the carbon emissions of vehicles in the park. Specifically, it includes:

[0134] The carbon emission measurement model for fuel - powered vehicles is constructed as: C fuel = D×E base ×(1 + αt emp )+β×t idle ; where D represents the driving distance in the park, E base represents the baseline carbon emission coefficient, αt emp represents the temperature correction factor, β represents the idle emission rate, and t idle represents the idle time;

[0135] The carbon emission measurement model for electric vehicles is constructed as: C elec = E charge / η charge ×EF grid ×(1 + α loss ); where E charge represents the charging power, η charge represents the charging efficiency of the charging pile, EF grid represents the grid emission factor, obtained through the national grid regional emission factor database, and α loss represents the transmission loss compensation coefficient, obtained through the national grid line loss rate report;

[0136] The carbon emission measurement model for hybrid vehicles is constructed as: Chybrid = γ × C fuel + × C elec ; where γ represents the fuel drive weight, obtained according to the formula obtained, m represents the calibration coefficient, obtained through practical experience, and E0 represents the preset power threshold;

[0137] Among them, the methods for obtaining the baseline carbon emission coefficient and the idle emission rate include:

[0138] When the visiting vehicle is a vehicle in the park or a vehicle with reservation registration, it is obtained through the built-in vehicle information database; among them, the built-in vehicle information database includes license plate number, vehicle type, carbon emission standard, baseline carbon emission coefficient, idle carbon emission rate, preset carbon emission limit, and vehicle photos;

[0139] When the visiting vehicle is an external vehicle, according to the vehicle type obtained by the license plate recognition model, several vehicle photos of the same vehicle type are screened out from the built-in information database,

[0140] Then, the pre-trained vehicle recognition model is used to extract the features of several vehicle photos and the image of the visiting vehicle respectively, obtaining several stored vehicle features and visiting vehicle features; among them, the pre-trained vehicle recognition model is constructed based on the deep learning algorithm, and in order to ensure that the vehicles in the photos are as much as possible at the same angle, both the vehicle photos and the image of the visiting vehicle are obtained from the camera at the same position;

[0141] Next, the cosine similarity function is used to calculate the similarity between several stored vehicle features and visiting vehicle features, and the baseline carbon emission coefficient and the idle emission rate of the vehicle photo with the largest similarity are selected as the baseline carbon emission coefficient and the idle emission rate of the external vehicle.

[0142] After each vehicle enters the park, the carbon emission analysis module can obtain the driving path of the vehicle in the park in real time through the UWB ultra-wideband positioning system, generate dynamic trajectory data in combination with the digital twin map, and calculate the accurate driving distance; or use the in-park camera to obtain the multi-view video stream of the vehicle, and use the license plate number to achieve cross-camera trajectory matching of the same vehicle, form a complete driving path, and obtain the driving distance;

[0143] For new energy vehicles, the module can conduct data interaction with the in-vehicle BMS system through the charging pile data interface, read the charging power of the vehicle in the park in real time, and synchronously obtain the charging efficiency parameters of the charging pile; for fuel vehicles, the idle time can be collected through the in-vehicle OBD interface, or the vehicle can be located through the target detection model of the in-park camera, and the time when the vehicle is in a stationary state and there is continuous exhaust emission is counted to obtain the idle time;

[0144] Then, the carbon emission output of each vehicle is obtained according to the carbon emission measurement model, stored in the carbon emission database, and the carbon emission data of the vehicles in the park and the reservation registration vehicles are associated with the vehicle information database to obtain the carbon footprint ledger of the vehicles, which is then transmitted to the vehicle warning module.

[0145] In the vehicle warning module, according to a number of carbon emission data, the total carbon emission of each vehicle entering the park within a preset period is counted. Vehicles with the difference between the total carbon emission of a single vehicle and the preset carbon emission limit less than a preset threshold (the preset threshold can be set to 10% of the preset carbon emission limit) are marked as restricted vehicles, and a warning signal is sent to the restricted vehicles.

[0146] For foreign vehicles, the access decision module is used to determine their access rights according to the vehicle type and the daily carbon emission capacity. Specifically, it may include:

[0147] D1, Determine whether the vehicle type is an electric vehicle; if so, send an allow access control signal; if not, jump to D2.

[0148] D2, According to a number of carbon emission data, count the total carbon emission in the park within the preset period to obtain the used carbon emissions.

[0149] D3, According to the carbon emission measurement model, count the predicted carbon emissions of the vehicles that did not leave the park the previous day in the park from the parking position to the section out of the park to obtain the reserved carbon emissions.

[0150] D4, After subtracting the used carbon emissions and the reserved carbon emissions from the park carbon emission limit within the preset period in turn, divide by the remaining days within the preset period to obtain the predicted daily available carbon emissions.

[0151] D5, According to a number of carbon emission data, count the total carbon emission of the day, and compare whether the difference between the predicted daily available carbon emissions and the total carbon emission of the day is greater than the preset reserved threshold; if so, send an allow access control signal; if not, send a prohibit access control signal.

[0152] The second aspect of the embodiments of the present invention provides a method for controlling the access of a park based on artificial intelligence, including:

[0153] S1, Use the license plate recognition model based on the artificial intelligence algorithm to perform license plate recognition and color classification on the visiting vehicle to obtain the license plate number and vehicle type, and determine the access right category of the visiting vehicle according to the license plate number; among them, the vehicle type includes electric vehicles, fuel vehicles, and hybrid vehicles, and the access right category includes vehicles in the park, reservation registration vehicles, and foreign vehicles.

[0154] S2, Construct a carbon emission measurement model according to the vehicle type, and use the carbon emission measurement model to analyze the carbon emissions of the vehicles in the park and the reservation registration vehicles after entering the park to obtain a number of carbon emission data.

[0155] S3. According to a number of carbon emission data, calculate the total carbon emission of each vehicle entering the park within a preset period. Mark the vehicles whose difference between the total carbon emission of a single vehicle and the preset carbon emission limit is less than the preset threshold as restricted vehicles, and send a warning signal to the restricted vehicles;

[0156] S4. Make a traffic decision for foreign vehicles according to the vehicle type and the daily carbon emission capacity.

[0157] Some of the data in the above formula are calculated by removing the dimension and taking their numerical values. The formula is obtained by software simulation of a large amount of collected data to get a formula closest to the actual situation; the preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0158] The working principle of the present invention:

[0159] The present invention first uses the license plate recognition module to collect vehicle images through a camera, and uses the improved ConvNeXt-Tiny architecture and the dual-branch network to perform character recognition and color analysis of the license plate number, and then combines the license plate color ratio threshold to complete vehicle classification;

[0160] Subsequently, the carbon emission analysis module constructs a differential model according to the vehicle type and dynamically calculates the carbon emission data of the vehicles entering the park;

[0161] Then the warning module calculates the total carbon emission of a single vehicle within a preset period and sends a warning to the vehicles approaching the limit; the traffic decision module implements hierarchical control for foreign vehicles: electric vehicles can be directly released, and fuel vehicles are dynamically evaluated through the remaining daily carbon emission capacity. If the remaining amount meets the standard, they are allowed to pass.

[0162] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An intelligent park access control system based on artificial intelligence, characterized in that, Including: License plate recognition module: It is used to perform license plate recognition and color classification on the visiting vehicle by using a dual-branch license plate recognition model based on artificial intelligence algorithms, obtain the license plate number and vehicle type, and judge the access permission category of the visiting vehicle according to the license plate number; wherein, the vehicle type includes electric vehicles, fuel vehicles, and hybrid vehicles, and the access permission category includes in-park vehicles, reservation-registered vehicles, and external vehicles; Carbon emission analysis module: It is used to construct a carbon emission measurement model according to the vehicle type, and use the carbon emission measurement model to analyze the carbon emissions of in-park vehicles and reservation-registered vehicles after entering the park, and obtain a number of carbon emission data; Vehicle warning module: It is used to calculate the total single-vehicle carbon emissions of each vehicle entering the park within a preset period according to a number of carbon emission data, mark the vehicle whose difference between the total single-vehicle carbon emissions and the preset carbon emission limit is less than the preset threshold as a restricted vehicle, and send a warning signal to the restricted vehicle; Access decision-making module: It is used to make access decisions for external vehicles according to the vehicle type and the daily carbon emission capacity.

2. The access control system for a park based on artificial intelligence according to claim 1, wherein The use of the license plate recognition model based on artificial intelligence algorithms to perform license plate recognition and color classification on the visiting vehicle includes: A1. Using a camera to capture an image of the visiting vehicle, and using an object detection model to perform license plate positioning on the image to obtain a license plate position box; wherein, the object detection model is constructed based on a deep learning algorithm and deployed on the camera; A2. Cropping the image of the visiting vehicle according to the license plate position box to obtain a license plate image; A3. Using the dual-branch license plate recognition model to perform license plate recognition and color classification on the license plate image to obtain the license plate number and the proportion of several preset color types; A4. Obtaining the vehicle type of the visiting vehicle according to the proportion of several preset color types.

3. An access control system for a park based on artificial intelligence according to claim 2, characterized in that The dual-branch license plate recognition model includes: a shared feature network, a character recognition branch, and a color analysis branch; wherein, The shared feature network is used to extract a mixed feature containing spatial semantic information and color distribution information from the input image, and obtain a character recognition-oriented feature and a color analysis-oriented feature; The character recognition branch is used to recognize license plate characters by using the character recognition-oriented feature to obtain the license plate number; The color analysis branch is used to recognize the license plate background color by using the color analysis-oriented feature and quantitatively count the color ratio to obtain the proportion of several preset color types.

4. An access control system for a park based on artificial intelligence according to claim 3, characterized in that, The shared feature network includes a multi-level feature extraction module and a feature decoupling module; wherein, The multi-level feature extraction module adopts an improved ConvNeXt-Tiny architecture, including: Stem layer: A 4×4 convolutional kernel, layer normalization, and GELU function are used to downsample the resolution of the input image to 1 / 4; A cascade module composed of several Stage modules, and each Stage module contains several dual-path feature retention units composed of a main path and a color retention path, where: The main path includes a depthwise separable convolution, a channel attention mechanism, and a GELU function, and is used to extract spatial semantic features; The color retention path includes 3×3 average pooling and 1×1 convolution, and is used to compress and retain the original color information; Moreover, the outputs of the main path and the color retention path are subjected to feature fusion through element-wise addition to obtain the mixed features of each Stage module, which are then input into the next Stage module; The feature decoupling module is connected to the last Stage module of the multi-stage feature extraction module and includes: Feature channel splitting layer: splitting the mixed features of the last Stage module into character recognition features and color parsing features; Channel rearrangement layer: performing a cross-channel random permutation operation on the character recognition features and the color parsing features to obtain the permuted character recognition-oriented features and color parsing-oriented features.

5. The access control system for a park based on artificial intelligence according to claim 4, wherein The character recognition branch includes a multi-scale feature fusion unit, a spatial attention enhancement module, a dynamic temporal modeling unit, and a multi-head classification output layer; among them, The input of the multi-scale feature fusion unit comes from the output features of several Stage modules and the character recognition-oriented features. Through 3×3 convolutional kernels and bilinear upsampling operations, the output features of several Stage modules are unified to the same resolution, and feature fusion is performed by cascading channel concatenation and 1×1 convolutions to obtain fused features; The spatial attention enhancement module utilizes the SE Block to generate the channel weights C of the fused features through global average pooling, and generates the initial parameter matrix P using a two-dimensional sine-cosine combination function init : P init [i, j] = sin(i / H · π) + cos(j / W · π). The initial parameter matrix is successively input into a 1×1 convolutional layer and the Sigmoid function to obtain the spatial position weight matrix S. The channel weights C are multiplied element-wise with the spatial position weight matrix S to obtain the spatially attention-enhanced features. Here, i and j represent the row index and column index in the initial parameter matrix, respectively The dynamic temporal modeling unit processes the spatially attention-enhanced features using a bidirectional LSTM architecture to obtain temporal features; The multi-head classification output layer constructs a digital recognition head, a letter recognition head, and a province abbreviation recognition head using a fully connected network and a Softmax function, and classifies and predicts the temporal features to obtain the license plate number.

6. The access control system for a park based on artificial intelligence according to claim 4, wherein The color parsing branch includes a shallow feature reuse unit, a mixed color space conversion layer, a multi-receptive field classification unit, and a color statistics output layer; among them, The shallow feature reuse unit performs channel concatenation on the output features of the first Stage module of the multi-stage feature extraction module and the color parsing-oriented features that are skip-connected to the feature decoupling module to obtain concatenated features; The mixed color space conversion layer uses an HSV transformation sub-network and an LAB transformation sub-network that are parallel including 3×3 convolutional kernels, batch normalization, and ReLU activation functions to perform color conversion on the concatenated features, and concatenates the outputs of the HSV transformation sub-network and the LAB transformation sub-network with the RGB3 channels of the input image to obtain mixed color features; among them, the HSV transformation sub-network is optimized using a loss function that minimizes the classification error of the hue channel, and the LAB transformation sub-network is optimized using a loss function that minimizes the regression error of the lightness channel; The multi-receptive field classification unit processes the mixed color features using a dilated convolutional pyramid structure, and inputs the processed features into a 1×1 convolutional layer for dimensionality reduction after channel concatenation to obtain a color probability map; where the shape of the color probability map is [B, N, H, W], B represents the batch size, N represents the preset number of color categories, and H and W are the length and width of the input image respectively; The color statistics output layer includes: The Softmax function is used to calculate along the category dimension N of the color probability map, and the color category probability distribution M of each pixel is obtained c,i,j ; where c represents the color category, and i and j respectively represent the row index and column index of the pixel Taking the mean of the output features of the first Stage module of the multi-stage feature extraction module along the channel dimension to obtain a single-channel feature map F; The single-channel feature map F is sequentially input into a 1×1 convolution and a Sigmoid function to generate a spatial weight matrix S’, and the color category probability distribution M c,i,j is sequentially element-wise multiplied with the spatial weight matrix S’ and summed along the spatial dimension to obtain the weighted probability sum P for each category c ; Normalize the weighted probability sum P c to obtain the proportions of several preset color types.

7. The park access control system based on artificial intelligence according to claim 1, characterized in that, The carbon emission measurement model constructed according to the vehicle type includes: The carbon emission measurement model for a fuel vehicle is: C fuel = D × E base × (1 + αt emp ) + β × t idle ; where D represents the driving distance within the park, E base represents the baseline carbon emission factor, αt emp represents the temperature correction factor, β represents the idling emission rate, and t idle represents the idling time; The carbon emission measurement model for electric vehicles is: C elec = E charge / η charge × EF grid × (1 + α loss ); where E charge represents the charging power, η charge represents the charging efficiency of the charging pile, EF grid represents the grid emission factor, obtained from the national grid regional emission factor database, and α loss represents the transmission loss compensation coefficient, obtained from the national grid line loss rate report; The carbon emission measurement model for a hybrid vehicle is: C hybrid = γ × C fuel + × C elec ; where γ represents the fuel drive weight, obtained according to the formula ; m represents the calibration coefficient, and E0 represents the preset power threshold.

8. The access control system for a park based on artificial intelligence according to claim 7, characterized in that, The methods for obtaining the reference carbon emission coefficient and the idle emission rate include: When the visiting vehicle is a vehicle within the park or a pre-registered vehicle, it is obtained through the built-in vehicle information database; wherein, the built-in vehicle information database includes license plate number, vehicle type, carbon emission standard, reference carbon emission coefficient, idle carbon emission rate, preset carbon emission limit, and vehicle photos; When the visiting vehicle is an external vehicle, according to the vehicle type obtained by the license plate recognition model, several vehicle photos of the same vehicle type are screened out from the built-in information database, and the cosine similarity function is used to calculate the similarity between the several vehicle photos and the image of the visiting vehicle, and the reference carbon emission coefficient and the idle carbon emission rate of the vehicle photo with the largest similarity are selected as the reference carbon emission coefficient and the idle carbon emission rate of the external vehicle.

9. The access control system for a park based on artificial intelligence according to claim 1, wherein, Making a traffic decision for external vehicles according to the vehicle type and the daily carbon emission capacity includes: D1, determining whether the vehicle type is an electric vehicle; if so, sending an allow-pass control signal; if not, jumping to D2; D2, according to several carbon emission data, statistically calculating the total carbon emissions in the park within a preset period to obtain the used carbon emissions; D3, according to the carbon emission measurement model, statistically calculating the predicted carbon emissions of the vehicles that did not leave the park the previous day in the park from the parking position to the section out of the park to obtain the reserved carbon emissions; D4, after successively subtracting the used carbon emissions and the reserved carbon emissions from the park carbon emission limit within the preset period, dividing by the remaining days within the preset period to obtain the estimated daily available carbon emissions; D5, according to several carbon emission data, statistically calculating the total carbon emissions on the day, and comparing whether the difference between the estimated daily available carbon emissions and the total carbon emissions on the day is greater than the preset reserved threshold; if so, sending an allow-pass control signal; if not, sending a prohibit-pass control signal.

10. A method for controlling access to a park based on artificial intelligence, which is applied to a system for controlling access to a park based on artificial intelligence according to any one of claims 1-9, and is characterized in that, Including: S1, using a license plate recognition model based on an artificial intelligence algorithm to perform license plate recognition and color classification on the visiting vehicle to obtain the license plate number and vehicle type, and judging the traffic permission category of the visiting vehicle according to the license plate number; wherein, the vehicle type includes electric vehicles, fuel vehicles, and hybrid vehicles, and the traffic permission category includes vehicles within the park, pre-registered vehicles, and external vehicles; S2, constructing a carbon emission measurement model according to the vehicle type, and using the carbon emission measurement model to analyze the carbon emissions of the vehicles within the park and the pre-registered vehicles after entering the park to obtain several carbon emission data; S3, according to several carbon emission data, statistically calculating the total carbon emissions per vehicle of each vehicle entering the park within a preset period, marking the vehicles with the difference between the total carbon emissions per vehicle and the preset carbon emission limit less than the preset threshold as restricted vehicles, and sending a warning signal to the restricted vehicles; S4, making a traffic decision for external vehicles according to the vehicle type and the daily carbon emission capacity.

Images