Traffic intelligent monitoring method and device based on Internet of Things

Abstract

The invention discloses a traffic intelligent monitoring method and device based on the Internet of Things, and the method comprises the steps: obtaining the traffic flow information through a plurality of vehicle quantity detectors disposed on a traffic road, outputting the interval intuitionistic fuzzy number of the traffic flow through a traffic flow detection module, and arranging a plurality of groups of environment parameter sensors on the traffic road, environmental parameters of temperature, visibility, rainfall, dust and wind speed are detected, the output value of the traffic flow detection module and the environmental parameters serve as input, and the traffic flow value of the detected road section influenced by the environmental parameters is determined through the road section traffic intelligent monitoring module; traffic flow values, detected by the road section traffic intelligent monitoring modules, of detected road sections serve as input, and four predicted values of traffic flow in the four directions of east, south, west and north are output through the regional traffic intelligent monitoring subsystem. Compared with the prior art, the traffic information in the monitoring area can be automatically obtained in real time, the vehicle flow of the road section can be effectively adjusted, and the traffic of each road section is smoother.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic environment and traffic monitoring automation equipment, and in particular to an intelligent traffic monitoring method and device of the Internet of Things. Background Art

[0002] Traffic signal control devices are the most widely used method of traffic diversion. Traffic signal control devices at intersections use fixed signal time to control traffic, which can lead to long waiting times and a large backlog of vehicles in the direction of heavy traffic, and no vehicles passing in the direction of light traffic. How to control traffic signals more intelligently and predict traffic flow more accurately has attracted the attention of many scholars.

[0003] The effective combination of the Internet of Things and intelligent technology will have a positive impact on the effective construction of an intelligent traffic management system. How to effectively combine sensor technology, communication technology, network technology, information processing technology, and intelligent control technology to design a system that can automatically obtain traffic environment information, traffic flow information, and vehicle information in the monitoring area in real time is a technical problem that needs to be solved in this field. Summary of the invention

[0004] Purpose of the invention: In view of the problems in the background technology, the present invention discloses an intelligent traffic monitoring method and device of the Internet of Things, which can automatically obtain the traffic environment information, vehicle flow information and vehicle information in the monitoring area in real time, and can provide reliable guarantee for the traffic management department to obtain the traffic flow conditions of each road section in real time and accurately, and can provide real-time and reliable travel services to more users through the Internet of Things based on the traffic flow conditions of each road section, and can effectively regulate the vehicle flow of the road section to make the traffic of each road section smoother.

[0005] Technical solution: The present invention discloses an intelligent traffic monitoring method of the Internet of Things, comprising the following steps:

[0006] Step 1: The vehicle flow information is obtained by setting up multiple vehicle volume detectors on the traffic road. The vehicle flow information is output as the interval intuitionistic fuzzy number of the vehicle flow through the vehicle flow detection module. The interval intuitionistic fuzzy number of the vehicle flow includes the interval fuzzy number, the credibility of the interval fuzzy number and the unreliability of the interval fuzzy number of the vehicle flow.

[0007] Step 2: Set up multiple sets of environmental parameter sensors on the traffic roads to detect environmental parameters such as temperature, visibility, rainfall, dust and wind speed. Take the output value of the vehicle flow detection module and the environmental parameters as input, and determine the traffic flow value of the detected road section under the influence of the environmental parameters through the road section traffic intelligent monitoring module;

[0008] Step 3: Taking the traffic flow values ​​of the detected road sections output by the multiple road section traffic intelligent monitoring modules as input, the regional traffic intelligent monitoring subsystem outputs 16 parameters to form four interval intuitive fuzzy number prediction values ​​of the traffic flow in the four directions of east, south, west and north, respectively. The interval intuitive fuzzy number prediction value of the traffic flow in one direction is [<s,t> ; m; r], s and t constitute the interval fuzzy number prediction value of the traffic flow in this direction:<s,t> , m is the credibility of the interval fuzzy number prediction value of the traffic flow in this direction, r is the unreliability of the interval fuzzy number prediction value of the traffic flow in this direction, and the sum of m and r is less than or equal to 1.

[0009] Further, in the step 1, the vehicle flow detection module includes NARX neural network model of FOA-metabolism gray GM (1,1) model 1, NARX neural network model of FOA-metabolism gray GM (1,1) model 2, denoising autoencoder-metabolism gray GM (1,1) model, NARX neural network model of FOA 1, NARX neural network model of FOA 2, NARX neural network model of FOA 3, GRNN neural network model A of FOA, NARX neural network model of interval intuitionistic fuzzy number of FOA-recursive fuzzy neural network model 1 of FOA, TDL beat delay 1, TDL beat delay 2;

[0010] The outputs of multiple vehicle detectors are used as the inputs of FOA's GRNN neural network model A, FOA's NARX neural network model 2 and the denoising autoencoder-metabolism gray GM (1,1) model respectively. The time series difference between the output of FOA's NARX neural network model 2 and the output of the denoising autoencoder-metabolism gray GM (1,1) model is used as the corresponding input of FOA's NARX neural network model 3. The output of the denoising autoencoder-metabolism gray GM (1,1) model is used as the corresponding input of FOA's NARX neural network model-metabolism gray GM (1,1) model. 2, the output of TDL beat delay device 1 is used as the corresponding input of FOA's NARX neural network model 3 and FOA's NARX neural network model-metabolism gray GM (1,1) model 2, the output of TDL beat delay device 2 is used as the input of FOA's NARX neural network model 1, the time series difference between the output of FOA's GRNN neural network model A and the output of FOA's NARX neural network model 1 is used as the input of FOA's NARX neural network model-metabolism gray GM (1,1) model 1, FOA's NARX neural network model -The output of metabolic grey GM (1,1) model 1 is used as the corresponding input of FOA's NARX neural network model - metabolic grey GM (1,1) model 2. The difference between the output of FOA's GRNN neural network model A and the output of FOA's NARX neural network model 1 minus the time series difference of the output value of FOA's NARX neural network model - metabolic grey GM (1,1) model 1 is used as the corresponding input of FOA's NARX neural network model 3. FOA's NARX neural network model 3 and FOA's NARX neural network model - metabolic grey GM (1,1) The output of model 2 is used as the corresponding input of NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number, and the output of NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number is used as the input of TDL beat delay device 1 and the corresponding input of TDL beat delay device 2. The four parameters a, b, u and v output by NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number constitute the interval intuitionistic fuzzy number of traffic flow: [<a,b> ; u; v], a and b constitute the interval fuzzy number of traffic flow:<a,b> ,u is the credibility of the interval fuzzy number of traffic flow, v is the unreliability of the interval fuzzy number of traffic flow, and the sum of u and v is less than or equal to 1.

[0011] Further, the road section traffic intelligent monitoring module includes a GRNN neural network model 1-n of FOA, an ESN neural network model 1-C of FOA, an environmental level subtraction clustering classifier, a NARX neural network model of FOA of interval intuitionistic fuzzy numbers-a recursive fuzzy neural network model 2 of FOA, a TDL beat delay device 3 and a vehicle flow detection module;

[0012] The outputs of multiple groups of environmental parameter sensors are respectively used as the inputs of the corresponding GRNN neural network models 1-n of FOA, the outputs of the GRNN neural network models 1-n of FOA are used as the inputs of the environmental level subtraction clustering classifier, the environmental level subtraction clustering classifier outputs multiple different types of FOA GRNN neural network model values ​​as the corresponding inputs of the corresponding ESN neural network models 1-C of FOA, the outputs of multiple vehicle detectors are used as the inputs of the vehicle detection module, the outputs of multiple ESN neural network models of FOA and the output of the vehicle detection module are respectively used as the corresponding inputs of the NARX neural network model of FOA with interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model 2 of FOA, the outputs of the NARX neural network model of FOA with interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model 2 of FOA are used as the inputs of the TDL beat delay device 3, the outputs of the TDL beat delay device 3 are used as the corresponding inputs of the NARX neural network model of FOA with interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model 2 of FOA, the outputs of the NARX neural network model of FOA with interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model 2 of FOA are used as the traffic flow value of the detected section, and the intelligent monitoring module of section traffic can be seen in Figure 2 .

[0013] Furthermore, the regional traffic intelligent monitoring subsystem includes a plurality of road section traffic intelligent monitoring modules, a TDL beat delay device 4 and an ESN neural network model of FOA of interval intuitionistic fuzzy numbers-a recursive fuzzy neural network model of FOA; the outputs of the plurality of road section traffic intelligent monitoring modules are respectively used as the corresponding inputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA, the outputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA are used as the inputs of the TDL beat delay device 4, the outputs of the TDL beat delay device 4 are used as the corresponding inputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA, and the outputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA of interval intuitionistic fuzzy numbers respectively constitute 4 interval intuitionistic fuzzy number prediction values ​​of the traffic flow in 4 directions of east, south, west and north, and the interval intuitionistic fuzzy number prediction value of the traffic flow in one direction is [<s,t> ; m; r], s and t constitute the interval fuzzy number prediction value of the traffic flow in this direction:<s,t> , m is the credibility of the interval fuzzy number prediction value of the traffic flow in this direction, r is the unreliability of the interval fuzzy number prediction value of the traffic flow in this direction, the sum of m and r is less than or equal to 1, the structure of the regional traffic intelligent monitoring subsystem is shown in Figure 3 shown.

[0014] Furthermore, the denoising autoencoder-metabolism grey GM (1,1) model is a denoising autoencoder and a metabolic grey GM (1,1) model connected in series, and the denoising autoencoder output is used as the metabolic grey GM (1,1) model input;

[0015] The NARX neural network model of FOA - metabolic grey GM (1,1) model 1 and the NARX neural network model of FOA - metabolic grey GM (1,1) model 2 have the same structure, that is, the NARX neural network model of FOA and the metabolic grey GM (1,1) model are connected in series, and the output of the NARX neural network model of FOA is used as the input of the metabolic grey GM (1,1) model;

[0016] GRNN neural network model A of FOA is the GRNN neural network model optimized by FOA;

[0017] NARX neural network model of FOA for interval intuitionistic fuzzy numbers - recursive fuzzy neural network model of FOA 1 is a series connection of NARX neural network model of FOA and recursive fuzzy neural network model of FOA, and the recursive fuzzy neural network model of FOA is an optimized recursive fuzzy neural network model of FOA;

[0018] The structures of NARX neural network model 1 of FOA, NARX neural network model 2 of FOA, and NARX neural network model 3 of FOA are the same as that of NARX neural network model of FOA, and all of them are FOA optimized NARX neural network models.

[0019] Furthermore, FOA's ESN neural network models 1-C are all FOA optimized ESN neural network models; FOA's GRNN neural network models 1-n are all FOA optimized GRNN neural network models; the environmental level subtraction clustering classifier takes each historical data point of the traffic environment parameter as a potential clustering center, so that the clustering result is independent of the dimension of the problem; FOA's NARX neural network model of interval intuitionistic fuzzy numbers - FOA's recursive fuzzy neural network model 2 is the NARX neural network model of FOA and the recursive fuzzy neural network model of FOA connected in series, the NARX neural network model of FOA is the FOA optimized NARX neural network model, and the recursive fuzzy neural network model of FOA is the FOA optimized recursive fuzzy neural network model.

[0020] Furthermore, the ESN neural network model of FOA of interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model of FOA is a series connection of the ESN neural network model of FOA and the recursive fuzzy neural network model of FOA, the output of the ESN neural network model of FOA is used as the input of the recursive fuzzy neural network model of FOA, the ESN neural network model of FOA is the FOA optimized ESN neural network model, and the recursive fuzzy neural network model of FOA is the FOA optimized recursive fuzzy neural network model.

[0021] The present invention also discloses an intelligent traffic monitoring device based on the intelligent traffic monitoring method of the Internet of Things, comprising an environmental parameter detection terminal, a traffic signal measurement and control terminal, a traffic scene monitoring terminal, a traffic monitoring gateway node, a traffic monitoring cloud platform, a traffic management monitoring terminal and a traffic monitoring mobile phone APP, wherein the environmental parameter detection terminal is responsible for collecting environmental parameter information of the detected traffic section, the traffic signal measurement and control terminal is responsible for measuring the traffic flow at the traffic intersection and controlling the traffic signal control device at the intersection, and multiple environmental parameter detection terminals interact with each other through the traffic monitoring gateway node and the traffic scene monitoring terminal, the traffic monitoring cloud platform, the traffic management monitoring terminal and the traffic monitoring mobile phone APP; The traffic scene monitoring terminal is equipped with a section traffic intelligent monitoring module to realize the intelligent monitoring of the detected section traffic and obtain the traffic flow value of the detected section under the influence of environmental parameters. The traffic management monitoring terminal is equipped with a regional traffic intelligent monitoring subsystem, which takes the traffic flow value of the detected section detected by multiple section traffic intelligent monitoring modules as input to realize the intelligent monitoring of the detected regional traffic, and outputs 16 parameters to form 4 interval intuitive fuzzy number prediction values ​​of the traffic flow in 4 directions of east, south, west and north respectively; the traffic monitoring mobile phone APP and the traffic management monitoring terminal access the traffic monitoring cloud platform through the 5G network to realize the collection and monitoring of traffic parameters. The structure of the traffic intelligent monitoring device of the Internet of Things is shown in Figure 4 shown.

[0022] Furthermore, the environmental parameter detection terminal includes temperature and humidity, visibility, wind speed, wind direction, rainfall, dust and harmful gas sensors for collecting traffic road environmental parameters and corresponding signal conditioning circuits, STM32 microcontrollers and 5G modules.

[0023] Compared with the prior art, the present invention has the following obvious advantages:

[0024] 1. The ESN neural network model of the present invention uses a large-scale random sparse network (reserve pool) as an information processing medium to map the input from a low-dimensional input space to a high-dimensional state space. The output connection weights of the ESN neural network model are trained using a linear regression method in the high-dimensional state space, while other connection weights are randomly generated and remain unchanged during the training process of the ESN neural network model. The output connection weights of the ESN neural network model change rapidly, while the internal connection weights only change slowly in a highly coupled manner. That is, if the internal connection weights of the ESN neural network model are appropriately selected, the changes in the internal connection weights can be ignored when the ESN neural network model is trained. As a newly emerging recursive neural network in recent years, the ESN neural network model has greatly improved stability, global optimality and convergence speed relative to traditional neural networks. The ESN neural network model excites the reserve pool so that the reserve pool generates continuous output state variable signals, and finally determines the ESN network output weights through a linear regression algorithm of the reserve pool state variables and the target output signal. The outstanding advantages of the ESN neural network are fast and simple training, and the input connection weights and the internal connection weights of the reserve pool remain basically unchanged during training and use. Therefore, the weights and thresholds of the ESN neural network model only require simple linear regression to calculate, which can ensure that its solution is the global optimal solution.

[0025] Second, the GRNN neural network model of the present invention has a strong nonlinear mapping ability and flexible network structure for input information, as well as a high degree of fault tolerance and robustness. The GRNN neural network model finally converges to the optimized regression surface with a large sample accumulation and has a good prediction effect when the input sample data is small and the input data is unstable. The GRNN neural network model has a strong generalization ability for input information, high prediction accuracy, and a stable algorithm. The GRNN neural network model also has the advantages of fast convergence speed, few adjustment parameters, and not easy to fall into local minima. The GRNN neural network model has a fast calculation speed and has good application prospects for the prediction of input information.

[0026] 3. According to the characteristics of the differences in traffic environment parameter samples, the present invention constructs an environment level subtraction clustering classifier to classify the multi-point historical sample parameters of traffic environment parameters. Each type of data is input into the corresponding FOA ESN neural network model, and the output of the corresponding FOA ESN neural network model is used as the prediction value of the corresponding type of traffic environment. In the continuous forecasting process of traffic environment parameters, the characteristics of traffic environment parameters in time and space are fully considered, and relatively homogeneous traffic environment data with similar causes are extracted from massive data, so as to establish a more targeted prediction model that can better reflect the level of traffic environment parameters at any time stage, thereby improving the prediction accuracy.

[0027] Fourth, the present invention adopts a recursive fuzzy neural network model to introduce internal variables in the fuzzy rule layer, so that the static network has dynamic characteristics; the activation of each rule of the network at time K not only includes the activation value calculated by the current input, but also includes the contribution of the activation values ​​of all rules at the previous moment, thereby improving the accuracy of network identification and better completing the dynamic identification of input information parameters. The recursive fuzzy neural network model is a typical dynamic recursive neural network, whose feedback connection is composed of a group of "structure" units, which are used to memorize the past state of the hidden layer, and together with the network input at the next moment as the input of the hidden layer unit. This property makes some recursive fuzzy neural networks have dynamic memory function, so they are suitable for establishing a time series traffic flow prediction model. Simulation experiments show that the model has good dynamic performance, high traffic flow prediction accuracy, and stable prediction performance.

[0028] 5. Since the traffic environment parameters have complex nonlinear characteristics and the traffic environment parameters vary greatly under different working conditions, it is difficult to establish an accurate mathematical model. The present invention uses the ESN neural network model of FOA with interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model of FOA and the NARX neural network model of FOA with interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model of FOA to scientifically quantify the traffic flow of regions and sections, including both interval fuzzy numbers and the credibility of interval fuzzy, thereby improving the scientificity and accuracy of quantifying the traffic flow of sections and regions. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 It is a vehicle flow detection module;

[0030] Figure 2 It is an intelligent monitoring module for road traffic;

[0031] Figure 3 It is the regional traffic intelligent monitoring subsystem;

[0032] Figure 4 It is an intelligent traffic monitoring device for the Internet of Things;

[0033] Figure 5 It is an environmental parameter detection terminal;

[0034] Figure 6 It is a traffic signal measurement and control terminal;

[0035] Figure 7 It is a traffic monitoring gateway node;

[0036] Figure 8 It is the traffic scene monitoring software. DETAILED DESCRIPTION

[0037] Combined with Figure 1-8 , further describe the technical solution of this application:

[0038] 1. Design of vehicle flow detection module

[0039] The vehicle flow detection module includes FOA's NARX neural network model-metabolism gray GM (1,1) model 1, FOA's NARX neural network model-metabolism gray GM (1,1) model 2, denoising autoencoder-metabolism gray GM (1,1) model, FOA's NARX neural network model 1, FOA's NARX neural network model 2, FOA's NARX neural network model 3, FOA's GRNN neural network model A, FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 1 and TDL beat delay 1, TDL beat delay 2. The vehicle flow detection module is shown in Figure 1 shown.

[0040] 1. Design of Denoising Autoencoder-Metabolic Gray GM (1,1) Model

[0041] The denoising autoencoder-metabolic grey GM (1,1) model is a denoising autoencoder and a metabolic grey GM (1,1) model connected in series. The output of the denoising autoencoder is used as the input of the metabolic grey GM (1,1) model. The denoising autoencoder is an autoencoder that increases the robustness of encoding by introducing noise. The core idea is to add random noise to the input traffic environment data layer for training, and restore the original unpolluted data by mapping the features of the polluted data, so that the encoder can learn a more robust and generalized feature expression.

[0042] Let X = (x1, x2, ..., x n ) is the original traffic signal, which is "polluted" by randomly setting zeros or adding noise, and X generates a noisy signal X = (x'1, x'2, ..., x' n ). Then, X′ is mapped to the hidden layer through the encoder network g, as shown in formula (1):

[0043] z=g(WX'+b) (1)

[0044] Where z is the effective feature of the hidden layer; W is the weight parameter mapped to the hidden layer network; b is the bias item mapped to the hidden layer network. After z is reversely mapped by the decoder, an output identical to the original traffic signal is finally constructed as shown in the following formula:

[0045]

[0046] Where: W′ is the weight parameter of the effective feature inverse mapping to the output layer network; b′ is the bias term inversely mapped to the network layer output. The target result is to make the output value as close as possible to the original traffic signal X, construct a square reconstruction error function, and minimize the function to obtain the optimized network parameters such as weights and bias terms.

[0047] The present invention adopts the metabolic gray GM (1, 1) model to predict the traffic flow parameter value at the future moment according to the traffic flow prediction value. After the traffic flow parameter is predicted by the above method, the traffic flow prediction value is added to the original sequence of the traffic flow parameter, and a data model is removed at the beginning of the sequence accordingly, and then the traffic flow parameter is predicted. Similarly, the traffic flow parameter is predicted. This method is called the metabolic gray GM (1, 1) model, which can achieve a long-term prediction, can more accurately grasp the change trend of the traffic flow parameter, and effectively make safety preparations for the traffic flow state.

[0048] The GM (1, 1) model has many advantages over traditional statistical prediction methods. It does not need to determine whether the prediction variables obey the normal distribution, does not require large sample statistics, and does not need to change the prediction model at any time according to the changes in the input variables of the traffic flow parameters. Through the cumulative generation technology, a unified differential equation model is established. The prediction results are obtained after the original values ​​of the cumulative traffic flow parameters are restored. The differential equation model has higher prediction accuracy. The essence of establishing the metabolic gray GM (1, 1) model is to perform a cumulative generation of the original data of the traffic flow parameters so that the generated series presents a certain regularity. By establishing a differential equation model, a fitting curve is obtained to predict the traffic flow parameters.

[0049] 2. Design of FOA's NARX neural network model - metabolic grey GM (1,1) model 1 and FOA's NARX neural network model - metabolic grey GM (1,1) model 2

[0050] The NARX neural network model of FOA - metabolism gray GM (1,1) model 1 and the NARX neural network model of FOA - metabolism gray GM (1,1) model 2 have the same structure, both of which are the NARX neural network model of FOA - metabolism gray GM (1,1) model, which are the NARX neural network model of FOA and the metabolism gray GM (1,1) model connected in series, and the output of the NARX neural network model of FOA is used as the input of the metabolism gray GM (1,1) model.

[0051] The NARX neural network model 1 of FOA, the NARX neural network model 2 of FOA, and the NARX neural network model 3 of FOA have the same structure as the NARX neural network model of FOA, which is a dynamic feedforward neural network. The NARX neural network model of FOA is a nonlinear autoregressive network with external input. It has a dynamic characteristic of multi-step delay and connects several layers of the closed network through feedback. The NARX neural network model of FOA is the most widely used dynamic neural network in nonlinear dynamic systems, and its performance is generally better than that of the full recurrent neural network. The NARX neural network model of FOA is mainly composed of input layer, hidden layer, output layer and input and output delay. Before application, the delay order of input and output and the number of hidden layer neurons should be determined in advance. The output of the NARX neural network model of FOA depends not only on the past output y(tn), but also on the input vector X(t) at that time and the delay order of the input vector. The structure of the NARX neural network model of FOA includes input layer, output layer, hidden layer and delay layer. The input signal of FOA's NARX neural network model is transmitted to the hidden layer through the delay layer. The hidden layer processes the input signal and transmits it to the output layer. The output layer linearly weights the hidden layer output signal to obtain the final FOA's NARX neural network model output signal. The delay layer delays the feedback signal and input signal of FOA's NARX neural network model and then transmits it to the hidden layer. x(t) represents the external input of FOA's NARX neural network model; m represents the delay order of the external input; y(t) is the output of FOA's NARX neural network model; n is the output delay order; s is the number of neurons in the hidden layer; thus, the output of the jth hidden unit can be obtained as:

[0052]

[0053] In the above formula, w ji is the connection weight between the ith input and the jth hidden neuron, b j is the bias value of the jth hidden neuron. The principle of FOA optimization of NARX neural network is as follows: according to the change of individual position in FOA iterative optimization, the updated value of weight and threshold in NARX neural network corresponds to the optimal value of individual position, which corresponds to the initial optimal value of weight and threshold, can eliminate the influence caused by empirical trial and error, thereby optimizing the learning ability and running speed of NARX neural network.

[0054] The main construction process of the FOA optimized NARX neural network model is as follows: (1) The initial weights and initial thresholds to be optimized are used as the position coordinates of individual fruit flies, the number of iterations, and the number of individuals in the population for initialization. (2) The mean square error between the measured and calculated load values ​​in the NARX neural network model is used as the flavor concentration judgment function, and the optimal position coordinates are iteratively updated according to the FOA optimization algorithm. (3) The weights are continuously updated and adjusted to reduce the error between the output and the expectation until the maximum number of iterations is reached or the convergence condition is met.

[0055] 3. Design of GRNN neural network model A of FOA

[0056] The GRNN neural network model is based on mathematical statistics and has a clear theoretical basis. After the learning sample is determined, the network structure and connection value are also determined. During the training process, only one variable, the smoothing parameter, needs to be determined. The learning of the GRNN neural network model depends entirely on the input data sample. It has strong nonlinear mapping and flexible network structure as well as high fault tolerance and robustness. It is particularly suitable for fast approximation of functions and processing unstable data. The GRNN neural network model has few artificial adjustment parameters, and the learning of the network depends entirely on data samples. This feature enables the network to minimize the impact of human subjective assumptions on the prediction results. The GRNN neural network model has strong prediction capabilities under small samples, and also has the characteristics of fast training and strong robustness, and is basically not troubled by multicollinearity of input data.

[0057] The GRNN neural network model of the present invention is composed of an input layer, a pattern layer, a summation layer and an output layer. The input vector X of the GRNN neural network model is an n-dimensional vector, and the output vector Y is a k-dimensional vector X={x1,x2,…,x n} T and Y={y1,y2,…,y k} T The number of neurons in the pattern layer is equal to the number of training samples m. Each neuron corresponds to a training sample one by one. The transfer function of the neuron in the pattern layer is p i for:

[0058] p i =exp{-[(x-xi) T (xx i )] / 2σ},(i=1,2,…,m) (4)

[0059] The neuron output in the above formula enters the summation layer for summation. The summation layer functions are divided into two categories:

[0060]

[0061]

[0062] Among them, y ij is the jth element value in the output vector of the i-th training sample. According to the aforementioned GRNN neural network model algorithm, the estimated value of the jth element of the network output vector Y is:

[0063] y j =s Nj / s D ,(j=1,2,…k) (7)

[0064] The GRNN neural network model can approximate its implicit mapping relationship according to the input sample data, and the output result of the GRNN neural network model can converge to the optimal regression surface, especially when the input data sample data is scarce, it can also obtain satisfactory prediction results. The GRNN neural network model has strong prediction ability and fast learning speed. It is mainly used to solve function approximation problems and has high parallelism in structure. The principle of FOA optimization of GRNN neural network is as follows: According to the change of individual position in FOA iterative optimization, the updated value of weight and threshold in GRNN neural network is corresponding to the optimal value of individual position, which can eliminate the influence caused by empirical trial and error, so as to optimize the learning ability and running speed of GRNN neural network. The main construction process of FOA optimization of GRNN neural network model is as follows: (1) The initial weight and initial threshold to be optimized are used as the coordinates of individual position of fruit flies, the number of iterations, and the number of individuals in the population for initialization. (2) The mean square error between the measured value and the calculated value of the load in the GRNN neural network model is used as the taste concentration judgment function, and the optimal position coordinate is iteratively updated according to the FOA optimization algorithm. (3) Continuously update and adjust the weights to reduce the error between the output and the expectation until the maximum number of iterations is reached or the convergence condition is met.

[0065] 4. Design of NARX neural network model of FOA for interval intuitionistic fuzzy numbers - Recursive fuzzy neural network model of FOA 1

[0066] NARX neural network model of FOA for interval intuitionistic fuzzy numbers - recursive fuzzy neural network model of FOA 1 is a series connection of NARX neural network model of FOA and recursive fuzzy neural network model of FOA. The four parameters a, b, u and v output by the recursive fuzzy neural network model of FOA constitute the interval intuitionistic fuzzy number of traffic flow:<a,b> ; u; v], a and b constitute the interval fuzzy number of traffic flow:<a,b> ,u is the credibility of the interval fuzzy number of traffic flow, v is the unreliability of the interval fuzzy number of traffic flow, and the sum of u and v is less than or equal to 1.

[0067] The design process of FOA's NARX neural network model refers to the above-mentioned relevant design steps of the present invention. The recursive fuzzy neural network model is a network topology structure with multiple inputs and a single output. The network consists of four layers: input layer, membership function layer, rule layer and output layer. The recursive fuzzy neural network model contains n input nodes, where each input node corresponds to m conditional nodes, m represents the number of rules, nm rule nodes, and 1 output node. The first layer introduces the input into the network; the second layer fuzzifies the input, and the membership function used is the Gaussian function; the third layer corresponds to fuzzy reasoning; and the fourth layer corresponds to defuzzification operations. Represent the input and output of the ith node of the kth layer respectively. The signal transmission process within the network and the input-output relationship between each layer can be described as follows. Layer I: Each input node of the input layer is directly connected to the input variable. The input and output of the network are expressed as:

[0068]

[0069] In the formula and is the input and output of the i-th node in the network input layer, and N represents the number of iterations.

[0070] Layer II: The nodes of the membership function layer fuzzify the input variables. Each node represents a membership function, and the Gaussian basis function is used as the membership function. The input and output of the network are expressed as:

[0071]

[0072] Where mij and σ ij They respectively represent the mean center and width value of the j-th Gaussian basis function of the ith linguistic variable in the II layer, and m is the total number of linguistic variables corresponding to the input node.

[0073] Layer III: Fuzzy reasoning layer, i.e. rule layer, adds dynamic feedback to make the network have better learning efficiency. The feedback link introduces internal variables h k , the sigmoid function is selected as the activation function of the internal variables of the feedback link. The input and output of the network are expressed as:

[0074]

[0075] Where ω jk is the connection weight of the recursive part. The neurons in this layer represent the antecedent part of the fuzzy logic rule. The nodes in this layer perform π operation on the output of the second layer and the feedback of the third layer. is the output of the third layer, and m represents the number of rules when fully connected. The feedback link mainly calculates the activation strength of the internal variable value and the corresponding membership function of the internal variable. The activation strength is related to the matching degree of the rule node of the third layer. The internal variables introduced in the feedback link include two types of nodes: receiving nodes and feedback nodes. The receiving node uses weighted summation to calculate the internal variable to achieve the defuzzification function, and the internal variable represents the result of fuzzy reasoning of the hidden rule. The feedback node uses the sigmoid function as the fuzzy membership function to realize the fuzzification of the internal variable. The membership function layer of the recursive fuzzy neural network model uses a local membership function, and the feedback part uses a global membership function on the domain of the internal variable to simplify the network structure and realize the feedback of global historical information. The number of receiving nodes is equal to the number of feedback nodes; the number of receiving nodes is equal to the number of nodes in the rule layer. The feedback is connected to the third layer as the input of the fuzzy rule layer, and the output of the feedback node contains the historical information of the activation strength of the fuzzy rule.

[0076] Layer IV: Defuzzification layer is the output layer. The nodes in this layer perform summation operations on the input quantities. The input and output of the network are expressed as:

[0077]

[0078] In the formula, j is the connection weight of the output layer. The recursive fuzzy neural network model has the performance of approximating highly nonlinear dynamic systems. The training error and test error of the recursive fuzzy neural network model with internal variables are significantly reduced. The prediction effect of this network is better than that of the recursive fuzzy neural network with self-feedback and the fuzzy neural network with dynamic modeling. This shows that the learning ability of the network has been enhanced after the internal variables are added. The recursive fuzzy neural network model introduces internal variables in the feedback link, takes the weighted sum of the output of the rule layer and then defuzzifies the output as the feedback, and takes the feedback and the output of the membership function layer as the input of the rule layer at the next moment. The output of the recursive fuzzy neural network model contains the historical information of the activation strength and output of the rule layer, which enhances the ability of the recursive fuzzy neural network model to adapt to nonlinear dynamic systems.

[0079] The recursive fuzzy neural network model of FOA includes three parts: recursive fuzzy neural network model structure determination, FOA algorithm optimization and recursive fuzzy neural network model prediction. M groups of input and output samples of the neural network are given as training sets, and the original data are standardized; the recursive fuzzy neural network model structure is determined according to the number of input and output parameters, thereby determining the length of the FOA algorithm particles; the indirect weights and thresholds between all neurons in the recursive fuzzy neural network model structure are encoded into individuals represented by real digital numbers. If the network contains N optimization weights and thresholds, each individual will be represented by an N-dimensional vector composed of N weights and threshold parameters to initialize the particle swarm; the absolute value and sum of the prediction error are used as the individual fitness value, and the individual extreme value and the global extreme value are obtained according to the fitness value; it is determined whether the global extreme value meets the FOA end condition, and if so, the FOA optimization search is exited; if not, the speed and position of each particle are updated; the particle corresponding to the global extreme value is decoded, and used as the initial weight and threshold of the recursive fuzzy neural network model; the optimal initial weight and threshold are assigned to the recursive fuzzy neural network model, the network model is trained and determined, and the trained recursive fuzzy neural network model is used to predict the input parameters.

[0080] The main construction process of the FOA optimized recursive fuzzy neural network model is as follows: (1) The initial weights and initial thresholds to be optimized are used as the position coordinates of individual fruit flies, the number of iterations, and the number of individuals in the population for initialization. (2) The mean square error between the measured and calculated load values ​​in the recursive fuzzy neural network model is used as the flavor concentration judgment function, and the optimal position coordinates are iteratively updated according to the FOA optimization algorithm. (3) The weights are continuously updated and adjusted to reduce the error between the output and the expectation until the maximum number of iterations is reached or the convergence condition is met.

[0081] 2. Design of intelligent monitoring module for road traffic

[0082] The intelligent monitoring module for road traffic includes GRNN neural network model 1-n of FOA, ESN neural network model 1-C of FOA, environmental level subtraction clustering classifier, NARX neural network model of interval intuitionistic fuzzy number of FOA-recursive fuzzy neural network model 2 of FOA, TDL beat delay device 3 and the above-mentioned vehicle flow detection module. The intelligent monitoring module for road traffic is shown in Figure 2 As shown. GRNN neural network model 1-n of FOA represents n GRNN neural network models of FOA, which are designed with reference to GRNN neural network model A of FOA of the vehicle flow detection module of the present invention, and the design method of NARX neural network model of FOA of interval intuitionistic fuzzy number - recursive fuzzy neural network model 2 of FOA refers to the design steps of NARX neural network model of FOA of the vehicle flow detection module - recursive fuzzy neural network model 1 of FOA of the present invention.

[0083] 1. Design of ESN neural network model 1-C of FOA

[0084] FOA's ESN neural network model 1-C is FOA's ESN neural network model 1, FOA's ESN neural network model 2, ..., FOA's ESN neural network model C are FOA's ESN neural network models.

[0085] The ESN neural network model (Echo state network, ESN) is a new type of dynamic neural network, which has all the advantages of dynamic neural networks. At the same time, because the echo state network introduces the concept of "reserve pool", this method can better adapt to nonlinear system identification than general dynamic neural networks. The "reserve pool" is to transform the middle connection part of the traditional dynamic neural network into a randomly connected "reserve pool". The whole learning process is actually the process of learning how to connect the "reserve pool". The "reserve pool" is actually a randomly generated large-scale recursive structure in which the neurons are sparsely connected. SD is usually used to represent the percentage of interconnected neurons in the total number of neurons N. The state equation of the ESN neural network model is:

[0086]

[0087] Where W is the state variable of the neural network, W in is the input variable of the neural network; W back is the output state variable connection weight matrix of the neural network; x(n) represents the internal state of the ESN neural network model; W out is the connection weight matrix between the core reserve pool of the ESN neural network model, the input of the RSN neural network model, and the output of the ESN neural network model; is the output deviation of the ESN neural network or can represent noise; f = f[f1,f2,…,f n ] is the n activation functions of the neurons inside the “reserve pool”; f i is the hyperbolic tangent function; f out is the ε output function of the ESN neural network model.

[0088] The main construction process of the FOA optimized ESN neural network model is as follows: (1) The initial weights and initial thresholds to be optimized are used as the position coordinates of individual fruit flies, the number of iterations, and the number of individuals in the population for initialization. (2) The mean square error between the measured and calculated load values ​​in the ESN neural network model is used as the flavor concentration judgment function, and the optimal position coordinates are iteratively updated according to the FOA optimization algorithm. (3) The weights are continuously updated and adjusted to reduce the error between the output and the expectation until the maximum number of iterations is reached or the convergence condition is met.

[0089] 2. Design of environmental level subtraction clustering classifier

[0090] Compared with other clustering methods, the environment level subtraction clustering classifier does not need to determine the number of clusters in advance. It can quickly determine the location of the cluster center and the number of clusters based on the density of the historical sample data of the traffic environment parameters. It takes each historical data point of the traffic environment parameter as a potential cluster center, which makes the clustering result independent of the dimension of the problem. Therefore, the traffic environment parameter subtraction clustering algorithm is a rule automatic extraction method suitable for data modeling. Set N data points (X1, X2, ... X N ), each historical data point of traffic environment parameters is a candidate for cluster center, i = 1, 2, ..., N, data point X i The density function of is defined as:

[0091]

[0092] Where the radius r a is a positive number, r a Defines an influential neighborhood of the point. Data points outside the radius contribute very little to the density index of the point and are generally negligible. Calculate each point X i The density value with the highest density index D is selected c1 The data point is taken as the first cluster center X c1 ; Then correct the density value to eliminate the influence of the previous cluster center. Correct the density value according to the following formula:

[0093]

[0094] Among them, D c1 is the highest density value corresponding to the initial cluster center, and the corrected radius r b The setting is to avoid the second cluster center being too close to the previous center, which is generally set to r b =ηr a ,1.25≤η≤1.5. After correcting the density index of each data point, when D ck With D c1 When the following formula is satisfied, the cluster center corresponding to the density index is the Kth cluster center. Repeat this process until the new cluster center X ck The corresponding density index D ck With D c1 Clustering is terminated when the following equation is satisfied:

[0095] D ck / D c1 <δ (15)

[0096] In the formula, δ is a threshold value set in advance according to the actual situation. The basic idea of ​​the environmental level subtraction clustering classifier of the present invention is as follows: if the distance from a point to the center of a group is less than the cluster radius r a , then the point belongs to this group; when new traffic environment parameter data is obtained, the group and the center of the group are changed accordingly. As the input traffic environment parameter space data continues to increase, the algorithm of the present invention obtains a better traffic environment parameter input space division by dynamically adjusting the cluster center and the number of clusters in real time. The steps are as follows:

[0097] Step 1: Data normalization, input data cluster radius r of each dimension a And the threshold δ and other parameters are set.

[0098] Step 2: Perform subtractive clustering on the historical training traffic environment parameter dataset to obtain c cluster centers and store v i (i=1,2,…,c) and its corresponding density value D(v i ).

[0099] Step 3: When the kth data in the newly added online dataset arrives, calculate x k (k=1,2,…,M) to i cluster centers v i The distance d ki =||x k -v i ||, if d ki >r a , go to step 4; if d ki ≤r a , go to step 5.

[0100] Step 4: Calculate x k The density value D(x k ), and D(x k )>ε, then the data x k If it does not belong to any existing cluster, a new cluster is created, the number of clusters in the input space c=c+1, and the process returns to step 3.

[0101] Step 5: Determine the data point x according to the minimum distance criterion k Belongs to the nearest cluster subset, further compare the new data x k The density value of the cluster center is the same as the density value of the cluster center. If D(x k )>D(v i ), then the data x k is closer to its nearest cluster center, x k Replace the original cluster center as the new cluster center of the subset; if D(x k )≤D(v i), the clustering result remains unchanged, and it is determined whether the newly added data group is finished. If it is finished, go to step 6; otherwise, return to step 3.

[0102] Step 6: Calculate the cluster center v i With v j The distance between them, if min||v i -v j ||≤(0.5-0.7)r a , and D(v i )>D(v j ), then the cluster subset v i With v j can be merged into a cluster with the cluster center v i ; otherwise, keep the clustering results unchanged.

[0103] 3. Design of Regional Traffic Intelligent Monitoring Subsystem

[0104] The regional traffic intelligent monitoring subsystem includes multiple road section traffic intelligent monitoring modules, a TDL beat delay device 4 and an ESN neural network model of FOA of interval intuitionistic fuzzy numbers - a recursive fuzzy neural network model of FOA.

[0105] The ESN neural network model of FOA of interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model of FOA is a series connection of the ESN neural network model of FOA and the recursive fuzzy neural network model of FOA. The output of the ESN neural network model of FOA is used as the input of the recursive fuzzy neural network model of FOA, and the output of the recursive fuzzy neural network model of FOA is 4 interval intuitionistic fuzzy numbers.

[0106] The NARX neural network model of FOA with interval intuitionistic fuzzy numbers - The recursive fuzzy neural network model of FOA outputs 16 parameters to form four interval intuitionistic fuzzy number prediction values ​​of traffic flow in four directions: east, south, west and north. The interval intuitionistic fuzzy number prediction value of traffic flow in one direction is [<s,t> ; m; r], s and t constitute the interval fuzzy number prediction value of the traffic flow in this direction:<s,t> , m is the credibility of the interval fuzzy number prediction value of the traffic flow in this direction, r is the unreliability of the interval fuzzy number prediction value of the traffic flow in this direction, and the sum of m and r is less than or equal to 1.

[0107] 4. Design of intelligent traffic monitoring device for the Internet of Things

[0108] The intelligent traffic monitoring device of the Internet of Things includes an environmental parameter detection terminal, a traffic signal measurement and control terminal, a traffic site monitoring terminal, a traffic monitoring gateway node, a traffic monitoring cloud platform, a traffic management monitoring terminal and a traffic monitoring mobile phone APP. The environmental parameter detection terminal is responsible for collecting environmental parameter information of the detected traffic section, and the traffic signal measurement and control terminal is responsible for measuring the traffic flow at the traffic intersection and controlling the traffic signal control device at the intersection. The two-way communication between the environmental parameter detection terminal, the traffic signal measurement and control terminal, the traffic site monitoring terminal, the traffic monitoring cloud platform, the traffic management monitoring terminal and the traffic monitoring mobile phone APP is realized through the traffic monitoring gateway node. The traffic site monitoring terminal has a section traffic intelligent monitoring module to realize intelligent monitoring of the detected section traffic, and the traffic management monitoring terminal has a regional traffic intelligent monitoring subsystem to realize intelligent monitoring of the detected regional traffic. The traffic monitoring mobile phone APP and the traffic management monitoring terminal access the traffic monitoring cloud platform through the 5G network to realize the collection and monitoring of traffic parameters. The structure of the intelligent traffic monitoring device of the Internet of Things is shown in Figure 4 shown.

[0109] 1. Design of environmental parameter detection terminal

[0110] A large number of environmental parameter detection terminals based on the Internet of Things are used as traffic road environmental parameter perception terminals. Multiple environmental parameter detection terminals interact with each other through traffic monitoring gateway nodes and traffic site monitoring terminals, traffic monitoring cloud platforms, traffic management monitoring terminals, and traffic monitoring mobile phone APPs. The environmental parameter detection terminal includes temperature and humidity, visibility, wind speed, wind direction, rainfall, dust and harmful gas sensors for collecting traffic road environmental parameters and corresponding signal conditioning circuits, STM32 microcontrollers and 5G modules; the software of the environmental parameter detection terminal mainly realizes wireless communication and the collection and preprocessing of traffic road parameters. The software is designed in C language with a high degree of compatibility, which greatly improves the work efficiency of software design and development and enhances the reliability, readability and portability of the program code. See the structure of the parameter detection terminal. Figure 5 .

[0111] 2. Design of traffic signal measurement and control terminal

[0112] The traffic signal measurement and control terminal includes an STM32 microcontroller, a 5G module, an electronic display device, a traffic signal control device, a lighting device, an RFID reader / writer, a vehicle detector, a MODEL2420 three-axis acceleration sensor and a camera. It is used to collect traffic vehicle information and traffic flow information at traffic intersections, control traffic information and display traffic signal information. The traffic signal measurement and control terminal realizes information interaction between the traffic site monitoring terminal, the traffic management monitoring terminal, the traffic monitoring mobile phone APP and the traffic monitoring cloud platform through the traffic monitoring gateway node. The MODEL2420 is used to measure the three-dimensional acceleration signal of the vehicle during activity, and the traffic signal control device is used to adjust the traffic flow at the traffic intersection. Traffic signal measurement and control terminals such as Figure 6 shown.

[0113] 3. Traffic monitoring gateway node design

[0114] The traffic monitoring gateway node includes a 5G module, an NB-IoT module, an STM32 microcontroller and a USB interface. The traffic monitoring gateway node uses the 5G module to realize a self-organizing communication network that communicates with the environmental parameter detection terminal and the traffic signal measurement and control terminal. The NB-IoT module realizes two-way data interaction between the gateway and the traffic monitoring cloud platform. The USB interface connects to the traffic site monitoring terminal to realize information interaction between the gateway and the traffic site monitoring terminal. Figure 7 .

[0115] 4. Traffic scene monitoring software design

[0116] The traffic scene monitoring terminal is an industrial control computer. It mainly collects traffic environment parameters and vehicle flow information and realizes intelligent monitoring of the detected road section through the intelligent monitoring module of the road section traffic. The traffic scene monitoring terminal realizes information interaction with the environmental parameter detection terminal, the traffic signal measurement and control terminal and the traffic monitoring cloud platform through the traffic monitoring gateway node. The main functions of the traffic scene monitoring terminal are traffic communication parameter setting, traffic data analysis and data management, and intelligent monitoring module of road section traffic. The management software uses Microsoft Visual++6.0 as the development tool. The functions of the traffic scene monitoring terminal software are shown in Figure 8 .

[0117] The above embodiments are only for illustrating the technical concept and features of the present invention, and their purpose is to enable people familiar with the technology to understand the content of the present invention and implement it accordingly, and they cannot be used to limit the protection scope of the present invention. Any equivalent transformation or modification made according to the spirit of the present invention should be included in the protection scope of the present invention.

Claims

1. A traffic intelligent monitoring method of the Internet of Things, characterized in that: The steps include: Step 1: The vehicle flow information is obtained by setting up multiple vehicle volume detectors on the traffic road. The vehicle flow information is output as the interval intuitionistic fuzzy number of the vehicle flow through the vehicle flow detection module. The interval intuitionistic fuzzy number of the vehicle flow includes the interval fuzzy number, the credibility of the interval fuzzy number and the unreliability of the interval fuzzy number of the vehicle flow. Step 2: Set up multiple sets of environmental parameter sensors on the traffic roads to detect environmental parameters such as temperature, visibility, rainfall, dust and wind speed. Take the output value of the vehicle flow detection module and the environmental parameters as input, and determine the traffic flow value of the detected road section under the influence of the environmental parameters through the road section traffic intelligent monitoring module; Step 3: Taking the traffic flow values ​​of the detected road sections output by the multiple road section traffic intelligent monitoring modules as input, the regional traffic intelligent monitoring subsystem outputs 16 parameters to form four interval intuitive fuzzy number prediction values ​​of the traffic flow in the four directions of east, south, west and north, respectively. The interval intuitive fuzzy number prediction value of the traffic flow in one direction is [<s,t> ; m; r], s and t constitute the interval fuzzy number prediction value of the traffic flow in this direction:<s,t> , m is the credibility of the interval fuzzy number prediction value of the traffic flow in this direction, r is the unreliability of the interval fuzzy number prediction value of the traffic flow in this direction, and the sum of m and r is less than or equal to 1; In the step 1, the vehicle flow detection module includes FOA's NARX neural network model-metabolism gray GM (1,1) model 1, FOA's NARX neural network model-metabolism gray GM (1,1) model 2, denoising autoencoder-metabolism gray GM (1,1) model, FOA's NARX neural network model 1, FOA's NARX neural network model 2, FOA's NARX neural network model 3, FOA's GRNN neural network model A, FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 1, TDL beat delay 1, TDL beat delay 2; The outputs of multiple vehicle detectors are used as inputs of FOA's GRNN neural network model A, FOA's NARX neural network model 2 and the denoising autoencoder-metabolism gray GM (1,1) model respectively. The time series difference between the output of FOA's NARX neural network model 2 and the output of the denoising autoencoder-metabolism gray GM (1,1) model is used as the corresponding input of FOA's NARX neural network model 3. The output of the denoising autoencoder-metabolism gray GM (1,1) model is used as the corresponding input of FOA's NARX neural network model-metabolism gray GM (1,1) model. 2, the output of TDL beat delay device 1 is used as the corresponding input of FOA's NARX neural network model 3 and FOA's NARX neural network model-metabolism gray GM (1,1) model 2 respectively, the output of TDL beat delay device 2 is used as the input of FOA's NARX neural network model 1, the time series difference between the output of FOA's GRNN neural network model A and the output of FOA's NARX neural network model 1 is used as the input of FOA's NARX neural network model-metabolism gray GM (1,1) model 1, FOA's NARX neural network model - Metabolic gray GM (1,1) model 1 output is used as the corresponding input of FOA's NARX neural network model - Metabolic gray GM (1,1) model 2, the difference between FOA's GRNN neural network model A output and FOA's NARX neural network model 1 output minus FOA's NARX neural network model - Metabolic gray GM (1,1) model 1 output time series difference is used as the corresponding input of FOA's NARX neural network model 3, FOA's NARX neural network model 3 and FOA's NARX neural network model - Metabolic gray GM (1,1) The output of model 2 is used as the corresponding input of the NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number, and the output of the NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number is used as the input of TDL beat delay device 1 and the corresponding input of TDL beat delay device 2. The four parameters a, b, u and v output by the NARX neural network model of FOA-FOA recursive fuzzy neural network model 1 of interval intuitionistic fuzzy number constitute the interval intuitionistic fuzzy number of traffic flow: [<a,b> ; u; v], a and b constitute the interval fuzzy number of traffic flow:<a,b> ,u is the credibility of the interval fuzzy number of traffic flow, v is the unreliability of the interval fuzzy number of traffic flow, and the sum of u and v is less than or equal to 1.

2. The method for intelligent traffic monitoring of the Internet of Things according to claim 1, characterized in that: The road section traffic intelligent monitoring module includes a GRNN neural network model 1-n of FOA, an ESN neural network model 1-C of FOA, an environmental level subtraction clustering classifier, a NARX neural network model of FOA of interval intuitionistic fuzzy numbers-a recursive fuzzy neural network model 2 of FOA, a TDL beat delay device 3 and a vehicle flow detection module; Multiple groups of environmental parameter sensor outputs are used as the corresponding inputs of FOA's GRNN neural network model 1-n, the outputs of FOA's GRNN neural network model 1-n are used as the inputs of the environmental level subtraction clustering classifier, the environmental level subtraction clustering classifier outputs multiple different types of FOA's GRNN neural network model values ​​as the corresponding inputs of FOA's ESN neural network model 1-C, multiple vehicle detector outputs are used as vehicle detection module inputs, multiple FOA's ESN neural network model outputs and vehicle detection module outputs are used as the corresponding inputs of FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 2, the outputs of FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 2 are used as TDL beat delay device 3 inputs, the outputs of TDL beat delay device 3 are used as the corresponding inputs of FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 2, and the outputs of FOA's NARX neural network model of interval intuitionistic fuzzy numbers-FOA's recursive fuzzy neural network model 2 are used as the detected road section traffic flow value.

3. The method for intelligent traffic monitoring of the Internet of Things according to claim 1, characterized in that: The regional traffic intelligent monitoring subsystem includes a plurality of road section traffic intelligent monitoring modules, a TDL beat delay device 4 and an ESN neural network model of FOA of interval intuitionistic fuzzy numbers-a recursive fuzzy neural network model of FOA; the outputs of the plurality of road section traffic intelligent monitoring modules are respectively used as the corresponding inputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA, the outputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA are used as the inputs of the TDL beat delay device 4, the outputs of the TDL beat delay device 4 are used as the corresponding inputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA, and the outputs of the ESN neural network model of FOA of interval intuitionistic fuzzy numbers-the recursive fuzzy neural network model of FOA of interval intuitionistic fuzzy numbers respectively constitute 4 interval intuitionistic fuzzy number prediction values ​​of the traffic flow in 4 directions of east, south, west and north, wherein the interval intuitionistic fuzzy number prediction value of the traffic flow in one direction is [<s,t> ; m; r], s and t constitute the interval fuzzy number prediction value of the traffic flow in this direction:<s,t> , m is the credibility of the interval fuzzy number prediction value of the traffic flow in this direction, r is the unreliability of the interval fuzzy number prediction value of the traffic flow in this direction, and the sum of m and r is less than or equal to 1.

4. The method for intelligent traffic monitoring of the Internet of Things according to claim 1, characterized in that: The denoising autoencoder-metabolism grey GM (1,1) model is a denoising autoencoder and a metabolic grey GM (1,1) model connected in series, and the denoising autoencoder output is used as the input of the metabolic grey GM (1,1) model; The NARX neural network model of FOA - metabolic gray GM (1,1) model 1 and the NARX neural network model of FOA - metabolic gray GM (1,1) model 2 have the same structure, that is, the NARX neural network model of FOA and the metabolic gray GM (1,1) model are connected in series, and the output of the NARX neural network model of FOA is used as the input of the metabolic gray GM (1,1) model; GRNN neural network model A of FOA is the GRNN neural network model optimized by FOA; NARX neural network model of FOA for interval intuitionistic fuzzy numbers - recursive fuzzy neural network model of FOA 1 is a series connection of NARX neural network model of FOA and recursive fuzzy neural network model of FOA, and the recursive fuzzy neural network model of FOA is an optimized recursive fuzzy neural network model of FOA; The structures of NARX neural network model 1 of FOA, NARX neural network model 2 of FOA, and NARX neural network model 3 of FOA are the same as that of NARX neural network model of FOA, and all of them are FOA optimized NARX neural network models.

5. The method for intelligent traffic monitoring of the Internet of Things according to claim 2, characterized in that: The ESN neural network models 1-C of FOA are all FOA optimized ESN neural network models; the GRNN neural network models 1-n of FOA are all FOA optimized GRNN neural network models; the environmental level subtraction clustering classifier takes each historical data point of the traffic environment parameter as a potential clustering center, so that the clustering result is independent of the dimension of the problem; The NARX neural network model of FOA for interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model 2 of FOA is a series connection of the NARX neural network model of FOA and the recursive fuzzy neural network model of FOA, the NARX neural network model of FOA is the FOA optimized NARX neural network model, and the recursive fuzzy neural network model of FOA is the FOA optimized recursive fuzzy neural network model.

6. The method for intelligent traffic monitoring of the Internet of Things according to claim 3, characterized in that: The ESN neural network model of FOA for interval intuitionistic fuzzy numbers - the recursive fuzzy neural network model of FOA is a series connection of the ESN neural network model of FOA and the recursive fuzzy neural network model of FOA, the output of the ESN neural network model of FOA is used as the input of the recursive fuzzy neural network model of FOA, the ESN neural network model of FOA is the FOA optimized ESN neural network model, and the recursive fuzzy neural network model of FOA is the FOA optimized recursive fuzzy neural network model.

7. A traffic intelligent monitoring device based on the traffic intelligent monitoring method of the Internet of Things according to any one of claims 1 to 6, characterized in that: It includes an environmental parameter detection terminal, a traffic signal measurement and control terminal, a traffic scene monitoring terminal, a traffic monitoring gateway node, a traffic monitoring cloud platform, a traffic management monitoring terminal and a traffic monitoring mobile phone APP. The environmental parameter detection terminal is responsible for collecting environmental parameter information of the detected traffic section, and the traffic signal measurement and control terminal is responsible for measuring the traffic flow at the traffic intersection and controlling the traffic signal control device at the intersection. Multiple environmental parameter detection terminals interact with each other through the traffic monitoring gateway node and the traffic scene monitoring terminal, the traffic monitoring cloud platform, the traffic management monitoring terminal and the traffic monitoring mobile phone APP; the traffic scene monitoring terminal is provided with a section traffic intelligent monitoring module to realize the intelligent monitoring of the detected section traffic and obtain the traffic flow value of the detected section under the influence of environmental parameters; the traffic management monitoring terminal is provided with a regional traffic intelligent monitoring subsystem, which takes the traffic flow value of the detected section detected by multiple section traffic intelligent monitoring modules as input to realize the intelligent monitoring of the detected regional traffic, and outputs 16 parameters to constitute 4 interval intuitive fuzzy number prediction values ​​of the traffic flow in 4 directions of east, south, west and north; the traffic monitoring mobile phone APP and the traffic management monitoring terminal access the traffic monitoring cloud platform through the 5G network to realize the collection and monitoring of traffic parameters.

8. The intelligent traffic monitoring device of the intelligent traffic monitoring method of the Internet of Things according to claim 7 is characterized in that: The environmental parameter detection terminal includes temperature and humidity, visibility, wind speed, wind direction, rainfall, dust and harmful gas sensors for collecting traffic road environmental parameters and corresponding signal conditioning circuits, STM32 microcontrollers and 5G modules.

Images