A signalized intersection illumination setting method

Abstract

The present application relates to a kind of signal intersection illumination setting method, belong to traffic management control technical field, solve the existing street lamp illumination according to fixed value setting, cannot adapt to different signal intersection and different traffic environment, lead to the speed variation of driver when entering and exiting signal intersection, affect driving safety Problem.The method includes: collecting the data information of driver and signal intersection, obtains the key factor of influencing driving behavior based on the data collected, forms training sample set based on speed variation and corresponding key factor, constructs speed variation and illumination correlation prediction model and obtains trained prediction model using training sample set, other key factor data except illumination of signal intersection is collected and input into correlation model, select the illumination when the speed variation is minimum as the optimal illumination of signal intersection.The method sets the optimal illumination of signal intersection according to signal intersection and traffic environment data, improves the safety of driving.

Description

Technical Field

[0001] This invention relates to the field of traffic management and control technology, and in particular to a method for setting illumination at signalized intersections. Background Technology

[0002] With my country's economic growth, urban transportation has developed rapidly, and urban roads have become increasingly complex. Urban roads are road networks connected by numerous signalized intersections, which are key to urban traffic and also areas prone to traffic accidents. Therefore, studying the key factors affecting driver behavior at signalized intersections and optimizing these factors can significantly reduce traffic accidents and is of great importance.

[0003] To improve driver safety at night, urban roads use streetlights spaced at regular intervals to enhance visibility. Currently, the illuminance of these streetlights is a fixed value set according to national standards, which cannot meet the driving needs of drivers at different signalized intersections and in varying traffic conditions. This results in significant speed variations when entering and exiting signalized intersections, impacting driving safety. Summary of the Invention

[0004] Based on the above analysis, the present invention aims to provide a method for setting the illuminance of signalized intersections, in order to solve the problem that the existing street light illuminance is set according to a fixed value, which cannot meet the driving needs of drivers at different signalized intersections and in different traffic environments, resulting in large changes in vehicle speed when entering and exiting signalized intersections, thus affecting driving safety.

[0005] This invention provides a method for setting illuminance at a signalized intersection, the method comprising the following steps:

[0006] Collect basic driver information, driver driving information, traffic environment data at signalized intersections, vehicle speed and illumination;

[0007] Based on the collected vehicle speed, the driver's driving behavior is obtained. The driver's basic information, driver's driving situation information, traffic environment data and illuminance corresponding to each driving behavior are used as the influencing factors of driving behavior. Logit regression is performed on the driver's driving behavior and the influencing factors of driving behavior to obtain the key factors affecting driving behavior.

[0008] The vehicle speed change is obtained based on each driving behavior. The vehicle speed change and the key factors corresponding to the driving behavior are used as a sample to form a training sample set.

[0009] A prediction model relating vehicle speed change and illumination is constructed, and the prediction model is trained using a training sample set to obtain a trained prediction model.

[0010] Collect data on other key factors besides illuminance at the signalized intersection, select several illuminance values ​​and other key factor data to input into the prediction model to obtain vehicle speed changes, obtain the relationship function between vehicle speed changes and illuminance, and select the illuminance at the minimum vehicle speed change as the optimal illuminance of the signalized intersection.

[0011] Furthermore, the sample set is divided according to weather conditions to form a training sample subset; the vehicle speed change and illumination correlation prediction model includes prediction sub-models under various weather conditions; during training, the prediction sub-models are trained using the training sample subsets under the corresponding weather conditions to obtain each trained prediction sub-model.

[0012] Furthermore, for each prediction sub-model, the sample subset is divided into a training subset and a test subset. After the prediction sub-model under the corresponding weather conditions is trained using the training subset, the sample data of the test subset is input into the trained prediction sub-model under the corresponding weather conditions to obtain the predicted value. The mean square error (MSE) between the true value and the predicted value is calculated. When the MSE is less than the error threshold, the trained prediction sub-model under that weather condition is obtained.

[0013] Furthermore, Logit regression analysis was performed on driver behavior and its influencing factors, revealing that the key factors affecting driving behavior include:

[0014] Obtain a utility function for driving behavior, the utility function including deterministic and random terms;

[0015] Assuming that the determinant has a linear relationship with each influencing factor of driving behavior, the functional relationship between the determinant and each influencing factor is obtained, that is, the determinant is the weighted sum of the influencing factors;

[0016] According to the utility maximization theory, the probability of a driver choosing a certain driving behavior is obtained;

[0017] Based on the probability, the log-likelihood function of the Logit model is obtained; the weight values ​​of each influencing factor when the log-likelihood function is maximized are calculated, and the influencing factors with weight values ​​greater than the parameter threshold are selected as key factors.

[0018] Furthermore, the data collected include: basic driver information, driver driving information, signalized intersection range, traffic environment, vehicle driving status, and illuminance at the signalized intersection.

[0019] A questionnaire survey was used to obtain basic information about drivers and their driving habits.

[0020] Traffic environment data and vehicle speed are obtained based on video surveillance;

[0021] Illuminance data at signal intersections was collected using a illuminometer during the period when streetlights were on.

[0022] Furthermore, the traffic environment data includes weather conditions, traffic flow, street light spacing, and distance data between street lights and signalized intersections; the vehicle speed includes vehicle speed data before, at, and after passing the signalized intersection.

[0023] Furthermore, the key factors include illuminance data at the signalized intersection, weather conditions, traffic flow, street light spacing, and the distance between the street lights and the signalized intersection.

[0024] Furthermore, the illuminance data collected at the signal intersection during the street light operation period based on the illuminance meter includes:

[0025] The signalized intersection area is divided into several connected 5*5 meter square regions. The illuminance value at the center of each region is collected, and the arithmetic mean of all illuminance values ​​within the signalized intersection area is obtained as the illuminance data at the signalized intersection.

[0026] Furthermore, the weather conditions include five types: sunny, cloudy, rainy, snowy, and foggy.

[0027] Furthermore, the set of driving behaviors includes decelerating through, passing at a constant speed, accelerating through, and decelerating first and then accelerating through.

[0028] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects:

[0029] 1. This invention changes the current mode of setting street light illuminance to a fixed value. With the goal of minimizing the change in vehicle speed when the driver crosses a signalized intersection, it sets the optimal illuminance of the signalized intersection based on different signalized intersections and different traffic environment data, thereby improving driving safety.

[0030] 2. Based on the collected vehicle driving status, a set of driver driving behaviors is obtained. Based on the collected driver basic information, driver driving situation information, traffic environment and illuminance data at signalized intersections, factors affecting driving behavior are obtained. The key factors affecting driving behavior during the operation of streetlights at signalized intersections are obtained through a Logit regression model. By using a Logit regression model to obtain the key factors affecting driving behavior, the key factors and the corresponding vehicle speed changes are used as samples to train the prediction model, thus making the trained model more accurate in prediction. At the same time, by obtaining the key factors instead of using all factors to train the model, the training efficiency of the model is also improved.

[0031] 3. Based on the analysis of the collected data, the set of driving behaviors chosen by drivers when passing through signalized intersections under green light conditions is obtained as follows: decelerating, maintaining a constant speed, accelerating, and decelerating first and then accelerating. By analyzing the collected data, driving behaviors with very low probability of being chosen by drivers are filtered out, and a set of driving behaviors is obtained. Logit regression is performed on the influencing factors corresponding to this set to obtain the key factors affecting driving behavior, thus making the obtained key factors more accurate. At the same time, Logit regression is only performed on the influencing factors corresponding to driving behaviors in this set, rather than performing Logit regression on all driving behaviors, thus improving the efficiency of obtaining key factors.

[0032] In this invention, the above-described technical solutions can be combined with each other to achieve more preferred combinations. Other features and advantages of this invention will be set forth in the following description, and some advantages may become apparent from the description or be learned by practicing the invention. The objects and other advantages of this invention can be realized and obtained from what is particularly pointed out in the description and drawings. Attached Figure Description

[0033] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0034] Figure 1 This is a flowchart of a method for setting illumination at signalized intersections, as described in an embodiment of the present invention. Detailed Implementation

[0035] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and are used together with the embodiments of the present invention to illustrate the principles of the present invention, but are not intended to limit the scope of the present invention.

[0036] A specific embodiment of the present invention discloses a method for setting the illumination at a signalized intersection.

[0037] like Figure 1 As shown, the method includes the following steps:

[0038] Step S1: Collect basic driver information, driver driving information, traffic environment data at the signalized intersection, vehicle speed and illumination.

[0039] Step S2: Based on the collected vehicle speed, obtain the driver's driving behavior. Take the driver's basic information, driver's driving situation information, traffic environment data and illuminance corresponding to each driving behavior as the influencing factors of driving behavior. Perform Logit regression on the driver's driving behavior and the influencing factors of driving behavior to obtain the key factors affecting driving behavior.

[0040] Step S3: Obtain the vehicle speed change based on the vehicle speed of each driving behavior, and take the vehicle speed change and the key factors corresponding to the driving behavior as a sample to form a training sample set.

[0041] Step S4: Construct a prediction model for the correlation between vehicle speed change and illumination, and train the prediction model using the training sample set to obtain a trained prediction model.

[0042] Step S5: Collect data on other key factors of the signalized intersection besides illuminance, select several illuminance values ​​and other key factor data and input them into the prediction model to obtain vehicle speed changes, obtain the relationship function between vehicle speed changes and illuminance, and select the illuminance with the smallest vehicle speed change as the optimal illuminance of the signalized intersection.

[0043] In step S1, the following steps are performed to obtain the collected data.

[0044] Step S1.1: Use a questionnaire to obtain basic driver information and driver driving information. Basic driver information mainly includes the driver's gender, age, and driving experience; driver driving information mainly includes the driver's daily driving time, weekly driving days, and type of vehicle driven.

[0045] Step S1.2: Acquire traffic environment data and vehicle speed based on video surveillance. The traffic environment data includes weather conditions, traffic flow, street light spacing, and distance data between street lights and signalized intersections; the vehicle speed includes vehicle speed data before, at, and after passing the signalized intersection.

[0046] Specifically, the weather conditions include five types: sunny, cloudy, rainy, snowy, and foggy.

[0047] The vehicle speed data before / after passing the signalized intersection is the average vehicle speed within a range of 80 to 60 meters from the driver at the signalized intersection; the vehicle speed data at the point where the vehicle passes the signalized intersection is the vehicle speed when the driver is 5 meters from the signalized intersection.

[0048] The traffic volume refers to the total traffic volume of all traffic in a one-way lane, including left turns, right turns, and straight-ahead traffic.

[0049] The street light spacing is the average value of the street light spacing within the video acquisition range of the four directions at the signal intersection.

[0050] The distance between the street light and the signalized intersection is the average of the distances between the street light and the intersection within the video capture range of the four directions at the signalized intersection.

[0051] Step S1.3: Collect illuminance data at signal intersections during the period when streetlights are on using an illuminance meter.

[0052] Specifically, the signalized intersection area is divided into several connected 5*5 meter square areas. The illuminance value at the center of each area is collected, and the arithmetic mean of all illuminance values ​​within the signalized intersection area is obtained as the illuminance data at the signalized intersection.

[0053] In step S2, the following steps are performed to obtain the key factors affecting driving behavior.

[0054] Step S2.1: Obtain the driver's driving behavior based on the collected vehicle speed.

[0055] Specifically, driving behaviors can be categorized as decelerating through, maintaining a constant speed, accelerating through, decelerating then accelerating through, and accelerating then decelerating through. It's understandable that since vehicle speed includes data from before, at, and after passing the signalized intersection, driving behaviors can be derived by comparing these speed data. After obtaining the driving behaviors, statistical analysis is performed, and behaviors with a percentage below a threshold are removed, resulting in a set of driving behaviors. This set consists of decelerating through, maintaining a constant speed, accelerating through, and decelerating then accelerating through.

[0056] Step S2.2: Take the driver's basic information, driver's driving situation information, traffic environment data and illuminance corresponding to each driving behavior as the influencing factors of driving behavior, and perform Logit regression on the driver's driving behavior and the influencing factors of driving behavior to obtain the key factors affecting driving behavior.

[0057] According to stochastic utility theory, different driving behaviors at a signalized intersection will result in different utilities for the driver, and drivers will tend to choose the behavior that maximizes utility. Each driver's behavior choice corresponds to a utility function, and each utility function consists of deterministic and stochastic terms.

[0058] Specifically, Logit regression was performed on driver behavior and the factors influencing it, revealing that the key factors affecting driving behavior include:

[0059] (1) Obtain the utility function of driving behavior, wherein the utility function includes deterministic terms and random terms.

[0060] For example, suppose driver n's set of driving behaviors when passing through a signalized intersection is Cn, and the utility function of the driver choosing the i-th behavior is...

[0061] U in =V in +ε in ,

[0062] Among them, V inThe term ε is a determination calculated based on collected data, including basic driver information, driver driving information, traffic environment data, and illuminance. in For random items, there are the effects of data that cannot be collected and the random errors caused by the collected data.

[0063] (2) Assuming that the definite term has a linear relationship with each driving behavior influencing factor, the functional relationship between the definite term and each influencing factor is obtained, that is, the definite term is the weighted sum of each influencing factor.

[0064] Assume V in =V(x) in ), where x in Factors influencing driving behavior include driver gender, age, driving experience, daily driving time, weekly driving days, type of vehicle driven, weather conditions, traffic volume, street light spacing, distance between street lights and signalized intersections, and illuminance.

[0065] We can obtain:

[0066] V in =β1x in1 +β2x in2 +β3x in3 +…+β K x inK ,

[0067] Where, x inK Let β be the value of the k-th influencing factor that determines whether driver n chooses the i-th behavior when crossing a signalized intersection. K These are its estimated parameters.

[0068] (3) Based on the utility maximization theory, the probability of a driver choosing a certain driving behavior is obtained.

[0069] According to the utility maximization theory, the driving behaviors chosen by drivers are those that maximize utility, therefore...

[0070] The probability P of driver n choosing the i-th driving behavior in for:

[0071]

[0072] make Then there is Assumption If it follows a double exponential distribution with parameters (0,1), then P in for:

[0073]

[0074] Since the difference between two independent double-distributed probability variables follows a double exponential distribution, we can obtain P. in The general form is:

[0075]

[0076] (4) Based on the probability, obtain the log-likelihood function of the Logit model; calculate the weight values ​​of each influencing factor when the log-likelihood function is maximized, and select the influencing factors with weight values ​​greater than the parameter threshold as key factors.

[0077] To determine the weight β of each influencing factor, we assume the sample size is N, and δ in δ represents the probability variable for the driver's choice of driving behavior. in The value can be 0 or 1. If driver n chooses the i-th driving behavior, δ in =1, otherwise, δ in =0.

[0078] Therefore, the likelihood function of the multinomial Logit model can be obtained as follows:

[0079]

[0080] For L * Taking the logarithm, we get L. * The log-likelihood function L:

[0081]

[0082] We need to find the value of L. * β is at its maximum, therefore, take the partial derivative of β with respect to L:

[0083]

[0084] in The above formula can be simplified to:

[0085]

[0086] Find the weight β when L is maximized, and then obtain the weight value of each influencing factor. Select the influencing factors whose weight value is greater than the parameter threshold as the key factors.

[0087] The parameter thresholds can be set according to actual needs.

[0088] By performing Logit regression on driver behavior and the factors influencing it, the key factors identified were illuminance data at the signalized intersection, weather conditions, traffic flow, street light spacing, and the distance between the street lights and the signalized intersection.

[0089] In step S3, the vehicle speed change ΔV = V s-V0, where V0 is the average speed of the vehicle before and after passing through the signalized intersection, V S This represents the vehicle speed at the signalized intersection. The speed changes and corresponding key factors of driving behavior are treated as a single sample, and all samples are used to form a training sample set. This can be understood as video recording of speed changes and key factors for multiple vehicles during peak or off-peak traffic periods, with each vehicle's speed change and key factors treated as a single sample.

[0090] In step S4, the following steps are performed to obtain the trained vehicle speed change and illumination correlation prediction model.

[0091] A predictive model for the relationship between vehicle speed change and illumination is constructed. Based on the backpropagation (BP) algorithm in neural networks, a latent function ΔV ~ f(x) is constructed, with vehicle speed change ΔV as the dependent variable and key factors as independent variables. i ), where x i These are key factors influencing driving behavior. The implicit function is the predictive model for the relationship between vehicle speed change and illumination.

[0092] The prediction model is trained using the training sample set to obtain a trained prediction model.

[0093] To improve the prediction accuracy of the model, preferably, the sample set is divided according to weather conditions to form multiple training sample subsets; the vehicle speed change and illumination correlation prediction model includes prediction sub-models under various weather conditions; during training, the prediction sub-models are trained using the training sample subsets under the corresponding weather conditions to obtain each trained prediction sub-model.

[0094] Specifically, the weather conditions include five types: sunny, cloudy, rainy, snowy, and foggy.

[0095] Specifically, the training sample set in step 3 can be divided according to weather conditions to form multiple sample subsets, including sample subsets under sunny weather, sample subsets under cloudy weather, sample subsets under rainy weather, sample subsets under snowy weather, and sample subsets under foggy weather.

[0096] Specifically, for each prediction sub-model, the sample subset is divided into a training subset and a test subset. After the prediction sub-model under the corresponding weather conditions is trained using the training subset, the sample data of the test subset is input into the trained prediction sub-model under the corresponding weather conditions to obtain the predicted value. The mean square error (MSE) between the true value and the predicted value is calculated. When the MSE is less than the error threshold, the trained prediction sub-model under that weather condition is obtained.

[0097] In implementation, the sample subset is divided into a training sample subset and a test sample subset in a 5:1 ratio. This allows for the generation of training and test sample subsets under different weather conditions. Because weather changes are unpredictable and significantly impact driving behavior, a prediction sub-model is constructed for conditions with fixed weather conditions to eliminate errors caused by weather variations and enable the prediction model to obtain more accurate predictions.

[0098] Based on the prediction model that correlates vehicle speed changes with illumination, prediction sub-models are constructed according to different weather conditions.

[0099] For example, by taking the training subsamples under sunny weather conditions and substituting them into the aforementioned hidden function, we obtain the prediction submodel ΔV~f'(x) under sunny weather conditions. i At this time, x i This includes illuminance data, traffic flow, street light spacing, and the distance between street lights and the signalized intersection.

[0100] The training sample subset data is input into the prediction sub-model for training, resulting in the trained prediction sub-model.

[0101] For example, a subset of training sample data under sunny weather conditions is input into the sunny weather prediction sub-model for training, resulting in the trained sunny weather prediction sub-model.

[0102] The test subset sample data is input into the trained prediction sub-model under the corresponding weather conditions to obtain the predicted value. The mean square error (MSE) between the true value and the predicted value is calculated. When the MSE is less than the error threshold, the trained prediction sub-model under the weather conditions is obtained.

[0103] For example,

[0104] Where N is the number of samples in the test subset, ΔV i To predict the true value of the i-th ΔV in the sub-model, E(i) is the predicted value of the i-th ΔV.

[0105] If MSE < 0.01, training ends and a trained prediction sub-model is obtained; otherwise, the MSE result is used as feedback value to update the parameters of the trained prediction sub-model, and training continues until MSE < 0.01.

[0106] Specifically, step S5 includes:

[0107] (1) Collect data on other key factors at the signal intersection, excluding illuminance.

[0108] (2) Select several illuminance values ​​evenly within the preset illuminance range, and input any illuminance value and other key factor data into the trained prediction model to obtain the vehicle speed change.

[0109] (3) Based on the vehicle speed change corresponding to each illuminance value, the relationship function between vehicle speed change and illuminance is obtained by fitting.

[0110] (4) Based on the relationship function, take the illuminance value when the vehicle speed changes the least as the optimal illuminance value.

[0111] Understandably, the greater the change in vehicle speed, the less safe it is to drive; therefore, the change in vehicle speed should be minimized. The illuminance corresponding to the minimum change in vehicle speed is selected as the optimal illuminance for the signalized intersection.

[0112] Preferably, in step (2), the weather conditions at the signalized intersection are determined based on key factors. The selected illuminance value, along with the collected traffic flow, street light spacing, and distance between street lights and the signalized intersection, are input into the prediction sub-model under the corresponding weather conditions to obtain vehicle speed changes. By obtaining the vehicle speed change corresponding to each selected illuminance value, the relationship function between vehicle speed change and illuminance can be obtained. The illuminance corresponding to the minimum vehicle speed change is selected as the optimal illuminance for the signalized intersection.

[0113] Compared with existing technologies, this invention provides a method for setting optimal illuminance at signalized intersections. First, it changes the current model of setting streetlight illuminance to a fixed value, aiming to minimize the change in vehicle speed when crossing a signalized intersection. It sets the optimal illuminance for each signalized intersection based on data from different intersections and traffic environments, thus improving driving safety. Second, it obtains a set of driver behaviors based on collected vehicle driving states. Based on collected driver basic information, driver driving situation information, traffic environment, and illuminance data at the signalized intersection, it identifies factors influencing driving behavior. A Logit regression model is used to identify key factors affecting driving behavior during the operation of streetlights at the signalized intersection. These key factors and their corresponding speed changes are used as samples to train a prediction model, thereby improving the accuracy of the trained model. The measurement is more accurate, and by acquiring key factors instead of using all factors to train the model, the training efficiency is improved. Furthermore, based on the analysis of collected data, the set of driving behaviors chosen by drivers when crossing signalized intersections under green light conditions is obtained: decelerating, maintaining a constant speed, accelerating, and decelerating then accelerating. By analyzing the collected data and filtering out driving behaviors with very low probabilities, a set of driving behaviors is obtained. Logit regression is then performed using the influencing factors corresponding to this set to obtain the key factors affecting driving behavior, thus making the obtained key factors more accurate. Moreover, Logit regression is only performed on the influencing factors corresponding to driving behaviors within this set, rather than performing Logit regression on all driving behaviors, thus improving the efficiency of acquiring key factors.

[0114] Those skilled in the art will understand that all or part of the processes of the methods described in the above embodiments can be implemented by a computer program instructing related hardware, and the program can be stored in a computer-readable storage medium. The computer-readable storage medium may be a disk, optical disk, read-only memory, or random access memory, etc.

[0115] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A signalized intersection illumination setting method, characterized by, The method includes the following steps: Collect basic driver information, driver driving information, traffic environment data at signalized intersections, vehicle speed and illumination; Based on the collected vehicle speed, the driver's driving behavior is obtained. The driver's basic information, driver's driving situation information, traffic environment data and illuminance corresponding to each driving behavior are used as the influencing factors of driving behavior. Logit regression is performed on the driver's driving behavior and the influencing factors of driving behavior to obtain the key factors affecting driving behavior. The vehicle speed change is obtained based on each driving behavior. The vehicle speed change and the key factors corresponding to the driving behavior are used as a sample to form a training sample set. A prediction model relating vehicle speed change and illumination is constructed, and the prediction model is trained using a training sample set to obtain a trained prediction model. Collect data on other key factors besides illuminance at the signalized intersection, select several illuminance values ​​and other key factor data and input them into the prediction model to obtain vehicle speed changes, obtain the relationship function between vehicle speed changes and illuminance, and select the illuminance with the smallest vehicle speed change as the optimal illuminance of the signalized intersection. Logit regression analysis of driver behavior and its influencing factors revealed key factors affecting driving behavior, including: Obtain a utility function for driving behavior, the utility function including deterministic and random terms; Assuming that the determinant has a linear relationship with each influencing factor of driving behavior, the functional relationship between the determinant and each influencing factor is obtained, that is, the determinant is the weighted sum of the influencing factors; According to the utility maximization theory, the probability of a driver choosing a certain driving behavior is obtained; Based on the probability, the log-likelihood function of the Logit model is obtained; the weight values ​​of each influencing factor when the log-likelihood function is maximized are calculated, and the influencing factors with weight values ​​greater than the parameter threshold are selected as key factors.

2. The signalized intersection illumination setting method of claim 1, wherein, The sample set is divided according to weather conditions to form multiple training sample subsets; the vehicle speed change and illumination correlation prediction model includes prediction sub-models under various weather conditions; during training, the prediction sub-models are trained using the training sample subsets under the corresponding weather conditions to obtain each trained prediction sub-model.

3. The signalized intersection illumination setting method of claim 2, wherein, For each prediction sub-model, the sample subset is divided into a training subset and a test subset. After the prediction sub-model under the corresponding weather conditions is trained using the training subset, the sample data of the test subset is input into the trained prediction sub-model under the corresponding weather conditions to obtain the predicted value. The mean square error (MSE) between the true value and the predicted value is calculated. When the MSE is less than the error threshold, the trained prediction sub-model under that weather condition is obtained.

4. The signalized intersection illumination setting method of claim 1, wherein The collection of basic driver information, driver driving information, traffic environment data at signalized intersections, vehicle speed, and illuminance includes: A questionnaire survey was used to obtain basic information about drivers and their driving habits. Traffic environment data and vehicle speed are obtained based on video surveillance; Illuminance at signal intersections was collected using a illuminometer while streetlights were on.

5. The signalized intersection illumination setting method of claim 4, wherein, The traffic environment data includes weather conditions, traffic flow, street light spacing, and distance data between street lights and signalized intersections; the vehicle speed includes vehicle speed data before, at, and after passing the signalized intersection.

6. The signalized intersection illumination setting method of claim 1, wherein, The key factors are illuminance data at the signalized intersection, weather conditions, traffic flow, street light spacing, and the distance between the street lights and the signalized intersection.

7. The signalized intersection illumination setting method of claim 4, wherein, The illuminance data collected at the signal intersection during the period when streetlights are on, based on the illuminance meter, includes: The signalized intersection area is divided into several connected 5*5 meter square regions. The illuminance value at the center of each region is collected, and the arithmetic mean of all illuminance values ​​within the signalized intersection area is obtained as the illuminance data at the signalized intersection.

8. The signalized intersection illumination setting method of claim 5, wherein, The weather conditions include five categories: sunny, cloudy, rainy, snowy, and foggy.

9. The signalized intersection illumination setting method of claim 1, wherein, The set of driving behaviors includes decelerating through, passing at a constant speed, accelerating through, and decelerating first and then accelerating through.

Images