Traffic environment noise data analysis method and system based on dynamic perception

Abstract

The invention discloses a traffic environment noise data analysis method and system based on dynamic perception, and belongs to the technical field of noise monitoring. The system comprises a dynamic sensing module, a noise analysis module, an influence evaluation module and a visualization module. The dynamic sensing module collects a traffic map, survey data and video images of different roads, builds a three-dimensional scene according to the traffic map and performs dynamic mapping in combination with the video images; the noise analysis module is used for extracting picture data and sound data in the video image and fitting an influence relational expression; drawing a noise curve according to the influence relational expression, calculating a difference coefficient and dividing an influence area; the influence evaluation module associates all questionnaire records according to influence areas to which filling addresses belong and then calculates the sum of noise scores; calculating an influence index of each influence area by combining the difference coefficient, and setting an attention area; and the visualization module displays the position of each attention area in the three-dimensional scene and the influence index, and at the same time, early warning is performed to prompt field staff to troubleshoot each attention area.

Description

Technical Field

[0001] This invention relates to the field of noise monitoring technology, specifically to a method and system for analyzing traffic environmental noise data based on dynamic sensing. Background Technology

[0002] Currently, the industry generally adopts physical sensing and standardized models to construct noise distribution maps through quantitative analysis and spatial mapping to guide noise reduction. However, such methods still have a series of systemic bottlenecks in dynamic and highly heterogeneous traffic scenarios, and cannot meet the needs of smart cities for precise noise control.

[0003] Currently, noise monitoring methods have certain structural limitations. First, data acquisition is limited by the spatiotemporal coverage blind spots of static hardware facilities, making it difficult to capture the noise mutation characteristics during the dynamic evolution of traffic flow in real time, resulting in a high rate of missed detections of abnormal events. Second, evaluation models adhere to the single-dimensional analysis of physical acoustic indicators, ignoring the spatiotemporal differences in human subjective perception data, causing a serious misalignment between the allocation of governance resources and the actual pain points of residents. Finally, traditional algorithms lack the ability to identify patterns in sudden road noise events and cannot distinguish the periodic anomalies caused by sudden road condition abnormalities, leaving road maintenance in a passive response mode for a long time. These shortcomings collectively lead to a predicament of high cost, low efficiency, and weak targeting in environmental noise governance, urgently requiring a breakthrough in a new generation of noise analysis paradigms. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for analyzing traffic environmental noise data based on dynamic perception, so as to solve the problems mentioned in the background art.

[0005] To address the aforementioned technical problems, this invention provides a method for analyzing traffic environmental noise data based on dynamic perception, comprising:

[0006] S100 collects traffic maps and survey data, and gathers video images of different road locations through monitoring equipment. Digital twin technology is used to build a 3D scene based on the traffic map, and then dynamically maps it using real-time video images.

[0007] A traffic map is a route map of a transportation network that includes the spatial layout of roads and the prescribed directions of travel.

[0008] The survey data includes different questionnaire records, each of which includes at least the address provided and noise ratings for each time period. Noise ratings range from 1 to 5, with higher scores indicating a greater negative perception of noise during the corresponding time period.

[0009] The specific times for each period are: early morning (6:00-9:00), morning (9:00-12:00), noon (12:00-14:00), afternoon (14:00-18:00), evening (18:00-22:00) and night (22:00-6:00).

[0010] The noise score is given by the questionnaire respondents and refers to an individual's adverse reaction to noise, including various situations such as "dissatisfaction", "disturbance", "annoyance" and "interference" caused by noise.

[0011] Noise rating is a quantitative representation of the degree of noise annoyance. It is an important psychological indicator that can more comprehensively reflect the impact of noise on human life.

[0012] Building a 3D scene and performing dynamic mapping specifically includes:

[0013] S101. Preset model library and put in 3D models containing different types and specifications of vehicles. Each model is labeled with actual size parameters and feature contour data.

[0014] Vehicle types can be categorized by function and purpose into dump trucks, buses, freight trucks, and transport vehicles. Specifications can be simply classified as small, medium, and large vehicles, or further subdivided.

[0015] S102. By accessing road topology data provided by traffic maps through digital twin technology, a 3D scene with terrain and road markings is generated in the 3D engine.

[0016] S103. Based on the road location of the monitoring equipment, mark the corresponding location in the three-dimensional scene and synchronously map the road area in the video images collected by each monitoring equipment.

[0017] S104. Identify all vehicle objects within the road area in the video image using a target detection algorithm, and extract the appearance size parameters, shape contour features, and real-time position coordinates of each vehicle object.

[0018] S105. Calculate the similarity between the extracted appearance size parameters and shape contour features and the actual size parameters and feature contour data of each 3D model in the model library.

[0019] S106. Match the three-dimensional model with the highest similarity to each vehicle object, and set the displacement direction and velocity of the corresponding matched three-dimensional model according to the extracted real-time position coordinate changes.

[0020] The maximum similarity can be calculated by summing the similarities of the actual size parameters and feature contour data, or by using a weighted average calculation. The weights of the size parameters and contour data are set in advance by the staff.

[0021] S107. Map the matched 3D models of each vehicle object to the corresponding coordinate positions in the 3D scene, and dynamically drive these 3D models to move along the path while maintaining position and attitude synchronization.

[0022] By using digital twin technology to construct a three-dimensional virtual model of the real traffic environment, dynamic real-time mapping of traffic noise sources can be achieved.

[0023] This reduces the delay and error in environmental data acquisition, providing a precise spatiotemporal basis for subsequent noise analysis and improving monitoring efficiency and data integration.

[0024] S200. Extract image and audio data from the video image and fit the influence relationship. Plot the predicted noise curve and the measured noise curve respectively, calculate the difference coefficient based on the deviation between the two noise curves, and divide the influence area. Specifically, this includes:

[0025] S201. Extract the historical data of each monitoring device. video clip The data is then analyzed and separated into video and audio data. The correlation between these two data points is analyzed in time sequence to fit influence relationships for each monitoring device. Specifically, this includes:

[0026] S2011. Bind the video data and audio data to the timeline separately, and set them evenly. Each time point. Preset conduction duration. Each time point on the audio timeline is delayed by a certain duration compared to the corresponding time point on the video timeline. .

[0027] The conduction time is used to compensate for the time error caused by the length of the propagation path of traffic noise from generation to collection. Each monitoring device has a different preset conduction time, which is based on the actual distance between the monitoring device installation location and the road being filmed, as well as the speed of sound propagation in different environments.

[0028] S2012, Video clips The image is parsed into continuous image frames, matching image frames at the same time point on the timeline, identifying vehicle objects contained in each image frame and matching them with 3D models in the model library.

[0029] S013, Obtain time points with the same ordinal number in the video and audio timelines. and Based on the time points in the video clips The instantaneous velocity of each vehicle object in front and behind is calculated based on its positional changes, and all vehicle objects are classified according to the matched 3D model.

[0030] S2014, Time Point The number of all vehicle objects in each category and their average instantaneous speed are used as independent variables. (Time point) The noise value of the sound is used as the dependent variable, and the independent and dependent variables at the same ordinal time points are packaged into a sample.

[0031] S2015. Establish the influence relationship formula, and input the following respectively. The independent variables in each sample are used, and the difference between the output result and the dependent variable is used as the difference coefficient for the corresponding sample. The influence relationship is as follows:

[0032] ;

[0033] In the formula, To predict noise, The number of classes at a given time point. For the first The total number of vehicle objects under this class. For the first The average instantaneous speed of all vehicle objects under this class.

[0034] and The first The corresponding three-dimensional model has preset aerodynamic noise coefficient and mechanical friction noise coefficient. This is a background noise correction term.

[0035] The term is used to reflect that aerodynamic noise is proportional to the cube of the speed. The term is used to characterize the linear relationship between vehicle tire-road friction noise and speed.

[0036] S2016, through the , and Adjustments are made until the sum of the difference coefficients of all samples is minimized, thus obtaining the influence relationship after training is completed.

[0037] The influence relationship formula is based on the principle of sound energy superposition. By analyzing the physical characteristics of vehicle type and operating status, it can achieve quantitative prediction of road traffic noise.

[0038] After classifying vehicles according to preset dynamic characteristics, the aerodynamic noise components and mechanical rolling friction noise components of each type of vehicle are aggregated.

[0039] The total sound energy is converted into an equivalent sound level value through logarithmic transformation, and an environmental background noise correction term is introduced to compensate for the road surface reflection effect.

[0040] By training parameters using historical data to fit the actual noise propagation pattern, a mapping relationship between multiple physical variable inputs and noise output is formed, providing a theoretical basis for subsequent abnormal noise identification.

[0041] Based on historical video clips Align with the time series, bind video data with audio data, and delay the transmission duration. To offset propagation errors.

[0042] S202. Real-time analysis of image data in the current video image. and sound data Extract image data The predicted noise is calculated by inputting the number and instantaneous speed of all vehicle objects into the influence relation.

[0043] The vehicle object needs to be matched with the corresponding 3D model in the model library first. Then, the influence relationship is established based on the aerodynamic noise coefficient and mechanical friction noise coefficient preset by the corresponding 3D model. The predicted noise is obtained by inputting the number of all vehicle objects and the average instantaneous speed under various 3D models.

[0044] S203, Transfer audio data The noise value of the sound is used as the measured noise. Predicted noise curves and measured noise curves are plotted according to their time evolution. Based on the deviation between these two noise curves, the difference coefficient for each monitoring device is calculated. Specifically, this includes:

[0045] S2031. Delay the measured noise curve according to the predicted noise curve duration. Then, time alignment is performed, a noise trend map is established, and the predicted noise curve and the measured noise curve are mapped respectively.

[0046] S2032, Set on the noise trend graph Analyze the measured noise at each time point. and prediction noise According to the formula: Calculate the fluctuation coefficient at each time point .

[0047] The fluctuation coefficient is defined as the relative percentage offset between the measured noise value and the predicted noise value at a single point in time. It is used to quantify the instantaneous deviation intensity of two noise curves at a specific sampling time after they have been synchronized along the time axis.

[0048] S2033. Mark the time points when the fluctuation coefficient is greater than the preset threshold, take the corresponding positions of each marked time point on the measured noise curve as sampling points, and extract the curve between each two adjacent sampling points as the abnormal curve.

[0049] S2034. Mark the turning point of the measured noise development trend in each abnormal curve, and divide each abnormal curve into different sub-curves according to the turning point. The measured noise development trend in each sub-curve remains unchanged.

[0050] S2035. Using all sub-curves showing an upward trend in measured noise as reference curves, analyze the duration of each reference curve and the measured noise range, as well as the interval duration between adjacent reference curves, and substitute them into the formula to calculate the difference coefficient. :

[0051] ;

[0052] In the formula, The standard deviation of the time interval between all adjacent reference curves. It is a constant. The standard deviation of the measured noise range for all reference curves. This represents the standard deviation of the duration of all reference curves.

[0053] The difference coefficient, as a comprehensive index for quantifying the evolution regularity of noise curves, is used to capture the periodic noise fluctuation patterns caused by sudden road obstacles.

[0054] By deconstructing the rising sub-curve sequence in the measured noise curve, three core temporal features are extracted:

[0055] Both the noise range dispersion and the duration dispersion are strongly positively correlated with regularity, and low dispersion indicates that the noise amplitude and duration are highly reproducible.

[0056] The dispersion of the interval between adjacent events is weakly correlated with regularity. High dispersion indicates that the triggering of events is affected by occasional factors, making it difficult for intermittent fluctuations to form a strict regularity and resulting in weak periodicity.

[0057] By identifying and extracting noise curves, recurring, regular noise events are separated from random background noise, providing mathematical criteria for targeted diagnosis of road anomalies. Its design fully aligns with the physical characteristics of sudden obstacle noise in traffic scenarios, achieving a deep coupling between theoretical models and empirical evidence.

[0058] S204. For monitoring devices with a difference coefficient less than a preset threshold, the road area in the video image collected by the monitoring device within the 3D scene is designated as the affected road segment, and the affected area is divided according to the difference index and the location of the affected road segment. Specifically, this includes:

[0059] S2041. Analyze the coverage area of ​​the video images collected by the monitoring equipment and map them to the corresponding virtual area in the three-dimensional scene, and take part of the road in the virtual area as the affected road section.

[0060] S2042. Evenly distribute the points in the affected road section, identify the two points with the largest straight-line distance and connect them to form a line segment, and use this line segment as the diameter to divide a circular area as a reference area.

[0061] S2043, Preset Standard Noise And obtain the current measured noise. The noise attenuation formula is used to calculate decay to Required distance .

[0062] S2044. Identify the center position of the reference area, and add the current radius to... As the new radius, a circular area is defined as the influence zone, which shares the same center as the reference zone.

[0063] Based on monitoring equipment with a small difference coefficient, road areas in video images are identified, and the required distance is calculated using a noise attenuation formula to expand the area into a circular influence zone.

[0064] Using the reference area diameter connection point and noise value attenuation model, ensure that the affected area is divided to cover the noise propagation range.

[0065] Through time series analysis and mathematical modeling, the influence relationship of noise correlation is fitted, noise curves are plotted, difference coefficients are calculated, and influence areas are divided.

[0066] Quantify noise deviation, accurately identify abnormal noise pollution areas, reduce human judgment errors, and improve the accuracy of noise analysis and the efficiency of priority classification.

[0067] S300. Analyze the address of each questionnaire record within the survey data, correlate them according to their respective impact areas, and calculate the sum of noise scores. Then, combine the noise data and the coefficient of variation to calculate the impact index of each impact area, thereby setting the areas of concern. Specifically, this includes:

[0068] S301. Mark the virtual location corresponding to the address filled in for each questionnaire record in the 3D scene, and associate the questionnaire records whose virtual locations are in the influence area with the influence area.

[0069] S302. Obtain all questionnaire records associated with the affected area, and extract the noise score for each time period in each questionnaire record. Calculate the sum of the noise scores for all time periods in all questionnaire records. .

[0070] S303. Mark the time period in which the current time is located, and calculate the sum of all noise scores within the marked time period. The impact index of each affected area was calculated by combining noise data. The area of ​​influence with an influence index exceeding a preset threshold is designated as the area of ​​interest. The formula is as follows:

[0071] ;

[0072] In the formula, The coefficient of variation is the largest among all affected areas. The coefficient of variation, and These are the measured noise and the predicted noise, respectively.

[0073] The impact index is a comprehensive evaluation index that integrates subjective noise perception data and objective noise monitoring data. It is used to assess the priority of the social negative impact of noise pollution within a specific spatial unit.

[0074] The results are normalized by taking the total noise score of the questionnaire in the current period as the numerator and combining it with the total noise score of all historical periods. Then, the results are multiplied by the relative weighting factor based on the acoustic model prediction bias and the maximum regional difference coefficient.

[0075] The computational mechanism enhances the immediate negative perception impact in the time dimension and highlights the relative severity of areas with abnormal prediction deviations in the spatial dimension, ultimately achieving quantitative ranking of noise concern areas and ensuring that governance resources are prioritized for areas with high social impact.

[0076] By integrating psychological indicators and noise curve data from questionnaire surveys, a precise fusion of objective monitoring and subjective perception is achieved. Questionnaire data is weighted based on the degree of negative perception at different time periods, with the influence index weighted difference coefficient highlighting abnormal areas. The sum of all noise scores for the marked time periods is calculated. This provides insights into the differences in perception at different times.

[0077] By linking questionnaire data to the affected areas, the sum of noise scores and the impact index are calculated to quantify the degree of noise's impact on human life.

[0078] By combining noise data with the output of the area of ​​interest, we can make scientific decisions and optimize resource allocation to prioritize high-priority areas.

[0079] The S400 displays the location and impact index of each area of ​​concern in the 3D scene in real time through the visualization screen of the monitoring center. At the same time, it provides early warning prompts to on-site staff to analyze and investigate each area of ​​concern in reverse order of impact index.

[0080] The monitoring center displays the spatial location and impact index of the area of ​​interest in the 3D scene in real time on the screen. Combined with the early warning system, it drives staff to conduct reverse investigations, improving response efficiency and operability, and reducing the risk of omissions in noise control.

[0081] The present invention also provides a traffic environmental noise data analysis system based on dynamic perception, including a dynamic perception module, a noise analysis module, an impact assessment module, and a visualization module.

[0082] The dynamic perception module collects traffic maps, survey data, and video images of different roads, builds a 3D scene based on the traffic map, and performs dynamic mapping by combining the video images.

[0083] Collect traffic maps, survey data, and video images collected through monitoring equipment.

[0084] A pre-built model library is used to generate 3D scenes with terrain and road markings using digital twin technology.

[0085] Vehicle objects in video images are identified using object detection algorithms, and similarity is calculated to match models in a model library.

[0086] The dynamically driven matching 3D model moves within the 3D scene, maintaining synchronization with its real-time position and pose.

[0087] It provides dynamic and real-time virtual environment mapping scenarios, making traffic noise analysis visualization possible. It effectively integrates physical world data into the digital space, providing an accurate spatiotemporal basis for subsequent noise analysis and improving the efficiency of data collection and the realism of environmental monitoring.

[0088] The noise analysis module extracts image and audio data from video images and fits an influence relationship formula. Based on the influence relationship formula, a noise curve is plotted, the difference coefficient is calculated, and the affected area is divided.

[0089] Analyze the historical and current segments of video images, extract image and audio data, and fit the influence relationship through time series analysis.

[0090] Plot the predicted and measured noise curves, analyze the deviations, and calculate the difference coefficient. Delineate the affected area based on the monitoring equipment with a small preset difference coefficient.

[0091] It enables accurate modeling and anomaly detection of noise data, and uses mathematical models to quantify curve deviations to identify key areas, thereby improving the accuracy of noise source analysis, reducing human error, and providing data support for prioritizing high-noise road sections.

[0092] The impact assessment module associates all questionnaire records according to the impact area corresponding to the entered address and calculates the sum of noise scores. It then calculates the impact index for each impact area based on the coefficient of variation and designates areas of concern.

[0093] The questionnaire records are mapped to the affected area in a 3D scene according to the filled-in address, and the sum of noise scores for all time periods of the associated questionnaires is calculated. Combining the noise data and the difference coefficient, the influence index is calculated using a formula, and finally, the area of ​​concern is set.

[0094] Quantifying the impact of noise on surrounding residents and prioritizing areas through index calculations allows noise management to focus more on high-impact areas, enhancing the scientific nature of decision-making and helping to optimize resource allocation and reduce the negative social impacts of noise pollution.

[0095] The visualization module displays the location and impact index of each area of ​​interest in the 3D scene, and simultaneously provides early warnings to on-site staff to investigate each area of ​​interest.

[0096] The monitoring center displays the location and impact index of each area of ​​interest in the 3D scene in real time on its screen, and triggers an early warning system to remind on-site staff to analyze and investigate in reverse order of the index.

[0097] It provides intuitive real-time visual feedback and early warning functions, enhancing the real-time response capability and operational efficiency of environmental supervision, helping staff to quickly locate and investigate problem areas, and improving the practicality of noise control.

[0098] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0099] Dynamic noise source holographic mapping: Breaking through the spatial limitations of static sensing devices, this mechanism reconstructs the 3D traffic flow scene in real time through a virtual mapping architecture, dynamically binding vehicle behavior and noise radiation characteristics. This mechanism overcomes the blind spots of physical monitoring and significantly improves the spatiotemporal accuracy of noise event capture.

[0100] Multi-dimensional influence on coupled decision-making: Integrating physical acoustic parameters with social perception data streams, a weighted evaluation model of time-based perceived intensity and noise anomaly deviation is constructed. Compared with traditional single physical indicator analysis, this achieves a paradigm shift in governance priority determination from "sound pressure-dominated" to "dual-driven by social pain points and physical anomalies".

[0101] Targeted Identification of Periodic Hazards: Based on the strong regularity of noise event amplitude stability and duration consistency, a difference coefficient algorithm is designed to extract periodic road hazard signals. Compared with threshold alarm mechanisms, this effectively distinguishes between occasional interference and regular abnormal noises caused by sudden events, improving the accuracy of hazard identification.

[0102] Proactive optimization of governance resources: Heat maps of areas of concern are dynamically generated based on impact indices, and a reverse-order screening mechanism for abnormal areas is implemented using a visualization system. This breaks through the passive response-based governance approach, forming a proactive intervention strategy guided by a dual weight of "high social impact - strong physical anomalies," correcting the hidden dangers caused by resource misallocation. Attached Figure Description

[0103] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0104] Figure 1 This is a flowchart illustrating the traffic environmental noise data analysis method based on dynamic perception of the present invention.

[0105] Figure 2 This is a schematic diagram of the traffic environment noise data analysis system based on dynamic perception, as described in this invention. Detailed Implementation

[0106] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0107] Example 1: Please refer to Figure 1 This invention provides a method for analyzing traffic environmental noise data based on dynamic perception, including:

[0108] S100 collects traffic maps and survey data, and gathers video images of different road locations through monitoring equipment. Digital twin technology is used to build a 3D scene based on the traffic map, and then dynamically maps it using real-time video images.

[0109] A traffic map is a route map of a transportation network that includes the spatial layout of roads and the prescribed directions of travel.

[0110] In the implementation process, the survey data included different questionnaire records. Each questionnaire record included at least the address where the questionnaire was filled in and the noise rating for each time period. The noise rating was 1-5 (1=no impact, 2=slight impact, 3=moderate impact, 4=significant impact, 5=severe impact), with higher scores indicating a greater negative perception of noise during the corresponding time period.

[0111] The specific times for each period are: early morning (6:00-9:00), morning (9:00-12:00), noon (12:00-14:00), afternoon (14:00-18:00), evening (18:00-22:00) and night (22:00-6:00).

[0112] The noise score is given by the questionnaire respondents and refers to an individual's adverse reaction to noise, including various situations such as "dissatisfaction", "disturbance", "annoyance" and "interference" caused by noise.

[0113] Noise rating is a quantitative representation of the degree of noise annoyance. It is an important psychological indicator that can more comprehensively reflect the impact of noise on human life.

[0114] Building a 3D scene and performing dynamic mapping specifically includes:

[0115] S101. Preset model library and put in 3D models containing different types and specifications of vehicles. Each model is labeled with actual size parameters and feature contour data.

[0116] In practice, vehicle types can be categorized according to function and purpose, such as dump trucks, buses, freight trucks, and transport vehicles. Specifications can be simply classified as small, medium, and large vehicles, or further subdivided.

[0117] S102. By accessing road topology data provided by traffic maps through digital twin technology, a 3D scene with terrain and road markings is generated in the 3D engine.

[0118] S103. Based on the road location of the monitoring equipment, mark the corresponding location in the three-dimensional scene and synchronously map the road area in the video images collected by each monitoring equipment.

[0119] S104. Identify all vehicle objects within the road area in the video image using a target detection algorithm, and extract the appearance size parameters, shape contour features, and real-time position coordinates of each vehicle object.

[0120] S105. Calculate the similarity between the extracted appearance size parameters and shape contour features and the actual size parameters and feature contour data of each 3D model in the model library.

[0121] S106. Match the three-dimensional model with the highest similarity to each vehicle object, and set the displacement direction and velocity of the corresponding matched three-dimensional model according to the extracted real-time position coordinate changes.

[0122] In the specific implementation process, the maximum similarity can be the sum of the similarity values ​​of the actual size parameters and feature contour data, or it can be calculated by distinguishing weighted average. The weights of the size parameters and contour data are set in advance by the staff.

[0123] S107. Map the matched 3D models of each vehicle object to the corresponding coordinate positions in the 3D scene, and dynamically drive these 3D models to move along the path while maintaining position and attitude synchronization.

[0124] By using digital twin technology to construct a three-dimensional virtual model of the real traffic environment, dynamic real-time mapping of traffic noise sources can be achieved.

[0125] This reduces the delay and error in environmental data acquisition, providing a precise spatiotemporal basis for subsequent noise analysis and improving monitoring efficiency and data integration.

[0126] S200. Extract image and audio data from the video image and fit the influence relationship. Plot the predicted noise curve and the measured noise curve respectively, calculate the difference coefficient based on the deviation between the two noise curves, and divide the influence area. Specifically, this includes:

[0127] S201. Extract the historical data of each monitoring device. video clip The data is then analyzed and separated into video and audio data. The correlation between these two data points is analyzed in time sequence to fit influence relationships for each monitoring device. Specifically, this includes:

[0128] S2011. Bind the video data and audio data to the timeline separately, and set them evenly. Each time point. Preset conduction duration. Each time point on the audio timeline is delayed by a certain duration compared to the corresponding time point on the video timeline. .

[0129] In the specific implementation process, the conduction time is used to offset the time error caused by the length of the propagation path of traffic noise from generation to collection. Each monitoring device is preset with a different conduction time, which is specifically determined by the actual distance between the monitoring device installation location and the road being filmed, as well as the sound propagation speed under different environments (such as road surface environment and weather environment).

[0130] S2012, Video clips The image is parsed into continuous image frames, matching image frames at the same time point on the timeline, identifying vehicle objects contained in each image frame and matching them with 3D models in the model library.

[0131] S013, Obtain time points with the same ordinal number in the video and audio timelines. and Based on the time points in the video clips The instantaneous velocity of each vehicle object in front and behind is calculated based on its positional changes, and all vehicle objects are classified according to the matched 3D model.

[0132] S2014, Time Point The number of all vehicle objects in each category and their average instantaneous speed are used as independent variables. (Time point) The noise value of the sound is used as the dependent variable, and the independent and dependent variables at the same ordinal time points are packaged into a sample.

[0133] S2015. Establish the influence relationship formula, and input the following respectively. The independent variables in each sample are used, and the difference between the output result and the dependent variable is used as the difference coefficient for the corresponding sample. The influence relationship is as follows:

[0134] ;

[0135] In the formula, To predict noise, The number of classes at a given time point. For the first The total number of all vehicle objects under this class. For the first The average instantaneous speed of all vehicle objects under this class.

[0136] and The first The corresponding three-dimensional model has preset aerodynamic noise coefficients (e.g., truck > bus > car) and mechanical friction noise coefficients (related to vehicle weight and tires). This is a background noise correction term (such as the effect of road surface reflecting and enhancing sound waves).

[0137] The term is used to reflect that aerodynamic noise (turbulent noise) is proportional to the cube of the velocity. The term is used to characterize the linear relationship between vehicle tire-road friction noise (rolling noise) and speed.

[0138] S2016, through the , and Adjustments are made until the sum of the difference coefficients of all samples is minimized, thus obtaining the influence relationship after training is completed.

[0139] The influence relationship formula is based on the principle of sound energy superposition. By analyzing the physical characteristics of vehicle type and operating status, it can achieve quantitative prediction of road traffic noise.

[0140] After classifying vehicles according to preset dynamic characteristics, the aerodynamic noise components (which are positively correlated with speed by a power factor) and the mechanical rolling friction noise components (which are positively correlated with speed by a linear factor) of each type of vehicle are aggregated.

[0141] The total sound energy is converted into an equivalent sound level value through logarithmic transformation, and an environmental background noise correction term is introduced to compensate for the road surface reflection effect.

[0142] By training parameters using historical data to fit the actual noise propagation pattern, a mapping relationship between multiple physical variable inputs and noise output is formed, providing a theoretical basis for subsequent abnormal noise identification.

[0143] In the specific implementation process, based on historical video clips Align with the time series, bind video data (number of vehicles and instantaneous speed) with audio data (noise level), and delay the transmission time. To offset propagation errors.

[0144] S202. Real-time analysis of image data in the current video image. and sound data Extract image data The predicted noise is calculated by inputting the number and instantaneous speed of all vehicle objects into the influence relation.

[0145] In the specific implementation process, the vehicle object needs to be matched with the corresponding three-dimensional model in the model library first. Then, the influence relationship is established based on the aerodynamic noise coefficient and mechanical friction noise coefficient preset by the corresponding three-dimensional model. The predicted noise is obtained by inputting the number of all vehicle objects and the average instantaneous speed under various three-dimensional models.

[0146] S203, Transfer audio data The noise value of the sound is used as the measured noise. Predicted noise curves and measured noise curves are plotted according to their time evolution. Based on the deviation between these two noise curves, the difference coefficient for each monitoring device is calculated. Specifically, this includes:

[0147] S2031. Delay the measured noise curve according to the predicted noise curve duration. Then, time alignment is performed, a noise trend map is established, and the predicted noise curve and the measured noise curve are mapped respectively.

[0148] S2032, Set on the noise trend graph Analyze the measured noise at each time point. and prediction noise According to the formula: Calculate the fluctuation coefficient at each time point .

[0149] In practical implementation, the fluctuation coefficient is defined as the relative percentage offset between the measured noise value and the predicted noise value at a single point in time. It is used to quantify the instantaneous deviation intensity of two noise curves at a specific sampling moment after they are synchronized along the time axis.

[0150] S2033. Mark the time points when the fluctuation coefficient is greater than the preset threshold, take the corresponding positions of each marked time point on the measured noise curve as sampling points, and extract the curve between each two adjacent sampling points as the abnormal curve.

[0151] S2034. Mark the turning point of the measured noise development trend in each abnormal curve, and divide each abnormal curve into different sub-curves according to the turning point. The measured noise development trend in each sub-curve remains unchanged.

[0152] S2035. Using all sub-curves showing an upward trend in measured noise as reference curves, analyze the duration of each reference curve and the measured noise range, as well as the interval duration between adjacent reference curves, and substitute them into the formula to calculate the difference coefficient. :

[0153] ;

[0154] In the formula, The standard deviation of the time interval between all adjacent reference curves. It is a constant. The standard deviation of the measured noise range for all reference curves. This represents the standard deviation of the duration of all reference curves.

[0155] The difference coefficient, as a comprehensive index for quantifying the evolution regularity of noise curves, is used to capture periodic noise fluctuation patterns caused by sudden road obstacles (such as loose manhole covers, piles of sand and gravel, or road surface damage).

[0156] By deconstructing the rising sub-curve sequence (i.e., the reference curve) in the measured noise curve, three core temporal features are extracted:

[0157] Both the noise range dispersion (event amplitude stability) and the duration dispersion (event evolution consistency) are strongly positively correlated with regularity. Low dispersion indicates that the noise amplitude and duration are highly reproducible (such as the stable vibration mode of manhole covers).

[0158] The dispersion of the interval between adjacent events is weakly correlated with regularity. High dispersion indicates that the triggering of events is affected by occasional factors (such as the uncertainty of traffic flow density weakening the periodicity), making it difficult for intermittent fluctuations to form a strict regularity and resulting in weak periodicity.

[0159] By identifying and extracting noise curves, recurring, regular noise events (such as periodic abnormal noises from road facilities) are separated from random background noise, providing mathematical criteria for targeted diagnosis of road anomalies. Its design fully aligns with the physical characteristics of sudden obstacle noises in traffic scenarios, achieving a deep coupling between theoretical models and empirical evidence.

[0160] S204. For monitoring devices with a difference coefficient less than a preset threshold, the road area in the video image collected by the monitoring device within the 3D scene is designated as the affected road segment, and the affected area is divided according to the difference index and the location of the affected road segment. Specifically, this includes:

[0161] S2041. Analyze the coverage area of ​​the video images collected by the monitoring equipment and map them to the corresponding virtual area in the three-dimensional scene, and take part of the road in the virtual area as the affected road section.

[0162] S2042. Evenly distribute the points in the affected road section, identify the two points with the largest straight-line distance and connect them to form a line segment, and use this line segment as the diameter to divide a circular area as a reference area.

[0163] S2043, Preset Standard Noise And obtain the current measured noise. The noise attenuation formula is used to calculate decay to Required distance .

[0164] S2044. Identify the center position of the reference area, and add the current radius to... As the new radius, a circular area is defined as the influence zone, which shares the same center as the reference zone.

[0165] In the specific implementation process, based on monitoring equipment with small difference coefficients, road areas (affected road sections) in video images are identified, and the required distance is calculated using the noise attenuation formula to expand into a circular affected area.

[0166] Using the reference area diameter connection point and noise value attenuation model (from measured noise attenuation to standard value), ensure that the affected area is delineated to cover the noise propagation range.

[0167] Through time series analysis and mathematical modeling, the influence relationship of noise correlation is fitted, noise curves are plotted, difference coefficients are calculated, and influence areas are divided.

[0168] Quantify noise deviation to accurately identify abnormal noise pollution areas (affected areas), reduce human judgment errors, and improve the accuracy of noise analysis and the efficiency of priority classification.

[0169] S300. Analyze the address of each questionnaire record within the survey data, correlate them according to their respective impact areas, and calculate the sum of noise scores. Then, combine the noise data and the coefficient of variation to calculate the impact index of each impact area, thereby setting the areas of concern. Specifically, this includes:

[0170] S301. Mark the virtual location corresponding to the address filled in for each questionnaire record in the 3D scene, and associate the questionnaire records whose virtual locations are in the influence area with the influence area.

[0171] S302. Obtain all questionnaire records associated with the affected area, and extract the noise score for each time period in each questionnaire record. Calculate the sum of the noise scores for all time periods in all questionnaire records. .

[0172] S303. Mark the time period in which the current time is located, and calculate the sum of all noise scores within the marked time period. The impact index of each affected area was calculated by combining noise data. The area of ​​influence with an influence index exceeding a preset threshold is designated as the area of ​​interest. The formula is as follows:

[0173] ;

[0174] In the formula, The coefficient of variation is the largest among all affected areas. The coefficient of variation is... and These are the measured noise and the predicted noise, respectively.

[0175] The impact index is a comprehensive evaluation index that integrates subjective noise perception data and objective noise monitoring data. It is used to assess the priority of the social negative impact of noise pollution within a specific spatial unit.

[0176] The results are normalized by taking the total noise score of the questionnaire in the current period as the numerator and combining it with the total noise score of all historical periods. Then, the results are multiplied by a relative weighting factor based on the acoustic model prediction bias (the absolute difference between the measured value and the predicted value) and the maximum regional difference coefficient.

[0177] The computational mechanism enhances the immediate negative perception impact in the time dimension and highlights the relative severity of areas with abnormal prediction deviations in the spatial dimension, ultimately achieving quantitative ranking of noise concern areas and ensuring that governance resources are prioritized for areas with high social impact.

[0178] By integrating psychological indicators and noise curve data from questionnaire surveys, a precise fusion of objective monitoring and subjective perception is achieved. Questionnaire data is weighted based on the degree of negative perception at different time periods, with the influence index weighted difference coefficient highlighting abnormal areas. The sum of all noise scores for the marked time periods is calculated. This provides insights into the differences in perception at different times.

[0179] In the specific implementation process, the impact area is associated with the questionnaire data, and the sum of noise scores and impact index are calculated to quantify the degree of noise impact on human life.

[0180] By combining noise data with the output of the area of ​​interest, we can make scientific decisions and optimize resource allocation to prioritize high-priority areas.

[0181] The S400 displays the location and impact index of each area of ​​concern in the 3D scene in real time through the visualization screen of the monitoring center. At the same time, it provides early warning prompts to on-site staff to analyze and investigate each area of ​​concern in reverse order of impact index.

[0182] The monitoring center displays the spatial location and impact index of the area of ​​interest in the 3D scene in real time on the screen. Combined with the early warning system, it drives staff to conduct reverse investigations, improving response efficiency and operability, and reducing the risk of omissions in noise control.

[0183] Example 2: Please refer to Figure 2 The present invention also provides a traffic environment noise data analysis system based on dynamic perception, including a dynamic perception module, a noise analysis module, an impact assessment module, and a visualization module.

[0184] The dynamic perception module collects traffic maps, survey data, and video images of different roads, builds a 3D scene based on the traffic map, and performs dynamic mapping by combining the video images.

[0185] In the specific implementation process, traffic maps (including road spatial layout and transportation network maps of driving directions), survey data (such as questionnaire records, including filled-in addresses and time period noise scores) and video images collected through monitoring equipment are collected.

[0186] A pre-set model library (containing 3D models of different vehicle types and their size parameters) is used to generate 3D scenes with terrain and road markings using digital twin technology.

[0187] Vehicle objects in video images are identified using object detection algorithms (extracting appearance size, contour features, and location coordinates), and similarity is calculated to match models in a model library.

[0188] The dynamically driven matching 3D model moves within the 3D scene, maintaining synchronization with its real-time position and pose.

[0189] It provides dynamic and real-time virtual environment mapping scenarios, making traffic noise analysis visualization possible. It effectively integrates physical world data into the digital space, providing an accurate spatiotemporal basis for subsequent noise analysis and improving the efficiency of data collection and the realism of environmental monitoring.

[0190] The noise analysis module extracts image and audio data from video images and fits an influence relationship formula. Based on the influence relationship formula, a noise curve is plotted, the difference coefficient is calculated, and the affected area is divided.

[0191] In the specific implementation process, the historical and current segments of video images are analyzed to extract image data (such as the number of vehicles and instantaneous speed) and sound data (noise value). Through time series analysis (such as binding the time axis and processing the delayed transmission time), the influence relationship is fitted (based on the sound energy superposition law, including the aerodynamic noise coefficient and mechanical friction noise coefficient formula).

[0192] Plot the predicted and measured noise curves (calculate the fluctuation coefficient and anomaly curve), analyze the deviation, and calculate the difference coefficient. Based on the small preset value of the difference coefficient, divide the influence zone of the monitoring equipment (e.g., by expanding the circular area, calculate the required distance using the noise attenuation formula).

[0193] It enables accurate modeling and anomaly detection of noise data, and uses mathematical models to quantify curve deviations to identify key areas (affected areas), thereby improving the accuracy of noise source analysis, reducing human error, and providing data support for prioritizing the treatment of high-noise road sections.

[0194] The impact assessment module associates all questionnaire records according to the impact area corresponding to the entered address and calculates the sum of noise scores. It then calculates the impact index for each impact area based on the coefficient of variation and designates areas of concern.

[0195] In the specific implementation process, questionnaire records are mapped to the influence area in the 3D scene according to the filled address, and the sum of noise scores for all time periods of the associated questionnaires is calculated (such as the total AR and the current time period's total NR). Combining noise data (measured and predicted noise) and the difference coefficient, the influence index is calculated using a formula, and finally, the focus area is set (the influence index is greater than a preset threshold).

[0196] Quantifying the impact of noise on surrounding residents and prioritizing areas through index calculations allows noise management to focus more on high-impact areas, enhancing the scientific nature of decision-making and helping to optimize resource allocation and reduce the negative social impacts of noise pollution.

[0197] The visualization module displays the location and impact index of each area of ​​interest in the 3D scene, and simultaneously provides early warnings to on-site staff to investigate each area of ​​interest.

[0198] During implementation, the location and impact index of each area of ​​interest in the 3D scene (output by the impact assessment module) are displayed in real time on the screen of the monitoring center. At the same time, the early warning system is triggered to remind on-site staff to analyze and investigate in reverse order of the index.

[0199] It provides intuitive real-time visual feedback and early warning functions, enhancing the real-time response capability and operational efficiency of environmental supervision, helping staff to quickly locate and investigate problem areas, and improving the practicality of noise control.

[0200] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0201] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A traffic environmental noise data analysis method based on dynamic perception, characterized in that: The method includes: S100: Collect traffic maps and survey data, and collect video images of different road locations through monitoring equipment; use digital twin technology to build a three-dimensional scene based on the traffic map and dynamically map it in combination with real-time video images; S200. Extract the image and audio data from the video image and fit the influence relationship formula; plot the predicted noise curve and the measured noise curve respectively, calculate the difference coefficient based on the deviation of the two noise curves and divide the influence area; S300: Analyze the filling address of each questionnaire record in the survey data, correlate them according to their respective impact areas, and calculate the sum of noise scores; then combine the noise data and the difference coefficient to calculate the impact index of each impact area, thereby setting the area of ​​concern; The S400 displays the location and impact index of each area of ​​concern in the 3D scene in real time through the visualization screen of the monitoring center. At the same time, it provides early warning prompts to on-site staff to analyze and investigate each area of ​​concern in reverse order of impact index.

2. The traffic environmental noise data analysis method based on dynamic perception according to claim 1, characterized in that: In S100, a traffic map refers to a transportation network route map that includes the spatial layout of roads and the prescribed driving directions; The survey data includes different questionnaire records. Each questionnaire record includes at least the address where the questionnaire was filled in and the noise rating for each time period. The noise rating is from 1 to 5, with a higher score indicating a greater negative perception of noise during the corresponding time period. Building a 3D scene and performing dynamic mapping specifically includes: S101, Preset model library, and put in 3D models containing different types and specifications of vehicles, with each model labeled with actual size parameters and feature contour data; S102. By accessing road topology data provided by traffic maps through digital twin technology, a 3D scene with terrain and road markings is generated in the 3D engine; S103. Based on the road location of the monitoring equipment, mark the corresponding location in the three-dimensional scene and synchronously map the road area in the video images collected by each monitoring equipment. S104. Identify all vehicle objects in the road area of ​​the video image using the target detection algorithm, and extract the appearance size parameters, shape contour features and real-time position coordinates of each vehicle object. S105. Calculate the similarity between the extracted appearance size parameters and shape contour features and the actual size parameters and feature contour data of each 3D model in the model library. S106. Match the three-dimensional model with the highest similarity to each vehicle object, and set the displacement direction and velocity of the corresponding matched three-dimensional model according to the extracted real-time position coordinate changes. S107. Map the matched 3D models of each vehicle object to the corresponding coordinate positions in the 3D scene, and dynamically drive these 3D models to move along the path while maintaining position and attitude synchronization.

3. The traffic environmental noise data analysis method based on dynamic perception according to claim 2, characterized in that: S200 includes: S201. Extract the historical data of each monitoring device. video clip The data is then parsed into video and audio data; the correlation between the two is analyzed in time sequence to fit the influence relationship formula for each monitoring device; S202. Real-time analysis of image data in the current video image. and sound data Extract image data The predicted noise is calculated by inputting the number and instantaneous speed of all vehicle objects into the influence relationship formula; S203, Transfer audio data The noise value of the sound is used as the measured noise. The predicted noise curve and the measured noise curve are plotted according to the time evolution. The difference coefficient of each monitoring device is calculated based on the deviation of these two noise curves. S204. For monitoring devices with a difference coefficient less than a preset threshold, the road area in the video image collected by the monitoring device in the three-dimensional scene is marked as the affected road segment, and the affected area is divided according to the difference index and the location of the affected road segment.

4. The traffic environmental noise data analysis method based on dynamic perception according to claim 3, characterized in that: S201 includes: S2011. Bind the video data and audio data to the timeline separately and set them evenly. A specific time point; preset conduction duration Each time point on the audio timeline is delayed by a certain duration compared to the corresponding time point on the video timeline. ; S2012, Video clips The image is parsed into continuous image frames, matching image frames at the same time point on the timeline, identifying vehicle objects contained in each image frame and matching them with 3D models in the model library. S013, Obtain time points with the same ordinal number in the video and audio timelines. and Based on the time points in the video clips The instantaneous velocity of the positional changes of each vehicle object in front and behind is calculated, and all vehicle objects are classified according to the matched 3D model; S2014, Time Point The number of all vehicle objects in each category and the average instantaneous speed are used as independent variables; time points. The noise value of the sound is used as the dependent variable, and the independent and dependent variables at the same ordinal time points are packaged into samples; S2015. Establish the influence relationship formula, and input the following respectively. The independent variables in the sample are used, and the difference between the output result and the dependent variable is used as the difference coefficient of the corresponding sample. S2016. By adjusting the parameters in the influence relationship formula to minimize the sum of the difference coefficients of all samples, the influence relationship formula after training is obtained.

5. The traffic environmental noise data analysis method based on dynamic perception according to claim 4, characterized in that: The relationship between the effects is as follows: ； In the formula, To predict noise, The number of classes at a given time point. For the first The total number of vehicle objects under this class. For the first The average instantaneous speed of all vehicle objects under this class; and The first The corresponding three-dimensional model has preset aerodynamic noise coefficient and mechanical friction noise coefficient; This is a background noise correction term; Through the , and Adjustments are made until the sum of the difference coefficients of all samples is minimized, thus obtaining the influence relationship after training is completed.

6. The traffic environmental noise data analysis method based on dynamic perception according to claim 3, characterized in that: S203 includes: S2031. Delay the measured noise curve according to the predicted noise curve duration. Then, time alignment is performed, a noise trend map is established, and the predicted noise curve and the measured noise curve are mapped respectively. S2032, Set on the noise trend graph Analyze the measured noise at each time point. and prediction noise According to the formula: Calculate the fluctuation coefficient at each time point ; S2033. Mark the time points when the fluctuation coefficient is greater than the preset threshold, take the corresponding positions of each marked time point on the measured noise curve as sampling points, and extract the curve between each two adjacent sampling points as the abnormal curve. S2034. Mark the turning point of the measured noise development trend in each abnormal curve, and divide each abnormal curve into different sub-curves according to the turning point. The measured noise development trend in each sub-curve remains unchanged. S2035. Using all sub-curves showing an upward trend in measured noise as reference curves, analyze the duration of each reference curve and the measured noise range, as well as the interval duration between adjacent reference curves, and substitute them into the formula to calculate the difference coefficient. : ； In the formula, The standard deviation of the time interval between all adjacent reference curves. It is a constant. The standard deviation of the measured noise range for all reference curves. This represents the standard deviation of the duration of all reference curves.

7. The traffic environmental noise data analysis method based on dynamic perception according to claim 3, characterized in that: S204 includes: S2041. Analyze the coverage area of ​​the video images collected by the monitoring equipment and map them to the corresponding virtual area in the three-dimensional scene, and take part of the road in the virtual area as the affected road section. S2042. Evenly lay out points in the affected road section, identify the two points with the largest straight-line distance and connect them into a line segment, and use this line segment as the diameter to divide a circular area as a reference area. S2043, Preset Standard Noise And obtain the current measured noise. The noise attenuation formula is used to calculate decay to Required distance ; S2044. Identify the center position of the reference area, and add the current radius to... As the new radius, a circular area is defined as the influence zone, which shares the same center as the reference zone.

8. The traffic environmental noise data analysis method based on dynamic perception according to claim 3, characterized in that: The S300 includes: S301. Mark the virtual location corresponding to the filling address of each questionnaire record in the 3D scene, and associate the questionnaire records whose virtual locations are in the influence area with the influence area; S302. Obtain all questionnaire records associated with the affected area, and extract the noise score for each time period in each questionnaire record; calculate the sum of the noise scores for all time periods in all questionnaire records. ; S303. Mark the time period in which the current time is located, and calculate the sum of all noise scores within the marked time period. The impact index of each affected area is calculated by combining noise data; the affected areas with an impact index greater than the preset threshold are designated as areas of interest.

9. The traffic environmental noise data analysis method based on dynamic perception according to claim 8, characterized in that: Impact Index The calculation formula is as follows: ； In the formula, The coefficient of variation is the largest among all affected areas. The coefficient of variation is... and These are the measured noise and the predicted noise, respectively.

10. A traffic environmental noise data analysis system based on dynamic perception, characterized in that: The system includes a dynamic sensing module, a noise analysis module, an impact assessment module, and a visualization module; The dynamic perception module collects traffic maps, survey data, and video images of different roads, builds a 3D scene based on the traffic map, and performs dynamic mapping by combining the video images. The noise analysis module extracts image and sound data from video images, fits the influence relationship formula, plots the noise curve based on the influence relationship formula, calculates the difference coefficient, and divides the influence area. The impact assessment module associates all questionnaire records according to the impact area to which the filled address belongs, calculates the sum of noise scores, and calculates the impact index of each impact area based on the difference coefficient and sets the areas of concern. The visualization module displays the location and impact index of each area of ​​interest in the 3D scene, and simultaneously provides early warnings to on-site staff to investigate each area of ​​interest.

Images