A method and device for estimating road network traffic flow parameters

Abstract

The present invention discloses a method and device for estimating road network traffic flow parameters. The method includes: obtaining basic data; wherein, the basic data includes the boundaries of a number of road segments; determining a road network research scope according to the boundaries of all road segments, and obtaining a road network image within the road network research scope; inputting the road network image into a preset convolutional neural network model to obtain a road network vehicle detection result output by the convolutional neural network model; and calculating road network traffic flow parameters within the road network research scope according to the road network vehicle detection result. Embodiments of the present invention can quickly identify the vehicle models and positions, so as to quickly calculate the average vehicle density of a road segment, the section interval speed and the average flow of a road segment, and can efficiently and flexibly obtain the all-day time-sharing traffic operation parameters of a large-scale road network at a lower cost.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular, to a method and device for estimating road network traffic flow parameters. Background Art

[0002] Traffic flow parameters (average speed, traffic density, and flow rate) are important bases for analyzing the traffic capacity of urban roads and evaluating traffic operation conditions, and are also the research basis for mastering the traffic operation rules of cities and carrying out traffic congestion control.

[0003] In the prior art, when estimating road network traffic flow parameters, traffic flow parameters are mainly obtained through manual investigation methods and loop detection methods. Among them, the manual investigation method is used for small-scale road networks, and data such as the number of vehicles, vehicle types, and headway time in a certain period on a road section are counted by manual counting or floating car following, and traffic flow parameters are calculated based on these data; the loop detection method sets a loop coil under the road. When a vehicle passes through or stops on the coil, it will cause a change in the inductance of the coil, and traffic flow parameters are calculated based on the data collected from the change in the inductance of the coil.

[0004] It can be seen that in the prior art, the manual investigation method has high costs and low efficiency, is difficult to apply to large-scale investigation scenarios of traffic flow parameters, and depends on the professional quality of investigators, resulting in difficulty in ensuring the accuracy of data; the hardware and construction costs of the loop detection method are relatively high. Summary of the Invention

[0005] To solve the above technical problems, an embodiment of the present invention provides a method and device for estimating road network traffic flow parameters. By inputting the road network image within the research scope of the road network obtained into a convolutional neural network model, a road network vehicle detection result is output, and road network traffic flow parameters are calculated based on the road network vehicle detection result, thereby realizing more efficient and lower-cost estimation of traffic flow parameters.

[0006] To achieve the above object, an embodiment of the present invention provides a method for estimating road network traffic flow parameters, including:

[0007] Obtain basic data; wherein, the basic data includes the boundaries of a number of road sections;

[0008] Determine the research scope of the road network according to the boundaries of all road sections, and obtain the road network image within the research scope of the road network;

[0009] Input the road network image into a preset convolutional neural network model to obtain the road network vehicle detection result output by the convolutional neural network model;

[0010] Based on the vehicle detection results of the road network, the road network traffic flow parameters within the research scope of the road network are calculated.

[0011] Further, determining the research scope of the road network according to the boundaries of all road segments and obtaining the road network image within the research scope of the road network specifically includes: reading the coordinates of the boundaries from the basic data; for each road segment, expanding to obtain a rectangle according to the corresponding boundary and obtaining the endpoint coordinates of the rectangle; selecting the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate according to the endpoint coordinates, and using the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate as the coordinates of the four endpoints of the research scope of the road network; using an aircraft to obtain the road network image within the research scope of the road network.

[0012] Furthermore, using an aircraft to obtain the road network image within the research scope of the road network specifically includes: dividing the research scope of the road network according to a preset division strategy to obtain a plurality of sampling grids; performing path planning according to the coordinates of the centroids in the sampling grids, so that the aircraft performs image sampling operations within each sampling grid according to the planned path, obtaining respective grid images corresponding to each sampling grid, and combining all grid images to form the road network image.

[0013] Furthermore, dividing the research scope of the road network according to a preset division strategy to obtain a plurality of sampling grids specifically includes: obtaining the effective visual range width of the aircraft and calculating the spacing according to the effective visual range width; dividing according to the coordinates of the four endpoints of the research scope of the road network and the spacing to obtain a plurality of sampling grids.

[0014] Further, inputting the road network image into a preset convolutional neural network model to obtain the road network vehicle detection results output by the convolutional neural network model specifically includes: inputting the road network image into the backbone network in the convolutional neural network model to obtain a number of pick-up box attributes output by the backbone network; wherein, the backbone network includes a convolutional layer, a group normalization layer, and a Mish activation function, and the pick-up box attributes include pick-up box encoding, pick-up box center point coordinates, pick-up box length, pick-up box width, and pick-up box confidence; for each pick-up box attribute, when judging whether the corresponding pick-up box confidence is greater than a preset confidence threshold, if so, then: determining that the pick-up object corresponding to this pick-up box attribute is a vehicle, and calculating the vehicle model according to the pick-up box length and pick-up box width in this pick-up box attribute; outputting all calculated vehicle models as the road network vehicle detection results.

[0015] Further, inputting the road network image into the backbone network of the convolutional neural network model to obtain several pick-up box attributes output by the backbone network specifically includes: inputting the road network image with three channels into the backbone network to obtain three-dimensional feature tensors corresponding to the three channels output by the backbone network; sorting the feature tensors according to the rule of from large to small in dimension to obtain a first feature tensor, a second feature tensor, and a third feature tensor; performing a deconvolution operation on the third feature tensor, and connecting the third feature tensor after the operation with the second feature tensor to obtain a new second feature tensor; performing a deconvolution operation on the new second feature tensor, and connecting the second feature tensor after the operation with the first feature tensor to obtain a new first feature tensor; performing a convolution operation on the new first feature tensor to obtain the pick-up box attributes.

[0016] Further, calculating the road network traffic flow parameters within the study scope of the road network according to the road network vehicle detection result specifically includes: for each road segment, calculating the road segment length according to the corresponding boundary, calculating the number of standard vehicles according to the number of pick-up objects determined as vehicles within this road segment, the preset standard vehicle coefficient, and the vehicle type, and calculating the average vehicle density of this road segment according to the number of standard vehicles and the road segment length; combining the average vehicle densities of all road segments calculated to obtain the average vehicle density of the road network in the road network traffic flow parameters.

[0017] Further, calculating the road network traffic flow parameters within the study scope of the road network according to the road network vehicle detection result specifically includes: for each road segment, calculating the corresponding road segment interval speed according to the pick-up box code and the pick-up box center point coordinates in the pick-up box attributes; combining the road segment interval speeds of all road segments calculated to obtain the road network interval speed in the road network traffic flow parameters.

[0018] Further, calculating the road network traffic flow parameters within the study scope of the road network according to the road network vehicle detection result specifically includes: for each road segment, calculating the corresponding road segment average flow according to the corresponding road segment average vehicle density and the road segment interval speed through a preset flow-density-speed formula; wherein, the flow-density-speed formula is that the road segment average flow is equal to the road segment average vehicle density multiplied by the road segment interval speed; combining the road segment average flows of all road segments calculated to obtain the road network average flow in the road network traffic flow parameters.

[0019] An embodiment of the present invention further provides a computer device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the road network traffic flow parameter estimation method described in any one of the above are implemented.

[0020] In summary, the present invention has the following beneficial effects:

[0021] By adopting the embodiment of the present invention, basic data is obtained; wherein, the basic data includes the boundaries of a number of road segments; the road network research scope is determined according to the boundaries of all road segments, and a road network image within the road network research scope is obtained; the road network image is input into a preset convolutional neural network model to obtain a road network vehicle detection result output by the convolutional neural network model; according to the road network vehicle detection result, the road network traffic flow parameters within the road network research scope are calculated; in particular, after dividing the road network research scope, a plurality of sampling grids are obtained, and path planning is performed according to the sampling grids, so that the aircraft performs image sampling along the planned path, and the grid images obtained form the road network image. The embodiment of the present invention can quickly identify the vehicle models and positions, thereby quickly calculating the average vehicle density of the road segment, the section interval speed, and the average traffic flow of the road segment, and can efficiently and flexibly obtain the all-day time-sharing traffic operation parameters of the large-scale road network at a lower cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a schematic flowchart of an embodiment of a road network traffic flow parameter estimation method provided by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] See Figure 1 , which is a schematic flowchart of an embodiment of the road network traffic flow parameter estimation method provided by the present invention. The method includes steps S1 to S4, which are specifically as follows:

[0025] S1, obtaining basic data; wherein, the basic data includes the boundaries of a number of road segments;

[0026] Exemplarily, the basic data includes the boundaries of a number of road segments and the coordinates of the boundaries.

[0027] Exemplarily, the basic data includes relevant aircraft parameters of the aircraft for obtaining road network images, aircraft management requirements, analysis accuracy requirements, effective aerial photography height of the aircraft, and effective visual range width corresponding to the effective aerial photography height.

[0028] S2. Determine the road network research scope according to the boundaries of all road segments, and obtain the road network images within the road network research scope;

[0029] Preferably, the step of determining the road network research scope according to the boundaries of all road segments and obtaining the road network images within the road network research scope specifically includes: reading the coordinates of the boundaries from the basic data; for each road segment, expanding a rectangle according to the corresponding boundary and obtaining the endpoint coordinates of the rectangle; selecting the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate according to the endpoint coordinates, and using the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate as the coordinates of the four endpoints of the road network research scope; using an aircraft to obtain the road network images within the road network research scope.

[0030] Specifically, the expanded rectangle is the rectangle with the smallest area that can cover the corresponding boundary.

[0031] Specifically, for each road segment, a corresponding rectangle is obtained respectively. Each rectangle has four endpoint coordinates. The minimum and maximum values of the abscissa and ordinate are selected from the endpoint coordinates of all rectangles, and the above minimum and maximum values are used as the coordinates of the four endpoints of the road network research scope, where the road network research scope is a rectangle.

[0032] As a further improvement of the above solution, the step of using an aircraft to obtain the road network images within the road network research scope specifically includes: dividing the road network research scope according to a preset division strategy to obtain a plurality of sampling grids; performing path planning according to the coordinates of the centroids in the sampling grids, so that the aircraft performs image sampling operations within each sampling grid according to the planned path, obtains grid images corresponding to each sampling grid, and combines all the grid images to form the road network image.

[0033] As a further improvement of the above solution, the step of dividing the road network research scope according to a preset division strategy to obtain a plurality of sampling grids specifically includes: obtaining the effective visual range width of the aircraft, and calculating the interval according to the effective visual range width; dividing according to the coordinates of the four endpoints of the road network research scope and the interval to obtain a plurality of sampling grids.

[0034] Exemplarily, the current research scope is divided into a*b sampling grids:

[0035]

[0036]

[0037] where w is the effective viewing distance width, and x min is the minimum abscissa of the road network research scope, and x max is the maximum abscissa of the road network research scope, and y min is the minimum ordinate of the road network research scope, and y max is the maximum ordinate of the road network research scope;

[0038] Taking the centroid in each sampling grid as the sampling point, the acquisition of the grid image within the grid can be realized. Among them, the centroid (x c，n , y c，m ) of the n*m sampling grid can be expressed as:

[0039]

[0040]

[0041] S3. Input the road network image into a preset convolutional neural network model to obtain the road network vehicle detection result output by the convolutional neural network model;

[0042] Preferably, the step of inputting the road network image into a preset convolutional neural network model to obtain the road network vehicle detection result output by the convolutional neural network model specifically includes: inputting the road network image into the backbone network in the convolutional neural network model to obtain a plurality of pick-up box attributes output by the backbone network; wherein, the backbone network includes a convolutional layer, a group normalization layer, and a Mish activation function, and the pick-up box attributes include pick-up box encoding, pick-up box center point coordinates, pick-up box length, pick-up box width, and pick-up box confidence; for each pick-up box attribute, when it is judged whether the corresponding pick-up box confidence is greater than a preset confidence threshold, if so: determine that the pick-up object corresponding to this pick-up box attribute is a vehicle, and calculate the vehicle model according to the pick-up box length and pick-up box width in this pick-up box attribute; output all the calculated vehicle models as the road network vehicle detection result.

[0043] As a further improvement of the above solution, inputting the road network image into the backbone network of the convolutional neural network model to obtain several pick-up box attributes output by the backbone network specifically includes: inputting a road network image with three channels into the backbone network to obtain three-dimensional feature tensors corresponding to the three channels output by the backbone network; sorting the feature tensors according to the rule of decreasing dimension to obtain a first feature tensor, a second feature tensor, and a third feature tensor; performing a deconvolution operation on the third feature tensor and connecting the deconvolution-operated third feature tensor with the second feature tensor to obtain a new second feature tensor; performing a deconvolution operation on the new second feature tensor and connecting the deconvolution-operated second feature tensor with the first feature tensor to obtain a new first feature tensor; performing a convolution operation on the new first feature tensor to obtain the pick-up box attributes.

[0044] Exemplarily, for the input road network image, construct a backbone network composed of multiple convolutional blocks (convolutional layer + group normalization layer + Mish activation function) to implement inputting a 3-channel image with different pixel sizes, outputting three-dimensional feature tensors, and recording them as Tensor1, Tensor2, and Tensor3 in descending order of dimension; perform deconvolution on Tensor3 and output a tensor with the same dimension as Tensor2, denoted as Tensor’3; connect Tensor’3 and Tensor2 to form a new tensor Tensor 2，3 ; perform deconvolution on Tensor 2，3 to obtain a tensor with the same dimension as Tensor1, denoted as Tensor’ 2，3 , connect Tensor’ 2，3 and Tensor1 to form a new tensor Tensor 1，2，3 , finally perform convolution on Tensor 1，2，3 to output a tensor of s×s×9, denoted as output (s represents the recognition accuracy, determined by the accuracy and size of the aerial image, and 9 represents the number of attributes), use the recognized pick-up box attributes to replace the vehicle attributes, and respectively obtain the pick-up box encoding id, the center point coordinates (x id , y id ) of the pick-up box, the length h id and width w id of the pick-up box, the confidence c id that the picked object is a vehicle, and calculate the probabilities that the vehicle type belongs to a small vehicle, a medium vehicle, and a large vehicle, respectively represented as p id and width w id of the pick-up box, calculate the probabilities that the vehicle type belongs to a small vehicle, a medium vehicle, and a large vehicle, respectively represented as p id，1 , p id，2 , p id，3 ; use max{p id，i} to judge the vehicle type Cartypeid 。

[0045] S4. Calculate the road network traffic flow parameters within the scope of the road network study based on the vehicle detection results of the road network.

[0046] As a further improvement of the above solution, calculating the road network traffic flow parameters within the scope of the road network study according to the vehicle detection results of the road network specifically includes: for each road section, calculate the road section length according to the corresponding boundary, calculate the number of standard vehicles based on the number of pickups determined to be vehicles within this road section, the preset standard vehicle coefficient, and the vehicle type, and calculate the average vehicle density of this road section based on the number of standard vehicles and the road section length; merge the average vehicle densities of all calculated road sections to obtain the average vehicle density of the road network in the road network traffic flow parameters.

[0047] Exemplarily, obtain the coordinates of the four endpoints of the rectangle expanded corresponding to any road section n, and denote [(x 1,n , y 1,n ), (x 2,n , y 2,n ), (x 3,n , y 3,n ), (x 4,n , y 4,n )] as this rectangle, then the road section length L n (unit: m) is:

[0048]

[0049] For the G road network images obtained by sampling, obtain the number of pickups determined to be vehicles located in road section n in the road network image and the corresponding vehicle types calculated. For the g-th road network image among them, count the numbers of small cars, medium cars, and large cars on road section n, which are respectively represented as and (the units are all: vehicles), and calculate the number of standard vehicles C n，g (unit: pcu) on road section n in road network image g according to the standard vehicle coefficient. The calculation formula is as follows:

[0050]

[0051] Then for all G road network images, calculate the average value of the number of standard vehicles of road section n to obtain the average vehicle density (unit: vehicles / km) of road section n, and there is

[0052]

[0053] As a further improvement of the above solution, calculating the road network traffic flow parameters within the scope of the road network research according to the road network vehicle detection results specifically includes: for each road section, calculating the corresponding section interval speed according to the pick-up frame code and the pick-up frame center point coordinates in the pick-up frame attributes; merging the calculated section interval speeds of all road sections to obtain the road network interval speed in the road network traffic flow parameters.

[0054] Exemplarily, for the G sampled road network images, the coordinates of each vehicle on road section n are obtained. For the g-th road network image, the coordinates of the vehicle with pick-up frame number id are denoted as (x g，id , y g，id ), and the average moving distance of the vehicle is denoted as d id . The calculation formula is as follows:

[0055]

[0056] Taking the average time vehicle speed to represent the section interval speed of road section n (unit: km / h), there is

[0057]

[0058] As a further improvement of the above solution, calculating the road network traffic flow parameters within the scope of the road network research according to the road network vehicle detection results specifically includes: for each road section, calculating the corresponding section average traffic flow according to the corresponding section average vehicle density and the section interval speed through a preset flow-density-speed formula; wherein, the flow-density-speed formula is that the section average traffic flow is equal to the section average vehicle density multiplied by the section interval speed; merging the calculated section average traffic flows of all road sections to obtain the road network average traffic flow in the road network traffic flow parameters.

[0059] Exemplarily, the section average vehicle density (unit: pcu / km) and the section interval speed (unit: km / h) are respectively calculated. According to the flow-density-speed formula, the section average traffic flow (unit: pcu / h) of road section n is obtained. There is

[0060]

[0061] An embodiment of the present invention further provides a computer device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, the steps of the road network traffic flow parameter estimation method described in any one of the above are implemented.

[0062] Wherein, when the processor executes the computer program, the steps in the above embodiments of the various road network traffic flow parameter estimation methods are implemented, such as Figure 1 the steps S1 to S4 shown.

[0063] Exemplarily, the computer program may be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program in the computer device.

[0064] The computer device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device may include, but is not limited to, a processor and a memory.

[0065] The processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the computer device and connects various parts of the entire computer device through various interfaces and lines.

[0066] The memory can be used to store the computer programs and / or modules. By running or executing the computer programs and / or modules stored in the memory, and by invoking the data stored in the memory, the processor realizes various functions of the computer device. The memory mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area can store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0067] Among them, if the modules / units integrated in the computer device are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above-described embodiment methods of the present invention, it can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-described various method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a mobile hard disk, a magnetic disk, an optical disc, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0068] Through the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus a necessary hardware platform, and of course, it can also be implemented entirely by hardware. Based on such an understanding, all or part of the technical solutions of the present invention that contribute to the background art can be embodied in the form of a software product. The computer software product can be stored in a storage medium, such as ROM / RAM, a magnetic disk, an optical disc, etc., and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0069] In summary, the present invention has the following beneficial effects:

[0070] By adopting the embodiment of the present invention, basic data is obtained, wherein the basic data includes the boundaries of several road sections; the road network research scope is determined according to the boundaries of all road sections, and a road network image within the road network research scope is obtained; the road network image is input into a preset convolutional neural network model to obtain a road network vehicle detection result output by the convolutional neural network model; according to the road network vehicle detection result, road network traffic flow parameters within the road network research scope are calculated; in particular, after dividing the road network research scope, a plurality of sampling grids are obtained, and path planning is performed according to the sampling grids, so that the aircraft performs image sampling along the planned path, and the obtained grid images form a road network image. The embodiment of the present invention can quickly identify the vehicle type and position, thereby quickly calculating the average vehicle density, section interval speed and section average flow of the road section, and can efficiently and flexibly obtain the all-day time-sharing traffic operation parameters of the large-scale road network at a lower cost.

[0071] The above is the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements are also regarded as the protection scope of the present invention.

Claims

1. A method for estimating road network traffic flow parameters, characterized in that Including: Obtain basic data; wherein, the basic data includes the boundaries of several road segments. Determine the road network research scope according to the boundaries of all road segments, and obtain the road network image within the road network research scope. Input the road network image into a preset convolutional neural network model to obtain the road network vehicle detection result output by the convolutional neural network model. Calculate the road network traffic flow parameters within the road network research scope according to the road network vehicle detection result. Among them, the determination of the road network research scope according to the boundaries of all road segments and the acquisition of the road network image within the road network research scope specifically include: Read the coordinates of the boundaries from the basic data. For each road segment, expand a rectangle according to the corresponding boundary and obtain the endpoint coordinates of the rectangle. Select the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate according to the endpoint coordinates, and use the minimum abscissa, maximum abscissa, minimum ordinate, and maximum ordinate as the coordinates of the four endpoints of the road network research scope. Use an aircraft to obtain the road network image within the road network research scope. Among them, the use of an aircraft to obtain the road network image within the road network research scope specifically includes: Divide the road network research scope according to a preset division strategy to obtain multiple sampling grids. Perform path planning according to the coordinates of the centroids in the sampling grids, so that the aircraft performs image sampling operations within each sampling grid according to the planned path, obtains each grid image corresponding to each sampling grid, and combines all grid images to form the road network image. Among them, the division of the road network research scope according to a preset division strategy to obtain multiple sampling grids specifically includes: Obtain the effective viewing distance width of the aircraft, and calculate the spacing according to the effective viewing distance width. Divide according to the coordinates of the four endpoints of the road network research scope and the spacing to obtain multiple sampling grids.

2. The method for estimating road network traffic flow parameters according to claim 1, wherein The input of the road network image into the backbone network of the preset convolutional neural network model to obtain several pick-up box attributes output by the backbone network specifically includes: Input the road network image into the backbone network of the convolutional neural network model to obtain several pick-up box attributes output by the backbone network; wherein, the backbone network includes a convolutional layer, a group normalization layer, and a Mish activation function, and the pick-up box attributes include pick-up box encoding, pick-up box center point coordinates, pick-up box length, pick-up box width, and pick-up box confidence. For each pick-up box attribute, when determining whether the corresponding pick-up box confidence is greater than a preset confidence threshold, if so, determine that the pick-up object corresponding to this pick-up box attribute is a vehicle, and calculate the vehicle model according to the pick-up box length and pick-up box width in this pick-up box attribute. Output all calculated vehicle models as the road network vehicle detection result.

3. The method for estimating road network traffic flow parameters according to claim 2, characterized in that, The input of the road network image into the backbone network of the convolutional neural network model to obtain several pick-up box attributes output by the backbone network specifically includes: Input the road network image with three channels into the backbone network to obtain three-dimensional feature tensors corresponding to the three channels output by the backbone network; Sort the feature tensors according to the rule of decreasing dimension to obtain the first feature tensor, the second feature tensor, and the third feature tensor; Perform a deconvolution operation on the third feature tensor, and concatenate the third feature tensor after the operation with the second feature tensor to obtain a new second feature tensor; Perform a deconvolution operation on the new second feature tensor, and concatenate the second feature tensor after the operation with the first feature tensor to obtain a new first feature tensor; Perform a convolution operation on the new first feature tensor to obtain the pick-up box attribute.

4. The method for estimating road network traffic flow parameters according to claim 2, wherein The road network traffic flow parameters within the research scope of the road network are calculated according to the road network vehicle detection result, specifically including: For each road segment, calculate the road segment length according to the corresponding boundary, calculate the standard vehicle number according to the number of pick-up objects determined as vehicles within this road segment, the preset standard vehicle coefficient, and the vehicle type, and calculate the average vehicle density of this road segment according to the standard vehicle number and the road segment length; Merge the average vehicle densities of all road segments calculated to obtain the average vehicle density of the road network in the road network traffic flow parameters.

5. The method for estimating road network traffic flow parameters according to claim 4, wherein The road network traffic flow parameters within the research scope of the road network are calculated according to the road network vehicle detection result, specifically including: For each road segment, calculate the corresponding road segment interval speed according to the pick-up box code and the pick-up box center point coordinates in the pick-up box attribute; Merge the road segment interval speeds of all road segments calculated to obtain the road network interval speed in the road network traffic flow parameters.

6. The method for estimating road network traffic flow parameters according to claim 5, characterized in that, The road network traffic flow parameters within the research scope of the road network are calculated according to the road network vehicle detection result, specifically including: For each road segment, calculate the corresponding road segment average flow according to the corresponding road segment average vehicle density and road segment interval speed through a preset flow-density-speed formula; where the flow-density-speed formula is that the road segment average flow is equal to the road segment average vehicle density multiplied by the road segment interval speed; Merge the road segment average flows of all road segments calculated to obtain the road network average flow in the road network traffic flow parameters.

7. A computer device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the road network traffic flow parameter estimation method according to any one of claims 1 to 6.

Images