UAV inspection data collection method and system for road construction sites
By establishing a spatio-temporal coordinate system and genetic algorithm to optimize the drone inspection plan, the problems of low data acquisition efficiency and large amount of data are solved, efficient data acquisition and coverage of key locations are achieved at the road construction site, and the On-site data processing pressure.
Patent Information
- Application Number
- CN202411091552.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-09
AI Technical Summary
The data collection of drones at road construction sites is large in volume and requires high computing resources, resulting in low patrol efficiency and high cost, and the fixed camera cannot cover all key locations, especially in long-distance construction scenarios.
Obtain construction location information and traffic flow data through map database and traffic history database, establish a spatio-temporal coordinate system, normalize and superimpose fitted traffic data, determine the construction impact interval, randomly generate the drone observation status, and optimize the observation information using genetic algorithms The ratio of entropy to existing states, optimize the drone inspection plan, reduce invalid flights, and improve data acquisition efficiency.
Data collection is carried out at key locations and time points, maximize observation information entropy, reduce invalid flights with low information volume, and preprocess data in the cloud, reduce on-site data processing pressure, and improve data acquisition efficiency and accuracy.
Smart Images

Figure CN118628943B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of engineering monitoring, and in particular to a method and system for collecting unmanned aerial vehicle inspection data at a road construction site. Background Art
[0002] In the current management and monitoring of road construction sites, manual inspections and fixed camera monitoring are usually used. However, manual inspections are inefficient and costly, and fixed cameras cannot cover all key locations, especially in long-distance road construction scenarios. With the development of drone technology, drones have shown significant advantages in monitoring road construction sites due to their high maneuverability, flexibility and wide coverage. Drones can be quickly deployed at construction sites to obtain high-resolution aerial images and video data in real time, making up for the shortcomings of traditional monitoring methods. However, in large-scale and long-distance data collection, drones still face the problem of frequent inspections of the entire construction section in order to grasp the on-site construction conditions and traffic impacts as much as possible, which results in high requirements for the life and endurance of drones. In addition, the large amount of image data collected over a long inspection time requires relatively expensive data analysis and processing computing resources. Summary of the invention
[0003] (1) Technical issues to be resolved
[0004] The purpose of the present invention is to provide a method and system for collecting data from drone inspections at road construction sites, so as to solve the problem that the amount of data collected by drone inspections is large and requires large computing resources to extract features of monitored objects.
[0005] (2) Technical solution
[0006] To achieve the above objectives, on the one hand, the present invention provides a method for collecting data from a drone inspection at a road construction site, the method comprising:
[0007] The construction location information of the road construction section is obtained through the map database, and the construction location information includes the construction time point and the corresponding construction location coordinates and the number of controlled lanes of the construction section; the traffic flow of the road construction section in the non-construction period is obtained through the traffic history database and recorded as the first traffic flow data, and the first traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point in the non-construction period;
[0008] In the early stage of road construction, the aerial images of the road construction section are collected through regular inspections by drones to establish the traffic flow during the construction period and record it as the second traffic flow data. The second traffic flow data is the time series data of the observed traffic flow obtained at the observation position and observation time point during the construction period;
[0009] Establish a spatiotemporal coordinate system, normalize and superimpose the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and find the construction impact interval in which the first traffic flow data and the second traffic flow data exceed the set distance threshold in the spatiotemporal coordinate system; establish a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, a construction impact time point, and a construction impact traffic flow that reflect the road construction impact reaching the traffic congestion threshold;
[0010] A single-frame aerial image of a UAV at a preset aerial photography height is obtained, and a single-observation road area for identifying traffic flow information is obtained based on the single-frame aerial image. The road construction section is divided into multiple observation blocks according to the single-observation road area and unit time; UAV observation states are randomly generated as the initial population in each observation block within the construction impact interval. Each UAV observation state includes an existence state and an observation information entropy, where the existence state includes two states, 0 and 1, and the observation information entropy reflects the change in traffic flow under the influence of road construction; the maximum value of the ratio between the observation information entropy and the number of existence states of 1 is obtained through a genetic algorithm; the traffic flow information at the observation location and observation time point where the existence state is 1 is used as the inspection and collection task of the UAV.
[0011] Furthermore, the method of establishing a spatiotemporal coordinate system, normalizing and superimposing the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and finding the construction impact interval in which the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system includes:
[0012] Defining space-time coordinate systems ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for:
[0013] ;
[0014] Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for:
[0015] ;
[0016] in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
[0017] Furthermore, each observation block in the construction-affected area randomly generates a drone observation state as an initial population, each drone observation state includes an existence state and an observation information entropy, wherein the existence state includes two states of 0 and 1, and the observation information entropy reflects the amount of change in traffic flow under the influence of road construction; the maximum value of the ratio between the observation information entropy and the number of existence states of 1 is obtained by a genetic algorithm; the method of using the traffic flow information of the observation position and observation time point of the existence state of 1 as the inspection and collection task of the drone includes:
[0018] Get the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
[0019] Furthermore, the individuals in the initial population The method of obtaining the observation information entropy under the observation state of the drone and calculating the ratio between the observation information entropy and the number of existence states of 1 and recording it as the observation information ratio includes:
[0020] Get the initial population Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for:
[0021] ;
[0022] in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for:
[0023] .
[0024] Furthermore, the method further comprises:
[0025] Set road construction section to Two-dimensional coordinate plane, if the existence status of the drone in the drone observation plan is 1, then its observation coordinates are recorded and the observation coordinates are The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme;
[0026] The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include:
[0027] The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
[0028] Based on the same inventive concept, on the other hand, the present invention also provides a drone inspection data collection system for a road construction site, the system comprising:
[0029] The first acquisition module is used to obtain the construction location information of the road construction section through the map database, and the construction location information includes the construction time point and the corresponding construction location coordinates and the number of controlled lanes in the construction section; the traffic flow of the road construction section in the non-construction period is obtained through the traffic history database and recorded as the first traffic flow data, and the first traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point in the non-construction period;
[0030] The second acquisition module is used to collect aerial images of the road construction section through regular inspections by drones in the early stage of road construction to establish the traffic flow during the construction period and record it as the second traffic flow data, wherein the second traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point during the construction period;
[0031] The construction impact assessment module is used to establish a spatiotemporal coordinate system, normalize and superimpose the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, find the construction impact interval in which the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system; establish a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, a construction impact time point, and a construction impact traffic flow that reflect the road construction impact reaching a traffic congestion threshold;
[0032] The UAV inspection planning module is used to obtain a single-frame aerial image of the UAV at a preset aerial photography height, obtain a single observation road area for identifying traffic information based on the single-frame aerial image, and divide the road construction section into multiple observation blocks according to the single observation road area and unit time; randomly generate UAV observation states as the initial population in each observation block in the construction impact area, each UAV observation state includes an existence state and an observation information entropy, where the existence state includes two states, 0 and 1, and the observation information entropy reflects the change in traffic under the influence of road construction; obtain the maximum value of the ratio between the observation information entropy and the number of existence states of 1 through a genetic algorithm; and use the traffic information at the observation location and observation time point where the existence state is 1 as the UAV inspection collection task.
[0033] Furthermore, the system further comprises:
[0034] Traffic flow fitting module, used to define the space-time coordinate system ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for:
[0035] ;
[0036] Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for:
[0037] ;
[0038] in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
[0039] Furthermore, the system further comprises:
[0040] The first iterative optimization module is used to obtain the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
[0041] Furthermore, the system further comprises:
[0042] The observation information ratio optimization module is used to obtain the first Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for:
[0043] ;
[0044] in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for:
[0045] .
[0046] Furthermore, the system further comprises:
[0047] The second iteration optimization module is used to set the road construction section as Two-dimensional coordinate plane, if the existence status of the drone in the drone observation plan is 1, then its observation coordinates are recorded and the observation coordinates are The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme;
[0048] The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include:
[0049] The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
[0050] (3) Beneficial effects
[0051] Compared with the prior art, the present invention has the following beneficial effects:
[0052] 1. By optimizing the drone inspection data collection method and optimizing the drone observation plan within the construction impact area, ensure data collection at key locations and time points, maximize observation information entropy, and reduce invalid flights with low information content.
[0053] 2. By moving a large amount of data processing work to the cloud in advance, a construction impact model is established through early construction data. In the later stage, only simple basic data of key locations and key times need to be transmitted, which greatly reduces the pressure of on-site data processing. BRIEF DESCRIPTION OF THE DRAWINGS
[0054] Figure 1 This is a flowchart of a method for collecting data from a drone inspection at a road construction site according to Embodiment 1 of the present invention;
[0055] Figure 2 This is a module block diagram of a drone inspection data collection system for a road construction site according to Embodiment 2 of the present invention. DETAILED DESCRIPTION
[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0057] Before giving examples, it is necessary to explain the application scenarios of the present invention. When a road construction site is constructing a long road such as a main road, multiple lanes are usually closed or traffic is controlled. However, in order to pursue efficiency, multiple sections of the road are divided for construction. When the construction is completed, the traffic control of the corresponding road section is cancelled. When drones are used to supervise on-site construction, it is impossible for drones to be available at any time and anywhere on the construction section. Another reason is that the processing volume of drones to extract target feature data through image analysis is very large.
[0058] Example 1: Figure 1 As shown, this embodiment provides a method for collecting drone inspection data at a road construction site, the method comprising:
[0059] S1. Obtaining construction location information of a road construction section through a map database, wherein the construction location information includes a construction time point and corresponding construction location coordinates and the number of controlled lanes of the construction section; obtaining the traffic flow of the road construction section during a non-construction period through a traffic history database as first traffic flow data, wherein the first traffic flow data is a time-series data of observed traffic flow obtained at an observation location and an observation time point during a non-construction period;
[0060] S2. In the early stage of road construction, the aerial images of the road construction section are collected by regular inspection by drones to establish the traffic flow during the construction period and record it as the second traffic flow data. The second traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point during the construction period;
[0061] S3, establishing a spatiotemporal coordinate system, normalizing and superimposing the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and finding the construction impact interval in which the first traffic flow data and the second traffic flow data exceed the set distance threshold in the spatiotemporal coordinate system; establishing a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, construction impact time point, and construction impact traffic flow that reflect the road construction impact reaching the traffic congestion threshold;
[0062] S4. Obtain a single-frame aerial image of the UAV at a preset aerial photography height, obtain a single-observation road area for identifying traffic flow information based on the single-frame aerial image, and divide the road construction section into multiple observation blocks according to the single-observation road area and unit time; randomly generate UAV observation states as the initial population in each observation block within the construction impact interval, each UAV observation state includes an existence state and an observation information entropy, wherein the existence state includes two states, 0 and 1, and the observation information entropy reflects the amount of traffic flow change under the influence of road construction; obtain the maximum value of the ratio between the observation information entropy and the number of existence states of 1 through a genetic algorithm; and use the traffic flow information of the observation position and observation time point at which the existence state is 1 as the inspection and collection task of the UAV.
[0063] For example, suppose a section of a main road is under construction. In order to improve construction efficiency, the main road construction is divided into several sections and carried out simultaneously. In each section, one or two lanes are closed for traffic control. To ensure smooth traffic and improve construction efficiency during the construction process, drone inspections are used for aerial monitoring.
[0064] Obtain detailed information about the construction section through the map database, including the specific time of construction, construction location coordinates and the number of controlled lanes. For example:
[0065] Construction location: from the 10th kilometer to the 20th kilometer of the road.
[0066] Construction hours: 9am to 5pm daily.
[0067] Controlled lanes: One lane is closed, leaving the remaining two lanes open to traffic.
[0068] The traffic flow data during the non-construction period is obtained through the traffic history database as the first traffic flow data. Existing traffic cameras or other monitoring methods such as map navigation software can be used as data sources. Assume that the traffic flow data for the past week is recorded in different time periods every day, for example:
[0069] Observation position: 10 km, 15 km, 20 km.
[0070] Observation time: 9:00, 12:00 and 15:00 every day.
[0071] Traffic flow data (during non-construction period): (unit: vehicles / hour)
[0072] 10th kilometer: 200 vehicles at 9 o'clock, 250 vehicles at 12 o'clock, and 230 vehicles at 15 o'clock.
[0073] 15th kilometer: 220 vehicles at 9 o'clock, 260 vehicles at 12 o'clock, 240 vehicles at 15 o'clock.
[0074] 20th kilometer: 210 vehicles at 9 o'clock, 255 vehicles at 12 o'clock, and 235 vehicles at 15 o'clock.
[0075] In the early stage of construction, drones are used to regularly patrol and collect aerial images of the construction section, and traffic flow data is extracted from the images as the second traffic flow data. Traffic flow is identified through aerial images, that is, the number of vehicles. There are many algorithms such as the Yolo algorithm that can achieve dynamic capture. Since drones cannot monitor every section of the road at all times, we set up drones to patrol at key time points every day.
[0076] Establish a spatiotemporal coordinate system, normalize the first traffic flow data and the second traffic flow data and map them to the spatiotemporal coordinate system; standardize the traffic flow data so that data at different time points and locations can be compared. For example, normalize the traffic flow data according to the maximum and minimum values:
[0077] Normalized first traffic flow data:
[0078] 10th km: 0.80 at 9 o'clock, 1.00 at 12 o'clock, and 0.92 at 15 o'clock.
[0079] 15th km: 0.85 at 9 o'clock, 1.00 at 12 o'clock, 0.92 at 15 o'clock.
[0080] 20th km: 0.82 at 9 o'clock, 1.00 at 12 o'clock, 0.92 at 15 o'clock.
[0081] Normalized second traffic flow data:
[0082] 10th km: 0.60 at 9 o'clock, 0.80 at 12 o'clock, 0.72 at 15 o'clock.
[0083] 15th km: 0.65 at 9 o'clock, 0.84 at 12 o'clock, 0.76 at 15 o'clock.
[0084] 20th km: 0.62 at 9 o'clock, 0.82 at 12 o'clock, 0.74 at 15 o'clock.
[0085] In the space-time coordinate system, the first traffic flow data and the second traffic flow data are superimposed and fitted to find the construction impact interval where the data exceeds the set distance threshold. For example, the threshold we set is a 30% decrease in traffic flow. Through superimposition and fitting calculation, we get:
[0086] The construction impact area is 50-300 meters before and after the 10th and 20th kilometers, and there are large traffic fluctuations at 9:00 and 15:00 every day, which affects the traffic efficiency and causes traffic jams. The construction impact model can be obtained by fitting the relationship between construction location information and traffic flow changes through a multivariate linear regression algorithm; it can also be trained and predicted using historical data and real-time data through machine learning methods such as random forests or support vector machines.
[0087] In order to improve the efficiency of data collection, a single-frame aerial image of a drone at a preset aerial photography height is obtained, and the single-time observation road area of the traffic flow information is identified based on the image. The road construction section is divided into multiple observation blocks according to the single-time observation road area and unit time. For example, the observation block size: one block per kilometer, one block per hour.
[0088] In each observation block of the construction impact zone, the drone observation state is randomly generated as the initial population. Each observation state includes the existence state (0 or 1) and the observation information entropy. The genetic algorithm is used to optimize the ratio between the observation information entropy and the number of existence states of 1, and the optimal drone inspection and collection task is determined. That is, the drone observation plan with the highest observation information efficiency is obtained by the amount of information obtained by observation and the number of drone observations (which may be a single or multiple drones). At this time, the observation information entropy has a preset minimum information entropy threshold to ensure that the minimum amount of observation information can be combined with the construction impact model to obtain the actual overall picture of the construction impact.
[0089] In the early stages of construction, a large amount of data will be transmitted to a cloud-based data processing and analysis computer to extract a model of the impact of road construction on traffic. Subsequent data processing only requires the transmission of simple basic data on key locations and key times, without the need to separately deploy data processing resources for data analysis and processing.
[0090] Furthermore, the method of establishing a spatiotemporal coordinate system, normalizing and superimposing the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and finding the construction impact interval in which the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system includes:
[0091] Defining space-time coordinate systems ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for:
[0092] ;
[0093] Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for:
[0094] ;
[0095] in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
[0096] Exemplarily, the first traffic flow data and the second traffic flow data are normalized and mapped to a spatiotemporal coordinate system. Assume that the normalized traffic flow data is as follows:
[0097] First traffic data:
[0098] 10th km: 0.8 at 9 o'clock, 1.00 at 12 o'clock, 0.92 at 15 o'clock.
[0099] 15th km: 0.85 at 9 o'clock, 1.00 at 12 o'clock, 0.92 at 15 o'clock.
[0100] 20th km: 0.82 at 9 o'clock, 1.00 at 12 o'clock, 0.92 at 15 o'clock.
[0101] Second traffic data:
[0102] 10th km: 0.60 at 9 o'clock, 0.80 at 12 o'clock, 0.72 at 15 o'clock.
[0103] 15th km: 0.65 at 9 o'clock, 0.84 at 12 o'clock, 0.76 at 15 o'clock.
[0104] 20th km: 0.62 at 9 o'clock, 0.82 at 12 o'clock, 0.74 at 15 o'clock.
[0105] The normalized first traffic flow data and the second traffic flow data are respectively mapped to the time-space coordinate system to form a point set and .For example:
[0106] : {(10, 9:00, 3, 0.8), (10, 12:00, 3, 1.0), (10, 15:00, 3, 0.92),(15, 9:00, 3, 0.85), (15, 12:00, 3, 1.0), (15, 15:00, 3, 0.92), (20, 9:00, 3,0.82), (20, 12:00, 3, 1.0), (20, 15:00, 3, 0.92)}.
[0107] : {(10, 9:00, 2, 0.6), (10, 12:00, 2, 0.8), (10, 15:00, 2, 0.72),(15, 9:00, 2, 0.65), (15, 12:00, 2, 0.84), (15, 15:00, 2, 0.76), (20, 9:00,2, 0.62), (20, 12:00, 2, 0.82), (20, 15:00, 2, 0.74)}.
[0108] The first and second curves are obtained by spline interpolation fitting. Three point sets {(10,9:00,3,0.8), (15,12:00,3,1.0), (20,15:00,3,0.92)} are sampled equidistantly on the first curve, and three point sets {(10,9:00,2,0.6), (15,12:00,2,0.84), (20,15:00,2,0.74)} are sampled equidistantly on the second curve.
[0109] Move the second curve to the first curve by moving the vector (0.5,0.5,0,0)(0.5, 0.5, 0, 0)(0.5,0.5,0,0) to perform superposition fitting. For example:
[0110] The point of the second curve after the shift at the time point 9:00 is (10.5,9:00,2)(10.5, 9:00, 2)(10.5,9:00,2).
[0111] The point of the second curve after the shift at the time point 12:00 is (15.5, 12:00, 2)(15.5, 12:00, 2)(15.5, 12:00, 2).
[0112] The point of the second curve after the shift at the time point 15:00 is (20.5,15:00,2)(20.5, 15:00, 2)(20.5,15:00,2).
[0113] Calculate the distance between the first curve and the second curve at the same construction time point. For example:
[0114] At time 9:00, the point on the first curve is (10,9:00,3)(10, 9:00, 3)(10,9:00,3), and the point on the second curve is (10.5,9:00,2)(10.5, 9:00, 2)(10.5,9:00,2), and the distance between the two points is 1.12.
[0115] At the time point 12:00, the point on the first curve is (15,12:00,3)(15, 12:00, 3)(15,12:00,3) and the point on the second curve is (15.5,12:00,2)(15.5, 12:00, 2)(15.5,12:00,2). The distance between the two points is 1.12.
[0116] At time 15:00, the point on the first curve is (20,15:00,3)(20, 15:00, 3)(20,15:00,3) and the point on the second curve is (20.5,15:00,2)(20.5, 15:00, 2)(20.5,15:00,2). The distance between the two points is 1.12.
[0117] The positions whose distance exceeds the set threshold are determined as the construction impact interval. For example, if the threshold is set to 1.0, the positions of all time points in the above calculations exceed the threshold. Therefore, the positions (10,2), (15,2), and (20,2) at the time points 9:00, 12:00, and 15:00 are all construction impact intervals. Please note that the above sampling time points are actually patrol sampling regularly based on unit time. For example, if the sampling unit time is 30 minutes, then 9:00 here actually represents the construction impact interval of the time period of 9:00-9:30.
[0118] Furthermore, each observation block in the construction-affected area randomly generates a drone observation state as an initial population, each drone observation state includes an existence state and an observation information entropy, wherein the existence state includes two states of 0 and 1, and the observation information entropy reflects the amount of change in traffic flow under the influence of road construction; the maximum value of the ratio between the observation information entropy and the number of existence states of 1 is obtained by a genetic algorithm; the method of using the traffic flow information of the observation position and observation time point of the existence state of 1 as the inspection and collection task of the drone includes:
[0119] Get the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
[0120] Exemplarily, drone observation states are randomly generated as the initial population in each observation block in the construction impact interval. Each observation state includes an existence state (0 means that the drone is not observed, 1 means that the drone is observed) and the observation information entropy. For example, we generate initial populations in the construction impact interval (10 km, 15 km, 20 km) and time points (9:00, 12:00, 15:00). The selection of observation blocks is due to the fact that the single aerial photography of the drone is affected by the field of view, that is, the traffic flow of a certain area can be distinguished at a time. Some roads that are wider may require drone scanning multiple times; the same construction road is regarded as different observation blocks at different times, such as rush hour or low peak at noon, and its impact on traffic flow is completely different.
[0121] Assume that the initial population contains 3 individuals, each individual has 9 observation blocks, and the state of each observation block is as follows:
[0122] Individual 1: (1, 0.6), (0, 0.8), (1, 0.72), (1, 0.65), (0, 0.84), (1, 0.76), (0, 0.62), (1, 0.82), (1, 0.74);
[0123] Individual 2: (0, 0.6), (1, 0.8), (1, 0.72), (1, 0.65), (1, 0.84), (0, 0.76), (1, 0.62), (0, 0.82), (0, 0.74);
[0124] Individual 3: (1, 0.6), (1, 0.8), (0, 0.72), (0, 0.65), (1, 0.84), (1, 0.76), (1, 0.62), (0, 0.82), (1, 0.74).
[0125] The observation information entropy of each individual drone under observation state is evaluated in turn, and the ratio between the observation information entropy and the number of existence states of 1 is calculated, which is recorded as the observation information ratio.
[0126] Instance 1:
[0127] Number of existence states 1: 6;
[0128] The total of observed information entropy: 0.6+0.72+0.65+0.76+0.82+0.74=4.29;
[0129] Observation information ratio: 4.29 / 6=0.715.
[0130] Instance 2:
[0131] Number of existence states of 1: 5;
[0132] The total of observed information entropy: 0.8+0.72+0.65+0.84+0.62=3.63;
[0133] Observation information ratio: 3.63 / 5=0.726.
[0134] Instance 3:
[0135] Number of existence states 1: 6;
[0136] The total of observed information entropy: 0.6+0.8+0.84+0.76+0.62+0.74=4.36;
[0137] Observation information ratio: 4.36 / 6=0.727.
[0138] Sort the individuals in the initial population according to the observation information ratio, and select the individuals with the highest observation information ratio for crossover and mutation. Suppose we select individuals 2 and 3 for crossover and mutation.
[0139] Crossover: Randomly select some genes of individual 2 and individual 3 to exchange. Mutation: Randomly change the state of some genes in individuals. Suppose the new individuals generated by crossover and mutation are as follows:
[0140] New individual 1: (1, 0.6), (1, 0.8), (1, 0.72), (1, 0.65), (1, 0.84), (1, 0.76), (1, 0.62), (1, 0.82), (1, 0.74);
[0141] New individual 2: (1, 0.6), (1, 0.8), (0, 0.72), (1, 0.65), (1, 0.84), (0, 0.76), (1, 0.62), (1, 0.82), (1, 0.74).
[0142] The newly generated offspring individuals replace the initial population and perform multiple iterations of evolution until the increase in the observed information ratio between the offspring individuals and the parent individuals is lower than the set threshold. Assume that after multiple iterations of evolution, the final winning individuals are as follows:
[0143] (1,0.6),(1,0.8),(1,0.72),(1,0.65),(1,0.84),(1,0.76),(1,0.62),(1,0.82),(1,0.74). This individual indicates that the drone makes observations in all observation blocks (10 km, 15 km, 20 km, time points 9:00, 12:00, 15:00), and can maximize the ratio between the observation information entropy and the number of existence states of 1.
[0144] Furthermore, the individuals in the initial population The method of obtaining the observation information entropy under the observation state of the drone and calculating the ratio between the observation information entropy and the number of existence states of 1 and recording it as the observation information ratio includes:
[0145] Get the initial population Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for:
[0146] ;
[0147] in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for:
[0148] .
[0149] For example, suppose we have an initial population of individuals whose drone observation schemes are for:
[0150] (1,0.6),(0,0.8),(1,0.72),(1,0.65),(0,0.84),(1,0.76),(0,0.62),(1,0.82),(1,0.74); The observation information entropy of each observation block is calculated from the traffic flow data. For example, the traffic flow data of the observation block is as follows:
[0151] Observation block 1: The traffic volume is 100 vehicles. Observation block 2: The traffic volume is 150 vehicles. Observation block 3: The traffic volume is 120 vehicles. Observation block 4: The traffic volume is 130 vehicles. Observation block 5: The traffic volume is 110 vehicles. Observation block 6: The traffic volume is 140 vehicles. Observation block 7: The traffic volume is 90 vehicles. Observation block 8: The traffic volume is 160 vehicles. Observation block 9: The traffic volume is 170 vehicles.
[0152] The observed information entropy is calculated from the probability distribution of traffic flow, for example:
[0153] Total traffic volume = 100+150+120+130+110+140+90+160+170=1170;
[0154] The probability of observing block 1 = 100 / 1170≈0.085;
[0155] The information entropy of observation block 1 is ≈0.6; the information entropy calculations of other observation blocks are similar.
[0156] For individuals , calculate the total information entropy and observation information ratio of the observation blocks with existence state 1:
[0157] Number of existence states 1: 6;
[0158] The total of observed information entropy: 0.6+0.72+0.65+0.76+0.82+0.74=4.29;
[0159] Observation information ratio: 4.29 / 6=0.715;
[0160] Assuming that individuals The observation scheme is:
[0161] Individual Y: (0, 0.6), (1, 0.8), (1, 0.72), (1, 0.65), (1, 0.84), (0, 0.76), (1, 0.62), (0, 0.82), (0, 0.74);
[0162] Number of existence states of 1: 5;
[0163] The total of observed information entropy: 0.8+0.72+0.65+0.84+0.62=3.63;
[0164] Observation information ratio: 3.63 / 5=0.726.
[0165] Sort the individuals according to the observation information ratio, and select the individual with the highest observation information ratio for crossover and mutation. Assuming that individual Y is selected for crossover and mutation, the generated new individual 1 is as follows:
[0166] (1,0.6),(1,0.8),(1,0.72),(1,0.65),(1,0.84),(1,0.76),(1,0.62),(1,0.82),(1,0.74).
[0167] Through the above steps, the winning individual finally generated can maximize the ratio between the observed information entropy and the number of existence states of 1, thereby improving the efficiency and information content of drone inspection data collection.
[0168] Furthermore, the method further comprises:
[0169] Set road construction section to Two-dimensional coordinate plane, if the existence status of the drone in the drone observation plan is 1, then its observation coordinates are recorded and the observation coordinates are The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme;
[0170] The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include:
[0171] The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value of and the corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value of and the corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
[0172] For example, it is assumed that the drone observation scheme in the winning individual is as follows:
[0173] Winning individuals: (1, 0.6), (1, 0.8), (1, 0.72), (1, 0.65), (1, 0.84), (1, 0.76), (1, 0.62), (1, 0.82), (1, 0.74);
[0174] We set the road construction section as Two-dimensional coordinate plane, and set the drone observation coordinates at Axis and The offset threshold in the axis direction. For example:
[0175] The offset threshold in the axis direction is 10 meters between the center and edge of the road (assuming that the drone can clearly see the traffic flow along the horizon). The axis direction is 30 meters, where the offset threshold refers to the boundary along the direction of road extension);
[0176] The offset threshold in the axis direction is 15 meters between the center and the edge of the road (assuming that the drone can clearly see the traffic flow along the horizon). The axis direction is 30 meters, where the offset threshold refers to the boundaries along both sides of the road).
[0177] The distance that the offset threshold is divided into several equal parts is set as the offset step size. The step length in the a-axis direction is 5 meters, and the step length in the y-axis direction is 3 meters. Axis and Each time the axis direction moves by an offset step, the corresponding observation information ratio is calculated. The extreme value of the observation information ratio is found by the gradient descent method.
[0178] along Axis movement: Initial observation coordinates (e.g., at 1 km) (1,0); observation coordinates after moving one step (1.005,0); calculate the new observation information ratio, for example, the new observation information ratio may be 0.72. Observation coordinates after moving one step again (1.01,0); calculate the new observation information ratio, for example, the new observation information ratio may be 0.74. Repeat the above process until the extreme value of the observation information ratio is found.
[0179] along Axis movement: Initial observation coordinates (e.g. at 0 km) (1,0); observation coordinates after moving one step (1,0.003); calculate the new observation information ratio, for example, the new observation information ratio may be 0.75. Observation coordinates after moving one step again (1,0.006); calculate the new observation information ratio, for example, the new observation information ratio may be 0.77. Repeat the above process until the extreme value of the observation information ratio is found.
[0180] Assume that When the axis moves to (1.03,0), the observation information ratio reaches an extreme value of 0.78; When the axis direction moves to (1.03, 0.02), the observation information ratio reaches an extreme value of 0.79. Combined with the offset step lengths in these two directions, the final observation coordinate position of the drone is determined as (1.03, 0.02); by correcting the drone inspection collection coordinate position in the drone observation plan, it is ensured that the drone collects data at the optimal position, thereby improving the effectiveness and accuracy of the inspection data.
[0181] Embodiment 2: Based on the same inventive concept, Figure 2 As shown, this embodiment also provides a drone inspection data collection system for a road construction site, the system comprising:
[0182] The first acquisition module is used to obtain the construction location information of the road construction section through the map database, and the construction location information includes the construction time point and the corresponding construction location coordinates and the number of controlled lanes in the construction section; the traffic flow of the road construction section in the non-construction period is obtained through the traffic history database and recorded as the first traffic flow data, and the first traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point in the non-construction period;
[0183] The second acquisition module is used to collect aerial images of the road construction section through regular inspections by drones in the early stage of road construction to establish the traffic flow during the construction period and record it as the second traffic flow data, wherein the second traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point during the construction period;
[0184] The construction impact assessment module is used to establish a spatiotemporal coordinate system, normalize and superimpose the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, find the construction impact interval in which the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system; establish a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, a construction impact time point, and a construction impact traffic flow that reflect the road construction impact reaching a traffic congestion threshold;
[0185] The UAV inspection planning module is used to obtain a single-frame aerial image of the UAV at a preset aerial photography height, obtain a single observation road area for identifying traffic information based on the single-frame aerial image, and divide the road construction section into multiple observation blocks according to the single observation road area and unit time; randomly generate UAV observation states as the initial population in each observation block in the construction impact area, each UAV observation state includes an existence state and an observation information entropy, where the existence state includes two states, 0 and 1, and the observation information entropy reflects the change in traffic under the influence of road construction; obtain the maximum value of the ratio between the observation information entropy and the number of existence states of 1 through a genetic algorithm; and use the traffic information at the observation location and observation time point where the existence state is 1 as the UAV inspection collection task.
[0186] Furthermore, the system further comprises:
[0187] Traffic flow fitting module, used to define the space-time coordinate system ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for:
[0188] ;
[0189] Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for:
[0190] ;
[0191] in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
[0192] Furthermore, the system further comprises:
[0193] The first iterative optimization module is used to obtain the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
[0194] Furthermore, the system further comprises:
[0195] The observation information ratio optimization module is used to obtain the first Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for:
[0196] ;
[0197] in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for:
[0198] .
[0199] Furthermore, the system further comprises:
[0200] The second iteration optimization module is used to set the road construction section as Two-dimensional coordinate plane, if the existence status of the drone in the drone observation plan is 1, then its observation coordinates are recorded and the observation coordinates are The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme;
[0201] The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include:
[0202] The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
[0203] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0204] Finally, it should be noted that: Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for collecting data from drone inspections at road construction sites, characterized in that: The method comprises: The construction location information of the road construction section is obtained through the map database, and the construction location information includes the construction time point and the corresponding construction location coordinates and the number of controlled lanes of the construction section; the traffic flow of the road construction section in the non-construction period is obtained through the traffic history database and recorded as the first traffic flow data, and the first traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point in the non-construction period; In the early stage of road construction, the aerial images of the road construction section are collected through regular inspections by drones to establish the traffic flow during the construction period and record it as the second traffic flow data. The second traffic flow data is the time series data of the observed traffic flow obtained at the observation position and observation time point during the construction period; Establish a spatiotemporal coordinate system, normalize and superimpose the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and find the construction impact interval in which the first traffic flow data and the second traffic flow data exceed the set distance threshold in the spatiotemporal coordinate system; establish a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, a construction impact time point, and a construction impact traffic flow that reflect the road construction impact reaching the traffic congestion threshold; Obtain a single-frame aerial image of a drone at a preset aerial photography height, obtain a single-time observed road area for identifying traffic flow information based on the single-frame aerial image, and divide the road construction section into multiple observation blocks according to the single-time observed road area and unit time; randomly generate drone observation states as the initial population in each observation block in the construction impact interval, and each drone observation state includes an existence state and an observation information entropy, where the existence state includes two states, 0 and 1, and the observation information entropy reflects the amount of traffic flow change under the influence of road construction; obtain the maximum value of the ratio between the observation information entropy and the number of existence states of 1 through a genetic algorithm; and use the traffic flow information at the observation location and observation time point with an existence state of 1 as the inspection and collection task of the drone; The method of establishing a spatiotemporal coordinate system, normalizing and superimposing the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, and finding the construction impact interval where the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system includes: Defining space-time coordinate systems ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for: ; Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for: ; in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
2. The method for collecting data from unmanned aerial vehicle inspections at road construction sites according to claim 1, characterized in that: The method randomly generates drone observation states as the initial population in each observation block in the construction impact zone, wherein each drone observation state includes an existence state and an observation information entropy, wherein the existence state includes two states, 0 and 1, and the observation information entropy reflects the amount of change in traffic flow under the influence of road construction; and obtains the maximum value of the ratio between the observation information entropy and the number of existence states of 1 by using a genetic algorithm; The method of using the traffic flow information of the observation location and observation time point with the existence state 1 as the inspection and collection task of the UAV includes: Get the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
3. The method for collecting data from unmanned aerial vehicle inspections at road construction sites according to claim 2, characterized in that: The individuals in the initial population The method of obtaining the observation information entropy under the observation state of the drone and calculating the ratio between the observation information entropy and the number of existence states of 1 and recording it as the observation information ratio includes: Get the initial population Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for: ; in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for: 。 4. The method for collecting data from unmanned aerial vehicle inspections at road construction sites according to claim 3 is characterized in that: The method further comprises: Set road construction section to Two-dimensional coordinate plane, if the existence status of the drone in the drone observation plan is 1, then its observation coordinates are recorded and the observation coordinates are The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. ; When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme; The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include: The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value and corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
5. The drone inspection data collection system at the road construction site is characterized by: The system comprises: The first acquisition module is used to obtain the construction location information of the road construction section through the map database, and the construction location information includes the construction time point and the corresponding construction location coordinates and the number of controlled lanes in the construction section; the traffic flow of the road construction section in the non-construction period is obtained through the traffic history database and recorded as the first traffic flow data, and the first traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point in the non-construction period; The second acquisition module is used to collect aerial images of the road construction section through regular inspections by drones in the early stage of road construction to establish the traffic flow during the construction period and record it as the second traffic flow data, wherein the second traffic flow data is the time sequence data of the observed traffic flow obtained at the observation position and observation time point during the construction period; The construction impact assessment module is used to establish a spatiotemporal coordinate system, normalize and superimpose the first traffic flow data and the second traffic flow data in the spatiotemporal coordinate system, find the construction impact interval in which the first traffic flow data and the second traffic flow data exceed a set distance threshold in the spatiotemporal coordinate system; establish a mapping relationship between the construction impact interval and the construction location information as a construction impact model; the construction impact interval is a construction impact location, a construction impact time point, and a construction impact traffic flow that reflect the road construction impact reaching a traffic congestion threshold; The UAV inspection planning module is used to obtain a single-frame aerial image of the UAV at a preset aerial photography height, obtain a single-time observation road area for identifying traffic flow information based on the single-frame aerial image, and divide the road construction section into multiple observation blocks according to the single-time observation road area and unit time; randomly generate UAV observation states as the initial population in each observation block in the construction impact interval, and each UAV observation state includes an existence state and an observation information entropy, where the existence state includes two states, 0 and 1, and the observation information entropy reflects the change in traffic flow under the influence of road construction; obtain the maximum value of the ratio between the observation information entropy and the number of existence states of 1 through a genetic algorithm; and use the traffic flow information at the observation location and observation time point with an existence state of 1 as the UAV inspection collection task; The system further comprises: Traffic flow fitting module, used to define the space-time coordinate system ,in and Represents the two-dimensional plane coordinate value of the construction location, represents the number of lanes, Represents the construction time point; normalize the first traffic flow data and the second traffic flow data in the time-space coordinate system and map them to the time-space coordinate system to form point sets and , the point set Hedianji The first curve and the second curve are obtained by respectively fitting the spline difference curve; The sampling point set is , and at the same time, take the equidistant Sampling point set ,in The value ranges from 1 to integer variable; by moving the vector Move the second curve to the first curve for superposition fitting, and the total sum of squares of the Euclidean distance between the first curve and the second curve for: ; Input the coordinate data of the sampling point set and obtain , , and Conditional , so as to superimpose and fit the second curve to the first curve; after the second curve is superimposed and fitted to the first curve, the Euclidean distance between the first curve and the second curve at the same construction time point is calculated for: ; in is a point on the first curve, is a point on the first curve Projected onto the second curve; Euclidean distance Exceeding the set distance threshold The value range is taken as the construction impact range.
6. The unmanned aerial vehicle inspection data collection system for road construction sites according to claim 5, characterized in that: The system further comprises: The first iterative optimization module is used to obtain the number of observation blocks , randomly generated in the construction influence area The observation states of the drones are arranged as the drone observation scheme. Each drone observation state includes the existence state and the observation information entropy. The existence state includes two states: 0 and 1. The observation information entropy reflects the change in traffic flow under the influence of road construction. The initial population of the drone observation scheme is generated. The initial population includes Individual , each individual represents a drone observation plan, and each drone observation plan has a randomly generated The initial population is iteratively evolved according to the observation state of the drones arranged in the order of the number of drones. The iterative evolution steps include: sequentially evaluating the individual The observation information entropy under the observation state of the drone is calculated, and the ratio between the observation information entropy and the number of existence states of 1 is calculated and recorded as the observation information ratio; the winning individuals are selected by the roulette selection method according to the observation information ratio, and the winning individuals are crossovered and mutated to obtain the offspring individuals ; Replace the initial population with offspring individuals , and iterate and evolve again until the improvement of the observation information ratio of the offspring individuals and the parent individuals is lower than the set observation information threshold; the drone observation plan of the final winning individual after the iterative evolution is used as the drone's inspection and collection task.
7. The unmanned aerial vehicle inspection data collection system for road construction sites according to claim 6, characterized in that: The system further comprises: The observation information ratio optimization module is used to obtain the first Individual drone observation programs ,in Representative UAV observation solutions No. The existence of drones, The value is 0 or 1; Representative UAV observation solutions No. The observation information entropy of a UAV, A variable whose value is a positive integer. The value ranges from 1 to the number of observation blocks. integer variable; UAV observation solutions No. The observation information entropy of observation blocks for: ; in For the UAV observation solutions No. The traffic flow of each observation block, For the UAV observation solutions No. The traffic flow of each observation block, The value ranges from 1 to integer variable; Individual drone observation programs The observation information ratio for: 。 8. The unmanned aerial vehicle inspection data collection system for road construction sites according to claim 7, characterized in that: The system further comprises: The second iteration optimization module is used to set the road construction section as In the two-dimensional coordinate plane, if the drone's existence status in the drone observation plan is 1, its observation coordinates are recorded and the observation coordinates are recorded in The offset threshold in the axis direction is set to The distance between the center of the road area and the edge of the road construction section in a single observation in the axial direction is calculated by converting the observation coordinates into The offset threshold in the axis direction is set to The distance between the center of the road area observed in a single axis and the edge of the road construction section; the distance of dividing the offset threshold into several equal parts is set as the offset step size, and the drones in the drone observation plan are moved along Axis or Each time the axis moves by one offset step, the corresponding observation information ratio is calculated. ; When the observed information is When it is the highest, the accumulated offset step length is used to correct the inspection and collection coordinate position of the UAV in the UAV observation scheme; The drone in the drone observation plan is placed along Axis or The methods of moving an axis direction by one offset step each time include: The drone in the drone observation plan is placed along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value of and the corresponding Axis cumulative offset; move the drone in the drone observation plan along The axis moves an offset step, obtained by gradient descent. Axis direction observation information ratio The extreme value of and the corresponding The axis cumulative offset; through The axis cumulative offset and The accumulated offset of the axis is taken as the accumulated offset step.
