Road intelligent management method and system based on unmanned aerial vehicle

Abstract

The invention relates to the technical field of road management, in particular to an intelligent road management method and system based on an unmanned aerial vehicle. Intersection videos are collected in real time through an unmanned aerial vehicle, graying, noise reduction and binarization processing are carried out on the intersection videos to extract vehicle features, vehicle movement is tracked, and the number of congested vehicles not passing through a stop line during a red light period and the number of passing vehicles passing through the stop line during a green light period at each phase are counted; and then comparing the difference of the congested vehicles in each phase to judge the setting reasonability of the signal lamps, dynamically calculating the green light duration of each phase and optimizing a timing scheme if the signal lamps are abnormal, and dynamically adjusting the duration of the signal lamps in each color through the high-altitude view angle of the unmanned aerial vehicle and a real-time processing technology, thereby improving the traffic efficiency of the intersection. The technical problem that in the prior art, an unmanned aerial vehicle is difficult to accurately count the number of vehicles in a complex traffic scene, so that dynamically collected traffic data is difficult to convert into an accurate signal timing strategy is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of road management, and particularly to a road intelligent management method and system based on an unmanned aerial vehicle. Background Art

[0002] With the acceleration of the urbanization process and the rapid growth of the motor vehicle ownership, traffic congestion has become a core problem in modern urban management. Traditional traffic signal control systems are mostly based on fixed-time timing or geomagnetic induction technology, adjusting the traffic light duration through preset time periods or local vehicle detection data. However, fixed timing schemes cannot adapt to the dynamic changes in traffic flow. Especially during peak hours or sudden congestion scenarios, it is easy to cause an imbalance in the waiting time of vehicles between phases. The geomagnetic induction technology is limited by the narrow detection range and can only obtain local information of a single lane, making it difficult to comprehensively reflect the actual traffic conditions of each phase at the intersection. In addition, the existing video surveillance systems based on fixed cameras have fixed perspectives and limited coverage ranges, and cannot flexibly capture multi-dimensional traffic data at complex intersections, resulting in a lack of global and real-time signal optimization.

[0003] In recent years, unmanned aerial vehicles have gradually been introduced into the field of traffic management due to their high mobility and wide-area coverage capabilities. However, existing unmanned aerial vehicle technologies are difficult to accurately count the number of vehicles in complex traffic scenarios. The problems of vehicle overlap and occlusion in complex traffic scenarios can lead to misdetection or missed detection of the number of vehicles by traditional image processing algorithms, and there is a lack of in-depth analysis of the causes of real-time congestion, making it difficult to achieve dynamic balance of the passing capacity between phases, and thus difficult to convert dynamically collected traffic data into accurate signal timing strategies. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a road intelligent management method based on an unmanned aerial vehicle, which solves the technical problem that existing unmanned aerial vehicles are difficult to accurately count the number of vehicles in complex traffic scenarios, and thus difficult to convert dynamically collected traffic data into accurate signal timing strategies.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: A road intelligent management method based on an unmanned aerial vehicle, including an unmanned aerial vehicle for collecting video data, and the management method specifically includes the following steps:

[0006] S1. Real-time obtain the intersection videos of each intersection on the road according to the inspection route;

[0007] S2. Obtain the vehicle data of the corresponding intersection according to the intersection video, and calculate the number of congested vehicles that have not passed the stop line in each phase during each red light period;

[0008] S3. Determine whether the traffic light data settings at the intersection are normal according to whether the maximum value of the difference in the number of congested vehicles between each phase is less than the number threshold;

[0009] If so, end;

[0010] If not, proceed to step S4;

[0011] S4. Calculate the green light duration for each phase based on the vehicle data and send it to the background management center.

[0012] Preferably, in step S2, it specifically includes the following steps:

[0013] S21. Mark each frame of the image in the intersection video as an intersection image and convert it into a grayscale image;

[0014] S22. Preprocess the grayscale image through smoothing processing and noise suppression to obtain a preprocessed image;

[0015] S23. Convert the preprocessed image into a binary image to extract vehicle features, and obtain the correspondence between the traffic light data and each phase at the intersection, the lanes included in each phase, and the number of vehicles on each lane according to the preprocessed image;

[0016] S24. Differentiate the preprocessed image of any frame from the preprocessed image of the previous frame to obtain a differential image, and correlate and track the vehicle features of any adjacent frames according to the differential image; the calculation formula of the differential image is:

[0017]

[0018] In the above formula, represents the differential image of the nth frame, and represent the preprocessed images of the nth frame and the (n - 1)th frame respectively;

[0019] S25. When the red light countdown is 1, obtain the number of stationary vehicles on all lanes in both directions corresponding to the phase respectively, and mark the larger value as the number of waiting vehicles in this phase;

[0020] S26. During the green light period, obtain the number of vehicles passing through the stop line on all lanes in both directions corresponding to the phase respectively, and mark the larger value as the number of passing vehicles passing through this phase;

[0021] S27. Calculate the difference in the number of waiting vehicles and passing vehicles in each phase within each cycle to obtain the number of congested vehicles; the calculation formula of the number of congested vehicles is:

[0022] N Congestion =N Stationary-N Passing

[0023] In the above formula, N Congestion , N Stationary and N Passing respectively represent the number of congested vehicles, waiting vehicles and passing vehicles in any phase within any cycle.

[0024] Preferably, in step S23, it specifically includes the following steps:

[0025] S231. Convert the preprocessed image into a binary image to extract vehicle features;

[0026] S232. Construct the minimum convex hull of each vehicle feature;

[0027] S233. Obtain the vehicle area of the vehicle feature and the convex hull area of the minimum convex hull;

[0028] S234. Calculate the area ratio of the vehicle area to the convex hull area; the calculation formula for the area ratio is

[0029]

[0030] In the above formula, S' represents the area ratio, S vehicle represents the vehicle area of the vehicle feature, and S ConvexHull represents the convex hull area of the vehicle feature;

[0031] S235. Determine whether the vehicle feature is a vehicle overlap feature according to whether the area ratio is less than the area ratio threshold;

[0032] If so, mark the vehicle feature as a vehicle overlap feature and enter step S236;

[0033] If not, enter step S24;

[0034] S236. Subtract the minimum convex hull from the corresponding vehicle feature to obtain a suspicious area, connect the two suspicious areas with the largest area and divide the vehicle overlap feature to obtain several vehicle features, and then return to step S232.

[0035] Preferably, in step S236, it specifically includes the following steps:

[0036] S2361. Subtract the minimum convex hull from the corresponding vehicle overlap feature to obtain several suspicious areas;

[0037] S2362. Establish a first segmentation plane in the vertical direction based on the lane dividing line;

[0038] S2363. Determine whether the first segmentation plane intersects with the vehicle overlap feature;

[0039] If so, divide the vehicle overlapping feature through the first dividing plane to obtain a number of vehicle features, and then return to step S232;

[0040] If not, proceed to the next step;

[0041] S2364. Obtain the centroid of the vehicle overlapping feature;

[0042] S2365. Obtain the two suspicious regions that are closest to the centroid of the vehicle overlapping feature;

[0043] S2366. Respectively obtain the two suspicious points on the two suspicious regions that are closest to the centroid, and connect the suspicious points to obtain a dividing line;

[0044] S2367. Divide the vehicle overlapping feature into a number of vehicle features through the dividing line, and then return to step S232.

[0045] Preferably, in step S3, it specifically includes the following steps:

[0046] S31. Calculate the absolute value of the difference in the number of congested vehicles between any two phases in any one cycle in sequence to obtain a first quantity difference;

[0047] S32. Calculate the average value of the first quantity differences between any two phases in each cycle in sequence to obtain a first average value;

[0048] S33. Set a quantity threshold, and determine whether the first average value is less than the quantity threshold;

[0049] If so, end;

[0050] If not, proceed to step S4.

[0051] Preferably, in step S4, it specifically includes the following steps:

[0052] S41. Calculate the average value of the sum of the number of congested vehicles and the number of passing vehicles in each phase in a number of cycles to obtain the number of vehicles to be passed;

[0053] S42. Calculate the ratio of the green light time of each phase to the number of passing vehicles to obtain the single vehicle passing time;

[0054] S43. Calculate the product of the single vehicle passing time of each phase and the number of vehicles to be passed to obtain the total passing time;

[0055] S44. Calculate the ratio of the total passing time of each phase to obtain the passing time ratio of each phase;

[0056] S45. Obtain the minimum green light time and the maximum green light time of the current intersection to construct a green light interval;

[0057] S46. Allocate the green light duration for each phase according to the passing time ratio and the green light interval, and send it to the background management center.

[0058] Preferably, in step S21, the calculation formula for the pixel value of each pixel point on the grayscale image is:

[0059] f(i, j) = α·R(i, j) + β·G(i, j) + γ·B(i, j)

[0060] In the above formula, f(i, j) is the grayscale value at the coordinate (i, j) on the grayscale image, α, β, and γ respectively represent the red pixel factor, the green pixel factor, and the blue pixel factor, and R(i, j), G(i, j), and B(i, j) respectively represent the red brightness value, the green brightness value, and the blue brightness value at the coordinate (i, j) on the intersection image.

[0061] Preferably, in step S22, the calculation formula for the pixel value of each pixel point on the preprocessed image is:

[0062]

[0063] In the above formula, f ip (i, j) represents the grayscale value at the coordinate (i, j) on the preprocessed image, σ is the set standard deviation, * is the convolution symbol, and f(i, j) is the grayscale value at the coordinate (i, j) on the grayscale image.

[0064] Preferably, in step S46, the calculation formula for the green light duration of each phase is:

[0065]

[0066] In the above formula, T k represents the green light duration of the k-th phase, T min and T max respectively represent the minimum value and the maximum value of the green light interval, and there are M phases in total.

[0067] A road intelligent management system based on an unmanned aerial vehicle includes a processor and a memory. The memory is used to store a computer program, and when the computer program is executed by the processor, it implements any one of the road intelligent management methods based on an unmanned aerial vehicle.

[0068] By means of the above technical solutions, the present invention provides a road intelligent management method based on an unmanned aerial vehicle, which at least has the following beneficial effects:

[0069] 1. The present invention collects intersection videos in real time through an unmanned aerial vehicle, grayscales, denoises, and binarizes the intersection videos to extract vehicle features, tracks vehicle movements, and counts the number of congested vehicles that have not passed the stop line during the red light and the number of passing vehicles that have passed the stop line during the green light for each phase. Subsequently, it compares the differences in congested vehicles for each phase to determine the rationality of signal light settings. If abnormal, it dynamically calculates the green light duration for each phase and optimizes the timing plan. Through the high-altitude perspective and real-time processing technology of the unmanned aerial vehicle, it dynamically adjusts the duration of each color signal light, improves the intersection traffic efficiency, and realizes intelligent road congestion management.

[0070] 2. The present invention converts the intersection video frames into grayscale images, reduces the data volume and retains the key structural information to improve the calculation efficiency. Then, it preprocesses the grayscale images to enhance the image quality and reduce environmental interference. Subsequently, it extracts vehicle features through binary segmentation to accurately obtain vehicle data.

[0071] 3. The present invention extracts vehicle features through binarization, constructs the minimum convex hull, and calculates the ratio of the vehicle area to the convex hull area to determine whether there are vehicle overlapping features. For vehicle overlapping features, based on vertical segmentation along the lane dividing line or centroid-driven adaptive segmentation, it generates a dividing line by connecting suspicious regions to identify the number of overlapping vehicles. Its function is to accurately identify and segment the merged contours caused by vehicle occlusion or side-by-side arrangement, solve the problems of false detection and missed detection of overlapping vehicles by traditional algorithms, and significantly improve the vehicle counting accuracy in complex traffic scenarios.

[0072] 4. The present invention first subtracts the minimum convex hull from the overlapping features to extract the suspicious regions. Subsequently, it attempts to establish a vertical segmentation plane based on the lane dividing line for segmentation, and by calculating the centroid of the overlapping features, locates the nearest suspicious region and connects its nearest points to generate a dividing line. Finally, it segments the overlapping features into independent vehicle features. Through the segmentation strategy of dynamic geometric analysis and scene adaptation, it accurately separates the merged contours caused by vehicle occlusion or side-by-side arrangement, significantly reduces the false detection and missed detection rates, and improves the accuracy and robustness of vehicle detection in complex traffic scenarios.

[0073] 5. The present invention calculates the number of vehicles waiting to pass and the single-vehicle passing time for each phase, and combines the number of vehicles waiting to pass to estimate the total passing time. Subsequently, it calculates the passing time ratio for each phase, and combines the green light interval to dynamically allocate the green light duration for each phase. Based on the real-time obtained vehicle data, it scientifically allocates the red and green light signal ratios, breaks the rigidity of the fixed timing mode, dynamically balances the passing demands of each phase, thereby reducing vehicle backlogs and improving the overall intersection traffic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] The accompanying drawings described herein are used to provide a further understanding of the present application, and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application, and do not constitute an improper limitation of the present application. In the drawings:

[0075] Figure 1 is a flowchart of the intelligent road management method based on an unmanned aerial vehicle of the present invention;

[0076] Figure 2 is a schematic diagram of the trajectory planning of the unmanned aerial vehicle of the present invention;

[0077] Figure 3 is a first schematic diagram of the intersection phase of the present invention;

[0078] Figure 4 is a second schematic diagram of the intersection phase of the present invention. Detailed implementation manners

[0079] In order to make the above objects, features, and advantages of the present invention more obvious and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific implementation manners. Thereby, a full understanding of how the present application uses technical means to solve technical problems and achieve the realization process of technical effects can be obtained and implemented accordingly.

[0080] The traffic conditions in urban roads have the characteristics of periodic fluctuations, suddenness, and uncertainty. For example, affected by commuting needs, traffic congestion breaks out concentratedly during the morning and evening rush hours on weekdays, showing an obvious tidal phenomenon. During holidays, the congested hours may be postponed or dispersed, related to travel demands such as tourism and shopping. Sudden situations such as traffic accidents, road closures due to construction, and bad weather will instantly exacerbate traffic congestion. It is difficult to predict situations such as traffic congestion caused by non-regular traffic flows (such as large-scale events and temporary controls). Therefore, in order to improve the overall traffic efficiency of the urban road network, data collection of the urban road network is carried out by using unmanned aerial vehicles to optimize the traffic signals at each intersection.

[0081] In order to solve the technical problem that it is difficult for unmanned aerial vehicles in the prior art to accurately count the number of vehicles in complex traffic scenarios, and thus it is difficult to convert the dynamically collected traffic data into accurate signal timing strategies, the present invention provides an intelligent road management method based on unmanned aerial vehicles, as Figure 1 shown, to solve problems such as insufficient real-time performance, incomplete data collection, and inaccurate signal timing in traffic signal control. Through periodic inspections of unmanned aerial vehicles, traffic congestion at each intersection can be detected in a timely manner, and the signal ratio of traffic lights can be adjusted in a timely manner, thereby improving the overall traffic efficiency of the urban road network. By using unmanned aerial vehicles to collect video data of intersections, the management method specifically includes the following steps:

[0082] S1. Obtain the intersection videos of each intersection on the road in real time according to the inspection route, as Figure 2As shown, the polyline in the figure is the inspection path of the unmanned aerial vehicle set through the unmanned aerial vehicle management system, which is used to inspect each road where congestion may occur, and other data collection can also be achieved by planning the inspection path.

[0083] S2. Obtain the vehicle data of the corresponding intersection according to the intersection video, and calculate the number of congested vehicles that have not passed the stop line during each red light period. The vehicle data includes the traffic light data on the lanes corresponding to each phase at the intersection, the number of stationary waiting vehicles, and the number of passing vehicles passing through each phase. As Figure 3 shown, it is a schematic diagram of a common crossroads. The traffic light data is the state of each phase being red, green, or yellow, and the remaining time of each state. This state can be obtained by connecting to the urban traffic management platform, collecting traffic light images by the unmanned aerial vehicle, vehicle-road coordination (V2X) technology, and third-party data service and open platform, etc. As Figure 4 shown, it is a schematic diagram of a phase. The arrow indicates the driving direction of the vehicles on the corresponding lane. The calculation method of obtaining vehicle data through the collected intersection video is further described in detail below, which specifically includes the following steps:

[0084] S21. Mark each frame of the image in the intersection video as an intersection image and convert it into a grayscale image. Since the computing power of the hardware carried by the unmanned aerial vehicle is limited, and the grayscale image only retains the luminance information (1 channel), compared with the RGB image (3 channels), the data volume is reduced to 1 / 3, which can significantly reduce the algorithm calculation burden and is suitable for the unmanned aerial vehicle to quickly complete image analysis, such as vehicle detection and signal light recognition, etc. Moreover, the grayscale image occupies less storage space and is suitable for the unmanned aerial vehicle to transmit the video stream in real time under limited bandwidth, reducing the communication pressure. Also, in the traffic scene, vehicle detection and congestion analysis rely on features such as shape and movement trajectory, rather than color details. Grayscale conversion can highlight the structural information, such as the vehicle edge and the position of the stop line, etc. The calculation formula for the pixel value of each pixel point on the grayscale image is:

[0085] f(i, j) = α·R(i, j) + β·G(i, j) + γ·B(i, j)

[0086] In the above formula, f(i, j) represents the grayscale value at the coordinate (i, j) on the grayscale image. α, β, and γ represent the red pixel factor, green pixel factor, and blue pixel factor respectively. R(i, j), G(i, j), and B(i, j) represent the red luminance value, green luminance value, and blue luminance value respectively at the coordinate (i, j) on the intersection image. When converting the intersection image into a grayscale image, α = 0.299, β = 0.587, and γ = 0.114 can be set. The human eye is most sensitive to green light (550nm), followed by red, and the least sensitive to blue. By assigning the largest weight (0.587) to green (G), the generated grayscale image can be closer to the human visual perception of luminance, enhancing the discernibility of image details. In road scenes, green elements such as vegetation and traffic signs account for a high proportion. Retaining green channel information with a high weight helps to effectively separate vehicles from the environment. Setting the blue channel weight to the lowest (0.114) can weaken the influence of blue noise such as sky reflection and shadows on the grayscale image, reducing false detections and missed detections. Thus, through scientific matching of human eye perception, standardized compatibility, efficient calculation, and anti-interference design, a high-precision and low-latency grayscale image foundation is provided for real-time vehicle flow detection.

[0087] S22. Preprocess the grayscale image through smoothing processing and noise suppression to obtain a preprocessed image. The calculation formula for the pixel value of each pixel point on the preprocessed image is:

[0088]

[0089] In the above formula, f ip (i, j) represents the grayscale value at the coordinate (i, j) on the preprocessed image. σ is the set standard deviation, which determines the smoothing degree of image processing. * is the convolution symbol, and f(i, j) is the grayscale value at the coordinate (i, j) on the grayscale image.

[0090] S23. Convert the preprocessed image into a binary image to extract vehicle features, and obtain the traffic light data of the intersection and the corresponding relationships of each phase, the lanes included in each phase, and the number of vehicles on each lane according to the preprocessed image. During the peak period of vehicle flow, vehicle overlap and occlusion problems will occur in complex traffic scenarios, which may lead to false detections or missed detections of the number of vehicles by traditional image processing algorithms. In view of this, the existing algorithms are optimized to accurately identify the number of vehicles when they overlap, specifically including the following steps:

[0091] S231. Convert the preprocessed image into a binary image to extract vehicle features, and initially separate vehicles from the background through threshold segmentation to provide a clear contour basis for subsequent analysis.

[0092] S232. Construct the minimum convex hull of each vehicle feature. The minimum convex hull can effectively cover the extreme points of the vehicle contour, helping to distinguish independent vehicles from merged contours, such as two vehicles side by side or overlapping front to back, and providing a geometric reference for subsequent area ratio analysis.

[0093] S233. Obtain the vehicle area of the vehicle feature and the convex hull area of the minimum convex hull. The area can be calculated based on the number of pixel points. For example, if there are a certain number of pixel points within the vehicle feature, and each pixel point has a certain area, then the vehicle area of the vehicle feature is the product of the number of pixel points of the vehicle feature and the area represented by a single pixel point.

[0094] S234. Calculate the area ratio of the vehicle area to the convex hull area. The formula for the area ratio is

[0095]

[0096] In the above formula, S' represents the area ratio, S vehicle represents the vehicle area of the vehicle feature, and S ConvexHull represents the convex hull area of the vehicle feature;

[0097] S235. Determine whether the vehicle feature is a vehicle overlap feature according to whether the area ratio is less than the area ratio threshold. Since vehicles are generally rectangular parallelepipeds, generally, a single vehicle feature is generally rectangular, and its minimum convex hull has a very high similarity to the vehicle feature and almost coincides. Generally, the area ratio threshold can be set to 0.8.

[0098] If so, mark the vehicle feature as a vehicle overlap feature and proceed to step S236;

[0099] If not, proceed to step S24;

[0100] S236. Subtract the minimum convex hull from the corresponding vehicle feature to obtain suspicious regions. Connect the two suspicious regions with the largest area and divide the vehicle overlap feature to obtain several vehicle features, and then return to step S232. After identifying the vehicle overlap feature, it is necessary to further identify it to determine the number of overlapping vehicles within the vehicle overlap feature. The specific steps are as follows:

[0101] S2361. Subtract the minimum convex hull from the corresponding vehicle overlap feature to obtain several suspicious regions;

[0102] S2362. Establish a first dividing plane in the vertical direction with the lane dividing line as the reference;

[0103] S2363. Determine whether the first dividing plane intersects with the vehicle overlap feature;

[0104] If so, divide the vehicle overlapping feature through the first dividing plane to obtain several vehicle features, and then return to step S232;

[0105] If not, proceed to the next step;

[0106] S2364. Obtain the centroid of the vehicle overlapping feature. The main methods for obtaining the centroid are as follows: 1. Geometric centroid calculation based on contour pixels: By traversing all foreground pixels, accumulating their horizontal and vertical coordinate values, and dividing the accumulated horizontal and vertical coordinate values by the total number of pixels respectively, the centroid coordinates can be obtained; 2. Centroid calculation based on image moments: By inputting the vehicle overlapping feature, calculate the zero-order moment and the first-order moment in sequence, and obtain the centroid coordinates through the moment ratio;

[0107] S2365. Obtain the two suspicious regions with the closest distance to the centroid of the vehicle overlapping feature;

[0108] S2366. Respectively obtain the two suspicious points on the two suspicious regions that are closest to the centroid, and connect the suspicious points to obtain a dividing line;

[0109] S2367. Divide the vehicle overlapping feature into several vehicle features through the dividing line, generally 2, and then return to step S232 to judge the divided vehicle features again to determine whether they are vehicle overlapping features until all vehicles are recognized.

[0110] S24. Differentiate the preprocessed image of any frame from the preprocessed image of the previous frame to obtain a differential image, and perform associated tracking on the vehicle features of any adjacent frames according to the differential image. Since the moving distance of the vehicle between any two adjacent frames is limited, the tracking of the vehicle can be realized according to the moving distance of the centroid. When the moving distance between two adjacent frames of images meets a certain threshold and is the closest, it can be judged that the vehicles represented by the centroids in the two frames of images are the same vehicle, avoiding some vehicles from being double-counted; The calculation formula of the differential image is:

[0111]

[0112] In the above formula, represents the differential image of the nth frame, and respectively represent the preprocessed images of the nth frame and the (n - 1)th frame;

[0113] S25. When the red light countdown is 1, respectively obtain the number of stationary vehicles on all lanes in the two directions corresponding to the phase, and mark the larger value as the number of waiting vehicles in this phase. At the next moment, the red light turns green and the vehicles start to move. Therefore, at this time, the number of waiting vehicles reaches the maximum value. One phase corresponds to two directions, such as Figure 3As shown, there may be multiple lanes in each direction, and the larger value of the number of stationary vehicles in a certain direction is marked as the number of waiting vehicles in this phase.

[0114] S26. During the green light period, respectively obtain the number of vehicles passing through the stop line on all lanes in two directions corresponding to the phase, and mark the larger value as the number of passing vehicles in this phase, that is, mark the larger value of the number of vehicles passing through the stop line in a certain direction of the phase as the number of passing vehicles. For example, when the number of vehicles passing through the stop line in two directions is 15 and 20 respectively, mark 20 as the number of passing vehicles in this phase.

[0115] S27. Respectively calculate the difference between the number of waiting vehicles and passing vehicles in each phase within each cycle to obtain the number of congested vehicles. Since in the congestion scenarios involved in the present invention, not all waiting vehicles pass through the stop line, the number of waiting vehicles must be greater than the number of passing vehicles. The calculation formula for the number of congested vehicles is:

[0116] N Congestion =N Stationary -N Passing

[0117] In the above formula, N Congestion 、N Stationary and N Passing respectively represent the number of congested vehicles, waiting vehicles, and passing vehicles in any phase within any cycle.

[0118] S3. Determine whether the traffic light data settings at this intersection are normal according to whether the maximum value of the difference between the number of congested vehicles between each phase is less than the quantity threshold;

[0119] If so, end. Since the traffic light signal time at each intersection has been scientifically calculated to meet most situations, if the difference is small, in order to reduce the adjustment of the traffic lights at the intersection, directly end, determine that the traffic light data settings at this intersection are normal, and thus no longer intervene, and proceed to the inspection of the next intersection;

[0120] If not, enter step S4;

[0121] To avoid the contingency of the calculated difference, it is best to collect vehicle data within multiple traffic light change cycles to ensure the accuracy of the difference between the number of congested vehicles between each phase. The specific steps are as follows:

[0122] S31. Calculate the absolute value of the difference between the number of congested vehicles between any two phases within any cycle in sequence to obtain the first quantity difference;

[0123] S32. Calculate the average value of the first quantity differences between any two phases in each cycle in sequence to obtain the first average value;

[0124] S33. Set a quantity threshold and determine whether the first average value is less than the quantity threshold;

[0125] If so, end;

[0126] If not, enter step S4 to adjust the duration of the traffic light signals.

[0127] S4. Calculate the green light duration of each phase based on the vehicle data and send it to the background management center. Since the number of vehicles passing through each phase is affected by whether there are pedestrians in each phase and the interference of the lane itself. For example, the passing time of a single vehicle in the straight lane may be less than that in the U-turn and left-turn lanes. Therefore, it is necessary to comprehensively consider and calculate the appropriate green light duration for each phase, which specifically includes the following steps:

[0128] S41. Calculate the average value of the sum of the number of congested vehicles and passing vehicles in each phase in several cycles, that is, first calculate the sum of the number of congested vehicles and passing vehicles in each cycle, then add up the sums of each cycle, and then divide by the number of cycles to obtain the number of vehicles waiting to pass. The waiting vehicles may just be temporarily parked in the lane corresponding to this phase and may change lanes when the green light is on. Through a large number of experimental comparisons, relatively speaking, the data accuracy is higher when calculating the number of vehicles waiting to pass through the congested vehicles and passing vehicles;

[0129] S42. Calculate the ratio of the green light time of each phase to the number of passing vehicles, that is, divide the green light time of this phase by the number of passing vehicles to obtain the passing time per vehicle;

[0130] S43. Calculate the product of the passing time per vehicle of each phase and the number of vehicles waiting to pass to obtain the total passing time required for all the vehicles waiting to pass through the intersection in each phase;

[0131] S44. Calculate the ratio of the total passing time of each phase to obtain the passing time ratio of each phase; for example, the expression of the passing time ratio of the first phase, the second phase, the third phase, and the fourth phase can be: τ1:τ2:τ3:τ4;

[0132] S45. Obtain the minimum green light time and the maximum green light time of the current intersection to construct a green light interval. The expression of the green light interval can be [T min , T max , where T min and T max respectively represent the minimum green light time and the maximum green light time of the current intersection. Generally, to ensure the passage of pedestrians, T minGenerally, it is not less than 15 s. To prevent the driver from getting impatient due to long waiting, T max is generally not greater than 90 s, which is specifically determined according to the actual road conditions.

[0133] S46. Allocate the green light duration of each phase according to the passing time ratio and the green light interval, and send it to the background management center for background review and recording, so as to realize the intelligent management of the traffic light information at the road intersection. The calculation formula for the green light duration of each phase is:

[0134]

[0135] In the above formula, T k represents the green light duration of the kth phase, T min and T max respectively represent the minimum value and the maximum value of the green light interval, and there are M phases in total.

[0136] The present invention collects the intersection video in real time through an unmanned aerial vehicle, performs grayscale processing, noise reduction and binarization on the intersection video to extract vehicle features, tracks the vehicle movement and counts the number of congested vehicles that have not passed the stop line during the red light. Subsequently, compare the differences in congested vehicles in each phase to judge the rationality of the signal lamp setting. If it is abnormal, dynamically calculate the green light duration of each phase and optimize the timing plan. Through the high-altitude perspective and real-time processing technology of the unmanned aerial vehicle, dynamically adjust the signal lamp to improve the intersection traffic efficiency and realize the intelligent management of road congestion.

[0137] The present invention also provides a road intelligent management system based on an unmanned aerial vehicle, including a processor and a memory. The memory is used to store a computer program, and when the computer program is executed by the processor, it realizes the road intelligent management method based on the unmanned aerial vehicle.

[0138] Those of ordinary skill in the art can understand that all or part of the steps in implementing the methods of the above embodiments can be completed by instructing relevant hardware through a program. Therefore, this application can adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0139] Each embodiment in this specification is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts between each embodiment can be referred to each other. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments.

[0140] The above embodiments have introduced the present invention in detail. Specific examples are used in this article to elaborate on the principle and implementation manner of the present invention. The description of the above embodiments is only used to help understand the method and its core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation to the present invention.

Claims

1. An intelligent road management method based on an unmanned aerial vehicle, including an unmanned aerial vehicle for collecting video data, characterized in that, The management method specifically includes the following steps: S1. Obtain the intersection videos of each intersection on the road in real time according to the inspection route; S2. Obtain the vehicle data of the corresponding intersection according to the intersection video, and calculate the number of congested vehicles that have not passed the stop line during each red light period for each phase; S3. Determine whether the traffic light data setting of this intersection is normal according to whether the maximum value of the difference in the number of congested vehicles between each phase is less than the number threshold; If so, end; If not, enter step S4; S4. Calculate the green light duration of each phase according to the vehicle data and send it to the background management center.

2. The management method according to claim 1, wherein In step S2, it specifically includes the following steps: S21. Mark each frame image in the intersection video as an intersection image and convert it into a grayscale image; S22. Perform preprocessing on the grayscale image through smoothing processing and noise suppression to obtain a preprocessed image; S23. Convert the preprocessed image into a binary image to extract vehicle features, and obtain the correspondence between the traffic lights and each phase of the intersection, the lanes included in each phase, and the number of vehicles on each lane according to the preprocessed image; S24. Differentiate the preprocessed image of any frame from the preprocessed image of the previous frame to obtain a differential image, and correlate and track the vehicle features of any adjacent frames according to the differential image; the calculation formula of the differential image is: In the above formula, represents the differential image of the nth frame, and respectively represent the preprocessed images of the nth frame and the (n - 1)th frame; S25. When the red light countdown is 1, obtain the number of stationary vehicles on all lanes in both directions corresponding to this phase respectively, and mark the larger value as the number of waiting vehicles for this phase; S26. During the green light period, obtain the number of vehicles passing the stop line on all lanes in both directions corresponding to this phase respectively, and mark the larger value as the number of passing vehicles for this phase; S27. Calculate the difference between the number of waiting vehicles and passing vehicles for each phase in each cycle respectively to obtain the number of congested vehicles; the calculation formula of the number of congested vehicles is: N Congestion = N Stationary -N Passing In the above formula, N Congestion , N Stationary and N Passing respectively represent the number of congested vehicles, waiting vehicles, and passing vehicles in any phase within any period.

3. The management method according to claim 2, wherein In step S23, it specifically includes the following steps: S231. Convert the preprocessed image into a binary image to extract vehicle features; S232. Construct the minimum convex hull of each vehicle feature; S233. Obtain the vehicle area of the vehicle feature and the convex hull area of the minimum convex hull; S234. Calculate the area ratio of the vehicle area to the convex hull area; the calculation formula of the area ratio is In the above formula, S' represents the area ratio, and S vehicle represents the vehicle area of the vehicle feature, and S ConvexHull represents the convex hull area of the vehicle feature; S235. Determine whether the vehicle feature is a vehicle overlap feature according to whether the area ratio is less than the area ratio threshold; If so, mark this vehicle feature as a vehicle overlap feature and enter step S236; If not, enter step S24; S236. Subtract the minimum convex hull from the corresponding vehicle feature to obtain a suspicious area, connect the two suspicious areas with the largest area and divide the vehicle overlap feature to obtain several vehicle features, and then return to step S232.

4. The management method according to claim 3, wherein In step S236, it specifically includes the following steps: S2361. Subtract the minimum convex hull from the corresponding vehicle overlap feature to obtain several suspicious areas; S2362. Establish a first dividing plane in the vertical direction with the lane dividing line as the reference; S2363. Determine whether the first dividing plane has an intersection with the vehicle overlap feature; If so, segment the vehicle overlapping feature through the first segmentation plane to obtain a number of vehicle features, and then return to step S232; If not, proceed to the next step; S2364. Obtain the centroid of the vehicle overlapping feature; S2365. Obtain the two suspicious regions that are closest to the centroid of the vehicle overlapping feature; S2366. Respectively obtain the two suspicious points on the two suspicious regions that are closest to the centroid, and connect the suspicious points to obtain a segmentation line; S2367. Segment the vehicle overlapping feature into a number of vehicle features through the segmentation line, and then return to step S232.

5. The management method according to claim 1, characterized in that, In step S3, it specifically includes the following steps: S31. Calculate the absolute value of the difference in the number of congested vehicles between any two phases in any one cycle in sequence to obtain a first quantity difference; S32. Calculate the average value of the first quantity differences between any two phases in each cycle in sequence to obtain a first average value; S33. Set a quantity threshold, and determine whether the first average value is less than the quantity threshold; If so, end; If not, proceed to step S4.

6. The management method according to claim 1, characterized in that In step S4, it specifically includes the following steps: S41. Calculate the average value of the sum of the number of congested vehicles and passing vehicles in each phase in a number of cycles to obtain the number of vehicles to be passed; S42. Calculate the ratio of the green light time of each phase to the number of passing vehicles to obtain the single-vehicle passing time; S43. Calculate the product of the single-vehicle passing time of each phase and the number of vehicles to be passed to obtain the total passing time; S44. Calculate the ratio of the total passing time of each phase to obtain the passing time ratio of each phase; S45. Obtain the minimum green light time and the maximum green light time of the current intersection to construct a green light interval; S46. Allocate the green light duration of each phase according to the passing time ratio and the green light interval, and send it to the background management center.

7. The management method according to claim 2, characterized in that, In step S21, the calculation formula for the pixel value of each pixel point on the grayscale image is: f(i, j) = α·R(i, j) + β·G(i, j) + γ·B(i, j) In the above formula, f(i, j) is the grayscale value at the coordinate (i, j) on the grayscale image, α, β, and γ respectively represent the red pixel factor, the green pixel factor, and the blue pixel factor, and R(i, j), G(i, j), and B(i, j) respectively represent the red brightness value, the green brightness value, and the blue brightness value at the coordinate (i, j) on the intersection image.

8. The management method according to claim 2, wherein In step S22, the calculation formula for the pixel value of each pixel point on the preprocessed image is: In the above formula, f ip (i, j) represents the gray value at the coordinate (i, j) on the preprocessed image, σ is the set standard deviation, * is the convolution symbol, and f(i, j) is the gray value at the coordinate (i, j) on the gray image.

9. The management method according to claim 6, characterized in that In step S46, the calculation formula for the green light duration of each phase is: In the above formula, T k represents the green light duration of the k-th phase, T min and T max represent the minimum value and the maximum value of the green light interval respectively, and there are M phases in total.

10. A system for implementing the unmanned aerial vehicle-based intelligent road management method according to any one of claims 1-9, characterized in that, It includes a processor and a memory, and the memory is used to store a computer program. When the computer program is executed by the processor, it implements the method for intelligent road management based on an unmanned aerial vehicle according to any one of claims 1-9.

Images