Road disaster early warning system and method using satellite remote sensing and complementary unmanned aerial vehicle data

Abstract

The present application relates to a road disaster early warning system and method using satellite remote sensing and complementary unmanned aerial vehicle data, and relates to the field of road disaster early warning, comprising: using satellite remote sensing, collecting remote sensing images of a target road area, and identifying a snow-covered area; according to the remote sensing images and the snow-covered area, analyzing the average brightness of the remote sensing images, and configuring a collection range for unmanned aerial vehicle image collection of the target road area; optimizing the collection position for unmanned aerial vehicle image collection to obtain an optimal collection position distribution, wherein in the optimization, the repeatability of the collection position is analyzed according to the collection range, and the collection positions are merged; according to the optimal collection position distribution and the collection range, unmanned aerial vehicle image collection is performed, and road disaster early warning is performed in combination with the remote sensing images. The present application solves the technical problems of insufficient comprehensive ground data collection and insufficient reliability in the prior art road disaster prediction, which leads to insufficient accuracy of road disaster prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of road disaster early warning, and in particular to a road disaster early warning system and method using satellite remote sensing and unmanned aerial vehicle data complementation. BACKGROUND

[0002] Currently, traditional road disaster early warning systems, such as snow disaster early warning systems, mainly rely on meteorological data and ground sensors, and have problems such as slow response speed and limited coverage. Satellite remote sensing technology can provide road information over a wide range, but due to the low resolution, it is difficult to accurately identify local disaster conditions. Although unmanned aerial vehicle technology has high-resolution image acquisition capabilities, it is limited by flight time and area and cannot cover a wide range of areas. Moreover, the current unmanned aerial vehicle ground data collection position is set by humans, and the collection position may be redundant or insufficient.

[0003] The prior art lacks an effective ground data collection method, resulting in an incomplete data basis for disaster risk prediction and insufficient reliability, which leads to insufficient accuracy of road disaster prediction and difficulty in dealing with road disaster prediction in complex environments. SUMMARY

[0004] The present application provides a road disaster early warning system and method using satellite remote sensing and unmanned aerial vehicle data complementation to solve the technical problems of incomplete ground data collection, insufficient reliability, and insufficient accuracy of road disaster prediction in the prior art.

[0005] The technical solution of the present application to solve the above technical problems is as follows:

[0006] In a first aspect, the present application provides a road disaster early warning system using satellite remote sensing and unmanned aerial vehicle data complementation, comprising:

[0007] A remote sensing identification module for collecting remote sensing images of a target road area using satellite remote sensing, and preliminarily identifying a snow-covered area;

[0008] A collection range configuration module for analyzing the average distinctness of the remote sensing images based on the remote sensing images and the snow-covered area, and configuring a collection range for unmanned aerial vehicle image collection of the target road area based on the average distinctness;

[0009] A collection position optimization module for optimizing the collection position of unmanned aerial vehicle image collection within the target road area to obtain an optimal collection position distribution, wherein the optimization analyzes the redundancy of the collection position based on the collection range and merges the collection positions;

[0010] A road risk early warning module is configured to perform unmanned aerial vehicle image collection according to the optimal collection position distribution and the collection range, perform road disaster risk prediction in combination with the remote sensing image, obtain a road disaster risk level, and perform early warning.

[0011] In a second aspect, the present application provides a road disaster early warning method using satellite remote sensing and complementary unmanned aerial vehicle data, comprising:

[0012] Remote sensing images of a target road region are collected using satellite remote sensing, and a snow-covered region is preliminarily identified and obtained;

[0013] The average distinctness of the remote sensing image is analyzed and obtained according to the remote sensing image and the snow-covered region, and a collection range for unmanned aerial vehicle image collection of the target road region is configured according to the average distinctness;

[0014] Optimization of the collection position of unmanned aerial vehicle image collection in the target road region is performed to obtain an optimal collection position distribution, wherein the degree of repetition of the collection position is analyzed according to the collection range during optimization, and the collection position is merged;

[0015] Unmanned aerial vehicle image collection is performed according to the optimal collection position distribution and the collection range, road disaster risk prediction is performed in combination with the remote sensing image, a road disaster risk level is obtained, and early warning is performed.

[0016] The present application has the following beneficial effects: by combining satellite remote sensing and complementary unmanned aerial vehicle data, the accuracy and efficiency of the road disaster early warning system are significantly improved. First, satellite remote sensing is used to widely collect images of the target road region, which can quickly identify the snow-covered region and provide basic data for subsequent disaster risk prediction. Unmanned aerial vehicle image collection is introduced, the collection range of the unmanned aerial vehicle is determined by analyzing the satellite remote sensing image and the snow-covered region, the accuracy of the unmanned aerial vehicle image collection range is improved, the efficiency and quality of data collection and disaster prediction are ensured, and the collection position distribution is optimized to avoid the problems of redundant or insufficient collection of ground image data, improve the reliability and comprehensiveness of ground image data collection, and further improve the reliability and timeliness of disaster prediction. Through the complementarity of high-definition images collected by unmanned aerial vehicles and satellite images, and the optimization of the collection range and collection position, it can help accurately identify disaster risks and ensure the timeliness and reliability of road snow disaster prediction. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A structure diagram of a road disaster early warning system using satellite remote sensing and complementary unmanned aerial vehicle data is provided for the present application;

[0018] Figure 2 A flowchart of a road disaster early warning method using satellite remote sensing and complementary unmanned aerial vehicle data is provided for the present application.

[0019] The components represented by the reference signs in the drawings are described as follows:

[0020] The remote sensing identification module 11, the acquisition range configuration module 12, the acquisition position optimization module 13, and the road risk early warning module 14. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by a person skilled in the art without creative work fall within the scope of protection of the present application.

[0022] In the description of the present application, the terms "first", "second" are used only for the purpose of description, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first", "second" can explicitly or implicitly include one or more of the features. In the description of the present application, the meaning of "multiple" is two or more, unless otherwise specifically limited.

[0023] In the description of the present application, the term "for example" is used to indicate "as an example, illustration or explanation". Any embodiment described as "for example" in the present application is not necessarily interpreted as more preferred or more advantageous than other embodiments. The following description is given in order to enable any person skilled in the art to implement and use the present application. In the following description, details are listed for the purpose of explanation. It should be understood that a person skilled in the art can realize the present application without using these specific details. In other examples, well-known structures and processes will not be described in detail in order to avoid unnecessary details making the description of the present application obscure. Therefore, the present application is not intended to be limited to the shown embodiments, but is consistent with the broadest scope of principles and features disclosed.

[0024] In one embodiment, as shown in the drawings, the present application provides a road disaster early warning system using satellite remote sensing and complementary unmanned aerial vehicle data, which specifically comprises: a remote sensing identification module 11, an acquisition range configuration module 12, an acquisition position optimization module 13, and a road risk early warning module 14. Figure 1 In one embodiment, as shown in the drawings, the present application provides a road disaster early warning system using satellite remote sensing and complementary unmanned aerial vehicle data, which specifically comprises: a remote sensing identification module 11, an acquisition range configuration module 12, an acquisition position optimization module 13, and a road risk early warning module 14.

[0025] The remote sensing identification module 11 is configured to perform step S10: using satellite remote sensing, acquiring remote sensing images of a target road area, and preliminarily identifying a snow-covered area.

[0026] In the embodiments of the present application, first, satellite remote sensing technology is used to collect remote sensing images of the target road area. The target road area is the area where the road to be snow disaster early warning is located, for example, the area within 5KM of the road and both sides of the road. Satellite remote sensing is a method of obtaining ground information through satellite sensors, which can provide wide-area ground data, and is especially suitable for large-scale monitoring and disaster assessment. In this process, the satellite sensor captures ground information through different spectral data of different wave bands to generate multispectral or panchromatic images. For example, the Landsat series of satellites are used to obtain the ground images of the target road area.

[0027] Further, by analyzing the spectral characteristics in the remote sensing image, we can preliminarily identify the snow-covered area. The reflectivity of the snow-covered area in the visible and near-infrared wave bands is higher, while other ground objects such as soil or roads reflect less, which can be identified by a threshold segmentation method, that is, according to the near-infrared wave band reflectivity in the remote sensing image, a specific near-infrared wave band reflectivity value is set, and the image is classified using this value, thereby preliminarily marking the possible snow-covered area. Alternatively, through machine learning, the snow-covered area in the remote sensing image is identified.

[0028] In the embodiments of the present application, step S10 performed by the remote sensing identification module 11 further includes:

[0029] Collecting remote sensing images of the target road area using satellite remote sensing;

[0030] Inputting the remote sensing images into a pre-constructed snow-covered area identification channel to identify and obtain the snow-covered area, wherein the snow-covered area identification channel is pre-trained and constructed based on a convolutional neural network, and the training data includes sample remote sensing images and sample snow-covered areas.

[0031] In the embodiments of the present application, first, satellite remote sensing technology is used to collect remote sensing images of the target road area. For example, Landsat or Sentinel-2 satellites are used to collect remote sensing images of the target road area.

[0032] Further, the collected remote sensing images are input into the snow-covered area identification channel for automatic identification of the snow-covered area. The snow-covered area identification channel is pre-trained and constructed based on a convolutional neural network, and the training data includes sample remote sensing images and sample snow-covered areas.

[0033] Exemplarily, the data record of the snow-covered area in the remote sensing data in the historical time can be used to collect a sample remote sensing image set, and the snow-covered area in each sample remote sensing image is positionally identified to obtain a sample snow-covered area set. Then, a network structure of the snow-covered area identification channel is built based on a convolutional neural network (CNN), which includes convolutional layers, pooling layers and the like, and can automatically extract useful features from images for image classification and identification. Specifically, the sample remote sensing images and the sample snow-covered areas are used as training data to supervise the training of the snow-covered area identification channel. The weights and other parameters of the network are adjusted during the training to improve the identification accuracy of the output snow-covered area. The training is continuously performed on multiple sets of training data until the accuracy meets the requirements, for example, the accuracy rate reaches 95%, and then the training is completed.

[0034] The current collected remote sensing image is input into the trained snow-covered area identification channel to identify the snow-covered area.

[0035] The above steps effectively improve the automated identification capability of the snow-covered area, greatly reducing manual intervention and improving the timeliness and accuracy of the monitoring system, providing a reliable data foundation for subsequent unmanned aerial vehicle data collection and disaster risk assessment.

[0036] The collection range configuration module 12 is configured to perform step S20: analyzing the average distinctness of the remote sensing image based on the remote sensing image and the snow-covered area, and configuring a collection range for unmanned aerial vehicle image collection of the target road area based on the average distinctness.

[0037] In the embodiments of the present application, the average distinctness of the remote sensing image is analyzed based on the remote sensing image and the snow-covered area. The average distinctness reflects the difference between all pixel points in the remote sensing image and the pixel points in the snow-covered area. The greater the difference, the greater the average distinctness, indicating that the pixels in the remote sensing image and the pixels in the snow-covered area are quite different, and the snowfall area is less, so the probability of snow disaster is smaller. Therefore, a larger collection range is configured for unmanned aerial vehicle ground image collection, for example, the flight height of the unmanned aerial vehicle is increased, the timeliness of snow disaster warning is improved, and at the same time, some accuracy is lost. However, since the probability of snow disaster is small, the warning demand can still be guaranteed.

[0038] On the contrary, the smaller the difference, the smaller the average distinctness, indicating that the pixels in the remote sensing image and the pixels in the snow-covered area are less different, and the snowfall area is more, so the probability of snow disaster is greater. Therefore, a smaller collection range is configured for unmanned aerial vehicle ground image collection, for example, the flight height of the unmanned aerial vehicle is reduced, and small-range image collection can improve the image resolution and quality, and improve the accuracy of snow disaster warning.

[0039] The step S20 performed by the collection range configuration module 12 in the embodiments of the present application includes:

[0040] graying the remote sensing image, and calculating a mean value of the gray scale of the snow-covered area remote sensing image to obtain a snow-covered area gray scale;

[0041] randomly extracting gray scale values of a plurality of pixel points in the gray remote sensing image, and calculating a mean value to obtain a random gray scale;

[0042] calculating a gray scale deviation ratio of the random gray scale and the snow-covered area gray scale as an average obviousness.

[0043] In the embodiments of the present application, the remote sensing image is grayed to convert the original multi-spectral remote sensing image into a single-channel gray image. Graying is to convert color information in the image into different gray levels of intensity, thereby simplifying the image analysis process. The gray scale value of each pixel point represents the brightness intensity of the position, and the gray scale value is usually between 0 and 255, wherein 0 represents black and 255 represents white, and the intermediate value represents different gray. The graying remote sensing image also includes the graying remote sensing image in the snow-covered area.

[0044] The mean value of the gray scale of the snow-covered area is calculated, which reflects the overall average gray scale of the snow-covered area in the remote sensing image, and the snow-covered area gray scale is obtained. The reflectivity of the snow-covered area is high, and it usually appears as a brighter area in the gray image, and the snow-covered area gray scale is large.

[0045] To further analyze the obviousness of the snow-covered area in the image, the gray scale values of a plurality of pixel points in the gray remote sensing image are randomly extracted next. For example, the gray scale values of 100 pixel points in the gray remote sensing image are randomly extracted, and the mean value is calculated to obtain a random gray scale. Random extraction ensures randomness and globality, and the random gray scale reflects the overall average gray scale of the gray remote sensing image. If there are more snow-covered areas in the remote sensing image, i.e., more snowfall, the random gray scale is larger, and vice versa, if there are fewer snow-covered areas in the remote sensing image, i.e., less snowfall, the random gray scale is smaller.

[0046] The gray scale deviation ratio of the random gray scale and the snow-covered area gray scale is calculated as an average obviousness. For example: the gray scale deviation ratio = |snow-covered area gray scale-random gray scale| / snow-covered area gray scale, that is, the ratio of the absolute value of the difference between the random gray scale and the snow-covered area gray scale to the snow-covered area gray scale. The larger the average obviousness, the less the snowfall in the remote sensing image, and the smaller the average obviousness, the more the snowfall in the remote sensing image. For example, the random gray scale is 100, the snow-covered area gray scale is 200, and the average obviousness is 50%, i.e., 0.5.

[0047] Through the above, the average obviousness of the snow-covered area in the remote sensing image can be quantitatively analyzed, the amount of snowfall in the target road area in the remote sensing image is reflected, and important basis is provided for subsequent unmanned aerial vehicle image collection range configuration.

[0048] The step S20 performed by the collection range configuration module 12 in the embodiment of the present application further includes:

[0049] acquiring a maximum collection range of image collection by the unmanned aerial vehicle;

[0050] multiplying the average distinctness by the maximum collection range of image collection by the unmanned aerial vehicle to obtain the collection range.

[0051] In the embodiment of the present application, firstly, the maximum collection range of image collection by the unmanned aerial vehicle needs to be determined. The maximum collection range of image collection by the unmanned aerial vehicle refers to the maximum ground area covered by the unmanned aerial vehicle in one flight. The maximum collection range is determined by factors such as the flight height of the unmanned aerial vehicle, the field of view (FOV) of the camera, the resolution of the camera, and the performance of the image collection device. For example, for a certain type of unmanned aerial vehicle, the maximum range of single collection of snow ground images meeting the accuracy requirement is 10000m 2 .

[0052] Further, the average distinctness is multiplied by the maximum collection range of image collection by the unmanned aerial vehicle to obtain the collection range, for example, 0.5*10000m 2 = 5000m 2 . The flight height and the field of view of the unmanned aerial vehicle can be adjusted to make the collection range 5000m 2 for image collection.

[0053] The average distinctness reflects how much the snowfall is in the remote sensing image. The larger the average distinctness, the less the snowfall, and the smaller the probability of snow disaster. Therefore, a larger collection range is configured for unmanned aerial vehicle ground image collection to improve the timeliness of snow disaster warning, while losing some accuracy. However, since the probability of snow disaster is small, the warning demand can still be guaranteed. The smaller the average distinctness, the more the snowfall, and the larger the probability of snow disaster. Therefore, a smaller collection range is configured for unmanned aerial vehicle ground image collection, for example, the flight height of the unmanned aerial vehicle is reduced, and small-range image collection can improve the image resolution and quality and improve the accuracy of snow disaster warning.

[0054] The collection position optimization module 13 is configured to perform step S30: optimizing the collection position of image collection by the unmanned aerial vehicle in the target road area to obtain an optimal collection position distribution, wherein in the optimization, the redundancy of the collection position is analyzed according to the collection range, and the collection position is merged.

[0055] In the embodiment of the present application, when the unmanned aerial vehicle performs image collection, the optimization of the collection position is a key step to ensure the efficiency and quality of data collection. In order to avoid redundancy or incompleteness of the collection position, the redundancy of the collection position is analyzed according to the collection range in the optimization, and the collection position is merged and optimized to reduce the overlap of the collection position, thereby improving the quality and value of the image collected by the unmanned aerial vehicle.

[0056] The step S30 performed by the collection position optimization module 13 in the embodiment of the application comprises:

[0057] dividing the area other than the snow-covered area in the remote sensing image to obtain a non-snow-covered area;

[0058] generating a plurality of first collection positions in the target road area randomly, and analyzing a first uniformity parameter of the plurality of first collection positions;

[0059] when the first uniformity parameter is greater than or equal to a uniformity threshold, dividing to obtain a plurality of first collection ranges according to the plurality of first collection positions and a collection range;

[0060] calculating a first collection fitness of the plurality of first collection ranges according to a repetition ratio of the plurality of first collection ranges and a coverage degree of a snow-covered boundary area;

[0061] calculating a repetition ratio of each first collection range and a neighboring first collection range, and merging the plurality of first collection ranges to obtain a plurality of merged second collection ranges;

[0062] generating a plurality of second collection positions in the target road area complementarily, combining collection positions of the plurality of merged second collection ranges, and calculating a second uniformity parameter, when the second uniformity parameter is greater than or equal to the uniformity threshold, continuing the optimization;

[0063] until the optimization converges, outputting a plurality of collection positions with the maximum collection fitness to obtain an optimal collection position distribution.

[0064] In the embodiment of the application, first, after the snow-covered area is identified from the remote sensing image, the remaining area of the remote sensing image is divided into a non-snow-covered area to obtain the non-snow-covered area.

[0065] In the target road area, a plurality of first collection positions are generated randomly, and the plurality of first collection positions are generated completely randomly, which can be any position in the target road area.

[0066] In order to ensure the uniformity of the distribution of the plurality of first collection positions in the target road area and avoid the plurality of first collection positions being concentrated in a local area, which affects the globality and accuracy of the analysis of the subsequent collection positions, the uniformity of the distribution of the plurality of first collection positions is analyzed to obtain a first uniformity parameter.

[0067] The step “in the target road area, a plurality of first collection positions are generated randomly, and a first uniformity parameter of the plurality of first collection positions is analyzed” performed by the collection position optimization module 13 in the embodiment of the application comprises:

[0068] calculate distances between each first collection position and adjacent first collection positions to obtain a plurality of first distances;

[0069] calculate a variance of the plurality of first distances and calculate an inverse of the variance to obtain a first uniformity parameter.

[0070] In the embodiments of the present application, each first collection position refers to a point determined in advance for collecting images in the target road area, and specifically includes a coordinate. For each first collection position, the spatial distance between it and other first collection positions is calculated, for example, the distance is calculated according to the coordinates, and then the first collection position with the minimum distance is determined as the adjacent first collection position. The distance between each first collection position and the adjacent first collection position is recorded to obtain a plurality of first distances.

[0071] Further, the variance of the plurality of first distances is calculated, which aims to quantify the uniformity of the distribution of the plurality of first collection positions. The greater the variance, the greater the deviation of the plurality of first distances, the more uneven the distribution of the plurality of first collection positions, and the worse the uniformity. Further, the inverse of the variance is calculated, and the inverse is taken as the first uniformity parameter. The smaller the variance, the greater the first uniformity parameter, and the more uniform the distribution of the plurality of first collection positions.

[0072] For example, the plurality of first distances are 50m, 60m, 40m and 70m, and the inverse of the variance is 0.008, that is, the first uniformity parameter is 0.008. The distribution

[0073] When the first uniformity parameter is greater than or equal to the uniformity threshold, the plurality of first collection positions are relatively uniform, and the subsequent steps are continued. The uniformity threshold can be set by a person skilled in the art based on the size of the target road area and the requirement for the uniformity distribution of the collection positions, for example, 0.01.

[0074] If the first uniformity parameter is less than the uniformity threshold, the plurality of first collection positions are randomly generated again in the target road area, and the first uniformity parameter is calculated until the first uniformity parameter is greater than or equal to the uniformity threshold, so as to ensure the uniformity and globality of the initial population in the optimization.

[0075] When the first uniformity parameter is greater than or equal to the uniformity threshold, according to the plurality of first collection positions, the collection range configured by the foregoing steps is used to divide a plurality of first collection ranges, respectively taking the plurality of first collection positions as the center points and taking the collection range as the division range. Among them, in combination with the scale of the remote sensing image, in combination with the area size of the plurality of first collection ranges and the coordinates of the plurality of first collection positions, the plurality of first collection ranges can be divided in the remote sensing image.

[0076] Further, according to the repetition ratio of the plurality of first collection ranges and the coverage degree of the snow-covered boundary region, a first collection fitness of the plurality of first collection ranges is calculated, with the purpose of reducing the repetition ratio and improving the coverage degree.

[0077] The step "calculating a first collection fitness of the plurality of first collection ranges according to the repetition ratio of the plurality of first collection ranges and the coverage degree of the snow-covered boundary region" performed by the collection position optimization module 13 in the embodiments of the present application comprises:

[0078] The range repetition ratio of each first collection range and the adjacent first collection range is calculated, and a plurality of first repetition ratios is obtained.

[0079] According to the snow-covered region and the snow-free covered region, a snow-covered boundary region is divided, and the coverage ratio of the plurality of first collection ranges to the snow-covered boundary region is obtained as the coverage degree.

[0080] According to the plurality of first repetition ratios and the coverage degree, a first collection fitness of the plurality of first collection ranges is calculated, as follows:

[0081]

[0082] wherein FIT cov is the collection fitness, w1 and w2 are the repetition weight and the coverage weight, the sum of w1 and w2 is 1, M is the number of the plurality of first collection ranges, M is an integer greater than 1, is the repetition ratio of the i-th first collection range and the adjacent first collection range, K c is the coverage degree.

[0083] In the embodiments of the present application, the range repetition ratio of each first collection range and the adjacent first collection range is calculated, and according to the adjacent first collection position of each first collection position in the foregoing, the adjacent first collection range of each first collection range can be determined.

[0084] Further, the range repetition ratio of each first collection range and the adjacent first collection range is calculated, for example, according to the range area of the plurality of first collection ranges divided in the remote sensing image, according to the scale of the remote sensing image, the area of the intersection of each first collection range and the adjacent first collection range is measured and calculated as the intersection area, and then the ratio of the intersection area and the area of the first collection range is calculated as the range repetition ratio. Exemplarily, the intersection area is 500m 2 , the area of the first collection range is 5000m 2 , and the range repetition ratio is 10%. If there is no intersection between the first collection range and the adjacent first collection range, the range repetition ratio is 0. In this way, a plurality of first repetition ratios is obtained.

[0085] Based on the previously defined snow-covered and snow-free areas, snow-covered boundary regions can be delineated within remote sensing images. These are the areas at the intersection of snow-covered and snow-free areas. The intersection of snow-covered and snow-free areas often provides early signs of snowstorms. For example, at the beginning of a snowfall, snow initially accumulates at the boundary region, but over time, it spreads to surrounding areas. Therefore, these boundary regions often represent the areas where snow cover changes most significantly, and monitoring these areas can help identify the risk of road snowstorms early on.

[0086] Exemplarily, the snow-covered boundary area includes the boundary line between the snow-covered area and the snow-non-covered area, specifically including the area within 1 m on both sides of the boundary line. In this way, the snow-covered boundary area in the remote sensing image is divided and the area of ​​the snow-covered boundary area can be measured.

[0087] Furthermore, based on the ranges of the multiple first acquisition ranges divided in the remote sensing image, the intersection area of ​​each first acquisition range and the snow-covered boundary area can be obtained, and the area of ​​the intersection area can be calculated based on the scale of the remote sensing image. The sum of the areas of the intersections of all the multiple first acquisition ranges and the snow-covered boundary area is calculated, and the ratio of the sum of the areas to the area of ​​the snow-covered boundary area is calculated as the coverage, that is, the ratio of the areas of the intersections of the multiple first acquisition ranges and the snow-covered boundary area to the area of ​​the snow-covered boundary area. For example, the sum of the areas of the intersections of the multiple first acquisition ranges and the snow-covered boundary area is 200m 2 The snow cover boundary area is 400m 2 , the coverage is 0.5.

[0088] The greater the coverage, the more the multiple first acquisition ranges can cover the junction of snow-covered areas and snow-free areas, thereby improving the reliability and accuracy of snow disaster road disaster prediction and warning.

[0089] Finally, the first acquisition fitness of the multiple first acquisition ranges is calculated based on the multiple first repetition ratios and coverages, as shown in the following formula:

[0090]

[0091] Among them, FIT cov is the acquisition fitness. The greater the acquisition fitness, the better the multiple first acquisition ranges and the higher the value of the acquired images. w1 and w2 are the repetition weight and coverage weight, respectively. The sum of w1 and w2 is 1. They can be set to 0.5 and 0.5, respectively, based on the importance of reducing the repetition ratio of acquisition ranges and improving coverage. M is the number of the multiple first acquisition ranges, an integer greater than 1, such as 10. The number of the multiple first acquisition positions is also M. The repetition ratio of the ith first collection range and the adjacent first collection range, the fitness decreases as the repetition ratio increases, K c The coverage, the fitness increases as the coverage increases.

[0092] In this way, the first collection fitness of the plurality of first collection ranges is calculated, the higher the first collection fitness, the better the plurality of first collection ranges, the higher the quality of the image of the target road area in the plurality of first collection ranges collected by the unmanned aerial vehicle, and the higher the reliability of the predicted snow disaster risk.

[0093] Further, in the optimization process, if the repetition ratio of a first collection range and an adjacent first collection range is large, the two first collection ranges are close, the content of the unmanned aerial vehicle image collected in the close collection range is similar, the first collection range and the adjacent first collection range with a large repetition ratio can be merged to retain one collection range, and the subsequent optimization in the similar collection range is avoided. Waste of computing power.

[0094] Specifically, the repetition ratio of each first collection range and the adjacent first collection range is calculated, and a plurality of groups of first collection ranges with large repetition ratios are merged to obtain a plurality of merged second collection ranges after merging.

[0095] The steps "calculating the repetition ratio of each first collection range and the adjacent first collection range, merging the plurality of first collection ranges, and obtaining a plurality of merged second collection ranges after merging" performed by the collection position optimization module 13 in the embodiment of the application include:

[0096] The range repetition ratio of each first collection range and the adjacent first collection range is calculated to obtain a plurality of first repetition ratios.

[0097] The first collection ranges corresponding to the N largest first repetition ratios are selected as N merged first collection ranges, and N is an integer greater than 1 and less than M.

[0098] The center points of the first collection positions of the N merged first collection ranges and the adjacent first collection positions are calculated respectively as N merged collection positions.

[0099] According to the N merged collection positions and the collection ranges, a plurality of merged second collection ranges are divided.

[0100] In the embodiment of the application, based on the steps in the foregoing content, the range repetition ratio of each first collection range and the adjacent first collection range is calculated to obtain a plurality of first repetition ratios.

[0101] Select the maximum N first repetition ratio corresponding to the first acquisition range, as the N merged first acquisition range, N is greater than 1 and less than M integer, for example, 5. The greater the repetition ratio, the greater the similarity of two adjacent first acquisition range, merge. The selected N merged first acquisition range is merged with its adjacent first acquisition range.

[0102] Specifically, the center point of the first acquisition position of the N merged first acquisition range and the acquisition position of the adjacent first acquisition range is calculated, that is, the center point between the two acquisition positions in each group of merged first acquisition range and adjacent first acquisition range is calculated as the acquisition position of the merged acquisition range, and N merged acquisition positions are obtained.

[0103] For example, the coordinates of the two first acquisition positions of the merged first acquisition range and the adjacent first acquisition range are (x1, y1) and (x2, y2) respectively, and the center point, that is, the merged acquisition position is ((x1+x2) / 2, (y1+y2) / 2). In this way, N merged acquisition positions are obtained by merging.

[0104] Further, according to the above acquisition range, the coordinates of the N merged acquisition positions are taken as the center point, and the division of the acquisition range is carried out, and the N merged acquisition ranges after merging are obtained, combined with the M-2N first acquisition ranges which are not merged, to obtain a plurality of merged second acquisition ranges, and complete the merging in the first round of optimization.

[0105] In the embodiment of the application, the second round of optimization is continued, and a plurality of second acquisition positions are generated in the target road area, and the number of the plurality of second acquisition positions is equal to the number of the merged first acquisition ranges, for example, N second acquisition positions are generated.

[0106] Based on the same way, combined with the plurality of second acquisition positions and the acquisition positions of the plurality of merged second acquisition ranges, the second uniformity parameter is continuously calculated and obtained, and when it is greater than or equal to the uniformity threshold, the optimization is continuously carried out, and if it is less than the uniformity threshold, the second acquisition position is re-generated.

[0107] Continue to optimize until the optimization converges, for example, reach the preset optimization number, for example, 50 times, then converge, output the plurality of acquisition positions with the maximum acquisition fitness, and obtain a plurality of optimal acquisition positions, that is, the optimal acquisition position distribution.

[0108] Through the above steps, not only the repeated collection positions are reduced, but also the complete coverage of the collection area is ensured. The merged collection range is more spatially uniform, and more images of the snow-covered interface area can be collected, which can improve the efficiency and quality of the unmanned aerial vehicle data collection, avoid unnecessary resource waste, improve the overall performance of the road disaster warning system, and ensure that the image data obtained through the unmanned aerial vehicle image collection has higher quality and more comprehensive geographical coverage.

[0109] The road risk warning module 14 is configured to perform step S40: performing unmanned aerial vehicle image collection according to the optimal collection position distribution and the collection range, performing road disaster risk prediction in combination with the remote sensing image, obtaining a road disaster risk level, and performing warning.

[0110] In the embodiment of the present application, the unmanned aerial vehicle is controlled to fly to the plurality of center points according to the plurality of optimal collection positions in the optimal collection position distribution, and then unmanned aerial vehicle images are collected according to the collection range to obtain ground images at the plurality of optimal collection positions, which include information of snowfall at the plurality of optimal collection positions. In combination with the remote sensing image, road disaster risk prediction is performed to obtain a road disaster risk level for warning.

[0111] Illustratively, based on a convolutional neural network in the machine learning technology in the prior art, sample unmanned aerial vehicle images and remote sensing images are collected, and are labeled according to the actual snowfall as training data to train a snowfall identification model to identify the snowfall in the unmanned aerial vehicle images and the remote sensing images, and to classify to obtain a corresponding road disaster risk level for warning. Artificial identification of the snowfall in the images by a person skilled in the art in combination with the snowfall area can also be used to evaluate and classify to obtain a corresponding road disaster risk level, such as blue, yellow, and red snow warning levels. For example, a snowfall of 0mm-5cm is a blue snow warning level, a snowfall of 5cm-15cm is a yellow snow warning level, and a snowfall of 15cm or more is a red snow warning level.

[0112] The road disaster warning system using satellite remote sensing and unmanned aerial vehicle data complementation provided by the embodiment of the present application has at least the following technical effects:

[0113] By combining satellite remote sensing and unmanned aerial vehicle data complement, the accuracy and efficiency of the road disaster early warning system are significantly improved. First, satellite remote sensing is used to collect images of the target road area, which can quickly identify snow-covered areas and provide basic data for subsequent disaster risk prediction. The introduction of unmanned aerial vehicle image collection determines the collection range of the unmanned aerial vehicle by analyzing satellite remote sensing images and snow-covered areas, improves the accuracy of unmanned aerial vehicle ground image collection, ensures the efficiency and quality of data collection and disaster prediction, optimizes the distribution of collection positions, avoids the problems of redundant or insufficient ground image data collection, improves the reliability and comprehensiveness of ground image data collection, and further improves the reliability and timeliness of disaster prediction. Through the complement of high-definition images collected by unmanned aerial vehicles and satellite images, and the optimization of the collection range and collection position, the disaster risk can be accurately identified, and the timeliness and reliability of the road snow disaster prediction can be ensured.

[0114] In the embodiment two, based on the same inventive concept of the road disaster early warning system using satellite remote sensing and unmanned aerial vehicle data complement provided in the embodiment one, the present embodiment also provides a road disaster early warning method using satellite remote sensing and unmanned aerial vehicle data complement, comprising: Figure 2

[0115] Using satellite remote sensing, collecting remote sensing images of the target road area, and preliminarily identifying snow-covered areas;

[0116] According to the remote sensing images and the snow-covered areas, the average apparentness of the remote sensing images is analyzed, and the collection range of the unmanned aerial vehicle image collection of the target road area is configured according to the average apparentness;

[0117] In the target road area, the optimization of the collection position of the unmanned aerial vehicle image collection is carried out to obtain the optimal collection position distribution, wherein the optimization is carried out according to the repetition of the collection position in the collection range to merge the collection positions;

[0118] According to the optimal collection position distribution and the collection range, the unmanned aerial vehicle image collection is carried out, the road disaster risk prediction is carried out in combination with the remote sensing images, the road disaster risk level is obtained, and the early warning is carried out.

[0119] Further, using satellite remote sensing, collecting remote sensing images of the target road area, and preliminarily identifying snow-covered areas, comprising:

[0120] Using satellite remote sensing, collecting remote sensing images of the target road area;

[0121] The remote sensing images are input into a pre-constructed snow-covered area identification channel to identify the snow-covered areas, wherein the snow-covered area identification channel is pre-trained and constructed based on a convolutional neural network, and the training data includes sample remote sensing images and sample snow-covered areas. ​

[0122] Further, according to the remote sensing image, the snow-covered area, the average obviousness of the remote sensing image is analyzed and obtained, including:

[0123] The remote sensing image is grayed, and the mean value of the gray scale of the snow-covered area remote sensing image is calculated to obtain the snow-covered area gray scale.

[0124] Randomly extract the gray scale values of a plurality of pixel points in the gray remote sensing image, and calculate the mean value to obtain a random gray scale.

[0125] The gray scale deviation ratio of the random gray scale and the snow-covered area gray scale is calculated as the average obviousness.

[0126] Further, according to the average obviousness, the collection range of the unmanned aerial vehicle image collection of the target road area is configured, including:

[0127] Obtain the maximum collection range of the unmanned aerial vehicle image collection;

[0128] The average obviousness is multiplied by the maximum collection range of the unmanned aerial vehicle image collection to obtain the collection range.

[0129] Further, in the target road area, the optimization of the collection position of the unmanned aerial vehicle image collection is performed to obtain the optimal collection position distribution, including:

[0130] Divide the area in the remote sensing image outside the snow-covered area to obtain a snow-free covered area;

[0131] Randomly generate a plurality of first collection positions in the target road area, and analyze the first uniformity parameters of the plurality of first collection positions;

[0132] When the first uniformity parameter is greater than or equal to the uniformity threshold, a plurality of first collection ranges are divided and obtained according to the plurality of first collection positions and the collection range;

[0133] According to the repetition ratio of the plurality of first collection ranges and the coverage of the snow-covered boundary area, the first collection fitness of the plurality of first collection ranges is calculated and obtained;

[0134] Calculate the repetition ratio of each first collection range and the adjacent first collection range, merge the plurality of first collection ranges, and obtain a plurality of merged second collection ranges after merging;

[0135] Supplement a plurality of second collection positions in the target road area, combine the collection positions of the plurality of merged second collection ranges, calculate the second uniformity parameter, and continue to optimize when it is greater than or equal to the uniformity threshold.

[0136] Until optimization converges, output the collection position with the maximum collection fitness, and obtain the optimal collection position distribution.

[0137] Further, within the target road area, a plurality of first collection positions are randomly generated, and a first uniformity parameter of the plurality of first collection positions is analyzed, including:

[0138] The distance between each first collection position and the adjacent first collection position is calculated to obtain a plurality of first distances.

[0139] The variance of the plurality of first distances is calculated and the reciprocal of the variance is calculated to obtain the first uniformity parameter.

[0140] Further, according to the repetition ratio of the plurality of first collection ranges and the coverage of the snow-covered boundary area, the first collection fitness of the plurality of first collection ranges is calculated, including:

[0141] The range repetition ratio between each first collection range and the adjacent first collection range is calculated to obtain a plurality of first repetition ratios.

[0142] According to the snow-covered area and the snow-free covered area, the snow-covered boundary area is divided to obtain the coverage ratio of the plurality of first collection ranges to the snow-covered boundary area as the coverage.

[0143] According to the plurality of first repetition ratios and the coverage, the first collection fitness of the plurality of first collection ranges is calculated as follows:

[0144]

[0145] wherein FIT cov is the collection fitness, w1 and w2 are the repetition weight and the coverage weight, the sum of w1 and w2 is 1, M is the number of the plurality of first collection ranges, M is an integer greater than 1, is the repetition ratio between the i-th first collection range and the adjacent first collection range, K c is the coverage.

[0146] Further, the repetition ratio between each first collection range and the adjacent first collection range is calculated, and the plurality of first collection ranges are merged to obtain a plurality of merged second collection ranges, including:

[0147] The range repetition ratio between each first collection range and the adjacent first collection range is calculated to obtain a plurality of first repetition ratios.

[0148] The first collection ranges corresponding to the maximum N first repetition ratios are selected as N merged first collection ranges, and N is an integer greater than 1 and less than M.

[0149] Calculate the center points of the first acquisition positions of the N merged first acquisition ranges and the acquisition positions adjacent to the first acquisition ranges respectively as N merged acquisition positions;

[0150] According to the N merged acquisition positions and the acquisition ranges, a plurality of merged second acquisition ranges are obtained by division.

[0151] It should be noted that in the above embodiments, the description of each embodiment has its own emphasis, and the parts not described in detail in a certain embodiment can be referred to the related description of other embodiments.

[0152] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take 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.

[0153] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as combinations of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded computer, or other programmable data processing devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in one or more flows and / or blocks.

[0154] These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that implements the functions specified in one or more flows and / or blocks.

[0155] These computer program instructions can also be loaded into a computer or other programmable data processing device, so that a series of operation steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocksFigure 1 the steps of the functions specified in the one or more blocks.

[0156] While the preferred embodiments of the application have been described, additional variations and modifications can be made to the embodiments by those skilled in the art once they learn of the basic inventive concepts taught in the disclosure.

[0157] It is apparent that those skilled in the art can make modifications and variations to the application without departing from the spirit and scope of the application. Accordingly, it is intended to include all such modifications and variations in the scope of the application and its equivalents.

Claims

1. A road disaster early warning system that utilizes satellite remote sensing and drone data to complement each other, characterized by: The system comprises: The remote sensing recognition module is used to collect remote sensing images of the target road area using satellite remote sensing and preliminarily identify the snow-covered area; a collection range configuration module, configured to analyze and obtain an average visibility of the remote sensing image based on the remote sensing image and the snow-covered area, and configure a collection range for drone image collection of the target road area based on the average visibility; A collection position optimization module is used to optimize the collection positions of drone image collection within the target road area to obtain an optimal collection position distribution, wherein the optimization analyzes the duplication of collection positions according to the collection range and merges the collection positions; The road risk warning module is used to collect drone images according to the optimal collection position distribution and collection range, and combine the remote sensing images to predict road disaster risks, obtain road disaster risk levels, and issue warnings.

2. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 1 is characterized in that: Using satellite remote sensing, collect remote sensing images of the target road area and preliminarily identify the snow-covered area, including: Using satellite remote sensing to collect remote sensing images of the target road area; The remote sensing image is input into a pre-constructed snow-covered area identification channel to identify the snow-covered area, wherein the snow-covered area identification channel is pre-trained and constructed based on a convolutional neural network, and the training data includes sample remote sensing images and sample snow-covered areas.

3. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 1 is characterized in that: The average visibility of the remote sensing image is obtained by analyzing the remote sensing image and the snow-covered area, including: grayscale processing is performed on the remote sensing image, and the mean grayscale of the remote sensing image of the snow-covered area is calculated to obtain the grayscale of the snow-covered area; Randomly extract the grayscale values ​​of multiple pixels in the grayscale remote sensing image and calculate the mean to obtain random grayscale; The grayscale deviation ratio between the random grayscale and the grayscale of the snow-covered area is calculated as the average conspicuity.

4. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 1 is characterized in that: According to the average visibility, the acquisition range of the drone image acquisition of the target road area is configured, including: Get the maximum acquisition range of the drone for image acquisition; The average conspicuity is multiplied by the maximum acquisition range of the UAV image acquisition to obtain the acquisition range.

5. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 1 is characterized in that: Optimizing the acquisition positions of drone images within the target road area to obtain an optimal acquisition position distribution includes: dividing the area outside the snow-covered area in the remote sensing image to obtain a snow-free area; randomly generating a plurality of first acquisition positions within the target road area, and analyzing first uniformity parameters of the plurality of first acquisition positions; When the first uniformity parameter is greater than or equal to a uniformity threshold, dividing the first acquisition ranges according to the multiple first acquisition positions and acquisition ranges; Calculating first acquisition adaptability of the plurality of first acquisition ranges according to the repetition ratios of the plurality of first acquisition ranges and the coverage of the snow-covered boundary area; Calculating a repetition ratio between each first acquisition range and an adjacent first acquisition range, merging the multiple first acquisition ranges to obtain a plurality of merged second acquisition ranges; Generating a plurality of additional second acquisition positions within the target road area, calculating and obtaining a second uniformity parameter by combining the plurality of acquisition positions that merge into the second acquisition range, and continuing the optimization when the second uniformity parameter is greater than or equal to the uniformity threshold; Until the optimization converges, multiple collection positions with the largest collection fitness are output to obtain the optimal collection position distribution.

6. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 5 is characterized in that: Randomly generating a plurality of first acquisition positions within the target road area, and analyzing first uniformity parameters of the plurality of first acquisition positions, including: Calculating the distance between each first acquisition position and an adjacent first acquisition position to obtain a plurality of first distances; The variance of the plurality of first distances is calculated and the inverse of the variance is calculated to obtain a first uniformity parameter.

7. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 5, characterized in that: Calculating first acquisition adaptability of the plurality of first acquisition ranges according to the repetition ratios of the plurality of first acquisition ranges and the coverage of the snow-covered boundary area includes: Calculating a range repetition ratio between each first acquisition range and an adjacent first acquisition range to obtain a plurality of first repetition ratios; Dividing the snow-covered area and the snow-free area into a snow-covered boundary area, and obtaining coverage ratios of the plurality of first acquisition ranges to the snow-covered boundary area as coverage; According to the multiple first repetition ratios and coverages, the first acquisition fitness of the multiple first acquisition ranges is calculated as follows: Among them, FIT cov is the acquisition fitness, w1 and w2 are the repetition weight and coverage weight, the sum of w1 and w2 is 1, M is the number of multiple first acquisition ranges, M is an integer greater than 1, is the repetition ratio of the i-th first acquisition range to the adjacent first acquisition range, K c For coverage.

8. The road disaster early warning system using satellite remote sensing and drone data complementation according to claim 7, characterized in that: Calculating a repetition ratio between each first acquisition range and an adjacent first acquisition range, merging the multiple first acquisition ranges to obtain multiple merged second acquisition ranges, including: Calculating a range repetition ratio between each first acquisition range and an adjacent first acquisition range to obtain a plurality of first repetition ratios; Select the first acquisition ranges corresponding to the largest N first repetition ratios as the N merged first acquisition ranges, where N is an integer greater than 1 and less than M; Calculating the center points of the first acquisition positions of the N merged first acquisition ranges and the acquisition positions adjacent to the first acquisition ranges as N merged acquisition positions; According to the N combined acquisition positions and the acquisition range, a plurality of combined second acquisition ranges are obtained by division.

9. A road disaster early warning method using satellite remote sensing and drone data complements each other, characterized in that: The method is applied to the road disaster early warning system using satellite remote sensing and drone data complementation as described in any one of claims 1 to 8, comprising: Using satellite remote sensing, collect remote sensing images of the target road area and preliminarily identify the snow-covered area; Analyze and obtain an average visibility of the remote sensing image based on the remote sensing image and the snow-covered area, and configure a collection range for drone image collection of the target road area based on the average visibility; Optimizing the acquisition positions of drone image acquisition within the target road area to obtain an optimal acquisition position distribution, wherein the optimization includes analyzing the duplication of acquisition positions according to the acquisition range and merging the acquisition positions; According to the optimal collection position distribution and collection range, drone image collection is carried out, and road disaster risk prediction is carried out in combination with the remote sensing image to obtain the road disaster risk level and issue an early warning.

Images