Urban inland inundation man-vehicle imbalance risk identification method
By constructing an 8-level waterlogging level standard and combining the RetinaNet algorithm with meteorological data, the risk of urban flooding and imbalance between people and vehicles is identified, solving the coverage and accuracy problems of traditional monitoring methods and achieving efficient waterlogging detection and risk assessment.
Patent Information
- Application Number
- CN202511335979.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-18
- Publication Date
- 2025-12-19
AI Technical Summary
Traditional urban waterlogging monitoring methods have limited coverage and slow response. Vision-based detection methods have poor anti-interference capabilities and cannot quantify waterlogging levels. Deep learning algorithms are not optimized for waterlogging scenarios. Weather warning and video surveillance systems lack data linkage, resulting in insufficient warning accuracy and low resource utilization.
An 8-level refined water accumulation level standard based on the characteristics of people and vehicles being submerged was constructed. Water accumulation detection was performed by combining the RetinaNet deep learning algorithm and a dynamic scheduling mechanism based on meteorological data. The risk of imbalance between people and vehicles was identified through video surveillance.
It significantly improves the accuracy of water accumulation detection, optimizes computing resources, and provides an efficient and reliable intelligent solution for urban flood control.
Smart Images
Figure CN121170701A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a method for identifying imbalance risk of people and vehicles in urban waterlogging. BACKGROUND
[0002] Traditional urban waterlogging monitoring relies on physical sensors and manual patrol, and has problems such as limited coverage and delayed response. The detection method based on vision mostly uses traditional image processing technology, which has poor anti-interference ability and cannot quantify the waterlogging grade. The existing deep learning algorithm is not optimized for waterlogging scenes, and the detection result has poor practicability and large algorithm consumption. In addition, the meteorological warning and the video monitoring system are independent of each other, and lack of data linkage mechanism, and the risk assessment of people and vehicles often ignores the water depth factor, resulting in low warning accuracy and low resource utilization. SUMMARY
[0003] The application aims to provide a method for identifying imbalance risk of people and vehicles in urban waterlogging, to solve the problems in the prior art. The technical problems to be solved by the application are solved by the following technical scheme.
[0004] A method for identifying imbalance risk of people and vehicles in urban waterlogging, comprising the following steps:
[0005] S1: obtaining urban road waterlogging pictures in a historical rainfall period to form a road waterlogging preliminary data set;
[0006] S2: screening the waterlogging pictures in the road waterlogging preliminary data set to make them at least include one of people and vehicles, and forming a road waterlogging data set;
[0007] S3: labeling the data in the road waterlogging data set according to waterlogging grades;
[0008] S4: dividing the labeled road waterlogging data set into a training set, a verification set and a test set, performing an enhancement operation on the pictures, training a model using a RetinaNet target detection algorithm by using the data of the training set, verifying and evaluating the model by using the data of the verification set and the test set, adjusting the training parameters until the model loss fitting is achieved, and obtaining a urban waterlogging grade identification model;
[0009] S5: applying the trained urban waterlogging grade identification model to video monitoring to obtain the waterlogging grade of a specific video monitoring point;
[0010] S6: calculating the imbalance state of people and vehicles at the video monitoring point respectively;
[0011] S7: combining the waterlogging grade of the video monitoring point and the state of people and vehicles at the point to obtain the imbalance risk of people and vehicles.
[0012] Preferably, in step S1, the urban road waterlogging picture comes from video monitoring data and network pictures obtained by network crawler.
[0013] Preferably, in step S2, the target detection model trained by coco2017 dataset is used to filter the pictures in the dataset, so that the pictures at least include one of people and vehicles.
[0014] Preferably, in step S3, according to the depth of road waterlogging in the picture, the waterlogging level is divided into multiple levels, the waterlogging level of the picture in the dataset is labeled by using a labeling tool Labelme, and the person or vehicle in the waterlogging is selected by a rectangular frame.
[0015] Preferably, in step S4, the image enhancement includes image splicing, image flipping and image scaling, the model includes ResNet for feature extraction, FPN for generating multi-scale feature maps, and SubNet for target classification and bounding box regression.
[0016] Preferably, in step S4, the evaluation index includes accuracy Precision, recall Recall, average precision AP and average precision mean mAP, and specifically:
[0017]
[0018] Wherein, TP is true positive, positive example is identified as positive example; FP is false positive, negative example is identified as positive example; FN is false negative, positive example is identified as negative example;
[0019]
[0020] Wherein, AP is average precision, the value is equal to the area under the precision-recall curve after interpolation, r is recall, P interpolated (r) is the interpolated precision value at a certain recall rate level r, and Δr is the step or small increment of recall rate r when summing;
[0021]
[0022] Wherein, K is the number of target detection categories.
[0023] Preferably, in step S6, the imbalance state of the person is divided into slipping and falling, and the imbalance speed is used to distinguish between the two;
[0024] The discrimination formula of slipping is as follows:
[0025]
[0026] The discrimination formula of falling is as follows:
[0027]
[0028] wherein, U c is the critical velocity at which the human body starts to be unstable, and alpha and beta are characteristic parameters of the human body and the experimental water tank respectively, rho is the density of water, h f is the accumulated water depth, h p is the height of the human body, m p is the mass of the human body, and a1 and b1 are characteristic parameters of the human body, and a2 and b2 are linear relationship coefficients of the volume and weight of the human body.
[0029] Preferably, the imbalance state of the vehicle is determined by the following formula:
[0030]
[0031] wherein, U c is the critical velocity at which the vehicle starts to be unstable, and alpha and beta are parameters related to the vehicle, h f is the submersion depth of the vehicle, h c is the height of the vehicle, g is the acceleration of gravity, b c is the width of the vehicle, rho c and rho f are the densities of the vehicle and water respectively, R f =(h c gamma c / h k gamma f ), R f is the force per unit area of the vehicle and flood, and hk is the critical depth at which the vehicle starts to float, gamma c and gamma f are the gravity per unit volume of the vehicle and flood respectively.
[0032] The urban waterlogging human-vehicle imbalance risk identification method of the present application significantly improves the waterlogging detection accuracy by constructing an 8-level refined accumulated water level standard based on human-vehicle submersion characteristics, combined with the RetinaNet deep learning algorithm, and simultaneously realizes the optimization of computing resources by combining the meteorological data dynamic scheduling mechanism, thereby providing an efficient and reliable intelligent solution for urban waterlogging prevention and control. BRIEF DESCRIPTION OF DRAWINGS
[0033] Figure 1 is the overall architecture diagram of the present application;
[0034] Figure 2 is the accumulated water level rule schematic diagram;
[0035] Figure 3 is the imbalance critical flow velocity relationship of adults and children at different water depths;
[0036] Figure 4 is the imbalance critical flow velocity relationship of vehicles at different water depths;
[0037] Figure 5 for video monitoring person detection example;
[0038] Figure 6 for video monitoring vehicle detection example. DETAILED DESCRIPTION
[0039] The technical solutions of the present application will be further described in detail below with the aid of the accompanying drawings and examples.
[0040] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with the aid of the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations.
[0041] Therefore, the detailed description of the embodiments of the present application provided below in the accompanying drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without making creative efforts fall within the scope of protection of the present application.
[0042] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.
[0043] In the description of the present application, it should be noted that the terms "upper", "lower", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, or the orientation or positional relationship in which the product of the present application is usually placed, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the indicated device or element must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.
[0044] In the description of the present application, it should also be noted that, unless otherwise explicitly specified and limited, the terms "arrangement", "installation", "connection" should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or integrally connected; it can be mechanically connected, or it can be electrically connected; it can be directly connected, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.
[0045] The technical solutions of the present application will be further described in detail below with the aid of the accompanying drawings and examples. Figures 1-5Detailed description of embodiments of the present application.
[0046] A method for identifying the risk of imbalance between people and vehicles in urban waterlogging, comprising the following steps:
[0047] Obtain the pictures of urban road waterlogging during the historical rainfall period to form a preliminary data set of road waterlogging.
[0048] Cooperate with relevant departments to obtain video monitoring data of road or bridge waterlogging during the rainfall period, and take pictures in the video with 1 second as a frame. Through network crawler technology, specify keywords such as urban waterlogging and flood to crawl photo data of Weibo, Douyin, Baidu pictures and other network media, and purchase and obtain picture data of urban waterlogging and flood through related platforms to form a preliminary data set of road waterlogging.
[0049] Use the target detection model trained by coco2017 dataset to filter the pictures in the preliminary data set of road waterlogging, so that it at least includes one of people and vehicles, and form a road waterlogging dataset.
[0050] Label the data in the road waterlogging dataset according to the waterlogging level. Divide the waterlogging into multiple levels and represent it with numbers, use the labeling tool Labelme to label the waterlogging level of the pictures in the dataset, and select the people or vehicles in the waterlogging by rectangular frame.
[0051] The following gives a way to divide the waterlogging level:
[0052] Waterlogging level 0 is 0 cm, i.e. no waterlogging; waterlogging level 1 is 1-10 cm, i.e. below the wrist, slight waterlogging; waterlogging level 2 is 10-30 cm, i.e. below half of the calf, below half of the tire; waterlogging level 3 is 30-50 cm, i.e. around the knees, more than half of the tire is submerged; waterlogging level 4 is 50-70 cm, i.e. thighs, almost completely submerged tire; waterlogging level 5 is 70-100 cm, i.e. below the waist, the car is submerged to the window; waterlogging level 6 is 100-120 cm, i.e. chest, half of the window is submerged; waterlogging level 7 is 120-150 cm, i.e. neck, almost all the car is submerged; waterlogging level 8 is > 150 cm.
[0053] The labeled road waterlogging dataset is divided into a training set, a validation set and a test set according to a ratio of 0.81:0.09:1, and an image enhancement operation is performed on the pictures, the image enhancement including image splicing, flipping and scaling, etc., a model adopting a RetinaNet target detection algorithm is trained using the data of the training set, the model including ResNet for feature extraction, FPN for generating multi-scale feature maps and SubNet for target classification and bounding box regression, the data of the validation set and the test set are used to verify and evaluate the model, and the training parameters are adjusted until the model loss fitting is achieved, at this time, the loss value reaches a stable minimum value, and a city waterlogging grade recognition model is obtained.
[0054] The evaluation indexes include accuracy Precision, recall Recall, average precision AP and average precision mean mAP, wherein,
[0055]
[0056] Wherein, TP is true positive, positive example is identified as positive example; FP is false positive, negative example is identified as positive example; FN is false negative, positive example is identified as negative example.
[0057]
[0058] Wherein, AP is average precision, the value thereof is equal to the area under the interpolated precision-recall curve, r is recall, P interpolated (r) is the interpolated precision value at a certain specific recall rate level r, and Δr is the step or small increment of recall r when summing.
[0059]
[0060] Wherein, K is the number of target detection categories.
[0061] The trained city waterlogging grade recognition model is applied to video monitoring, and the waterlogging grade of a specific video monitoring point is obtained.
[0062] The imbalance state of people and vehicles at the video monitoring point is calculated respectively.
[0063] The imbalance state of people is divided into slipping and falling, which is distinguished by imbalance speed.
[0064] The discrimination method of slipping is as follows:
[0065]
[0066] The discrimination method of falling is as follows:
[0067]
[0068] where U c is the critical speed at which the human body starts to become unstable, a and b are characteristic parameters of the human body and the experimental water tank, respectively, p is the density of water, h f is the depth of the accumulated water, h p is the height of the human body, m p is the mass of the human body, a1 and b1 are characteristic parameters of the human body, and a2 and b2 are linear relationship coefficients of the volume and weight of the human body. Table 1 is the characteristic parameters of the human body.
[0069]
[0070] Table 1
[0071] The imbalance phenomena of the vehicle in the urban waterlogging include three kinds: sliding, rolling and floating. The occurrence of rolling indicates that sliding has occurred and the occurrence probability of floating is small. When the vehicle and the water flow have an angle of 90°, the vehicle is more likely to be washed away, and thus the imbalance of the vehicle is determined according to the method under the condition.
[0072]
[0073] where U c is the critical speed at which the vehicle starts to become unstable, a and b are parameters related to the vehicle, h f is the depth of the vehicle being submerged, h c is the height of the vehicle, g is the acceleration of gravity, b c is the width of the vehicle, p c and p f are the densities of the vehicle and water, respectively, R f = (hcyc / hkycf), R f is the force ratio per unit area of the vehicle and the flood, hk is the critical depth at which the vehicle starts to float, yc and yf are the gravity per unit volume of the vehicle and the flood, respectively. Table 2 is the characteristic parameters of the vehicle.
[0074]
[0075] Table 2
[0076] The imbalance risk of the human body and the vehicle is obtained by combining the waterlogging level of the video monitoring point and the state of the human body and the vehicle at the point.
[0077] It should be noted that the above detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present application belongs.
[0078] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments described herein. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof.
[0079] In addition, the terms "comprise" and "comprising" and any variations thereof are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements that are expressly identified as being included in the process, method, system, product, or apparatus.
[0080] For purposes of the description hereinafter, the terms "upper", "lower", "right", "left", "vertical", "horizontal", "top", "bottom", "lateral", "medial", "superior", "inferior", "anterior", "posterior" and derivatives thereof shall relate to the application as it is shown in the drawings, unless stated otherwise. The terms "on", "side-by-side", "top of", "bottom of", "over", "under", "adjacent to" and "against" encompass both direct and indirect contact.
[0081] In the detailed description herein, reference is made to the accompanying drawings, which form a part hereof. In the drawings, similar symbols typically identify similar components, unless context dictates otherwise. The illustrative embodiments described in the detailed description, drawings, and claims are not meant to be limiting. Other embodiments can be used, and other changes can be made, without departing from the spirit or scope of the subject matter presented herein.
[0082] The specific embodiments described herein have been chosen for purposes of illustration and example only. Numerous alternative embodiments can be utilized by persons of ordinary skill in the art without departing from the spirit and scope of the subject matter presented herein, and any equivalents thereol. Additional modifications and changes will readily occur to persons skilled in the art. Persons skilled in the art will readily understand that additional changes might be made to the embodiments described herein, without departing from the spirit and scope of the subject matter presenting herein. Accordingly, changes and modifications can be made to the described embodiments without departing from the spirit or scope of the application.
Claims
1. An urban waterlogging vehicle imbalance risk identification method, characterized in that: The method comprises the following steps: S1: obtaining urban road waterlogging pictures in a historical rainfall period to form a road waterlogging preliminary data set; S2: screening the waterlogging pictures in the road waterlogging preliminary data set to include at least one of people and vehicles, to form a road waterlogging data set; S3: labeling the data in the road waterlogging data set according to waterlogging grades; S4: dividing the labeled road waterlogging data set into a training set, a verification set and a test set, performing an enhancement operation on the pictures, training a model adopting a RetinaNet target detection algorithm by using the data of the training set, verifying and evaluating the model by using the data of the verification set and the test set, and adjusting training parameters until the model loss fitting is achieved, to obtain an urban waterlogging grade recognition model; S5: applying the trained urban waterlogging grade recognition model to video monitoring to obtain waterlogging grades of specific video monitoring points; S6: calculating the imbalance states of people and vehicles at the video monitoring points respectively; S7: combining the waterlogging grades of the video monitoring points with the states of people and vehicles at the points to obtain imbalance risks of people and vehicles.
2. The urban waterlogging vehicle imbalance risk identification method according to claim 1, characterized in that: In step S1, the urban road waterlogging pictures come from video monitoring data and network pictures obtained through network crawling.
3. The urban waterlogging vehicle imbalance risk identification method of claim 1, wherein: In step S2, the target detection model trained by using the coco2017 data set is used to screen the pictures in the data set to include at least one of people and vehicles.
4. The urban waterlogging vehicle imbalance risk identification method of claim 1, wherein: In step S3, the waterlogging grades are divided into multiple grades according to the depths of road waterlogging in the pictures, the waterlogging grades of the pictures in the data set are labeled by using a labeling tool Labelme, and the people or vehicles in the waterlogging are selected by rectangular frames.
5. The urban waterlogging vehicle imbalance risk identification method of claim 1, wherein: In step S4, the image enhancement includes image splicing, flipping and scaling, the model includes ResNet for feature extraction, FPN for generating multi-scale feature maps and SubNet for target classification and bounding box regression.
6. The urban waterlogging vehicle imbalance risk identification method of claim 5, wherein: In step S4, the evaluation indexes include accuracy Precision, recall rate Recall, average precision AP and average precision mean mAP, and specifically: Wherein, TP is true positive, a positive example is identified as a positive example; FP is false positive, a negative example is identified as a positive example; FN is false negative, a positive example is identified as a negative example; where AP is the average precision, whose value is equal to the area under the interpolated precision-recall curve, r is the recall, and P interpolated (r) is the interpolated precision value at some particular recall level r, and Δr is the step or infinitesimal increment in recall r when summing. Wherein, K is the number of target detection categories.
7. The urban waterlogging vehicle imbalance risk identification method of claim 1, wherein: In step S6, the imbalance state of people is divided into slipping and falling, and the imbalance speed is used to distinguish between them; The discrimination formula of slipping is as follows: The discrimination formula of falling is as follows: where U c is the critical speed at which the human body starts to become unstable, a and b are characteristic parameters of the human body and the experimental water tank, respectively, p is the density of water, h f is the accumulated water depth, h p is the height of the human body, m p is the mass of the human body, a1 and b1 are characteristic parameters of the human body, and a2 and b2 are linear relationship coefficients of the volume and weight of the human body.
8. The urban waterlogging vehicle imbalance risk identification method of claim 1, wherein: The imbalance state of a vehicle is discriminated by the following formula: where U c is the critical speed at which the vehicle starts to become unstable, a and b are parameters related to the vehicle, h f is the submergence depth of the vehicle, h c is the height of the vehicle, g is the acceleration due to gravity, b c is the width of the vehicle, p c and p f are the densities of the vehicle and water respectively, R f = (h c g c / h k g f ), R f is the force per unit area on the vehicle due to the flood, hk is the critical depth at which the vehicle starts to float, g c and g f are the gravitational forces per unit volume on the vehicle and flood respectively.