Disaster assessment method and system based on aerial image target detection

Abstract

The invention provides a disaster assessment method and system based on aerial image target detection, and relates to the technical field of data processing, and the method comprises the steps: obtaining a to-be-detected aerial image; through a YOLO target detection algorithm, multiple targets of the aerial image are detected; classifying the target detection results according to target types to obtain a plurality of target sets; determining the pixel position, the pixel size and the detection confidence of each target in different target sets; calculating a target weighted quantitative index according to the target type and the corresponding detection confidence; constructing a time sequence according to the target weighted quantitative index; carrying out stationarity judgment on the time sequence in different judgment modes; based on the judgment result, extracting a time sequence with a trend state by using an adaptive difference mode; and according to the trend state, assessing the disaster. According to the invention, assessment and mining of disaster development laws and trends can be realized, and the cost of manual rescue is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a disaster assessment method and system based on aerial image target detection. Background Art

[0002] An aerial image refers to a ground image taken by an aerial device such as a drone, an aircraft, or a satellite. Target detection is a task in computer vision, aiming to identify the targets of interest in the image and accurately locate these targets. And the disaster assessment based on aerial image target detection is a process of combining aerial images with target detection technology to quickly and accurately identify important targets in the disaster scene, such as ruins, crowds, and vehicles, etc., so as to achieve the high efficiency and precision of disaster assessment.

[0003] Emergency response personnel can use the aerial images (including data frames in the video stream) of the disaster scene to quickly understand and master the key information of the disaster. Therefore, using the method based on aerial image target detection can obtain the key data of the disaster area in the first time, avoid the rescue delay caused by traffic interruption and information transmission delay, and is of great significance for quickly obtaining disaster information, reducing the error of manual recognition, and improving the disaster assessment efficiency, providing a solid data basis for scientific rescue and disaster recovery.

[0004] Traditional disaster assessment methods are based on aerial images and perform target detection manually. Especially under the condition of no or few people, obtaining disaster information from aerial images leads to low efficiency and high labor costs. Secondly, traditional disaster assessment methods mainly rely on multispectral data, neural networks, and pixel changes to assess specific disasters. The adaptability of the scheme itself is relatively poor and it is usually difficult to be widely used in various types of disaster scenarios, resulting in limited assessment value, reduced reliability of detection results, untimely and inaccurate acquisition of disaster information, thus leading to rescue delays and increasing the cost of the rescue work. Summary of the Invention

[0005] In order to solve the technical problems existing in the prior art that, based on aerial images, target detection is performed manually to assess specific disasters, the adaptability of the scheme itself is relatively poor, it is usually difficult to be widely used in various types of disaster scenarios, resulting in limited assessment value, reduced reliability of detection results, untimely and inaccurate acquisition of disaster information, thus leading to rescue delays and increasing the cost of the rescue work, the present invention provides a disaster assessment method and system based on aerial image target detection.

[0006] The technical solutions provided by the embodiments of the present invention are as follows:

[0007] In the first aspect

[0008] A disaster assessment method based on aerial image target detection provided by an embodiment of the present invention includes:

[0009] S1: Obtain the aerial image to be detected;

[0010] S2: Detect multiple targets in the aerial image through the YOLO target detection algorithm;

[0011] S3: Classify the target detection results according to the target type to obtain multiple target sets;

[0012] S4: Determine the pixel position, pixel size, and detection confidence of each target in different target sets;

[0013] S5: Calculate the target weighted quantization index according to the target type and the corresponding detection confidence;

[0014] S6: Construct a time series according to the target weighted quantization index;

[0015] S7: Perform stationarity discrimination on the time series in different discrimination methods, where the discrimination methods include the autocorrelation graph discrimination method and the ADF test discrimination method;

[0016] S8: Based on the discrimination result, use the adaptive difference method to extract the time series with a trend state;

[0017] S9: Evaluate the disaster according to the trend state.

[0018] Second aspect

[0019] A disaster assessment system based on aerial image target detection provided by an embodiment of the present invention includes:

[0020] A processor;

[0021] A memory, on which computer-readable instructions are stored. When the computer-readable instructions are executed by the processor, the disaster assessment method based on aerial image target detection as described in the first aspect is implemented.

[0022] Third aspect

[0023] A computer-readable storage medium provided by an embodiment of the present invention, on which a computer program is stored. When the program is executed by the processor, the disaster assessment method based on aerial image target detection as described in the first aspect is implemented.

[0024] The beneficial effects brought by the technical solutions provided by the embodiments of the present invention at least include:

[0025] In the embodiments of the present invention, based on aerial images, multiple targets in the aerial images are detected through the YOLO target detection algorithm, and the target detection results are classified according to the target types to extract key information. Then, the pixel positions, pixel sizes, and detection confidence levels of each target in different target sets are determined, and the target weighted quantization index is calculated according to the target type and the corresponding detection confidence level, thereby providing a unified quantitative standard for evaluation. According to the target weighted quantization index, a time series is constructed, and the stationarity of the time series is discriminated by different discrimination methods. Based on the discrimination results, an adaptive difference method is used to extract the time series with a trend state. Finally, according to the trend state, the disaster is evaluated, reducing the cost and time of manual target detection, improving the reliability and accuracy of the detection results, enhancing the efficiency of disaster evaluation, providing sufficient time for rescue work, and saving rescue time. The present invention is widely applicable to disaster scenarios such as earthquakes, fires, floods, rainstorms, frosts, and traffic accidents, and can be deployed and implemented in scenarios where aerial image target detection can operate, meeting diverse rescue needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0027] Figure 1 It is a schematic flowchart of a disaster evaluation method based on aerial image target detection provided by an embodiment of the present invention;

[0028] Figure 2 It is a schematic diagram of the quantity value of vehicle targets provided by an embodiment of the present invention;

[0029] Figure 3 It is a schematic diagram of the sample autocorrelation function of a time series provided by an embodiment of the present invention;

[0030] Figure 4 It is a schematic structural diagram of a disaster evaluation system based on aerial image target detection provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0031] The following will describe the technical solutions in the present invention with reference to the drawings.

[0032] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to give examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or more advantageous than other embodiments or design solutions. Rather, the use of the word "example" is intended to present concepts in a specific manner. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one of the two.

[0033] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0034] Refer to the attached Figure 1 figures, which show a schematic flow chart of a disaster assessment method based on aerial image target detection provided by an embodiment of the present invention.

[0035] The embodiments of the present invention provide a disaster assessment method based on aerial image target detection. This method can be implemented by a disaster assessment device based on aerial image target detection. The disaster assessment device based on aerial image target detection can be a terminal or a server. The processing flow of the disaster assessment method based on aerial image target detection may include the following steps:

[0036] S1: Obtain the aerial image to be detected.

[0037] Among them, the aerial image refers to a ground image taken from the air by a drone, satellite, helicopter or other flying devices. By obtaining the aerial image, it provides a basic data source for disaster assessment. The aerial image can quickly cover a large area of the disaster area, especially in the case where ground traffic is blocked or the disaster area is difficult to access, ensuring the timeliness and comprehensiveness of information collection. The aerial image has a high resolution and dynamic capture ability, and can show the real-time state of the disaster in detail, providing high-quality input for subsequent target detection and analysis.

[0038] S2: Detect multiple targets in the aerial image through the YOLO target detection algorithm.

[0039] Among them, the YOLO target detection algorithm is an efficient target detection algorithm, which can complete the detection of all targets in the image in a single forward propagation, and has the characteristics of strong real-time performance, high detection accuracy and wide application range.

[0040] It should be noted that detecting multiple targets in aerial images through the YOLO object detection algorithm has significant advantages such as high efficiency, accuracy, and real-time performance. The YOLO algorithm can process the entire image at once without multiple scans, thus significantly improving the detection speed, and is particularly suitable for situations that require rapid response in disaster scenarios. At the same time, the YOLO algorithm has a high detection accuracy and can identify targets that are sparsely distributed or partially occluded in aerial images, such as trapped people, rescue vehicles, and equipment.

[0041] In a possible implementation, the targets include people, vehicles, equipment, smoke and fire, and floods.

[0042] S3: Classify the object detection results according to the object type to obtain multiple object sets.

[0043] Among them, the object type refers to the category attribute of the object, which is used to distinguish different objects, such as people, vehicles, equipment, smoke and fire, floods, etc. The object set refers to the set of objects after classification, and each set contains all the detected objects of the same category, such as the people set, the vehicle set, etc.

[0044] It should be noted that by classifying the object detection results according to the object type, the scale of the disaster, the number of people or equipment to be rescued or being rescued can be effectively classified and characterized, which can reflect the refined and quantitative characteristics of the disaster from different perspectives, and significantly improve the organization and efficiency of subsequent analysis. The classification process integrates complex and diverse object detection results into multiple clear object sets, making data management and processing more systematic. This classification method makes the quantity statistics and feature analysis of different types of objects (such as people, vehicles, equipment, etc.) more convenient, thus providing accurate support for the quantitative characterization of different types of objects.

[0045] S4: Determine the pixel positions, pixel sizes, and detection confidence levels of each object in different object sets.

[0046] Among them, the pixel position refers to the specific position of the object in the image, represented by pixel coordinates, which is used to locate the area of the object in the aerial image. The pixel size refers to the area size of the object in the image, usually measured by the number of pixels covered by the object's bounding box, which reflects the physical scale of the object (such as building area or smoke diffusion range). The detection confidence level refers to the credibility of the object detection algorithm's prediction result for a certain object, usually represented by a value between 0 and 1.

[0047] It should be noted that by extracting the pixel position, pixel size, and detection confidence of the target, it provides key data support for the refined analysis of disaster information, can intuitively reflect the spatial distribution and scale characteristics of the target in the image, and provides an accurate basis for the subsequent quantitative characterization of disasters. At the same time, the introduction of detection confidence can effectively measure the reliability of the target detection result, making the subsequent calculations more scientific and controllable.

[0048] S5: Calculate the target weighted quantization index according to the target type and the corresponding detection confidence.

[0049] Among them, the target type refers to the classification of the targets detected in the aerial image, such as personnel, vehicles, equipment, etc. The target weighted quantization index is a quantization method that combines the target detection result and the target type, and uses the detection confidence to weight each target, so as to generate an index value that can reflect the importance of the target.

[0050] It should be noted that by weighting different target types and their detection confidences, the key targets in the disaster scene can be accurately quantified, providing a flexible evaluation method that can adapt to both countable targets (such as personnel and vehicles) and regional targets (such as flood area), ensuring the accuracy and reliability of disaster assessment.

[0051] In one possible implementation, S5 specifically includes:

[0052] S501: When the target type is a countable target, select the direct quantity summation method or the confidence-weighted quantity summation method to determine the weighted quantization index; otherwise, go to step S502.

[0053] Among them, the countable target refers to the target that can be directly quantified by counting the quantity, such as personnel, vehicles, etc. The direct quantity summation method is to directly accumulate and sum the quantities of all detected targets. The confidence-weighted quantity summation method means that when counting the target quantity, the detection confidence of each target is considered and the quantity is weighted.

[0054] S502: When the target type is a regional target, select the direct area summation method or the confidence-weighted area summation method to determine the weighted quantization index.

[0055] It should be noted that by selecting an appropriate quantization method according to the target type, it is ensured that the calculation result can effectively reflect the target characteristics and importance in the disaster. Through this weighting method, the detection results of different targets can be effectively integrated, making the disaster assessment more comprehensive and having high timeliness and operability.

[0056] In one possible implementation, the formulas of the direct quantity summation method and the confidence-weighted quantity summation method are specifically:

[0057]

[0058] Among them, X J represents a weighted quantity value type index, M represents the total number of countable targets, m = 1, …, M, P m represents the confidence level of the m-th countable target.

[0059] The formulas for the direct area summation method and the confidence-weighted area summation method are specifically as follows:

[0060]

[0061] Among them, X Z represents a weighted area value type index, N represents the total number of regional targets, S n represents the area of the n-th regional target, n = 1, …, N, P n represents the confidence level of the n-th regional target.

[0062] S6: Construct a time series according to the target weighted quantization index.

[0063] Among them, the time series is a set of data points arranged in chronological order, reflecting the change trend of the target weighted quantization index at different time points. By constructing the time series, the dynamic change information of the targets in the disaster scenario is systematically organized with time as the clue. This way can capture the trend characteristics of the targets at different time points, which helps to analyze the development law and influence range of the disaster.

[0064] In a possible implementation manner, the time series is specifically as follows:

[0065]

[0066] Among them, represents the weighted quantity value type index at the i-th sampling, represents the weighted area value type index at the i-th sampling, i = 1, …, L, and L represents the length of the time series.

[0067] It should be noted that the establishment of the time series provides a basis for subsequent stationarity discrimination and trend extraction, enabling the evaluation method to dynamically respond to disaster changes, thereby achieving more efficient and accurate disaster response and resource allocation. This dynamic analysis ability significantly improves the scientificity and adaptability of the method.

[0068] S7: Conduct stationarity discrimination on the time series using different discrimination methods, where the discrimination methods include the autocorrelation plot discrimination method and the ADF test discrimination method.

[0069] Among them, the stationarity discrimination refers to the weighted quantization index data arranged in chronological order, which is used to describe the dynamic characteristics of the target changing over time. For example, the change trend of the number of rescued people. The autocorrelation graph discrimination method refers to calculating the autocorrelation coefficient of the time series to judge the correlation between it and its lag value, and analyzing whether the data has periodicity or trend. The ADF test discrimination method, namely the Augmented Dickey-Fuller Test, is used to judge whether there is a unit root (non-stationarity) in the time series, and provides statistical significance results through the t-test value and p-value.

[0070] It should be noted that the stationarity of the time series is discriminated through the autocorrelation graph and the ADF test, which provides a scientific basis for trend extraction and disaster assessment. The autocorrelation graph discrimination method can intuitively display the periodicity and trend characteristics of the time series, helping to identify the regular changes affected by disasters. The ADF test, on the other hand, quantifies the stationarity of the time series through strict statistical analysis to ensure that the discrimination results are statistically reliable.

[0071] In a possible implementation manner, the autocorrelation graph method in S7 is specifically as follows:

[0072] Calculate the autocorrelation coefficient of the time series:

[0073]

[0074] Among them, represents the autocorrelation coefficient at time k, X i represents the observed value at the i-th time in the time series, represents the mean value of the time series, |k| represents the absolute value of the lag value, and M represents the maximum range of the lag value.

[0075] Generate a plane two-dimensional coordinate vertical line graph according to the autocorrelation coefficient. Among them, the abscissa represents the delay, the ordinate represents the autocorrelation coefficient, and the length of the vertical line represents the magnitude of the autocorrelation coefficient.

[0076] Judge the stationarity of the time series according to the plane two-dimensional coordinate vertical line graph.

[0077] It should be noted that if the autocorrelation coefficient decays rapidly, the series has obvious stationarity, and its physical meaning is the target distribution at the disaster site. If the decay rate of the autocorrelation coefficient is slow, it indicates that the series has non-stationarity, and its physical meaning is that the target distribution at the disaster site may have a certain trend or periodic development direction.

[0078] Specifically, if the plane two-dimensional coordinate plumb line diagram shows obvious triangular symmetry, it usually indicates that the sequence has a certain trend. In a disaster scenario, it means that the number and area of the aerial photography targets are continuously increasing or decreasing, which can reflect that the number of rescue targets is continuously decreasing as the rescue progresses or that the area of the disaster (smoke or flood) is continuously expanding; if the plane two-dimensional coordinate plumb line diagram shows obvious triangular relationship (sine or cosine) fluctuation law, it usually indicates that the sequence has a certain periodicity.

[0079] The ADF test method in S7 is specifically as follows:

[0080] Extract the deterministic information of the time series.

[0081] Set the t-test value and p-value of the ADF test method.

[0082] According to the t-test value and p-value, judge the stationarity of the time series.

[0083] It should be noted that the ADF test method directly judges whether there is a unit root in the sequence. Among them, if the sequence is stationary, there is no unit root, otherwise there will be a unit root. Through the stationarity judgment, the time series with trends can be effectively screened out, laying a foundation for subsequent difference analysis and trend extraction, which makes the disaster assessment more scientific and accurate, and can better support the dynamic change analysis and decision-making in complex disaster scenarios.

[0084] S8: Based on the discrimination result, use the adaptive difference method to extract the time series with trend state.

[0085] Among them, the adaptive difference method refers to a method that dynamically adjusts the difference order according to the stationarity and trend state of the time series. The purpose is to eliminate non-stationarity through differencing while retaining the core trend information of the data. The trend state refers to the change characteristics of the time series after eliminating non-stationarity, such as the trend of growth, decline or periodic fluctuation, which is used to describe the dynamic evolution of the disaster.

[0086] It should be noted that by using the adaptive difference method to extract the time series with trend state, the problem that non-stationary data is difficult to directly analyze is effectively solved, and the scientificity and flexibility of trend extraction are significantly improved. The adaptive difference dynamically adjusts the order according to the data characteristics to ensure that the differenced time series not only eliminates noise and interference, but also completely retains the trend characteristics of the data.

[0087] In a possible implementation manner, the calculation formula of the adaptive difference method is specifically as follows:

[0088] ▽X i =X i -X i-1

[0089] ▽ 2 X i =▽X i -▽X i-1

[0090] wherein,▽X i represents the first-order difference value at the i-th moment, and▽ 2 X i represents the second-order difference value at the i-th moment, and X i-1 represents the observed value at the (i - 1)-th moment.

[0091] It should be noted that the adaptive differencing method avoids information loss or misjudgment that may be caused by excessive or insufficient differencing, making data processing more refined and intelligent. By extracting the trend state, it provides accurate input for subsequent disaster assessment and prediction, and significantly enhances the adaptability and practicality of the assessment method in complex scenarios.

[0092] S9: Evaluate the disaster according to the trend state.

[0093] It should be noted that by combining the trend state, autocorrelation plot, and ADF test results, a scientific and systematic assessment of the disaster is carried out, providing comprehensive data support for disaster emergency management. First, the extraction of the trend state enables the assessment to not be limited to static indicators, but to dynamically analyze the change process of the disaster, thereby accurately capturing the critical timing of rescue and resource allocation. Second, the dual discrimination method combining the autocorrelation plot and ADF test improves the reliability and accuracy of the assessment.

[0094] In a possible implementation manner, S9 specifically includes:

[0095] S901: According to the discrimination results of the autocorrelation plot discrimination method and the ADF test discrimination method, use the adaptive differencing method for non-stationary time series to extract the trend state of the time series, wherein the trend state includes: the decreasing rate trend of the number of rescue personnel, the increasing trend of the number of rescue vehicles, and the expanding trend of the target area of the disaster area.

[0096] S902: Based on the decreasing rate trend of the number of rescue personnel, evaluate the rescue effectiveness; based on the increasing trend of the number of rescue vehicles, evaluate the resource allocation efficiency; based on the expanding trend of the target area of the disaster area, evaluate the disaster severity.

[0097] It should be noted that judging the emergency response requirements in different disaster situations based on the trend state, for example, evaluating the rescue effectiveness when the number of rescue personnel is steadily decreasing, and optimizing resource allocation when the number of vehicles is increasing. This assessment method based on dynamic data makes disaster management more forward-looking and adaptable. It not only improves the efficiency of emergency decision-making but also provides an important basis for post-disaster recovery planning, greatly enhancing the practical value of the assessment method.

[0098] In a specific embodiment, taking a trapped vehicle as an example. First, direct summation weighted quantization is performed on the vehicle target. Figure 2 is the value of the weighted quantization of the vehicle target quantity. Among them, the visible value fluctuates continuously and shows fluctuations at the frame. After comparison, it is found that there is a jump in the middle of the video (at the 219th frame). Therefore, it shows that the value after weighted quantization can clearly display the distribution increase and decrease of the target and the abnormal events in the aerial photography.

[0099] Refer to the appendix of the specification Figure 2 , which shows a schematic diagram of the quantity value of a vehicle target provided by the present invention.

[0100] Figure 2 Among them, the abscissa represents the number of frames in the time series, usually corresponding to the time points of the aerial photography video or data acquisition. The ordinate represents the number of vehicle targets detected in each frame, which is the value output by the target detection algorithm. It can be seen from the figure that the number of vehicles detected in the initial stage is small, and the value remains at a low level with slight fluctuations. This reflects that in the early stage of the disaster, traffic is restricted or rescue vehicles have not arrived in large numbers. In the middle stage, the number of vehicles detected begins to increase rapidly, showing an obvious upward trend. This indicates that the rescue work has started to intensify and more vehicles have arrived at the disaster site. In the later stage, the number of vehicles detected reaches a peak and then gradually decreases, meaning that the rescue operation is gradually approaching the end and the number of vehicles on the site decreases.

[0101] After that, since the video is taken by a drone flying along the road, the time series of the classification weighted index is constructed according to the drone flight route, and time series analysis is carried out to evaluate the situation of the trapped vehicle.

[0102] Finally, the situation assessment of the trapped vehicle is carried out based on the time series analysis. Figure 3 is the sample autocorrelation function diagram of the time series. Among them, the autocorrelation coefficient drops to zero after the 65th frame, indicating that there is an inherent correlation within the sequence and it is not a simple stationary sequence.

[0103] Refer to the appendix of the specification Figure 3 , which shows a schematic diagram of the sample autocorrelation function of a time series provided by the present invention.

[0104] Figure 3Among them, the abscissa represents the lag term of the time series, that is, the distance between the current data point and the data point several steps ago. The ordinate represents the correlation between the current time series and the lagged series, with a range of [-1, 1]. An autocorrelation coefficient of 1 indicates a perfect positive correlation, 0 indicates no correlation, and -1 indicates a perfect negative correlation. In the initial stage (lag order 0 - 20), the autocorrelation coefficient is close to 1, indicating that the time series has a strong positive correlation at a small lag order, that is, the relationship between the current value and the recent past value is strong. In the middle stage (lag order 20 - 100), the autocorrelation coefficient drops rapidly and tends to zero or even becomes negative. This shows that as the lag order increases, the correlation between the current value of the time series and the more distant past value gradually weakens and even shows a negative correlation. In the later stage (after lag order 100), the autocorrelation coefficient tends to zero and fluctuates randomly within the confidence interval. This indicates that there is almost no correlation between the long-term lag terms of the time series and the current value, showing random characteristics.

[0105] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0106] In the embodiments of the present invention, based on aerial images, through the YOLO object detection algorithm, multiple objects in the aerial images are detected, and the object detection results are classified according to the object types to extract key information. Then, the pixel positions, pixel sizes, and detection confidence levels of each object in different object sets are determined. According to the object types and the corresponding detection confidence levels, the object weighted quantization index is calculated, and thus a unified quantitative standard is provided for evaluation. According to the object weighted quantization index, a time series is constructed, and the stationarity of the time series is discriminated by different discrimination methods. Based on the discrimination results, an adaptive difference method is used to extract the time series with a trend state. Finally, according to the trend state, the disaster is evaluated, reducing the cost and time of manual object detection, improving the reliability and accuracy of the detection results, enhancing the efficiency of disaster assessment, providing sufficient time for rescue work, and saving rescue time. The present invention is widely applicable to disaster scenarios such as earthquakes, fires, floods, rainstorms, frosts, and traffic accidents, and can be deployed and implemented in scenarios where object detection in aerial images can run, meeting diverse rescue needs.

[0107] Refer to the attached Figure 4 illustrates a schematic structural diagram of a disaster assessment system based on object detection in aerial images provided by the present invention.

[0108] The present invention also provides a disaster assessment system 20 based on object detection in aerial images, which is applied to the above-mentioned disaster assessment method based on object detection in aerial images, and includes:

[0109] A processor 201.

[0110] Memory 202 stores computer-readable instructions thereon. When the computer-readable instructions are executed by the processor 201, a disaster assessment method based on aerial image target detection as in the method embodiment is implemented.

[0111] The disaster assessment system 20 based on aerial image target detection provided by the present invention can execute the above-mentioned disaster assessment method based on aerial image target detection and achieve the same or similar technical effects. To avoid repetition, the present invention will not elaborate further.

[0112] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0113] In the embodiments of the present invention, based on aerial images, multiple targets are detected in the aerial images through the YOLO target detection algorithm, and the target detection results are classified according to the target types to extract key information. Then, the pixel positions, pixel sizes, and detection confidence levels of each target in different target sets are determined, and according to the target types and the corresponding detection confidence levels, a target weighted quantization index is calculated, thereby providing a unified quantitative standard for evaluation. According to the target weighted quantization index, a time series is constructed, and the stationarity of the time series is discriminated by different discrimination methods. Based on the discrimination results, an adaptive difference method is used to extract the time series with a trend state. Finally, according to the trend state, the disaster is evaluated, reducing the cost and time of manual target detection, improving the reliability and accuracy of the detection results, enhancing the efficiency of disaster assessment, providing sufficient time for rescue work, and saving rescue time. The present invention is widely applicable to disaster scenarios such as earthquakes, fires, floods, rainstorms, frosts, and traffic accidents, and can be deployed and implemented in any scenario where aerial image target detection can run, meeting diverse rescue needs.

[0114] It should be understood that the processor in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0115] It should also be understood that the memory in the embodiments of the present invention may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0116] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0117] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood with reference to the context.

[0118] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0119] It should be understood that in various embodiments of the present invention, the magnitudes of the serial numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0120] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0121] Those skilled in the art can clearly understand that for the convenience and conciseness of description, the specific working processes of the above-described devices, apparatuses, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be described herein again.

[0122] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some interfaces. The indirect coupling or communication connection of devices or units can be in an electrical, mechanical, or other form.

[0123] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0124] In addition, the functional units in various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit.

[0125] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: various media such as USB flash drives, external hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0126] An embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the disaster assessment method based on aerial image target detection as in the method embodiment.

[0127] The computer-readable storage medium provided by the present invention can implement the steps and effects of the disaster assessment method based on aerial image target detection in the above method embodiment. To avoid repetition, the present invention will not elaborate further.

[0128] The beneficial effects brought by the technical solution provided by the embodiments of the present invention at least include:

[0129] In the embodiments of the present invention, based on aerial images, through the YOLO target detection algorithm, multiple targets in the aerial images are detected, and the target detection results are classified according to the target types, and key information is extracted. Then, the pixel positions, pixel sizes, and detection confidence levels of each target in different target sets are determined. According to the target type and the corresponding detection confidence level, a target weighted quantization index is calculated, thereby providing a unified quantitative standard for evaluation. According to the target weighted quantization index, a time series is constructed, and the stationarity of the time series is discriminated by different discrimination methods. Based on the discrimination results, an adaptive difference method is used to extract the time series with a trend state. Finally, according to the trend state, the disaster is evaluated, reducing the cost and time of manual target detection, improving the reliability and accuracy of the detection results, enhancing the efficiency of disaster assessment, providing sufficient time for rescue work, and saving rescue time. The present invention is widely applicable to disaster scenarios such as earthquakes, fires, floods, rainstorms, frosts, and traffic accidents, and can be deployed and implemented in any scenario where aerial image target detection can run, meeting diverse rescue needs.

[0130] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

[0131] The following points need to be explained:

[0132] (1) The accompanying drawings of the embodiments of the present invention only relate to the structures involved in the embodiments of the present invention, and other structures can refer to the general design.

[0133] (2) For clarity, in the accompanying drawings used to describe the embodiments of the present invention, the thickness of layers or regions is enlarged or reduced, that is, these drawings are not drawn to actual scale. It can be understood that when an element such as a layer, film, region or substrate is referred to as being "on" or "under" another element, the element can be "directly" on or under the other element or there can be intervening elements.

[0134] (3) Without conflict, the embodiments of the present invention and the features in the embodiments can be combined with each other to obtain new embodiments.

[0135] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. The protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A disaster assessment method based on aerial image target detection, characterized in that: include: S1: Acquire the aerial image to be detected; S2: Detect multiple targets in the aerial image using the YOLO target detection algorithm; S3: classify the target detection results according to the target type to obtain multiple target sets; S4: Determine the pixel position, pixel size and detection confidence of each target in different target sets; S5: Calculate the target weighted quantitative index according to the target type and the corresponding detection confidence; S6: constructing a time series according to the target weighted quantitative indicators; S7: performing stationarity determination on the time series using different determination methods, wherein the determination methods include an autocorrelation diagram determination method and an ADF test determination method; S8: Based on the discrimination results, the time series with trend state is extracted using adaptive difference method; S9: Evaluate the disaster according to the trend status.

2. The method for disaster assessment based on aerial image target detection according to claim 1, characterized in that: Said targets include people, vehicles, equipment, fireworks and flooding.

3. The disaster assessment method based on aerial image target detection according to claim 1, characterized in that: The S5 specifically includes: S501: When the target type is a countable target, a direct quantity summation method or a confidence weighted quantity summation method is selected to determine the weighted quantitative index; otherwise, proceed to step S502; S502: When the target type is an area target, a direct area summation method or a confidence weighted area summation method is selected to determine the weighted quantitative index.

4. The method for disaster assessment based on aerial image target detection according to claim 3 is characterized in that: The formulas of the direct quantity summation method and the confidence weighted quantity summation method are specifically as follows: Among them, X J Represents a weighted quantitative indicator, M represents the total number of countable targets, m = 1, ..., M, P m Represents the confidence of the mth countable target; The formulas of the direct area summation method and the confidence weighted area summation method are specifically: Among them, X Z represents a weighted area value indicator, N represents the total number of regional targets, S n Represents the area of ​​the nth regional target, n=1,…,N,P n Represents the confidence of the nth region target.

5. The method for disaster assessment based on aerial image target detection according to claim 1, characterized in that: The time series is specifically: in, Represents the weighted quantitative index at the i-th sampling time, It represents the weighted area value index at the i-th sampling time, i=1,…,L, and L represents the length of the time series.

6. The method for disaster assessment based on aerial image target detection according to claim 1, characterized in that: The autocorrelation diagram method in S7 is specifically: Calculate the autocorrelation coefficient of the time series: in, represents the autocorrelation coefficient at time k, X i represents the observed value at the i-th moment in the time series, represents the mean of the time series, |k| represents the absolute value of the lagged value, and M represents the maximum range of the lagged value; According to the autocorrelation coefficient, a two-dimensional coordinate suspension line diagram is generated, wherein the abscissa represents the delay, the ordinate represents the autocorrelation coefficient, and the length of the suspension line represents the magnitude of the autocorrelation coefficient; Judging the stationarity of the time series according to the two-dimensional coordinate hanging line graph; The ADF test method in S7 is specifically as follows: extracting deterministic information of the time series; Set the t-test value and p-value of the ADF test method; The stationarity of the time series is judged according to the t-test value and the p-value.

7. The method for disaster assessment based on aerial image target detection according to claim 1, characterized in that: The calculation formula of the adaptive difference method is specifically: in, represents the first-order difference value at the i-th moment, represents the second-order difference value at the i-th moment, X i-1 Represents the observation value at the i-1th moment.

8. The method for disaster assessment based on aerial image target detection according to claim 1, characterized in that: The S9 specifically includes: S901: according to the determination results of the autocorrelation diagram determination method and the ADF test determination method, an adaptive difference method is used for the non-stationary time series to extract the trend state of the time series, wherein the trend state includes: the rate trend of the reduction of the number of rescuers, the growth trend of the number of rescue vehicles, and the expansion trend of the target area of ​​the disaster area; S902: Based on the rate trend of the reduction in the number of rescued people, the rescue effectiveness is evaluated; based on the growth trend of the number of vehicles seeking help, the resource scheduling efficiency is evaluated; based on the expansion trend of the target area of ​​the disaster area, the disaster extent is evaluated.

9. A disaster assessment system based on aerial image target detection, characterized in that: include: processor; A memory having computer-readable instructions stored thereon, wherein when the computer-readable instructions are executed by the processor, the disaster assessment method based on aerial image target detection as described in any one of claims 1 to 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the disaster assessment method based on aerial image target detection as described in any one of claims 1 to 8 is implemented.

Images