Earth surface anomaly instant remote sensing detection method, detection system and early warning system
The method constructs a priori knowledge base from long-term data to enhance satellite-based real-time detection of surface anomalies, improving accuracy and adaptability across various types of anomalies, facilitating rapid and automated emergency responses.
Patent Information
- Application Number
- CN202310663062.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-06-06
- Publication Date
- 2024-12-06
AI Technical Summary
Current satellite remote sensing detection methods for surface anomalies are inadequate in timeliness, accuracy, and adaptability, often missing or misidentifying anomalies due to complex processes, low automation, and lack of universal applicability across different types of anomalies.
A method utilizing long-term surface remote sensing feature information to construct a priori knowledge base, identify potential anomalies through real-time data comparison, and calculate anomaly intensity, incorporating vegetation, water body, ice, and exposed area indicators, enabling on-board detection and early warning.
Enhances anomaly detection accuracy and adaptability, allowing real-time identification and quantification of surface anomalies with reduced response times and minimal human intervention, supporting timely emergency responses.
Smart Images

Figure CN119091308A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of real-time remote sensing services, and in particular to a real-time remote sensing detection method, a detection system and an early warning system for surface anomalies. Background Art
[0002] In recent years, with the intensification of global climate change and the increasing number of human activities, various types of sudden surface abnormal events (such as natural disasters, environmental pollution, ecological damage, safety accidents, illegal development, etc.) have occurred frequently, and have shown the characteristics of wide distribution, high frequency, rapid evolution, and heavy economic losses, which have seriously affected the normal evolution of the natural environment and human well-being, and restricted the healthy and sustainable development of human society and economy. Therefore, the real-time monitoring and early warning of various surface abnormal events have become the top priority for ensuring ecological security, human health, and high-quality and stable development of social economy. It is urgent to build a full-process automated real-time detection method system for all surface abnormal events.
[0003] Research shows that satellite remote sensing is the most effective method for real-time detection of large-scale surface anomalies. However, due to the suddenness, diversity, randomness, concealment and complexity of surface anomalies, a large number of surface anomalies are often not discovered during the actual detection process, especially not detected and warned in the first time. The timeliness and accuracy of satellite detection are far from meeting the real-time remote sensing service needs characterized by minute-level response in disaster relief, accident disposal, pollution control, supervision and law enforcement. The main reasons for the above phenomenon are as follows:
[0004] First, currently, satellite remote sensing warnings of surface anomalies usually have to go through a relatively long process from mission planning, satellite operation and control, satellite-to-ground data transmission, to monitoring information acquisition and early warning knowledge extraction. For example, the natural resources department’s remote sensing emergency monitoring first needs to formulate a multi-satellite collaborative shooting plan, and then obtain satellite images through the ground satellite data center. After processing by the application system, relevant surface anomaly remote sensing information is extracted. It takes an average of 1-3 days from remote sensing image acquisition to extracting early warning knowledge such as the location, range, and intensity of the target event. Therefore, from satellite images to real-time detection and real-time early warning of surface anomalies, there are problems such as long service chains, many links, complicated processes, low intelligence levels, and slow responses, which are far from meeting the actual needs of emergency response.
[0005] Second, almost all current studies on surface anomaly detection only focus on certain specific surface anomaly events, such as surface thermal anomalies (Rembold et al., 2019), floods (DeVries et al., 2020), wildfires (Zhang et al., 2021), earthquakes (Liu et al., 2020), forest disturbances (Tang et al., 2019), etc. There is a lack of remote sensing detection systems that can be applied to all types of surface anomalies, and it is impossible to effectively detect all surface anomalies at the same time, resulting in a large number of surface anomalies being ignored. If only the existing detection methods of different surface anomaly events are combined and superimposed, the complexity of the entire surface anomaly detection method will be greatly increased, and the computational efficiency will be seriously reduced. At the same time, due to the logical conflicts and application scenario limitations between different surface anomaly event detection methods, surface anomalies in some areas will be misdetected or missed, reducing the accuracy of surface anomaly identification and failing to meet the needs of real-time remote sensing services.
[0006] Third, most of the current surface anomaly remote sensing detection methods generally face problems such as low automation level, complex model structure, and low accuracy in some scenarios, making it difficult to meet the high timeliness requirements of real-time remote sensing services. At present, the surface anomaly remote sensing detection methods can be roughly divided into manual visual interpretation, threshold segmentation, superposition and multi-model integration, and machine learning. The manual visual interpretation method is relatively simple, but it often requires a lot of manual intervention, low automation, and long calculation cycle. The threshold segmentation method mainly recognizes surface anomaly areas by segmenting remote sensing indices suitable for surface anomaly evaluation. However, due to the complexity and diversity of surface anomaly types, some surface anomaly events are similar to some normal landscapes, which can easily lead to low accuracy of surface anomaly detection results in some scenarios. Although the superposition and multi-model integration methods have high accuracy results, their model structures are relatively complex and often require a large amount of ground measurement or other auxiliary data support. Compared with all the previous methods, the machine learning method has a higher degree of intelligence. However, it not only faces the same problems as the threshold segmentation method, but also faces problems such as massive sample selection and model generalization, which will increase the difficulty of real-time and accurate detection of surface anomalies to a certain extent. Summary of the invention
[0007] (I) Purpose of the invention
[0008] The purpose of the present invention is to provide a method, a detection system and an early warning system for real-time remote sensing detection of surface anomalies that solve the problem that the effectiveness of traditional satellite remote sensing detection seriously lags behind the actual needs of surface anomaly monitoring and emergency disposal, and the problem that it cannot be applied to all different types of surface anomalies.
[0009] (II) Technical solution
[0010] In order to solve the above problems, the present invention provides a method for real-time remote sensing detection of surface anomalies, comprising:
[0011] Acquiring first long-term surface remote sensing feature information to form a first long-term surface remote sensing feature data set;
[0012] Constructing a priori knowledge base of surface remote sensing features according to the first long-term surface remote sensing feature data set;
[0013] Acquire real-time remote sensing data on the satellite, calculate surface remote sensing feature information, and identify potential surface anomaly areas based on current surface remote sensing feature information;
[0014] Identifying the surface abnormal area according to the surface abnormal potential area and the surface remote sensing feature prior knowledge base;
[0015] According to the identified surface anomaly area, the surface anomaly intensity of the surface anomaly area is calculated.
[0016] Preferably, the construction of the prior knowledge base of surface remote sensing features includes: respectively calculating and obtaining threshold values of various types of surface abnormal remote sensing feature indicators, and determining whether there are abnormal points or discrete values in the first long-term surface remote sensing feature data set:
[0017] max=Q3+1.5*(Q3-Q1)
[0018] min=Q1-1.5*(Q3-Q1)
[0019] Among them, max represents the maximum value of the threshold of the surface abnormal remote sensing characteristic index, min represents the minimum value of the threshold of the surface abnormal remote sensing characteristic index, all the values in the first long time series surface remote sensing characteristic data set are arranged from small to large, and divided into the first range, the second range, the third range and the fourth range, Q1 represents the largest value in the first range; Q3 represents the largest value in the third range;
[0020] According to the various types of surface abnormal remote sensing characteristic index thresholds and the first long-term surface remote sensing characteristic data set, outliers are eliminated to obtain a second long-term surface remote sensing characteristic data set;
[0021] The values in the second long-term surface remote sensing characteristic data set are all greater than or equal to the minimum value of the surface abnormal remote sensing characteristic index threshold, and less than or equal to the maximum value of the surface abnormal remote sensing characteristic index threshold;
[0022] The various types of surface abnormal remote sensing characteristic indicators include vegetation characteristic indicators, water body characteristic indicators, ice and snow characteristic indicators and exposed area characteristic indicators.
[0023] Preferably, the construction of the prior knowledge base of surface remote sensing features further comprises: calculating the mean and standard deviation of the surface abnormal remote sensing feature indexes based on the second long-time series surface remote sensing feature data set:
[0024]
[0025]
[0026] Among them, f mean (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The average value of remote sensing features, v(x i ,y i ,d k ,t n ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The tth of the remote sensing features n A value, n represents the number of points in the prior knowledge base of surface remote sensing features located at (x i ,y i ) is the pixel at the dth k The number of values in a remote sensing feature, f sd (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The standard deviation of a remote sensing feature.
[0027] Preferably, identifying potential areas of surface anomaly includes:
[0028] It is determined whether the vegetation characteristic index in the current surface remote sensing characteristic information meets the first standard. If not, it is identified as a potential surface abnormality area.
[0029] Preferably, identifying the surface abnormal area includes determining whether the surface abnormal potential area meets the second standard, and if so, determining it as the surface abnormal area;
[0030] Wherein, the second standard includes:
[0031] |F(x i ,y i , d k )-f menn(x i ,y i , d k )|>15*f sd (x i ,y i , d k )
[0032] (F(x i ,y i , d NDVI )>f mean (x i ,y i , d NDVI ))&(f mean (x i ,y i , d NDWI )<0)
[0033] (F(x i ,y i ,d NDWI )>f mean (x i ,y i , d NDWI ))&(f mean (x i ,y i , d NDWI )>=0)
[0034] Among them, F(x i ,y i ,d k ) indicates real-time acquisition of the remote sensing image at (x i ,y i ) is the pixel at the dth k The value of the remote sensing feature, F(x i ,y i ,d NDVI ) represents the vegetation characteristic index data obtained based on real-time remote sensing image inversion located at (x i ,y i ), F(x i ,y i ,d NDWI ) represents the water body characteristic index data obtained based on real-time remote sensing image inversion, located at (x i ,y i ), f mean (x i ,y i ,d NDVI ) represents the vegetation characteristic index in the prior knowledge base of surface remote sensing characteristics (x i ,y i) is the average value of the pixels at mean (x i ,y i ,d NDWI ) represents the water body characteristic index in the prior knowledge base of surface remote sensing characteristics (x i ,y i ) is the average value of the pixels at .
[0035] Preferably, calculating the surface anomaly intensity of the surface anomaly area comprises:
[0036] s1(x i ,y i )=(|F(x i ,y i , d max )-f meann (x i ,y i , d max )|) / f sd (x i ,y i , d max )
[0037] Among them, s1(x i ,y i ) represents the position at (x i ,y i ) is the surface anomaly intensity of the pixel at i ,y i ,d max ) represents the target image at (x i ,y i ) is the value of the remote sensing feature that changes the most after the surface anomaly occurs, f mean (x i ,y i ,d max ) represents the prior knowledge base located at (x i ,y i ) is the average value of the remote sensing feature that changes the most after the surface anomaly occurs, f sd (x i ,y i ,d max ) represents the prior knowledge base located at (x i ,y i ) is the standard deviation of the remote sensing feature that changes the most after the surface anomaly occurs.
[0038] Preferably, the calculating the surface anomaly intensity of the surface anomaly area further comprises:
[0039]
[0040] Among them, S2 represents the intensity of surface anomalies at the patch scale, aera plaque Represents the area of independent patches of surface anomalies, in square meters.
[0041] Preferably, the calculating of the surface anomaly intensity of the surface anomaly area further comprises:
[0042] S=a*s1+b*s2
[0043] Among them, S is the surface anomaly intensity value, S1 represents the surface anomaly intensity at the pixel scale, S2 represents the surface anomaly intensity at the patch scale, and a and b are the weight coefficients of the surface anomaly intensity at the pixel and patch scales, respectively.
[0044] Preferably, a real-time remote sensing detection system for surface anomalies comprises:
[0045] Acquisition module: acquiring first long-term surface remote sensing feature information to form a first long-term surface remote sensing feature data set;
[0046] Construction module: constructing a priori knowledge base of surface remote sensing features based on the first long-term surface remote sensing feature data set;
[0047] Identification module: obtains real-time remote sensing data on the satellite, calculates surface remote sensing feature information, identifies potential surface anomaly areas based on current surface remote sensing feature information; identifies surface anomaly areas based on the potential surface anomaly areas and a priori knowledge base of surface remote sensing features;
[0048] Calculation module: Calculate the surface anomaly intensity of the surface anomaly area based on the identified surface anomaly area.
[0049] Preferably, a real-time remote sensing early warning system for surface anomalies comprises a satellite terminal and a local terminal, wherein the satellite terminal is arranged on an in-orbit satellite and the local terminal is arranged on the surface;
[0050] The satellite end calculates the surface anomaly detection result and the surface anomaly intensity data by using the detection method as described in any one of the above items, and transmits the surface anomaly detection result and the surface anomaly intensity to the local end;
[0051] The local end receives the surface anomaly detection result and the surface anomaly intensity data, and issues a surface anomaly warning according to the surface anomaly detection result and the surface anomaly intensity data.
[0052] (III) Beneficial effects
[0053] The above technical solution of the present invention has the following beneficial technical effects:
[0054] The present invention utilizes long-term surface remote sensing feature information to construct a priori knowledge base of surface remote sensing features, thereby effectively capturing the normal surface state; performs two surface abnormal area identifications based on surface remote sensing feature information calculated from real-time remote sensing data and a priori knowledge base of surface remote sensing features, with high identification accuracy; calculates the intensity of the surface abnormal area, thereby facilitating corresponding emergency treatment based on the intensity, and has strong adaptability and is conducive to promotion; the present invention can be used as an on-board real-time remote sensing detection technology system for surface anomalies, and can provide important support for real-time detection and early warning of surface anomalies; the entire method can achieve on-board storage and on-orbit calculation, and can realize real-time on-orbit detection of surface anomalies. BRIEF DESCRIPTION OF THE DRAWINGS
[0055] Figure 1 is an overall flow chart of an embodiment of the present invention;
[0056] Figure 2 is an overall conceptual diagram of an embodiment of the present invention;
[0057] Figure 3 is a flow chart of the identification process of another embodiment of the present invention;
[0058] Figure 4 It is a spatial distribution map of the surface anomaly area and the surface anomaly intensity in the experimental area in one embodiment of the present invention. DETAILED DESCRIPTION
[0059] In order to make the purpose, technical scheme and advantages of the present invention clearer, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are only exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the description of well-known structures and technologies is omitted to avoid unnecessary confusion of the concept of the present invention.
[0060] Obviously, the described embodiments are only some embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0061] In the description of the present invention, it should be noted that the terms “first”, “second” and “third” are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.
[0062] Embodiment 1
[0063] A method for real-time remote sensing detection of surface anomalies. Figure 1 The overall process of an embodiment of the present invention is shown as follows: Figure 1 As shown, including:
[0064] Acquire the first long-term surface remote sensing characteristic information to form the first long-term surface remote sensing characteristic data set; the specific method and content of acquiring the first long-term surface remote sensing characteristic information are not limited here. Optionally, in this embodiment, the acquisition method includes using one or more satellites such as hyperspectral, microwave, infrared or high-resolution to obtain information from comprehensive observations of the target area. The specific time of acquiring the first long-term surface remote sensing characteristic information is also not limited here. Optionally, several years of remote sensing images are the first long-term surface remote sensing characteristic information. Here, the imaging time can be selected as the specific imaging days as close to the target area as possible. Further, in this embodiment, the remote sensing images of the target area in the first 4 years are selected as the first long-term surface remote sensing characteristic information. Optionally, in this embodiment, the first long-term surface remote sensing characteristic information is preprocessed; the specific content of the preprocessing is not limited here, and it can be cloud removal, correction or alignment processing of the remote sensing image. In this embodiment, optionally, the preprocessing includes screening the first long-term surface remote sensing characteristic information and eliminating the first long-term surface remote sensing characteristic information with a cloud content exceeding 60%.
[0065] The first long-term surface remote sensing feature data set is constituted; the specific content of the first long-term surface remote sensing feature data set is not limited here. Optionally, it is classified according to the surface remote sensing features. From the perspective of land cover type, the global surface can be roughly divided into four main land cover types: vegetation covered areas, water bodies, ice and snow, and bare areas without vegetation cover (such as bare land, impermeable surface, etc.). In this embodiment, the first long-term surface remote sensing feature data set is optionally constituted, including: vegetation remote sensing features, water body remote sensing features, ice and snow remote sensing features, and bare area remote sensing features; and each land cover type has remote sensing features that can describe its characteristics in detail. When a surface anomaly occurs, it often causes different degrees of damage to the original land cover type, and this damage is reflected in the remote sensing features that can describe different land cover characteristics, which is manifested as an obvious change in the remote sensing feature value. In order to describe the first long-term surface remote sensing feature data set, a surface anomaly remote sensing feature index system is constructed: the specific content of the index system is not limited here. Optionally, in this embodiment, constructing the surface anomaly remote sensing feature index system includes:
[0066] Vegetation characteristic indicators:
[0067] Water body characteristic indicators:
[0068] Ice and snow characteristic indicators:
[0069] Exposed area characteristic indicators:
[0070] Among them: NIR represents the near infrared band, RED represents the red wave, GREEN represents the green wave, BLUE represents the blue wave, and SWIR1 represents the shortwave infrared band.
[0071] The first long-time series surface remote sensing feature information is obtained and calculated based on the surface anomaly remote sensing feature indicator system to obtain the first long-time series surface remote sensing feature data set, where the first long-time series surface remote sensing feature data set is the above four indicator sets.
[0072] Based on the first long-term surface remote sensing feature data set, a priori knowledge base of surface remote sensing features is constructed; in the surface abnormal remote sensing feature index system, the surface abnormal area can be determined by simply finding the area where the remote sensing features have changed significantly. Prior to this, the first long-term surface remote sensing feature data set is screened and processed, and the thresholds of various types of surface abnormal remote sensing feature indexes are calculated separately:
[0073] max=Q3+1.5*(Q3-Q1)
[0074] min=Q1-1.5*(Q3-Q1)
[0075] Among them, max represents the maximum value of the threshold of the surface abnormal remote sensing characteristic index, min represents the minimum value of the threshold of the surface abnormal remote sensing characteristic index, all the values in the first long time series surface remote sensing characteristic data set are arranged from small to large, and divided into the first range, the second range, the third range and the fourth range, Q1 represents the largest value in the first range; Q3 represents the largest value in the third range;
[0076] According to the threshold values of various surface anomaly remote sensing characteristic indicators and the first long-time surface remote sensing characteristic data set, a second long-time surface remote sensing characteristic data set is obtained; the values in the second long-time surface remote sensing characteristic data set satisfy the minimum value greater than or equal to the threshold value of the surface anomaly remote sensing characteristic indicator and are less than or equal to the maximum value of the threshold value of the surface anomaly remote sensing characteristic indicator; the various surface anomaly remote sensing characteristic indicators include vegetation characteristic indicators, water body characteristic indicators, ice and snow characteristic indicators and exposed area characteristic indicators.
[0077] Here, all the values in the first long-time series surface remote sensing feature data set are arranged from small to large and divided into the first range, the second range, the third range and the fourth range. They can be equally divided into four ranges or unequally divided into four ranges. Optionally, in this embodiment, they are equally divided into four ranges, Q1 represents the 25th% number after all the values in the set are arranged from small to large; Q3 refers to the 75th% number after all the values in the set are arranged from small to large.
[0078] After the outliers are removed, the mean and standard deviation are calculated feature by feature and pixel by pixel. Based on the second long-time series surface remote sensing feature data set, the mean and standard deviation of the surface abnormal remote sensing feature indicators are calculated respectively:
[0079]
[0080]
[0081] Among them, f mean (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The average value of remote sensing features, v(x i ,y i ,d k ,t n ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The tth of the remote sensing features n A value, n represents the number of points in the prior knowledge base of surface remote sensing features located at (x i ,y i ) is the pixel at the dth k The number of values in a remote sensing feature, f sd (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The standard deviation of a remote sensing feature.
[0082] By integrating and storing the multi-year average and standard deviation data of these four main remote sensing features, we can obtain a priori knowledge base of surface remote sensing features, which characterizes the normal state of each remote sensing feature of each pixel point and its reasonable fluctuation range. The spatial scope of the construction is not limited here. Optionally, a comprehensive priori knowledge base that covers the globe in space and is continuous in time can be constructed and stored on the satellite. In this way, any place where surface anomaly detection is needed can realize instant detection of surface anomalies by simply calling the priori knowledge base.
[0083] Acquire real-time remote sensing data on the satellite, calculate surface remote sensing feature information, and identify potential areas of surface anomalies; the spatial position of the computing subject is not limited here, and the calculation can be performed by an in-orbit satellite or a local end on the surface. In this embodiment, the calculation can be performed by an in-orbit satellite; the acquisition method is not limited here, and the optional acquisition method includes information obtained by comprehensive observation using one or more satellites such as hyperspectral, microwave, infrared or high-resolution satellites.
[0084] Identify potential areas of surface anomalies; the specific content of obtaining potential areas of surface anomalies is not limited here. Figure 3 The identification process flow chart is shown as Figure 3 As shown, optionally, in this embodiment, according to the meaning of surface anomaly, surface anomaly often brings different degrees of surface damage after it occurs, breaking the relatively stable surface and the fluctuating equilibrium state that conforms to the law of natural evolution. That is, the surface anomaly brings a trend towards "bad". According to the real-time acquisition of remote sensing feature information, it is judged whether the vegetation feature index in the current surface remote sensing feature information meets the first standard. If not, it is identified as a potential area of surface anomaly. The specific content of the first standard is not limited here. Optionally, in this embodiment, the first standard is whether the NDVI value is greater than 0.4. If it is greater than 0.4, it is divided into an area without surface anomaly. If it is less than or equal to 0.4, it is preliminarily divided into an area with potential surface anomaly. When the vegetation growth condition of a certain place is good, the growth is lush, and the vegetation coverage is high, it does not meet any of the characteristics of the surface anomaly at all, and it must be an area with obvious no surface anomaly. When the NDVI is greater than 0.4, the vegetation growth condition of the area is often good, the vegetation is lush and the coverage is high. Therefore, the NDVI is obtained by real-time remote sensing image inversion, and then the NDVI value is judged to be greater than 0.4. If it is greater than 0.4, it is classified as an area without surface anomalies, and if it is less than or equal to 0.4, it is identified as a potential area of surface anomalies.
[0085] According to the prior knowledge base of surface abnormal potential areas and surface remote sensing features, the identification of surface abnormal areas includes judging whether the surface abnormal potential areas meet the second standard. If they meet the second standard, they are judged as surface abnormal areas. The average values and standard deviations of four remote sensing features, such as NDVI, NDWI, NDSI, and BSI, are stored in the prior knowledge base of surface remote sensing features. The NDVI, NDWI, NDSI, and BSI data of real-time remote sensing images are obtained by inversion (using F(x i ,y i ,d k)). Then, the pixel-by-pixel comparison is performed with the average value data in the prior knowledge base of the surface remote sensing features. If the absolute value of the difference of one or more remote sensing features is greater than the set fixed threshold, it can be preliminarily determined as a surface anomaly. In this embodiment, the fixed threshold is set to 15 times the standard deviation, which has the best effect.
[0086] When there is a significant difference between the remote sensing feature values of the real-time remote sensing image and the remote sensing feature values in the prior knowledge base of surface remote sensing features, it may be a surface anomaly or a land restoration or surface ecological quality improvement phenomenon, which does not belong to the surface abnormality phenomenon. Therefore, some restrictions are added to eliminate the land restoration or surface ecological quality improvement phenomenon. The land restoration or surface ecological quality improvement phenomenon specifically includes: (1) The vegetation growth condition in non-water areas improves and the vegetation coverage increases. That is, the NDVI value of the area with relatively small NDWI in the prior knowledge base of surface remote sensing features increases; (2) The water purity in the water area increases and the water level rises, that is, the NDWI value of the area with relatively high NDWI in the prior knowledge base of surface remote sensing features continues to rise. The second standard is expressed by the formula:
[0087] |F(x i ,y i , d k )-f mean (x i ,y i , d k )|>15*f sd (x i ,y i , d k )
[0088] (F(x i ,y i , d NDVI )>f mean (x i ,y i , d NDVI ))&(f mean (x i ,y i , d NDWI )<0)
[0089] (F*x i ,y i , d NDWI )>f mean (x i ,y i , d NDWI ))&(f mean (x i ,y i , d NDWI)>=0)
[0090] Among them, F(x i ,y i ,d k ) indicates real-time acquisition of the remote sensing image at (x i ,y i ) is the pixel at the dth k The value of the remote sensing feature, F(x i ,y i ,d NDVI ) represents the NDVI data obtained based on real-time remote sensing image inversion, located at (x i ,y i ), F(x i ,y i ,d NDWI ) represents the NDWI data obtained based on the real-time acquisition of remote sensing images inversion located at (x i ,y i ), f mean (x i ,y i ,d NDVI ) represents the (x i ,y i ) is the average value of the pixels at mean (x i ,y i ,d NDWI ) represents the (x i ,y i ). In this embodiment, when the NDWI threshold is set to 0, water bodies and non-water body areas can be better distinguished, thereby promoting the effective identification of land restoration or surface ecological quality improvement phenomena.
[0091] According to the identified surface anomaly area, the surface anomaly intensity of the surface anomaly area is calculated; after the surface anomaly is identified, the surface anomaly intensity of different areas is further described quantitatively. From the pixel scale, the surface anomaly intensity is expressed as the degree of difference of the remote sensing characteristics before and after the surface anomaly occurs. That is, the greater the degree of difference, the greater the surface anomaly intensity. Therefore, the surface anomaly intensity of a single pixel point is determined by judging the ratio of the change value of the remote sensing feature to the standard deviation of the remote sensing feature in the prior knowledge base of the surface remote sensing feature. That is, the greater the ratio of the change value to the standard deviation, the greater the surface anomaly intensity. In addition, the larger the distribution range or the influence range of the surface anomaly event, the higher the surface anomaly intensity is judged to be. Therefore, from the patch scale, the surface anomaly intensity is closely related to the patch area, that is, the larger the patch area, the greater the surface anomaly intensity.
[0092] Pixel-scale intensity calculations include:
[0093] s1(x i ,y i )=(|F(x i ,y i , d max )-f mean (x i ,y i , d max )|) / f sd (x i ,y i , d max )
[0094] Among them, s1(x i ,y i ) represents the position at (x i ,y i ) is the surface anomaly intensity of the pixel at i ,y i ,d max ) represents the real-time acquisition of the remote sensing image at (x i ,y i ) is the value of the remote sensing feature that changes the most after the surface anomaly occurs, f mean (x i ,y i ,d max ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the average value of the remote sensing feature that changes the most after the surface anomaly occurs, f sd (x i ,y i ,d max ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the standard deviation of the remote sensing feature that changes the most after the surface anomaly occurs.
[0095] Patch-scale intensity calculations include:
[0096]
[0097] Among them, S2 represents the intensity of surface anomalies at the patch scale, aera plaque Represents the area of independent patches of surface anomalies, in square meters.
[0098] After completing the evaluation of the surface anomaly intensity at the pixel and patch scales, the two are weighted averaged to obtain the final surface anomaly intensity estimate. However, due to the huge difference between the surface anomaly intensity values at the pixel scale and the patch scale, they are not at the same order of magnitude. The two are uniformly standardized. In this embodiment, the surface anomaly intensity values at most pixel scales are within 100, so the surface anomaly intensity values at the pixel scale are retained; in addition, a complete surface anomaly event must be composed of independent surface anomaly patches. Although there are differences in the areas of these independent patches, in this embodiment, in order to match the surface anomaly intensity value at the patch scale with the pixel scale intensity value, the patch area is divided by 250 square kilometers and then multiplied by 100. But it is undeniable that there are still a small number of surface anomaly events that involve a very large scope, such as extremely severe snowstorms, extremely severe sandstorms, extremely severe floods, extremely severe fires, etc., and the area of a single patch will also be very likely to exceed the set threshold (250 square kilometers). But in general, most of the above situations are low-probability events. And once such an event occurs, people in real life tend to judge it as a high-intensity surface anomaly event. Similarly, at the pixel scale, there may be some pixels whose surface anomaly intensity values exceed 100. But this does not affect the subsequent judgment of the surface anomaly intensity, which will only be used as one of the proofs that it is an ultra-high-intensity surface anomaly event. After unifying the surface anomaly intensity at the pixel and patch scales to the same order of magnitude as much as possible, the surface anomaly intensity at the pixel and patch scales is weighted. The specific content of the weight is not limited here. Optionally, in this embodiment, the same weight is assigned, and then the final surface anomaly intensity estimate is obtained by comprehensive calculation.
[0099] Calculating the intensity of the surface abnormal area also includes:
[0100] S=a*s1+b*s2
[0101] Wherein, S1 represents the pixel-scale surface anomaly intensity, S2 represents the patch-scale surface anomaly intensity, S is the surface anomaly intensity value, and a and b are weight coefficients of the pixel-scale and patch-scale surface anomaly intensities, respectively. The weight coefficients of the pixel-scale and patch-scale surface anomaly intensities are not limited here, and optionally, the weight coefficients of the two are the same in this embodiment.
[0102] In this embodiment, Figure 4 The spatial distribution of the surface anomaly area and the surface anomaly intensity in the experimental area is shown. Figure 4As shown in the figure, four remote sensing images were randomly selected for surface anomaly identification and accuracy verification, including: building a priori knowledge base for the images of the experimental area, obtaining the mean and standard deviation of four remote sensing characteristics such as NDVI, NDWI, NDSI, and BSI; performing data preprocessing such as cloud removal on the real-time remote sensing images, and preliminarily determining potential surface anomaly areas; completing surface anomaly identification by setting reasonable thresholds; and comprehensively considering the pixel scale and patch scale to calculate the estimated value of the surface anomaly intensity.
[0103] like Figure 4 As shown in Figure 1, the real-time remote sensing images and surface anomaly detection results of experimental area 1 are shown in Figure 1. Figure 4 (a) and (b) show that a serious flood disaster occurred in the region. In particular, the river in the west of the image burst its banks, causing a large amount of flood water to spread to both sides of the river and submerge a large amount of land. At the same time, a large amount of rainfall accumulated in the lowland farmland area in the east of the image. Figure 4 (b) It can be found that the surface anomaly areas obtained in this study are mainly concentrated in the eastern part of the experimental area in block shape. At the same time, there are also a small number of surface anomaly areas in the western part of the experimental area that are discretely distributed in strips or scattered points. This is basically consistent with the spatial distribution pattern of the visual interpretation results, and the similarity is high. In addition, 1,000 verification points were randomly and evenly arranged throughout the experimental area, and the accuracy evaluation was completed with the help of corresponding historical data and higher-resolution remote sensing images. The overall classification accuracy of the surface anomaly detection results in Experimental Area 1 is 92.20%.
[0104] like Figure 4 As shown in Figure 2, the real-time remote sensing images and surface anomaly detection results of experimental area 2 are as follows: Figure 4 (d) and (e) show that forest fires occurred in the eastern and southeastern parts of the region, leaving behind large areas of burnt land. Figure 4 (e) It can be found that the surface anomaly areas are mainly concentrated in the east and southeast in blocks. This is basically consistent with the spatial distribution pattern of the visual interpretation results. In order to quantitatively evaluate it, 900 verification points were randomly and evenly arranged in the entire experimental area. The overall classification accuracy of the surface anomaly detection results in experimental area 2 is 94.33%.
[0105] like Figure 4 As shown in Figure 2, the real-time remote sensing images and surface anomaly detection results of experimental area 3 are as follows: Figure 4 (g) and (h) show that deforestation and ecological land destruction occurred in the north and south of the region. Figure 4 (h) It can be found that the surface anomaly areas are mainly distributed in blocks in the north and scattered areas in the south. This is basically consistent with the spatial distribution pattern of the visual interpretation results. The overall classification accuracy of the surface anomaly detection results in experimental area 3 is 89.75%.
[0106] like Figure 4 As shown in Figure 2, the real-time remote sensing images and surface anomaly detection results of experimental area 4 are as follows: Figure 4 (j) and (k) show that there is no surface anomaly in this area. Figure 4 (k) It can be found that there is basically no surface anomaly in the results obtained by the present invention. Only a few small and scattered surface anomaly areas were detected in the northwest of the experimental area. The overall classification accuracy of the surface anomaly detection results of experimental area 4 is 97.38%.
[0107] The results of surface anomaly intensity in experimental areas 1-4 are as follows Figure 4 (c), (f), (i), and (l), where the color represents the stronger the intensity of the surface anomaly. It can be found that the high-intensity surface anomaly areas in experimental area 1 are mainly distributed on both sides of the original river channel in the west of the experimental area, while the intensity of the surface anomaly areas in the east of the experimental area is relatively small. This is because the heavy rainfall caused the river to burst its banks and quickly rushed to both banks. The closer the area is to the original river ancient channel, the more serious the disaster, and the stronger the intensity of the surface anomaly. The eastern area is more due to the relatively low terrain, which has accumulated some precipitation. However, its degree of disaster is far less than that of the river bursting its banks, and its surface anomaly intensity is also relatively low. In experimental area 2, the intensity of the surface anomaly area in the middle and east is higher than that of the surface anomaly areas in the north and south of the area. Through analysis, it is found that the key area of forest fires is in the middle and east of the experimental area. This area is most affected by the fire, and basically all the vegetation has been burned. The relatively small surface anomaly areas in the south and north of the area still survive a certain number of trees and vegetation, and are relatively less affected by the fire. In experimental area 3, high-intensity surface anomaly areas are mainly concentrated in the largest surface anomaly patches in the central and northern parts. It can be found that this area was previously a forest area with a high vegetation coverage. However, due to human logging, its land cover type has been directly converted from forest land to bare land. It is precisely because of the drastic changes in this area and its relatively large area that its surface anomaly intensity is relatively high. Since the surface anomaly areas detected in experimental area 4 are very few and not obvious, the surface anomaly intensity obtained is also relatively small.
[0108] This embodiment determines the main remote sensing features that can effectively capture the degree of damage caused by surface anomalies by analyzing the main land cover types on the surface, and forms a surface anomaly remote sensing feature indicator system that can be used for all types of surface anomalies. However, traditional methods for remote sensing detection of surface anomalies are mostly carried out under the premise of known types of surface anomaly events, and the remote sensing characteristic indicators constructed therefrom are more inclined to reflect the surface characteristics after the surface anomaly occurs, which means that these remote sensing characteristics are only applicable to the selected types of surface anomaly events. This embodiment takes surface anomaly receptors as the starting point, and determines the surface anomaly by capturing the change information of surface anomaly receptors pixel by pixel, breaking the limitation that the previous remote sensing detection of surface anomalies is limited to a single type of surface anomaly, and forming a more universal method for remote sensing instant detection of surface anomalies. After constructing the surface anomaly remote sensing characteristic indicator system, the present invention obtains the corresponding remote sensing characteristics based on the inversion of multi-year surface images, calculates the average value and standard deviation, and constructs a priori knowledge base of the surface stable state. Since multi-year remote sensing images are selected in the construction of the prior knowledge base, and the outliers therein are excluded, the prior knowledge base reflects the relatively stable surface and the fluctuating equilibrium state that conforms to the law of natural evolution, which provides an important basis for the subsequent instant and accurate judgment of surface anomalies.
[0109] Due to the complexity and diversity of surface anomaly types, the spectral characteristics, texture characteristics, and radiation characteristics of some surface abnormal events and some normal landforms are relatively similar, such as landslides, debris flows, and bare soil areas are relatively similar, and barrier lakes are relatively similar to normal water bodies. This also leads to the fact that it is difficult to better distinguish normal landforms from surface anomalies when detecting surface anomalies in traditional threshold segmentation methods and machine learning and deep learning methods that have gradually emerged in recent years. The comparative evaluation based on the prior knowledge base proposed in this embodiment can correspond the target image to the prior knowledge one by one, and it focuses more on the analysis of the variation in the same area on a time scale, which avoids the misclassification phenomenon caused by the similarity of remote sensing characteristics to a certain extent.
[0110] Traditional change monitoring is mostly based on the premise that surface anomalies are known to occur, and then new remote sensing images and historical images are searched to determine the surface anomaly area. However, this embodiment can realize surface anomaly detection based on real-time remote sensing images after completing the prior knowledge base and judgment rule settings. It already has the ability to autonomously diagnose surface anomalies on orbit, and can effectively and autonomously distinguish between surface anomalies and non-surface anomaly areas without human intervention. At the same time, it is no longer necessary to re-search for historical remote sensing images of the area where real-time remote sensing data is acquired, download the latest images and off-satellite anomaly detection. It only needs to execute the surface anomaly instant remote sensing detection algorithm on the real-time acquired remote sensing images to directly transmit the surface anomaly detection results to the local end, greatly reducing the surface anomaly detection cycle.
[0111] Embodiment 2
[0112] A real-time remote sensing detection system for surface anomalies, such as Figure 1 As shown, including:
[0113] Acquisition module: acquiring first long-term surface remote sensing feature information to form a first long-term surface remote sensing feature data set;
[0114] Construction module: constructing a priori knowledge base of surface remote sensing features according to the first long-time series surface remote sensing feature data set. The storage location of the priori knowledge base of surface remote sensing features is not limited here. Optionally, in this embodiment, the priori knowledge base of surface remote sensing features is stored on the satellite.
[0115] Identification module: obtains real-time remote sensing data on the satellite, calculates surface remote sensing feature information, identifies potential surface anomaly areas based on current surface remote sensing feature information; identifies surface anomaly areas based on the potential surface anomaly areas and a priori knowledge base of surface remote sensing features;
[0116] Calculation module: Calculate the surface anomaly intensity of the surface anomaly area based on the identified surface anomaly area. The spatial position of the calculation subject is not limited here. Optionally, the calculation is performed by an in-orbit satellite or a local surface terminal. In this embodiment, the calculation is optionally performed by an in-orbit satellite.
[0117] In this embodiment, the prior knowledge base of surface remote sensing features is stored on-board, which makes it possible to process the real-time remote sensing detection of surface anomalies on-orbit, and directly calculates the intensity of surface anomalies through satellites, which is highly efficient and can provide important support for real-time detection and early warning of surface anomalies.
[0118] Embodiment 3
[0119] A real-time remote sensing early warning system for surface anomalies. Figure 2 The overall concept diagram of an embodiment of the present invention is shown as follows: Figure 2 As shown, it includes a satellite terminal and a local terminal, the satellite terminal is set on an in-orbit satellite, and the local terminal is set on the surface of the earth; the specific content of the local terminal is not limited here, and it can be a handheld terminal, a computer terminal or a TV terminal, etc. In this embodiment, it can be a handheld terminal;
[0120] The satellite end calculates the surface anomaly detection result and the surface anomaly intensity data by using the detection method as described in any one of the above items, and transmits the surface anomaly detection result and the surface anomaly intensity to the local end;
[0121] The local end receives the surface anomaly detection result and the surface anomaly intensity data, and performs surface anomaly warning according to the surface anomaly detection result and the surface anomaly intensity data, wherein the surface anomaly detection result includes at least one of the calculated surface remote sensing feature information, the identified surface anomaly potential area, and the identified surface anomaly area.
[0122] This embodiment takes the surface anomaly receptors as the entry point, and based on the representative remote sensing characteristics of different land covers reflecting surface anomalies, further constructs a priori knowledge base that can effectively describe the normal surface state; by constructing the priori knowledge base and storing it on board, it provides the possibility for on-orbit processing of real-time remote sensing detection of surface anomalies, and directly calculates the intensity of surface anomalies through satellites and transmits it to the local end, so that the local end can directly issue early warnings based on the intensity of surface anomalies.
[0123] It should be understood that the above specific embodiments of the present invention are only used to illustrate or explain the principles of the present invention, and do not constitute a limitation of the present invention. Therefore, any modifications, equivalent substitutions, improvements, etc. made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modifications that fall within the scope and boundaries of the appended claims, or the equivalent forms of such scope and boundaries.
Claims
1. A method for real-time remote sensing detection of surface anomalies, characterized in that: include: Acquiring first long-term surface remote sensing feature information to form a first long-term surface remote sensing feature data set; Constructing a priori knowledge base of surface remote sensing features according to the first long-term surface remote sensing feature data set; Acquire real-time remote sensing data on the satellite, calculate surface remote sensing feature information, and identify potential abnormal areas on the surface; Identifying the surface abnormal area according to the surface abnormal potential area and the surface remote sensing feature prior knowledge base; According to the identified surface anomaly area, the surface anomaly intensity of the surface anomaly area is calculated.
2. The method for real-time remote sensing detection of surface anomalies according to claim 1, characterized in that: The construction of the prior knowledge base of surface remote sensing features includes: respectively calculating and obtaining the threshold values of various types of surface abnormal remote sensing feature indicators, and determining whether there are abnormal points or discrete values in the first long-term surface remote sensing feature data set: max=Q3+1.5*(Q3-Q1) min=Q1-1.5*(Q3-Q1) Among them, max represents the maximum value of the threshold of the surface abnormal remote sensing characteristic index, min represents the minimum value of the threshold of the surface abnormal remote sensing characteristic index, all the values in the first long time series surface remote sensing characteristic data set are arranged from small to large, and divided into the first range, the second range, the third range and the fourth range, Q1 represents the largest value in the first range; Q3 represents the largest value in the third range; According to the various types of surface abnormal remote sensing characteristic index thresholds and the first long-term surface remote sensing characteristic data set, the remote sensing characteristic abnormal values are eliminated to obtain the second long-term surface remote sensing characteristic data set; The values in the second long-term surface remote sensing characteristic data set are all greater than or equal to the minimum value of the surface abnormal remote sensing characteristic index threshold, and less than or equal to the maximum value of the surface abnormal remote sensing characteristic index threshold; The various types of surface abnormal remote sensing characteristic indicators include vegetation characteristic indicators, water body characteristic indicators, ice and snow characteristic indicators and bare surface characteristic indicators.
3. The method for real-time remote sensing detection of surface anomalies according to claim 2, characterized in that: The constructing of the prior knowledge base of surface remote sensing features further includes: calculating the mean and standard deviation of the surface abnormal remote sensing feature indexes based on the second long-time series surface remote sensing feature data set: Among them, f mean (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The average value of remote sensing features, v(x i ,y i ,d k ,t n ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The tth of the remote sensing features n A value, n represents the number of points in the prior knowledge base of surface remote sensing features located at (x i ,y i ) is the pixel at the dth k The number of values in a remote sensing feature, f sd (x i ,y i ,d k ) represents the surface remote sensing feature prior knowledge base located at (x i ,y i ) is the pixel at the dth k The standard deviation of a remote sensing feature.
4. The method for real-time remote sensing detection of surface anomalies according to claim 3, characterized in that: The identification of potential surface anomaly areas includes: It is determined whether the vegetation characteristic index in the surface remote sensing characteristic information calculated based on the acquired real-time remote sensing data on the satellite meets the first standard. If not, it is identified as a potential surface abnormality area.
5. The method for real-time remote sensing detection of surface anomalies according to claim 4, characterized in that: Identifying the surface abnormal area includes determining whether the surface abnormal potential area meets the second standard, and if so, determining it as the surface abnormal area; Wherein, the second standard includes: |F(x i ,y i ,d k )-f mean (x i ,y i ,d k )|>15*f sd (x i ,y i ,d k ) (F(x i ,y i ,d NDVI )>f mean (x i ,y i ,d NDVI ))&(f mean (x i ,y i ,d NDWI )<0) (F(x i ,y i ,d NDWI )>f mean (x i ,y i ,d NDWI ))&(f mean (x i ,y i ,d NDWI )>=0) Among them, F(x i ,y i ,d k ) indicates real-time acquisition of the remote sensing image at (x i ,y i ) is the pixel at the dth k The value of the remote sensing feature, F(x i ,y i ,d NDVI ) represents the vegetation characteristic index data obtained based on real-time remote sensing image inversion located at (x i ,y i ), F(x i ,y i ,d NDWI ) represents the water body feature data obtained based on real-time remote sensing image inversion located at (x i ,y i ), f mean (x i ,y i ,d NDVI ) represents the vegetation characteristic index in the prior knowledge base of surface remote sensing characteristics (x i ,y i ) is the average value of the pixels at mean (x i ,y i ,d NDWI ) represents the water body characteristic index in the prior knowledge base of surface remote sensing characteristics (x i ,y i ) is the average value of the pixels at .
6. The method for real-time remote sensing detection of surface anomalies according to claim 5, characterized in that: The calculation of the surface anomaly intensity of the surface anomaly area comprises: s1(x i ,y i )=(|F(x i ,y i ,d max )-f menn (x i ,y i ;,d max )|) / f sd (x i ,y i ,d max ) Among them, s1(x i ,y i ) represents the position at (x i ,y i ) is the surface anomaly intensity of the pixel at i ,y i ,d max ) represents the target image at (x i ,y i ) is the value of the remote sensing feature that changes the most after the surface anomaly occurs, f mean (x i ,y i ,d max ) represents the prior knowledge base located at (x i ,y i ) is the average value of the remote sensing feature that changes the most after the surface anomaly occurs, f sd (x i ,y i ,d max ) represents the prior knowledge base located at (x i ,y i ) is the standard deviation of the remote sensing feature that changes the most after the surface anomaly occurs.
7. The method for real-time remote sensing detection of surface anomalies according to claim 6, characterized in that: The calculating of the surface anomaly intensity of the surface anomaly area further comprises: Among them, S2 represents the intensity of surface anomalies at the patch scale, aera plaque Represents the area of independent patches of surface anomalies, in square meters.
8. The method for real-time remote sensing detection of surface anomalies according to claim 7, characterized in that: The calculating of the surface anomaly intensity of the surface anomaly area further comprises: S=a*s1+b*s2 Among them, S is the surface anomaly intensity value, S1 represents the surface anomaly intensity at the pixel scale, S2 represents the surface anomaly intensity at the patch scale, and a and b are the weight coefficients of the surface anomaly intensity at the pixel and patch scales, respectively.
9. A real-time remote sensing detection system for surface anomalies, characterized in that: include: Acquisition module: acquiring first long-term surface remote sensing feature information to form a first long-term surface remote sensing feature data set; Construction module: constructing a priori knowledge base of surface remote sensing features based on the first long-term surface remote sensing feature data set; Identification module: obtains real-time remote sensing data on the satellite, calculates surface remote sensing feature information, and identifies potential surface anomaly areas; identifies surface anomaly areas based on the potential surface anomaly areas and a priori knowledge base of surface remote sensing features; Calculation module: Calculate the surface anomaly intensity of the surface anomaly area based on the identified surface anomaly area.
10. A real-time remote sensing early warning system for surface anomalies, characterized in that: It includes a satellite end and a local end, wherein the satellite end is arranged on an in-orbit satellite, and the local end is arranged on the surface of the earth; The satellite end calculates the surface anomaly detection result and the surface anomaly intensity data by using the detection method according to any one of claims 1 to 8, and transmits the surface anomaly detection result and the surface anomaly intensity to the local end; The local end receives the surface anomaly detection result and the surface anomaly intensity data, and issues a surface anomaly warning according to the surface anomaly detection result and the surface anomaly intensity data.
Citation Information
Patent Citations
Abnormal submerging area detection method based on remote sensing vegetation index time sequence
CN104239884A
Method and device for rapidly positioning water color abnormity based remote sensing image
CN108169142A
Land utilization / covering information space-time monitoring method based on priori knowledge
CN110263111A
High-resolution FAPAR estimation method based on downscaling of low-resolution remote sensing product
CN111242022A
Remote sensing image water area automatic extraction method and system based on deep learning
CN111767801A
Cited By
Instant remote sensing satellite constellation system
CN121366364A
