A target and background spatio-temporal spectral similarity analysis system and method

Abstract

The application discloses a target and background spatio-temporal spectrum similarity analysis system and method, which comprises a spectrum pretreatment module, a spectrum feature extraction module, a spatio-temporal variation rule analysis module and a similarity analysis module; target and background spectra are transmitted to the spectrum feature extraction module after being processed by the spectrum pretreatment module; the spectrum feature extraction module extracts features of the spectra and quantifies differences, and then transmits the features to the spatio-temporal variation rule analysis module; the spatio-temporal variation rule analysis module analyzes the spectrum feature differences with the spatio-temporal variation rule, and transmits the variation rule to the similarity analysis module; the similarity analysis module combines the spectrum feature differences with the spatio-temporal variation rule, assigns weights to each feature, and comprehensively obtains a similarity value to grade the spectrum similarity. The application considers the influence of the time dimension and the space dimension on the spectrum features, combines the spectrum features with the variation rule to assign weights, and realizes the expansion of traditional static spectrum features to the analysis system and method of a dynamic scene.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of spectral analysis, and relates to a target and background spatio-temporal spectral similarity analysis system and method, in particular to a target and background spatio-temporal spectral similarity analysis system and method under dynamic spatio-temporal conditions. BACKGROUND

[0002] Thanks to high resolution in the spectral dimension, hyperspectral imaging technology has strong target detection capability, which can image a target and a background at high resolution from a visible light band to an infrared band, obtain complete and continuous spectral curves of the target and the background, and enable a slight difference between the target and the background in the spectrum to be perceived, but this is only for static hyperspectral images, and when the target and the background change over time, whether the hyperspectral detection technology can still achieve good results needs to be researched on the dynamic hyperspectral detection technology in time and space.

[0003] The basis of the dynamic hyperspectral detection technology is to analyze the spatio-temporal spectral consistency of the target and the background, that is, to dynamically extract spectral features of the target and the background, quantitatively represent the spectral features, and analyze the dynamic change law of the spectral features over time and space. The spectral feature extraction methods at the present stage are mostly for static targets or objects, such as extracting spectral features of a gemstone to identify the gemstone, extracting spectral features of a drug to analyze the drug composition, extracting spectral features of a ground object to conduct mineral exploration, and the like. Although these application scenarios are various, the research targets are mostly static and fixed, and do not involve dynamic changes in the time and space dimensions.

[0004] To solve the problem of target and background spectral similarity analysis under dynamic spatio-temporal changes, not only the spectral feature difference between the target and the background needs to be considered, but also the influence of the spatio-temporal change on the spectral feature difference between the target and the background needs to be considered, and how to effectively combine the two is a technical problem to be solved for the development of the dynamic hyperspectral detection technology in time and space. SUMMARY

[0005] In view of the above prior art, the technical problem to be solved by the application is to provide a target and background spatio-temporal spectral similarity analysis system and method under dynamic spatio-temporal conditions, which can analyze the change law of spectral features over time and space and give a comprehensive similarity judgment in combination with the law.

[0006] To solve the above technical problem, a target and background spatio-temporal spectral similarity analysis system according to the application comprises a spectral preprocessing module, a spectral feature extraction module, a spatio-temporal change law analysis module, and a similarity analysis module.

[0007] The target spectrum and the background spectrum are input into the spectral preprocessing module for preprocessing and then transmitted to the spectral feature extraction module.

[0008] The spectral feature extraction module extracts spectral features of the target and the background respectively and quantifies the difference, the number of the extracted spectral features is n, meanwhile, the time and spatial position information of the target is recorded, the data vector of [feature difference, time, position] corresponding to each spectral feature is obtained and input to the time-space variation rule analysis module;

[0009] The time-space variation rule analysis module records the input [feature difference, time, position] vector in a period of time, projects the vector to the feature difference-time, feature difference-longitude and feature difference-latitude coordinate system respectively, and adopts the fitting regression method to fit the feature difference-time, feature difference-longitude and feature difference-latitude regression curve corresponding to each spectral feature respectively; the derivative of the regression curve is calculated to obtain the variation rate k ti , k lngi and k lati of the i-th spectral feature with time, longitude and latitude respectively, i=1, 2, …, n, the quantified rules k ti , k lngi and k lati of each spectral feature with time, longitude and latitude are input to the similarity analysis module;

[0010] The similarity analysis module assigns weights to the feature difference values of each input spectral feature according to k ti , k lngi and k lati , i=1, 2, …, n, performs weighted summation on the feature difference values of all spectral features according to the assigned weights w i , and obtains the comprehensive difference value Δ, classifies the similarity of the target and the background according to the size of the weighted difference value, and outputs the similarity analysis result.

[0011] Further, the preprocessing includes filtering, baseline drift elimination and data normalization processing.

[0012] Further, the extraction of the spectral features of the target and the background and the quantification of the difference specifically includes that the spectral features include one or more of spectral curve features, spectral transformation features and spectral similarity features; when the spectral features are the spectral curve features and the spectral transformation features, the corresponding features are extracted from the target and the background respectively, and then the difference quantification is performed to obtain the feature difference of the target and the background, the difference quantification method includes but is not limited to the difference method and the contrast method, and when the spectral features are the spectral similarity features, the difference quantification processing is not performed.

[0013] Further, the similarity analysis module assigns weights to the feature difference values of each input spectral feature according to k ti , k lngi and k lati, i=1, 2, …, n, the weight of the feature difference value of each spectral feature of the input is assigned, which is specifically:

[0014] The weight value of each spectral feature value is:

[0015]

[0016] Wherein, w i represents the weight value of the feature difference value of the i-th spectral feature, k Ti represents the comprehensive change rate of the i-th spectral feature, and is specifically:

[0017] k Ti =|k ti |+|k lngi |+|k lati |.

[0018] Further, the feature difference values of all spectral features are weighted and summed according to the assigned weights w i , to obtain the comprehensive difference value Δ, which is specifically:

[0019] Δ=w1s1+w2s2+…+w n s n

[0020] In the formula, s i represents the feature difference value of the i-th spectral feature of the target and the background, w i represents the weight value of the feature difference value of the i-th spectral feature, i=1, 2, …, n.

[0021] The application also includes a target and background spatiotemporal spectral similarity analysis method using any of the foregoing systems, comprising the following steps:

[0022] Step 1: Obtain the spectra of the target and the background of the target at different times and different places, and transmit the obtained spectra to a spectral pretreatment module;

[0023] Step 2: The spectral pretreatment module pre-processes the spectral features of the target and the background, and transmits the pre-processed spectra to a spectral feature extraction module;

[0024] Step 3: The spectral feature extraction module extracts the spectral features of the target and the background and quantifies the difference, the number of extracted spectral features is n, simultaneously records the time and spatial position information of the target, obtains the data vector of [feature difference, time, position] corresponding to each spectral feature of the target and the background and inputs it to a spatiotemporal change rule analysis module;

[0025] Step 4: The time-space variation rule analysis module records the input [feature difference, time, location] vector in a period of time, projects the vector to the feature difference-time, feature difference-longitude, and feature difference-latitude coordinate system respectively, and uses fitting regression to respectively fit the feature difference-time, feature difference-longitude, and feature difference-latitude regression curves corresponding to each spectral feature; and respectively derives the rate of change of the ith spectral feature with time k ti , the rate of change of the ith spectral feature with longitude k lngi , and the rate of change of the ith spectral feature with latitude k lati , i = 1, 2, …, n, and inputs each spectral feature with the quantification rule of the change of space and time k ti , k lngi , and k lati , i = 1, 2, …, n, to the similarity analysis module.

[0026] Step 5: The similarity analysis module assigns weights to the feature difference values of each spectral feature according to k ti , k lngi , and k lati , i = 1, 2, …, n, and performs weighted summation on the feature difference values of all spectral features according to the assigned weights w i to obtain a comprehensive difference value Δ, classifies the similarity of the target and the background according to the size of the weighted difference value, and outputs the similarity analysis result.

[0027] Further, the preprocessing includes moving window average smoothing filtering, multivariate scattering correction, and min-max standardization processing.

[0028] Further, the spectral features extracted in step 3 include one spectral transformation feature and two spectral similarity features, specifically, a vegetation index NDVI, a spectral Euclidean distance, and a spectral angle, wherein the vegetation index NDVI is extracted from the target and the background respectively and subjected to difference quantization processing, and the difference quantization adopts a difference method, and the formula is as follows:

[0029] ΔNDVI = |NDVI t - NDVI s |

[0030] In the formula, ΔNDVI represents the vegetation index difference value, NDVI t represents the target vegetation index, and NDVI s represents the background vegetation index.

[0031] The beneficial effects of the present application: the purpose of the present application is to solve the problem of spectral similarity analysis of target and background in space-time variation, and a dynamic spectral analysis system and method combining time and space dimensions are proposed. The spectral similarity analysis system and method proposed in the present application considers the influence of time and space variation on spectral characteristics, and combines the change rule to allocate weights to the spectral characteristics, finally realizes an analysis system and method which extends the application of traditional static spectral characteristics to dynamic scene, and provides basic support for the realization of space-time dynamic hyperspectral detection technology. BRIEF DESCRIPTION OF DRAWINGS

[0032] Figure 1 is a camouflage target and background space-time spectral similarity analysis system block diagram proposed by the present application;

[0033] Figure 2 is a general schematic diagram of a camouflage target and background space-time spectral similarity analysis method proposed by the present application;

[0034] Figure 3 is a spectral preprocessing module schematic diagram of a camouflage target and background space-time spectral similarity analysis method proposed by the present application;

[0035] Figure 4 is a space-time variation rule analysis module schematic diagram of a camouflage target and background space-time spectral similarity analysis method proposed by the present application. DETAILED DESCRIPTION

[0036] The present application will be further described below in conjunction with the drawings and examples in the specification.

[0037] In conjunction with Figure 1 , a target and background space-time spectral similarity analysis system under space-time variation is proposed by the present application, which comprises: a spectral preprocessing module, a spectral feature extraction module, a space-time variation rule analysis module, and a similarity analysis module; the spectral preprocessing module processes the input camouflage target and background spectrum and transmits it to the spectral feature extraction module; the spectral feature extraction module extracts the features of the input target and background spectrum and quantifies the difference between them, and transmits the obtained quantized feature difference to the space-time variation rule analysis module; the space-time variation rule analysis module analyzes the variation rule of the target and background spectral feature difference with space-time, and transmits the variation rule to the similarity analysis module; the similarity analysis module combines the variation rule of the feature difference with space-time, and allocates weights to each feature, and then obtains a similarity value, and grades the spectral similarity according to the similarity value.

[0038] In one embodiment, the input spectral information includes the spectral information of the target and its background at different times and spaces, which can be obtained by a spectral imager.

[0039] In one embodiment, the spectral pretreatment module filters, removes baseline drift, and normalizes the input spectral information, facilitating subsequent processing and analysis by the module; the spectral pretreatment module transmits the pretreated data to the spectral feature extraction module.

[0040] In one embodiment, the filtering process aims to remove spectral noise, and specific methods include but are not limited to moving window average smoothing filter, convolution smoothing filter, etc.

[0041] In one embodiment, the baseline drift removal method specifically includes derivation and light scattering correction methods; the derivation algorithm includes but is not limited to direct difference method, convolution derivation method, etc.; the light scattering correction algorithm includes but is not limited to multivariate scattering correction, standard normal variable transformation, etc.

[0042] In one embodiment, the data normalization process scales and translates the data according to the proportion, so that the data falls within a small specific interval, and specific methods include but are not limited to min-max standardization method, z-score standardization method, logistic transformation method, etc.

[0043] In one embodiment, the spectral feature extraction module extracts the spectral features of the camouflage target and background and quantifies the differences, while recording the time and spatial position information of the target, obtaining the data vector of "feature difference-time-position" of the target and background; the spectral features include spectral curve features, spectral transformation features, and spectral similarity features, wherein the spectral curve features and spectral transformation features are characteristics of the target and background themselves, which need to be extracted from the target and background respectively, and then the target and background features need to be quantified to obtain the feature difference of the target and background; the difference quantification method includes but is not limited to difference method and contrast method; for the spectral similarity features, it is a difference feature of the target and background itself, so it does not need to be quantified; the time data is recorded in the format of hours-minutes; the position data is recorded in the format of longitude-latitude; finally, a multi-dimensional vector of [feature difference, time, space] is obtained, which is input into the spatiotemporal variation rule analysis module.

[0044] In one embodiment, the spatio-temporal variation analysis module records the input feature difference-time-location vectors in a period of time, which form discrete points in the feature difference-time-location coordinate system, and the relationship between the feature difference and the time and location is fitted from the discrete points by using fitting regression. Specifically, the points are projected into the feature difference-time, feature difference-longitude, and feature difference-latitude coordinate systems respectively, and linear or nonlinear regression is performed in the projected coordinate systems respectively. Qualitatively, the regression curve can roughly show the variation trend of the target and background spectral feature difference with time and location, and quantitatively, the regression curve is differentiated to obtain the rate k t of change of the target and background spectral feature difference with time, the rate k lng of change of the target and background spectral feature difference with longitude, and the rate k lat of change of the target and background spectral feature difference with latitude. t , k lng , k lat , which are the quantitative representations of the spatio-temporal variation of the target and background spectral feature difference. The above quantitative rules are input into the similarity analysis module.

[0045] In one embodiment, the similarity analysis module assigns weights to the input target and background spectral feature difference values according to the input spatio-temporal quantitative rules of the spectral feature difference. The weight assignment methods include but are not limited to entropy weight method, information weight method, etc., i.e., the weight of each feature in the final evaluation is determined according to the information amount or the degree of spatio-temporal variation. A comprehensive difference value is obtained by weighted summing of each spectral feature difference, which comprehensively reflects the spectral feature difference between the target and the background and is regulated by the spatio-temporal variation. According to the size of the difference value, the similarity between the target and the background is divided into four levels: completely consistent, consistent, inconsistent, and completely inconsistent.

[0046] In combination with Figure 2 , the present application proposes a target and background spatio-temporal spectral similarity analysis method under spatio-temporal variation, which includes spectral preprocessing, spectral feature extraction, spatio-temporal variation analysis, and similarity analysis.

[0047] The input camouflage target and background spectrum are transmitted to the spectrum feature extraction module after being processed by the spectrum preprocessing module, the spectrum feature extraction module extracts features from the input spectrum, and the extracted features are transmitted to the space-time change rule analysis module. The space-time change rule analysis module analyzes the change rule of the spectrum features with time and space and transmits the change rule to the similarity analysis module. The similarity analysis module combines the change rule of the features with time and space, gives appropriate weights to the spectrum feature difference values of the target and the background, and obtains a comprehensive similarity evaluation by weighted summation. Finally, the similarity level of the target and the background and the weight of each spectrum feature in the evaluation are output.

[0048] The target and background space-time spectrum similarity analysis method of the embodiment analyzes the spectrum features of the target and the background in the space-time dynamic change in the summer sunny noon grassland scene, and the implementation steps are as follows:

[0049] Step (1): Obtain the spectrum of the target and the background of the target at different times and different places through the imaging spectrometer, and transmit the obtained spectrum to the spectrum preprocessing module.

[0050] Step (2): The spectrum preprocessing module performs moving window average smoothing filtering, multivariate scattering correction, and min-max standardization processing on the obtained spectrum information of the target and the background, as shown in Figure 3 , and transmits the preprocessed spectrum to the feature extraction module.

[0051] In this embodiment, the window size of the moving window average smoothing filter is set to 5 (the window size is an odd number), five points on the spectrum data are selected from the beginning: x -2 ,x -1 ,x0,x1,x2, and then the average value is assigned to x0:

[0052] x0=(x -2 +x -1 +x0+x1+x2) / 5

[0053] Then move the window so that the center point of the window traverses the entire spectrum data, that is, the moving window average smoothing is completed.

[0054] In this embodiment, the specific implementation method of multivariate scattering correction is as follows:

[0055] Step 1: Obtain the average value of all spectrum data as the "ideal spectrum".

[0056]

[0057] In the formula:

[0058] — the average value of the target or background spectrum data;

[0059] data i — target or background i-th spectrum data;

[0060] n — target or background spectrum data number.

[0061] Step2: Linear regression of each sample spectrum to the average spectrum, solve the least square problem to get the baseline shift and offset of each sample.

[0062]

[0063] In the formula:

[0064] k i — slope of linear regression;

[0065] b i — intercept of linear regression.

[0066] Step3: Correct each sample spectrum: subtract the baseline shift and divide by the offset to get the corrected spectrum.

[0067]

[0068] In the formula:

[0069] data i(MSC) — corrected spectrum data.

[0070] In this embodiment, the min-max standardization processing formula is as follows:

[0071]

[0072] In the formula:

[0073] — normalized spectrum value;

[0074] x — target or background spectrum value;

[0075] min — minimum value of target or background spectrum value;

[0076] max — maximum value of target or background spectrum value.

[0077] Step (3) The feature extraction module extracts features of the target and the background. In this embodiment, the extracted features include one spectral transformation feature and two spectral similarity features, specifically, a vegetation index NDVI, a spectral Euclidean distance, and a spectral angle. The vegetation index NDVI is extracted from the target and the background respectively and is subjected to difference quantification processing. The features are extracted and quantified, and at the same time, the time and spatial position information is recorded to obtain a multi-dimensional vector [feature difference, time, space].

[0078] In this example, the feature calculation formula is as follows:

[0079] Vegetation index NDVI:

[0080]

[0081] In the formula:

[0082] NIR - target or background near-infrared band spectral value

[0083] R - target or background red band spectral value

[0084] Spectral Euclidean distance:

[0085]

[0086] In the formula:

[0087] A i - the i-th band target spectral reflectance

[0088] B i - the i-th band background spectral reflectance

[0089] N - the number of bands

[0090] Spectral angle:

[0091]

[0092] In the formula:

[0093] A - A = (A1, A2,..., A N ) is a target spectral vector

[0094] B - B = (B1, B2,..., B N ) is a background spectral vector

[0095] N - the number of bands

[0096] In this embodiment, the difference quantification method adopted is the difference method, and the formula is as follows:

[0097] ΔNDVI = |NDVI t - NDVI s |

[0098] wherein:

[0099] ΔNDVI - difference value of vegetation index

[0100] NDVI t - target vegetation index

[0101] NDVI s - background vegetation index

[0102] Step (4) obtains discrete sequences of feature difference-time, feature difference-longitude, and feature difference-latitude, respectively, from the spatiotemporal feature difference vectors stored in the storage unit for a period of time, and then fits these discrete points to obtain a function relationship between the target and background spectral feature difference and time, longitude, and latitude by linear fitting, as shown in the following formula: Figure 4

[0103]

[0104]

[0105]

[0106] wherein:

[0107] - spectral feature difference

[0108] t - time variable

[0109] g - longitude variable

[0110] l - latitude variable

[0111] k - fitting slope

[0112] b - fitting intercept

[0113] Further, the absolute value of the slope of the fitted straight line, |k t |, |k lng |, |k lat | is used to represent the rate of change of the spectral feature difference with space and time, and the size of the absolute value reflects the degree of change of the feature difference with space and time. The greater the absolute value, the more obvious the change of the feature with space and time, which is an object that needs to be focused on. This module uses the slope of the fitted straight line to quantitatively represent the spatiotemporal variation law of the target and background spectral feature difference, and transmits the quantified feature to the similarity analysis module.

[0114] ​Step (5) The similarity analysis module assigns weights to the spectral feature difference values ​​at the current moment based on the spatiotemporal variation patterns of the target and background spectral feature differences over a previous period. In this example, the spectral feature extraction module extracts three spectral feature differences. The spatiotemporal variation pattern of each feature difference can be represented by |k t |、|k lng |、|k lat | to represent |k t |、|k lng |、|k lat |The sum of these yields a comprehensive rate of change k T ,Right now

[0115] k T =|k t |+|k lng |+|k lat |

[0116] In the formula:

[0117] k T —Comprehensive change rate of spectral feature differences

[0118] |k t |——Absolute value of the rate of change of spectral characteristic differences over time

[0119] |k lng |——Absolute value of the rate of change of spectral characteristic differences with longitude

[0120] |k lat |——Absolute value of the rate of change of spectral characteristic differences with latitude

[0121] The combined rate of change k of the three characteristic differences was obtained respectively. T1 k T2 k T3 Then the weights of the differences in each feature are:

[0122]

[0123] In the formula:

[0124] w i —Weight of feature differences

[0125] k T1 —Comprehensive rate of change of vegetation index characteristics

[0126] k T2 —Comprehensive rate of change of spectral angular characteristics

[0127] k T3 —Comprehensive rate of change of spectral distance characteristics

[0128] According to the calculated weight, the features are weighted and summed to obtain a comprehensive difference value of the target and background features:

[0129] Δ = w1s1 + w2s2 + w3s3

[0130] In the formula:

[0131] Δ — comprehensive difference value of the target and background features

[0132] s i — difference value of each spectral feature of the target and background

[0133] Finally, according to the size of the comprehensive difference value Δ, the similarity of the target and background spectra is rated:

[0134] Completely consistent Δ < 0.2 Consistent 0.2 < Δ < 0.5 Not consistent 0.5 < Δ < 0.9 Not consistent at all Δ > 0.9

Claims

1. A target and background spatio-temporal spectral similarity analysis system, characterized in that: The method comprises the following steps: The target spectrum and the background spectrum are input into the spectrum preprocessing module for preprocessing and then transmitted to the spectrum feature extraction module; The spectrum feature extraction module extracts the spectrum features of the target and the background respectively and quantifies the difference, the number of extracted spectrum features is n, and the time and spatial position information of the target is recorded at the same time, so as to obtain the data vector of [feature difference, time, position] corresponding to the target and the background of each spectrum feature and input into the time-space variation law analysis module; The spatio-temporal variation rule analysis module records the input [feature difference, time, location] vector in a period of time, projects the vector to a feature difference-time, feature difference-longitude, and feature difference-latitude coordinate system respectively, and obtains the feature difference-time, feature difference-longitude, and feature difference-latitude regression curves corresponding to each spectral feature by fitting regression respectively; and derives the change rate of the i-th spectral feature with time , the change rate of the i-th spectral feature with longitude , and the change rate of the i-th spectral feature with latitude , , , , , , and inputs the quantification rule of the change of each spectral feature with space and time to the similarity analysis module; The similarity analysis module is based on , , , Assign weights to the feature difference values ​​of each input spectral feature, and then... The weighted sum of the characteristic difference values ​​of all spectral features is used to obtain the comprehensive difference value. The similarity between the target and the background is graded based on the magnitude of the comprehensive difference value, and the similarity analysis results are output. The weight distribution method comprises an entropy weight method and an information weight method, that is, the weight of each spectrum feature in the final evaluation is determined according to the information amount contained in the spectrum feature or the degree of variation with time and space, a comprehensive difference value is obtained by weighted summation of the spectrum feature differences, the value comprehensively reflects the spectrum feature difference between the target and the background and is regulated by the time-space variation law, and the similarity between the target and the background is divided into four levels according to the size of the difference value: complete consistency, consistency, inconsistency and complete inconsistency.

2. The system for target and background spatio-temporal spectral similarity analysis of claim 1, wherein: The preprocessing comprises filtering, baseline drift elimination and data normalization processing.

3. A target and background spatio-temporal spectral similarity analysis system according to claim 2, wherein: The spectrum feature extraction module extracts the spectrum features of the target and the background respectively and quantifies the difference, the number of extracted spectrum features is n, and the time and spatial position information of the target is recorded at the same time, so as to obtain the data vector of [feature difference, time, position] corresponding to the target and the background of each spectrum feature and input into the time-space variation law analysis module; 4. A system for target and background spatio-temporal spectral similarity analysis according to claim 3, characterized in that: The similarity analysis module assigns weights to the feature difference values of each input spectral feature according to , , , , The weight value of each spectrum feature value is: ； wherein, a weight value representing a feature difference value of the i-th spectral feature, a comprehensive change rate of the i-th spectral feature, specifically: 。 5. A target and background spatio-temporal spectral similarity analysis system according to claim 4, characterized in that: The assigned weights The feature difference values of all spectral features are weighted and summed to obtain a comprehensive difference value Specifically: ； wherein denotes the feature difference value of the i-th spectral feature of the target and the background, denotes the weight value of the feature difference value of the i-th spectral feature, .

6. The method of claim 5, wherein the target and background spatiotemporal spectral similarity analysis system is characterized by, The method comprises the following steps: Step 1: obtaining the spectra of the target and the background of the target at different times and different locations, and transmitting the obtained spectra to the spectrum preprocessing module; Step 2: the spectrum preprocessing module pre-processes the spectrum features of the target and the background, and transmits the pre-processed spectrum to the spectrum feature extraction module; Step 3: the spectrum feature extraction module extracts the spectrum features of the target and the background respectively and quantifies the difference, the number of extracted spectrum features is n, and the time and spatial position information of the target is recorded at the same time, so as to obtain the data vector of [feature difference, time, position] corresponding to the target and the background of each spectrum feature and input into the time-space variation law analysis module; Step 4: The spatiotemporal variation pattern analysis module records the input [feature difference, time, location] vector over a period of time. This vector is then projected onto feature difference-time, feature difference-longitude, and feature difference-latitude coordinate systems, respectively. A regression fitting method is used to obtain the feature difference-time, feature difference-longitude, and feature difference-latitude regression curves for each spectral feature. The derivative of each regression curve is then calculated to obtain the rate of change of the i-th spectral feature over time. Rate of change with longitude and the rate of change with latitude , Quantifying the spatiotemporal variation of each spectral feature , , , The data is then input into the similarity analysis module. Step 5: The similarity analysis module assigns a weight to the feature difference value of each input spectral feature according to the assigned weight , , , , and performs a weighted summation of the feature difference values of all spectral features according to the assigned weights to obtain a comprehensive difference value , and grades the similarity of the target and the background according to the size of the weighted difference value, and outputs the similarity analysis result.

7. The method of claim 6, wherein: The preprocessing comprises moving window average smoothing filtering, multivariate scattering correction and min-max standardization processing.

8. The method of claim 6, wherein the method further comprises: determining a spatial similarity between the target and the background. The extracted spectrum features in step 3 comprise one spectrum transformation feature and two spectrum similarity features, specifically, a vegetation index NDVI, a spectral Euclidean distance and a spectral angle, wherein the vegetation index NDVI is extracted from the target and the background respectively and subjected to difference quantization, and the difference quantization adopts a difference value method, and the formula is as follows: ； In the formulae: denotes a vegetation index difference value, denotes a target vegetation index, denotes a background vegetation index.

Images