Method and device for detecting ground feature changes

Abstract

The present disclosure provides a method and device for detecting changes in land objects, which belong to the field of remote sensing image processing and target detection technology. The detection method includes: obtaining remote sensing image data of the target detection area in multiple phases; performing preprocessing operations on the acquired remote sensing image data of multiple phases to make the spectral statistical characteristics of the remote sensing image data of multiple phases consistent; establishing original difference images of the remote sensing image data of multiple phases based on the preprocessed remote sensing image data; establishing a low-rank sparse expression model based on the original difference image, and obtaining the low-rank part of the low-rank sparse expression model; clustering the low-rank part of the low-rank sparse expression model using a clustering algorithm to determine the changed part in the target detection area. Among them, the low-rank sparse expression model established by the original difference image can represent redundant information with simple information, thereby removing the redundancy between data.

Description

Technical Field

[0001] The present disclosure relates to the technical fields of remote sensing image processing and target detection, and in particular to a method and device for detecting ground object changes. Background Art

[0002] Land use / land cover change detection is an important aspect of resource and environmental research. Accurate and timely change information of land use / land cover provides decision support for the production and management of human resources and environment.

[0003] Remote sensing technology, with its real-time, rapid, wide-area, multispectral, and periodic characteristics, provides strong technical support for land use / land cover change detection. Remote sensing technology primarily uses various sensing instruments to collect, process, and ultimately image electromagnetic wave information radiated and reflected by distant targets, thereby detecting and identifying various ground features. Change detection technology is a primary means of extracting change information from remote sensing images. Change detection involves multi-temporal observation of targets or phenomena in the same area to identify areas of change. It primarily compares the radiation values ​​of targets at the same geographic location at different times.

[0004] However, remote sensing images contain multiple bands. While the large number of bands provides rich information about land objects, it also brings difficulties to the analysis and interpretation of land object changes. Moreover, each band contains surface information, and the information is correlated, which can easily lead to data redundancy. Summary of the Invention

[0005] The present disclosure provides a method and device for detecting changes in ground objects, which introduces a low-rank expression model and considers the low-rank structure of remote sensing images, thereby minimizing redundancy between data. The technical solution is as follows:

[0006] In one aspect, a method for detecting a ground feature change is provided, the method comprising:

[0007] Acquire remote sensing image data of the target detection location in multiple time phases;

[0008] Preprocessing the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent;

[0009] Creating original difference images of the remote sensing image data of the multiple time phases according to the pre-processed remote sensing image data;

[0010] Establishing a low-rank sparse expression model according to the original difference image, and obtaining a low-rank part in the low-rank sparse expression model;

[0011] The low-rank part of the low-rank sparse expression model is clustered using a clustering algorithm to determine the changed part in the target detection area.

[0012] Optionally, the step of establishing original difference images of the remote sensing image data of the multiple time phases based on the pre-processed remote sensing image data includes:

[0013] Converting the pre-processed remote sensing image data into a matrix, wherein each column of the matrix corresponds to the remote sensing image data of a band;

[0014] The original difference images of the remote sensing image data of the multiple time phases are established according to the matrix.

[0015] Optionally, converting the pre-processed remote sensing image data into a matrix includes:

[0016] The image of each band in the pre-processed remote sensing image data is converted into a column vector in sequence to obtain a matrix with b columns, where b is the number of bands of the remote sensing image data.

[0017] Optionally, the step of establishing original difference images of the remote sensing image data of the multiple time phases according to the matrix includes:

[0018] performing mean filtering on the remote sensing image data of the multiple time phases;

[0019] The first original difference image D of the remote sensing image data of the multiple time phases is obtained according to the difference operator s ;

[0020] The second original difference image D of the remote sensing image data of the multiple time phases is obtained according to the logarithmic ratio operator l ;

[0021] Perform low-rank expression decomposition on the matrix to obtain a sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of the multiple time phases. lrr ;

[0022] The first original difference image D s , the second original difference image D l and the third original difference image D lrr Combined to obtain the original difference image.

[0023] Optionally, acquiring remote sensing image data of the target detection location at multiple time phases includes:

[0024] Remote sensing image data of multiple time phases detected at a target detection location within a set time period is acquired, where the set time period is at least one continuous month.

[0025] Optionally, preprocessing the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent includes:

[0026] Performing geometric correction and radiometric correction on the remote sensing image data of the multiple time phases.

[0027] Optionally, clustering the low-rank part of the low-rank sparse expression model using a clustering algorithm includes:

[0028] The low-rank part is clustered to obtain two clusters of change regions and background values, and the change regions are used as the change parts in the target detection area.

[0029] In another aspect, a device for detecting a change in a ground feature is provided, the device comprising:

[0030] A data acquisition module is used to acquire remote sensing image data of the target detection area in multiple time phases;

[0031] A preprocessing module, configured to preprocess the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent;

[0032] An image creation module, configured to create original difference images of the remote sensing image data of the multiple time phases based on the pre-processed remote sensing image data;

[0033] A low-rank acquisition module, configured to establish a low-rank sparse expression model according to the original difference image, and obtain a low-rank part in the low-rank sparse expression model;

[0034] A clustering module is used to cluster the low-rank part of the low-rank sparse expression model using a clustering algorithm to determine the changed part in the target detection area.

[0035] Optionally, the image creation module includes:

[0036] A matrix conversion unit, configured to convert the pre-processed remote sensing image data into a matrix, wherein each column of the matrix corresponds to the remote sensing image data of a band;

[0037] The original image establishing unit is used to establish original difference images of the remote sensing image data of the multiple time phases according to the matrix.

[0038] Optionally, the matrix conversion unit is used to:

[0039] The image of each band in the pre-processed remote sensing image data is converted into a column vector in sequence to obtain a matrix with b columns, where b is the number of bands of the remote sensing image data.

[0040] Optionally, the original image creation unit is used to:

[0041] performing mean filtering on the remote sensing image data of the multiple time phases;

[0042] The first original difference image D of the remote sensing image data of the multiple time phases is obtained according to the difference operator s ;

[0043] The second original difference image D of the remote sensing image data of the multiple time phases is obtained according to the logarithmic ratio operator l ;

[0044] Perform low-rank expression decomposition on the matrix to obtain a sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of the multiple time phases. lrr ;

[0045] The first original difference image D s , the second original difference image D l and the third original difference image D lrr Combined to obtain the original difference image.

[0046] Optionally, the data acquisition module is used to:

[0047] Remote sensing image data of multiple time phases detected at a target detection location within a set time period is acquired, where the set time period is at least one continuous month.

[0048] Optionally, the preprocessing module is used to:

[0049] Performing geometric correction and radiometric correction on the remote sensing image data of the multiple time phases.

[0050] Optionally, the clustering module is used to:

[0051] The low-rank part is clustered to obtain two clusters of change regions and background values, and the change regions are used as the change parts in the target detection area.

[0052] The technical solutions provided by the embodiments of the present disclosure have the following beneficial effects:

[0053] By acquiring remote sensing image data from multiple time phases of the target detection location and preprocessing it to make the spectral statistical characteristics of the remote sensing image data consistent across multiple time phases, interference factors in the remote sensing image data from multiple time phases can be suppressed or filtered out. The low-rank sparse representation model established by the original difference image can represent redundant information with simple information, thus eliminating redundancy between data. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0055] Figure 1 is a method flow chart of a method for detecting ground feature changes provided by an embodiment of the present disclosure;

[0056] Figure 2 is a flow chart of another method for detecting ground feature changes provided by an embodiment of the present disclosure;

[0057] Figure 3 This is a data conversion diagram provided by an embodiment of the present disclosure;

[0058] Figure 4 It is a structural diagram of a ground feature change detection device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present disclosure more clear, the embodiments of the present disclosure will be further described in detail below with reference to the accompanying drawings.

[0060] The following briefly describes an application scenario of a method for detecting ground feature changes provided by an embodiment of the present disclosure:

[0061] In this embodiment, the detection of ground feature changes is the process of conducting multi-temporal observations of targets or phenomena in the same area to identify the areas of change. This is mainly achieved by comparing the radiation values ​​of targets in the same geographical location at different time phases. The disclosed embodiment is described using an oil pipeline corridor as an example to detect ground feature changes along the oil pipeline corridor. For example, changes in vegetation along the oil pipeline corridor or changes in surrounding buildings along the oil pipeline corridor can be detected to determine whether there are any factors affecting pipeline safety, such as trenching and canal construction, pipeline occupation, engineering construction, and vegetation destruction, along the oil pipeline corridor.

[0062] Figure 1 is a flow chart of a method for detecting a ground feature change provided by an embodiment of the present disclosure, such as Figure 1 As shown, the detection method includes:

[0063] Step 101: Acquire remote sensing image data of a target detection location in multiple time phases.

[0064] Step 102: pre-process the acquired remote sensing image data of multiple time phases to make the spectral statistical characteristics of the remote sensing image data of multiple time phases consistent.

[0065] For example, remote sensing image data of multiple time phases may be subjected to geometric correction and radiometric correction, with the accuracy of geometric correction being one-fifth of a pixel.

[0066] Step 103: Create original difference images of remote sensing image data of multiple time phases based on the pre-processed remote sensing image data.

[0067] For example, the pre-processed remote sensing image data can be converted into a matrix, where each column of the matrix corresponds to the image data of a band, and original difference images of remote sensing image data of multiple time phases are established based on the matrix.

[0068] Step 104: Establish a low-rank sparse expression model based on the original difference image, and obtain the low-rank part of the low-rank sparse expression model.

[0069] Step 105: cluster the low-rank part of the low-rank sparse expression model using a clustering algorithm to determine the changed part in the target detection area.

[0070] The disclosed embodiments provide a method for detecting ground feature changes. This method first acquires remote sensing image data of the target detection area at multiple time phases and preprocesses the data to make the spectral statistical characteristics of the remote sensing image data consistent across multiple time phases. This method can suppress or filter out interference factors in the remote sensing image data across multiple time phases. A low-rank sparse representation model established using the original difference image can represent redundant information with simple information, thus eliminating redundancy between the data.

[0071] Figure 2 is a flow chart of another method for detecting ground feature changes provided by an embodiment of the present disclosure. Figure 2 As shown, the detection method includes:

[0072] Step 201: Acquire remote sensing image data of a target detection location in multiple time phases.

[0073] In this embodiment, the target detection location is along the pipeline in the southwest region. The disclosed embodiment can detect changes in ground objects along the pipeline in the southwest region by performing change detection processing on satellite remote sensing image data of multiple phases along the pipeline in the southwest region.

[0074] Optionally, step 201 includes:

[0075] Acquire remote sensing image data of multiple time phases detected at the target detection location within a set time period.

[0076] Optionally, the time period is set to be at least one consecutive month.

[0077] Since change detection is primarily based on the characteristics of changes in the spectral reflectance curves of ground objects, and other factors that can cause changes in spectral reflectance curves include environmental factors such as climate conditions and solar angle, selecting appropriate data can, to a certain extent, reduce the impact of these interfering factors. The present disclosure selects remote sensing image data of multiple time phases detected within a month or within a consecutive number of similar months. At this time, the environments in which the remote sensing image data of multiple time phases are acquired are not much different, thereby reducing the adverse effects of changes in radiation values ​​caused by atmospheric errors, etc., achieving effective analysis of multi-temporal images and obtaining accurate results. This, to a certain extent, reduces the impact of "pseudo-changes" caused by seasonal changes on vegetation and building height differences, thereby improving the accuracy of change detection.

[0078] For example, in this embodiment, raw multi-temporal satellite remote sensing image data of the target detection area can be obtained through satellite observation. The multi-temporal satellite remote sensing image data can be stored in a device with data storage capabilities (e.g., a storage medium). In step S201, the remote sensing image data of multiple temporal phases can be directly obtained from the device with data storage capabilities. Multi-temporal generally refers to the characteristics of a set of remote sensing images in a time series. Broadly speaking, any set of images, maps, or geographic data of the same area acquired at different times can be considered "multi-temporal" data.

[0079] Step 202: Perform geometric correction on the remote sensing image data of multiple time phases.

[0080] In multi-temporal remote sensing imagery change detection, factors such as satellite sensor motion and terrain fluctuations can cause variations in the spectral intensity of image pixels, which are difficult for the human eye to detect. Therefore, geometric correction of multi-temporal remote sensing image data is a crucial step in change detection preprocessing.

[0081] The basic method of image geometric correction is to establish a mathematical model for geometric correction, then determine the model parameters using known conditions, and finally perform geometric correction on the image based on the model. This is existing technology and will not be described in detail in the embodiments of the present disclosure.

[0082] Step 203: Perform radiometric correction on the remote sensing image data of multiple time phases to make the spectral statistical characteristics of the remote sensing image data of multiple time phases consistent.

[0083] The difference in spectral statistical characteristics refers to the difference in image grayscale or color on the remote sensing image.

[0084] The radiation intensity of remote sensing images is mainly affected by two physical quantities: solar radiation and the spectral reflectivity of the ground objects. When the light radiation is the same, different radiation intensities reflect the differences in ground object types.

[0085] The goal of remote sensing image change detection algorithms is to detect significant changes and ignore insignificant ones. However, the complex remote sensing imaging environment causes variations in radiation intensity during the process of electromagnetic wave transmission and reception, and these variations are often caused by insignificant factors. Therefore, radiometric correction of images is often required before change detection to suppress or filter out interference factors that cause insignificant changes.

[0086] Step 204: Convert the pre-processed remote sensing image data into a matrix.

[0087] Remote sensing image data is an image cube H∈R composed of b band images S superimposed together m×n×b , the size of each band image is m×n, and the matrix F∈R can be obtained after conversion k×b (k=m×n).

[0088] Figure 3 This is a data conversion diagram provided by an embodiment of the present disclosure. Figure 3 As shown, Figure 3 The middle (a) figure is remote sensing image data, that is, the image cube H∈R m×n×b , Figure 3 Figure (b) is a decomposition diagram of Figure (a), consisting of Figure 3 As can be seen from Figure (b), the image cube H∈R m×n×b To form an image cube by superimposing multispectral images S, each image S may include m×n pixel blocks. Figure 3 Figures (b) to (c) show the process of columnarizing each band, converting the 3D remote sensing image data cube into a matrix. Figure 3 Figure (d) shows the matrix F∈R after transformation k×b (k=m×n).

[0089] Exemplarily, step 204 includes:

[0090] Convert the images of each band in the preprocessed remote sensing image data into a column vector in turn, and obtain a matrix F∈R with a column number of b k×b (k=m×n), b is the number of bands of remote sensing image data.

[0091] Through step 204 , the remote sensing image data acquired in multiple time phases can be converted from 3D data into a 2D matrix to achieve data format conversion.

[0092] Step 205: Create original difference images of remote sensing image data of multiple time phases according to the matrix.

[0093] Optionally, step 205 includes:

[0094] The first step is to perform mean filtering on remote sensing image data of multiple time phases;

[0095] The second step is to obtain the first original difference image D of the remote sensing image data of multiple time phases according to the difference operator. s ;

[0096] The third step is to obtain the second original difference image D of the remote sensing image data of multiple time phases according to the logarithmic ratio operator. l .

[0097] For example, the first original difference image D of the remote sensing image data of multiple time phases can be obtained according to the following formulas (1) and (2): s and the second original difference image D l :

[0098] D s =|X1-X2| (1)

[0099]

[0100] Among them, D s Denotes the first original difference image obtained by the difference operator, D l represents the second original difference image generated by the logarithmic ratio operator. X1 and X2 are two remote sensing images acquired at different times.

[0101] The fourth step is to perform low-rank expression decomposition on the matrix to obtain the sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of multiple time phases. lrr .

[0102] For example, the remote sensing image data matrix X = [x1, x2] can be decomposed into a low-rank expression to obtain the sparse part thereof, that is, the original difference image D of the remote sensing image. lrr .

[0103] The fourth step is to convert the first original difference image D s , the second original difference image D l and the third original difference image D lrr Combine them to get the original difference image.

[0104] For example, the original difference image D s 、D l and D lrr Joint, can include:

[0105] The first original difference image D s , the second original difference image D l and the third original difference image D lrrThe corresponding two-dimensional matrices are generated according to the matrix conversion form in step 204, and finally the three two-dimensional matrices are arranged and merged in sequence to form a matrix.

[0106] Step 206: Establish a low-rank sparse expression model based on the original difference image, and obtain the low-rank part of the low-rank sparse expression model.

[0107] Among them, the low-rank part corresponds to the changing area of ​​the remote sensing image.

[0108] In this embodiment, the low-rank expression decomposition may be performed again on the low-rank sparse expression model to obtain the low-rank part of the low-rank sparse expression model.

[0109] Optionally, step 206 may include:

[0110] The following low-rank sparse expression model Z is established based on the original difference images of remote sensing images of multiple time phases * :

[0111]

[0112] Among them, Z * The size is n i ×n i , the rank is This model can effectively achieve the purpose of change detection from a global perspective.

[0113] Alternatively, assume that the vector matrix of the low-rank expression model is X = [x1 x2 … x n ]∈R, and each column vector can be obtained through a dictionary basis A=[a1 a2 … a m ] linear expression is obtained, and formula (3) can be obtained:

[0114] X=AZ (3)

[0115] Where Z=[z1 z2 … z n ] is the coefficient matrix, each z i Corresponding to x i The linear representation of . The dictionary base A in formula (3) is [a1 a2 … a m ] is usually overcomplete, so there are many ways to solve the above optimization problem. We seek a representation of the coefficient matrix Z to solve the following constraint problem:

[0116] min rank(Z)

[0117] stX=AZ (4)

[0118] We call the optimal solution of the above constraint problem Z *is the lowest-rank representation of the vector matrix X with respect to the dictionary A. The rank function in the above problem is discrete, making the optimization problem of Equation (5) an NP-hard problem. Fortunately, the convex optimization method for the matrix completion problem provides a good alternative to Equation (4), see Equation (5):

[0119] min||Z|| *

[0120] stX=AZ (5)

[0121] Among them, ||Z|| * is the kernel function of the matrix, and is the sum of the singular values ​​obtained by SVD (Singular Value Decomposition) of the matrix.

[0122] Without loss of generality, we assume that X=[x1 x2 … x k ], in order to ensure that the data are independently divided into their respective subspaces, the matrix X is used as the dictionary, so Equation (5) can be converted to Equation (6):

[0123] min||Z|| *

[0124] stX=XZ (6)

[0125] From the above formula, we can see that since the data vector itself can be used to express itself in low-rank expression, even in the case of insufficient sampling data, the optimization problem of formula (6) has a feasible solution. When the sampling amount of data is sufficient, that is, d i =rank(X i ) <n i If the data subspaces are independent, then the optimization problem (6) has a set of optimal solutions Z * , and the optimal solution Z * is a size of n i ×n i , the rank is A block diagonal matrix, that is, a low-rank sparse expression model:

[0126]

[0127] Step 207: cluster the low-rank part of the low-rank sparse expression model using a clustering algorithm to determine the changed part in the target detection area.

[0128] Optionally, step 207 includes:

[0129] The low-rank part is clustered to obtain two families: the change region and the background value. The change region is used as the change part in the target detection area.

[0130] For example, the K-Means algorithm can be used to cluster the low-rank part of the low-rank sparse expression model. The main idea of ​​the K-Means algorithm is to divide the low-rank part into K clusters according to the distance between each region in the low-rank part for a given low-rank part, so that the points in the cluster are connected as closely as possible, and the distance between clusters is as large as possible.

[0131] In this embodiment, the K-Means algorithm can be used to divide the low-rank part of the low-rank sparse expression model into two parts: a changing part and a background value. The data in the changing part represents the changing part along the oil pipeline corridor, such as the changed parts of the vegetation and buildings along the oil pipeline corridor. The data in the background value family represents the unchanged parts along the oil pipeline corridor, such as the unchanged parts of the vegetation and buildings along the oil pipeline corridor. The changed area is then used as the changed part in the target detection area.

[0132] It should be noted that the disclosed embodiments are particularly suitable for detecting terrain changes along pipelines in the southwest region based on high-resolution satellite remote sensing imagery, such as trenching and canal construction, pipeline occupation, construction, and vegetation destruction, which can impact pipeline safety. This approach has broad applicability in the southwest region. Furthermore, by establishing new sample libraries in new areas, this method can be effectively extended beyond the southwest region.

[0133] The disclosed embodiments provide a method for detecting ground feature changes. This method first acquires remote sensing image data of the target detection area at multiple time phases and preprocesses the data to make the spectral statistical characteristics of the remote sensing image data consistent across multiple time phases. This method can suppress or filter out interference factors in the remote sensing image data across multiple time phases. A low-rank sparse representation model established using the original difference image can represent redundant information with simple information, thus eliminating redundancy between the data.

[0134] Figure 4 FIG. 1 is a schematic diagram of a structure of a device for detecting ground feature changes provided by an embodiment of the present disclosure. Figure 4 As shown, the detection device includes a data acquisition module 410 , a pre-processing module 420 , an image creation module 430 , a low-rank acquisition module 440 and a clustering module 450 .

[0135] The data acquisition module 410 is used to acquire remote sensing image data of the target detection location in multiple time phases.

[0136] The pre-processing module 420 is used to pre-process the acquired remote sensing image data of multiple time phases to make the spectral statistical characteristics of the remote sensing image data of multiple time phases consistent.

[0137] The image creation module 430 is used to create original difference images of remote sensing image data of multiple time phases according to the matrix.

[0138] The low-rank acquisition module 440 is used to establish a low-rank sparse expression model according to the original difference image, and obtain the low-rank part in the low-rank sparse expression model.

[0139] The clustering module 450 is used to cluster the low-rank part of the low-rank sparse expression model using a clustering algorithm to determine the changed part in the target detection area.

[0140] Optionally, the data acquisition module 410 is used to:

[0141] Acquire remote sensing image data of multiple time phases detected at the target detection location within a set time period.

[0142] Exemplarily, the time period is set to be at least one consecutive month.

[0143] The pre-processing module 420 is used to:

[0144] Perform geometric and radiometric correction on remote sensing image data of multiple time phases.

[0145] Optionally, the image creation module 430 includes:

[0146] The matrix conversion unit 431 is used to convert the pre-processed remote sensing image data into a matrix, where each column of the matrix corresponds to the remote sensing image data of a band;

[0147] The original image creation unit 432 is used to create original difference images of remote sensing image data of multiple time phases according to the matrix.

[0148] Exemplarily, the matrix conversion unit 431 is used to:

[0149] The images of each band in the preprocessed remote sensing image data are converted into a column vector in sequence to obtain a matrix with b columns, where b is the number of bands of the remote sensing image data.

[0150] Exemplarily, the original image creation unit 432 is used to:

[0151] Perform mean filtering on remote sensing image data of multiple time phases;

[0152] The first original difference image D of the remote sensing image data of multiple time phases is obtained according to the difference operator s ;

[0153] The second original difference image D of the remote sensing image data of multiple time phases is obtained according to the logarithmic ratio operator l ;

[0154] Perform low-rank expression decomposition on the matrix to obtain the sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of multiple phases lrr ;

[0155] The first original difference image D s , the second original difference image D l and the third original difference image D lrr Combine them to get the original difference image.

[0156] Optionally, the clustering module 450 is configured to:

[0157] The low-rank part is clustered to obtain two families: the change region and the background value. The change region is used as the change part in the target detection area.

[0158] It should be noted that the above-mentioned embodiments provide a detection device for detecting ground feature changes, using only the division of the above-mentioned functional modules as an example. In actual applications, the above-mentioned functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the detection device and the detection method embodiments provided in the above-mentioned embodiments are based on the same concept. The specific implementation process is detailed in the method embodiments and will not be repeated here.

[0159] The disclosed embodiments provide a device for detecting ground feature changes. This device can acquire remote sensing image data of a target location at multiple time phases and preprocess it to make the spectral statistical characteristics of the remote sensing image data consistent across multiple time phases. This can suppress or filter out interference factors in the remote sensing image data across multiple time phases. A low-rank sparse representation model established using the original difference image can represent redundant information with simple information, eliminating redundancy between data.

[0160] The above description is merely an optional embodiment of the present disclosure and is not intended to limit the present disclosure. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the scope of protection of the present disclosure.

Claims

1. A method for detecting ground feature changes, characterized in that: The detection method comprises: Acquire remote sensing image data of multiple time phases detected at a target detection location within a set time period, where the set time period is at least one consecutive month and the target detection location is along a pipeline in the southwest region; Preprocessing the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent; Converting the pre-processed remote sensing image data into a matrix, wherein each column of the matrix corresponds to the remote sensing image data of a band; performing mean filtering on the remote sensing image data of the multiple time phases; The first original difference image D of the remote sensing image data of the multiple time phases is obtained according to the difference operator s ; The second original difference image D of the remote sensing image data of the multiple time phases is obtained according to the logarithmic ratio operator l ; Perform low-rank expression decomposition on the matrix to obtain a sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of the multiple time phases. lrr ; The first original difference image D s , the second original difference image D l and the third original difference image D lrr , generating a corresponding two-dimensional matrix according to the matrix transformation form, and then arranging and merging the three two-dimensional matrices in sequence to form a matrix to obtain the original difference image; Establishing a low-rank sparse expression model according to the original difference image, and obtaining a low-rank part in the low-rank sparse expression model; A clustering algorithm is used to cluster the low-rank part of the low-rank sparse expression model to obtain two families of change areas and background values, and the change areas are used as the change parts in the target detection area; the data in the change parts represent the change parts along the oil pipeline corridor, including the vegetation and buildings along the oil pipeline corridor that have changed; the data in the background values ​​represent the unchanged parts along the oil pipeline corridor, including the vegetation and buildings along the oil pipeline corridor that have not changed.

2. The detection method according to claim 1, wherein The step of converting the pre-processed remote sensing image data into a matrix comprises: The image of each band in the pre-processed remote sensing image data is converted into a column vector in sequence to obtain a matrix with b columns, where b is the number of bands of the remote sensing image data.

3. The detection method according to claim 1 or 2, characterized in that The preprocessing of the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent includes: Performing geometric correction and radiometric correction on the remote sensing image data of the multiple time phases.

4. A device for detecting changes in ground features, characterized in that: The detection device comprises: A data acquisition module is used to acquire remote sensing image data of multiple time phases detected at a target detection location within a set time period, wherein the set time period is at least one continuous month and the target detection location is along a pipeline in the southwest region; A preprocessing module, configured to preprocess the acquired remote sensing image data of the multiple time phases to make the spectral statistical characteristics of the remote sensing image data of the multiple time phases consistent; An image creation module, configured to create original difference images of the remote sensing image data of the multiple time phases based on the pre-processed remote sensing image data; A low-rank acquisition module, configured to establish a low-rank sparse expression model according to the original difference image, and obtain a low-rank part in the low-rank sparse expression model; A clustering module is used to cluster the low-rank part of the low-rank sparse expression model using a clustering algorithm to obtain two families of change areas and background values, with the change areas being the change parts in the target detection area; the data in the change parts represent the change parts along the oil pipeline corridor, including the vegetation and buildings along the oil pipeline corridor that have changed; the data in the background values ​​represent the unchanged parts along the oil pipeline corridor, including the vegetation and buildings along the oil pipeline corridor that have not changed; The image creation module includes: A matrix conversion unit, configured to convert the pre-processed remote sensing image data into a matrix, wherein each column of the matrix corresponds to the remote sensing image data of a band; An original image creation unit, configured to perform mean filtering on the remote sensing image data of the plurality of time phases; The first original difference image D of the remote sensing image data of the multiple time phases is obtained according to the difference operator s ; The second original difference image D of the remote sensing image data of the multiple time phases is obtained according to the logarithmic ratio operator l ; Perform low-rank expression decomposition on the matrix to obtain a sparse part of the matrix, and use the sparse part of the matrix as the third original difference image D of the remote sensing image data of the multiple time phases. lrr ; The first original difference image D s , the second original difference image D l and the third original difference image D lrr , generate a corresponding two-dimensional matrix according to the matrix transformation form, and then arrange and merge the three two-dimensional matrices in sequence to form a matrix to obtain the original difference image.

5. The detection device according to claim 4, characterized in that The matrix conversion unit is used for: The image of each band in the pre-processed remote sensing image data is converted into a column vector in sequence to obtain a matrix with b columns, where b is the number of bands of the remote sensing image data.

6. The detection device according to claim 4 or 5, characterized in that: The pre-processing module is used for: Performing geometric correction and radiometric correction on the remote sensing image data of the multiple time phases.

Images