A sample consistency inspection method and system for remote sensing images

Abstract

The application discloses a sample consistency inspection method and system for remote sensing images, and the existing remote sensing image samples only rely on manual inspection, when the task amount is too large, the number of personnel is large, and time is urgent, the samples cannot be quickly inspected, and service classification detection and other tasks cannot be performed. The application creates different remote sensing images, only uses vector samples and remote sensing images, and gives consistency inspection results based on sample data distribution characteristics, and can adaptively adjust the abnormal value detection rate, combines with actual sample inspection, improves the efficiency, and optimizes the related work flow.

Description

TECHNICAL FIELD

[0001] The present application relates to remote sensing data processing, in particular to a sample consistency inspection method and system for remote sensing images BACKGROUND

[0002] Remote sensing image samples are the basis for realizing remote sensing image information extraction, and their quality directly affects the precision and generalization ability of interpretation models. With the combination of natural resources and remote sensing fields with Internet and computer technologies, and the expansion of classification needs in real tasks, the amount of sample data collected for remote sensing images is increasing. Due to inevitable human misinterpretation and other factors, these data often contain abnormal data, which need to be distinguished to better serve classification and other applications. Therefore, it is necessary to check and evaluate the quality of remote sensing image samples, and to establish a scientific and reasonable remote sensing image sample quality checking and evaluation standard and method, which plays an important role in improving the efficiency and quality of remote sensing image intelligent interpretation. SUMMARY

[0003] The present application aims to solve the problems of different collection standards of remote sensing image classification samples, difficulty and heavy workload of manual verification, and provides a sample consistency inspection method and system for remote sensing images.

[0004] To achieve the above-mentioned purpose, the technical solution of the present application is:

[0005] In a first aspect, the present application provides a sample consistency inspection method for remote sensing images, which comprises:

[0006] Grouping and classifying samples based on the classification field of the attribute of the vector sample;

[0007] After coordinate consistency conversion of the vector sample range with reference to the original remote sensing image, sample cropping is performed on the remote sensing image to obtain pixel multi-band information of the sample range;

[0008] Based on the pixel multi-band information of the sample range, a multi-band multi-dimensional feature of the sample is generated using a statistical method;

[0009] Based on the sample grouping and classification and the abnormal rate or screening rate, a consistency detection rate is designed to detect samples or model self-adaptively determine the detection rate to detect abnormal samples, so as to group and classify the multi-band multi-dimensional features of the samples;

[0010] The sample multi-band multi-dimensional features after grouping and classification are subjected to sample consistency detection using a sample consistency detection model, and the results after consistency detection are respectively assigned and output as result vectors to provide the consistency results of the samples for subsequent viewing.

[0011] In a second aspect, the present application provides a sample consistency inspection method for remote sensing images, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the method as described above when executing the computer program.

[0012] In a third aspect, the present application provides a computer readable storage medium storing a computer program, wherein the computer program implements the steps of any of the methods as described above when executed by a processor.

[0013] Compared with the prior art, the present application has the following advantages:

[0014] The existing remote sensing image samples only rely on manual inspection, and when the task volume is too large, the number of personnel is large, and the time is urgent, the samples cannot be quickly inspected, and the service classification detection and other tasks cannot be performed. The present application is suitable for different remote sensing images, only uses vector samples and remote sensing images based on sample data distribution characteristics to give consistency inspection results, and can automatically adjust the abnormal value detection rate, combines the actual sample inspection to improve the efficiency and optimize the related work flow. BRIEF DESCRIPTION OF DRAWINGS

[0015] Figure 1 A flowchart of a sample consistency inspection method for remote sensing images provided by the present application for example 1 is shown in the figure.

[0016] Figure 2 A software operation interface diagram of a sample consistency inspection method for remote sensing images provided by the present application for example 1 is shown in the figure.

[0017] Figure 3 A schematic diagram of a sample consistency inspection method system for remote sensing images provided by the present application for example 2 is shown in the figure. DETAILED DESCRIPTION

[0018] The technical solutions of the present application will be further described below in combination with the drawings and examples.

[0019] Example 1

[0020] Referring to Figure 1 The sample consistency inspection method for remote sensing images provided by the present application mainly comprises the following steps:

[0021] Grouping and classifying the samples based on the classification field of the vector sample attribute;

[0022] After the coordinate consistency conversion of the vector sample range with reference to the original remote sensing image, the sample range is cut to obtain the pixel multi-band information of the sample range;

[0023] Based on the pixel multi-band information of the sample range, a statistical method is used to generate the multi-band multi-dimensional features of the sample, i.e., sample statistical feature calculation;

[0024] Based on the sample grouping classification and the abnormal rate or screening rate, a consistency detection rate is designed to detect samples or models according to a specified percentage, and the abnormal samples are determined by the detection rate to group and classify the multi-band multi-dimensional features of the samples.

[0025] In this step, in the case of known approximate abnormal rate or screening rate of the sample, the most abnormal data can be labeled according to the specified percentage of different sample consistency detection models. If the detection rate is not specified, the automatic detection is selected, and then the abnormal values are selected according to the characteristics of the data and the characteristics of different detection models. For example, LOF will determine the factor threshold according to the data distribution after calculating the local outlier factor of all data points, and select the outlier factor that is not in the normal range. Other models also do so, so as to adapt to different application conditions.

[0026] After the grouping and classification of the sample multi-band multi-dimensional features, the sample consistency detection model is used for sample consistency detection, and the results after consistency detection are respectively assigned and output as result vectors to provide the consistency results of the samples for subsequent viewing.

[0027] As can be seen, the method only uses vector samples and remote sensing images to give consistency test results based on the sample data distribution characteristics, and can automatically adjust the abnormal value detection rate, combine the actual sample test, improve the efficiency, and optimize the related work flow.

[0028] Specifically, the multi-band includes all bands of the sample cropped image without specific restrictions; the multi-dimensional features include a plurality of statistical features describing the number of samples, including: minimum value, mean value, maximum value, median, mode, range, standard deviation, 10th percentile, 90th percentile, skewness, kurtosis.

[0029] Specifically, the sample consistency detection model includes LOF (Local Outlier Factor), Isolation Forest, and Elliptic Envelope. The three models will calculate the sample abnormality degree in numerical form. When the detection rate is set, the most abnormal samples of the specified detection rate number are marked as abnormal; when the detection rate is not set, all abnormal values are marked according to the characteristics of the samples calculated by different models, such as the default Elliptic Envelope model marking all data outside the ellipsoid as abnormal values. When the detection rate is set, the specified number of abnormal values are marked, so that the abnormal value detection rate can be automatically adjusted.

[0030] The principles of the three sample consistency detection models are as follows:

[0031] (1) LOF (Local Outlier Factor)

[0032] The LOF algorithm is a density-based anomaly detection algorithm. The concepts involved include: k-distance neighborhood, reachable distance, local reachable density, and local outlier factor (LOF).

[0033] k-distance neighborhood: The k-th distance neighborhood Nk(x) of point x (x∈X), that is, all data points within the k-th distance (including the k-th distance) of x, so the number of k-th neighborhood points of x is N k (x)≥k.

[0034] Reachable distance: The reachable distance from any point o' to point x in the dataset is defined as:

[0035] reach-dist k (x, o′)=max{d k (o′), d(x, o′)}

[0036] Local reachability density: The local reachability density of a data point x is defined as:

[0037]

[0038] Local Outlier Factor (LOF): The local outlier factor of a data point x is expressed as:

[0039]

[0040] The LOF algorithm first determines the k-distance neighborhood of a data point, then calculates the reachable distance from any point in the dataset to the data points within its neighborhood. It then calculates the local reachable density of that point, and finally calculates the local outlier factor (LOF value) for that point. The LOF value for each data point is calculated using the above steps and sorted in reverse order. The larger the value, the more likely it is an outlier. The algorithm determines the degree of anomaly of each data point by comparing the density of the data points in its adjacent area. The lower the density in the adjacent area, the more likely the point is to be an outlier.

[0041] (2)IsolationForest

[0042] The IsolationForest algorithm can be roughly divided into two stages. In the first stage, we need to train t isolated trees to form an isolation forest. Then, each sample point is introduced into each isolated tree in the forest, the average height is calculated, and then the outlier score of each sample point is calculated.

[0043] Step 1: X = {x1, ..., x n} For a given dataset, x i =(x i1 ,..., x id ), randomly select ψ sample points from X to form a subset X' of X and put them into the root node.

[0044] Step2: Randomly select a dimension q from the d dimensions, and randomly generate a cutting point p in the current data, min(x ij , j = q, x ij ∈ X') < p < max(x ij , j = q, x ij ∈ X').

[0045] Step3: This cutting point p generates a hyperplane, which divides the current data space into two subspaces: sample points with specified dimensions less than p are put into the left child node, and those greater than or equal to p are put into the right child node.

[0046] Step4: Recursively execute Step2 and Step3 until all leaf nodes have only one sample point or the isolated tree (iTree) has reached the specified height.

[0047] Step5: Loop Step1 to Step4 until t isolated trees (iTree) are generated.

[0048] In the second stage, for each data point xi, let it traverse each isolated tree (iTree) to calculate the average height h(xi) of the point xi in the forest, and normalize the average height of all points. The formula for calculating the anomaly score is as follows:

[0049]

[0050] where,

[0051]

[0052] (3) EllipticEnvelope

[0053] EllipticEnvelope is used to detect outliers in Gaussian distributed data. Its algorithm flow is to find the center of the data sample, and then draw an ellipsoid around the center. The radius of the ellipsoid is measured by the Mahalanobis distance as the Euclidean distance. After the ellipsoid is drawn, the elements drawn outside the ellipsoid are called outliers.

[0054] The specific software program operation interface of the method is shown in Figure 2 .

[0055] Example 2:

[0056] Referring to Figure 3 As shown in the accompanying drawings, the embodiment provides a sample consistency inspection method for remote sensing images, which comprises a processor 31, a memory 32, and a computer program 33 stored in the memory 32 and capable of running on the processor 31, for example, a sample consistency inspection method for remote sensing images. The processor 31 implements the steps of the above-mentioned embodiment 1 when executing the computer program 33, for example Figure 1 The steps are shown in the accompanying drawings.

[0057] For example, the computer program 33 can be divided into one or more modules / units, which are stored in the memory 32 and executed by the processor 31 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 33 in the sample consistency inspection method for remote sensing images. For example, the computer program 33 can be divided into a conversion module and a matching operation module.

[0058] The sample consistency inspection method for remote sensing images can be a desktop computer, a notebook computer, a palm computer, a cloud server, and other computing devices. The sample consistency inspection method for remote sensing images can include, but is not limited to, a processor 31 and a memory 32. Those skilled in the art can understand that Figure 3 The sample consistency inspection method for remote sensing images is only an example and does not constitute a limitation on the sample consistency inspection method for remote sensing images, which can include more or fewer components than those shown, or combine certain components, or different components, for example, the sample consistency inspection method for remote sensing images can also include input / output devices, network access devices, buses, etc.

[0059] The processor 31 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0060] The memory 32 can be an internal storage unit of the sample consistency inspection method system for remote sensing images, such as a hard disk or a memory of the sample consistency inspection method system for remote sensing images. The memory 32 can also be an external storage device of the sample consistency inspection method system for remote sensing images, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, and the like. Further, the memory 32 can include both an internal storage unit and an external storage device of the sample consistency inspection method system for remote sensing images. The memory 32 is used to store the computer program and other programs and data required by the sample consistency inspection method system for remote sensing images. The memory 32 can also be used to temporarily store data that has been output or will be output.

[0061] Embodiment 3

[0062] The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the method of embodiment 1.

[0063] The computer readable medium can be any means that can contain, store, communicate, propagate or transport the program for use by or in connection with the instruction execution system, apparatus or device. More specific examples (a non-exhaustive list) of the computer readable medium include the following: an electrical connection having one or more wires (electrical devices), a portable computer diskette (magnetic devices), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). In addition, the computer readable medium can even be paper or other suitable medium on which the program is printed, such as by optical scanning, then editing, interpreting or otherwise processing the program to be electronically obtained, and then stored in the computer memory. The computer readable program code can be transmitted or received by a modem, a laser, or other suitable means.

[0064] The above embodiments are only for the purpose of illustrating the technical concept and characteristics of the present application, and the purpose is to enable those skilled in the art to understand the content of the present application and to implement it, and cannot limit the protection scope of the present application. Any equivalent changes or modifications made according to the essence of the present application should be covered within the protection scope of the present application.

Claims

1. A sample consistency verification method for remote sensing images, characterized in that: The method comprises: Group and classify samples based on the classification fields of vector sample attributes; The vector sample range is converted into a coordinate-consistent one with reference to the original remote sensing image, and then the remote sensing image is sampled and cropped to obtain the multi-band pixel information of the sample range; Based on the multi-band information of pixels in the sample range, a statistical method is used to generate the multi-band and multi-dimensional features of the sample; Based on the sample grouping and classification and the abnormality rate or screening rate, the consistency detection rate is designed to detect samples according to a specified percentage or the model adaptively determines the detection rate to detect abnormal samples, so as to group and classify the multi-band and multi-dimensional features of the samples; The multi-band and multi-dimensional features of the grouped and classified samples are tested for sample consistency using a sample consistency detection model. The results after consistency detection are assigned values ​​according to the model results and output as result vectors, providing sample consistency results for subsequent review.

2. The remote sensing image sample consistency verification method according to claim 1, wherein: The multi-band includes all bands of the sample cropped image without specific restrictions.

3. The sample consistency verification method for remote sensing images according to claim 1, characterized in that: The multidimensional features include a variety of statistical features that describe the sample quantity, including: minimum value, mean, maximum value, median, mode, range, standard deviation, 10th percentile, 90th percentile, skewness, and kurtosis.

4. The remote sensing image sample consistency verification method according to claim 1, wherein: The sample consistency detection model is the LOF algorithm, which includes k-distance neighborhood, reachable distance, local reachable density, and local outlier factor LOF; Distance neighborhood: the kth distance neighborhood N of point x (x∈X) k (x), that is, all data points within the kth distance (including the kth distance) of x, so the number of kth neighborhood points of x is N k (x)≥k; Reachable distance: Any point in the dataset The reachable distance to point x is defined as: Local reachability density: The local reachability density of a data point x is defined as: Local outlier factor LOF: The local outlier factor of a data point x is expressed as: The LOF algorithm first determines the k-distance neighborhood of the data point, then calculates the reachable distance from any point in the data set to the data points in its area, then calculates the local reachable density of the point, and finally calculates the local outlier factor LOF value of the point; calculate the LOF value of each data point according to the above steps and arrange them in reverse order.

5. The remote sensing image sample consistency verification method according to claim 1, wherein: The sample consistency detection model is the IsolationForest algorithm, which includes two stages. The first stage is to train t isolated trees to form an isolation forest. The second stage is to bring each sample point into each isolated tree in the forest, calculate the average height, and then calculate the outlier score of each sample point.

6. The remote sensing image sample consistency verification method according to claim 5, wherein: The first stage includes: Step1: X={x1,...,x n } is a given data set, , , randomly selected from X The sample points constitute a subset of X Put in the root node; Step 2: Randomly specify a dimension q from the d dimensions and randomly generate a cutting point p in the current data. ; Step 3: This cutting point p generates a hyperplane, which divides the current data space into two subspaces: the sample points with a specified dimension less than p are placed in the left child node, and the sample points with a dimension greater than or equal to p are placed in the right child node; Step 4: Recursively repeat Step 2 and Step 3 until all leaf nodes have only one sample point or the isolated tree has reached the specified height; Step 5: Loop Step 1 to Step 4 until t isolated trees are generated; The second phase includes: For each data point xi, traverse each isolated tree and calculate the average height h(xi) of point xi in the forest. Normalize the average height of all points. The calculation formula of the outlier score is as follows: in, 。 7. The remote sensing image sample consistency verification method according to claim 1, wherein: The sample consistency detection model is the EllipticEnvelope algorithm. The EllipticEnvelope algorithm process is to find the center of the data sample and then draw an ellipsoid around the center; the radius of the ellipsoid is measured using the Mahalanobis distance as the Euclidean distance; after the ellipsoid is drawn, the elements drawn outside the ellipsoid are called outliers.

8. A sample consistency verification method system for remote sensing images, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images