WO2022252799A1

WO2022252799A1 - Model training method, woodland change detection method, system, and apparatus, and medium

Info

Publication number: WO2022252799A1
Application number: PCT/CN2022/084976
Authority: WO
Inventors: 贾鸿顺; 韩飞; 胡佳辉; 胡兴林
Original assignee: 成都数之联科技股份有限公司
Priority date: 2021-06-04
Filing date: 2022-04-02
Publication date: 2022-12-08
Also published as: CN113095303B; CN113095303A

Abstract

Disclosed are a model training method, a woodland change detection method, system, and apparatus, and a medium, which relate to the field of remote sensing image processing. The present method comprises: obtaining a remote sensing image of a target region; extracting a vegetation coverage region from the remote sensing image, dividing the vegetation coverage region into a woodland region and a non-woodland region, and classifying and labelling the woodland region and the non-woodland region to obtain a classification label map; constructing a classification model, wherein an input of the classification model is the remote sensing image of a preset region, and an output of the classification model is the woodland region and the non-woodland region in the remote sensing image of the preset region; collecting several pieces of feature data of the woodland region and the non-woodland region from the classification label map, and training the classification model on the basis of the feature data. The classification model trained according to the present invention can quickly and accurately obtain the woodland region and the non-woodland region in the remote sensing image.

Description

Model training method and forest land change detection method and system and device and medium

technical field

The invention relates to the field of remote sensing image processing, in particular to a model training method, a forest land change detection method, a system, a device, and a medium.

Background technique

Forest resource change detection technology is a technology to obtain the area of forest land change caused by external factors. At present, the main technology is to rely on manual visual interpretation of high-resolution remote sensing images, compare the two phases of remote sensing images, and manually outline the areas of forest changes. This method is inefficient, and the interpretation results are easily affected by subjective consciousness, resulting in The accuracy of the interpretation results is not high, and it cannot meet the needs of long-term and high-frequency change detection for large-scale forest resources.

Contents of the invention

In order to solve the above problems, the present invention provides a model training method, a forest change detection method, a system, a device, and a medium. The classification model trained by the present invention can quickly and accurately obtain forest areas and non-forest areas in remote sensing images.

To achieve the above object, the present invention provides a model training method, the method comprising:

Obtain remote sensing images of the target area;

Extracting a vegetation coverage area from the remote sensing image, dividing the vegetation coverage area into a woodland area and a non-woodland area, classifying and marking the woodland area and the non-woodland area to obtain a classification label map;

Constructing a classification model, the input of the classification model is the remote sensing image of the preset area, and the output of the classification model is the woodland area and the non-woodland area in the remote sensing image of the preset area;

A number of feature data of the woodland area and the non-woodland area are collected from the classification label map, and the classification model is trained based on the feature data.

Among them, the principle of this method is to firstly obtain the remote sensing image of the target area; then extract the vegetation coverage area from the remote sensing image, divide the vegetation coverage area into a woodland area and a non-woodland area, divide the woodland area and the Classify and mark the non-woodland area to obtain a classification label map, which is convenient for model training after marking, collect some feature data of the woodland area and the non-woodland area from the classification label map, and train based on the feature data In the classification model, there are many kinds of data in the classification label map, and the method extracts feature data reflecting woodland areas and non-woodland areas, and the classification model obtained by training the feature data can accurately obtain remote sensing images in subsequent applications The woodland area and non-woodland area in the forest area, the classification model can be obtained through the above training method, and the classification model can quickly and accurately output the woodland area and non-woodland area in the remote sensing image, which can replace the traditional artificial outline of the forest change. area, improving efficiency and accuracy.

Wherein, in this method, the characteristic data include one or more of the following types of data: spectral characteristic data, normalized difference vegetation index NDVI data, normalized difference water body index NDWI data and terrain characteristic data.

Among them, any one of spectral feature data, normalized difference vegetation index NDVI data, normalized difference water index NDWI data and topographic feature data can determine the woodland area and non-woodland area in the remote sensing image.

Wherein, the feature data is spectral feature data. Spectral feature data is the most accurate reflection of the woodland area and non-woodland area in remote sensing images among the above data. Therefore, when the feature data is spectral feature data, it can accurately reflect the woodland area and non-woodland area in remote sensing images.

Among them, it is easy to have the risk of misjudgment when only relying on one kind of data to judge the woodland area and non-woodland area in the remote sensing image. The accuracy of non-woodland areas, such as spectral feature data, normalized difference vegetation index NDVI data, normalized difference water index NDWI data and topographic feature data, can be combined to improve the accuracy of the overall judgment.

Preferably, in this method, after the step of obtaining the remote sensing image of the target area, and before the step of extracting the vegetation coverage area from the remote sensing image, the method further includes the step of: atmospherically analyzing the remote sensing image of the target area Correction processing. This method can correct the radiation error in the remote sensing image of the sub-cavity through atmospheric correction, and improve the accuracy of the training method.

Preferably, after the step of obtaining the remote sensing image of the target area and before the step of extracting the vegetation coverage area from the remote sensing image, the method further includes the step of removing clouds and cloud shadows in the remote sensing image of the target area. Among them, when there are clouds and cloud shadows in the remote sensing image, it will block the woodland and non-woodland areas, resulting in inaccurate model training. In order to improve the accuracy of model training, this method removes the remote sensing image of the target area after obtaining the remote sensing image. Clouds and cloud shadows in images.

Preferably, the method adopts a mask method to remove clouds and cloud shadows in the remote sensing image of the target area. Among them, the mask method is to first obtain the position of the cloud and cloud shadow, and then obtain the corresponding mask, and then use the mask to cover the cloud and cloud shadow, and then use the historical cloud-free and cloud-free image to replace the original cloud and cloud shadowed areas, and then obtained new cloud-free and cloud-free images. The cloud and cloud shadow in the remote sensing image can be quickly and accurately removed through the mask method, and the data of the remote sensing image can be retained to the greatest extent.

Preferably, the cloud and cloud shadows in the remote sensing image of the target area are removed by using a mask in the method, which specifically includes:

Performing cloud and cloud shadow detection on the remote sensing image to obtain a cloud and cloud shadow mask;

From the historical remote sensing image data corresponding to the target area, select the historical remote sensing image with the lowest cloud and cloud shadow coverage as the base map;

Use standard remote sensing images of other phases in the same period as the remote sensing image to fill the cloud and cloud shadow area corresponding to the cloud and cloud shadow mask in the base map, and the standard remote sensing image is the same as the remote sensing image The remote sensing image with the lowest cloud and cloud shadow coverage among the remote sensing images of other time phases in the same period.

Preferably, the method performs relative radiation correction processing on remote sensing images of other phases in the same period as the remote sensing images before filling the base map, in order to reduce radiation differences between remote sensing images of different phases.

Preferably, the method includes: calculating the normalized normalized vegetation index NDVI of the remote sensing image, determining a threshold based on the normalized normalized vegetation index NDVI, performing binarization based on the threshold to obtain a vegetation mask, using the The vegetation mask performs mask processing on the remote sensing image to obtain the vegetation coverage area. This method obtains the vegetation mask by calculating the normalized difference vegetation index NDVI of the remote sensing image, and then uses the vegetation mask to mask the remote sensing image to obtain the vegetation coverage area.

Preferably, the calculation method of the normalized difference vegetation index NDVI of the remote sensing image is:

Among them, NIR is the near-infrared band, and RED is the red band.

Preferably, the method comprises:

Carry out multi-scale segmentation of the vegetation coverage area to obtain a segmentation map, convert the segmentation map into a vector to obtain a segmentation vector; create a new classification field in the segmentation vector, and select forest land and non-forest map spots based on the segmentation vector , and assign different values to the forest map spots and non-forest map spots according to the category, convert the segmentation vector into a raster map according to the classification field, and obtain a classification label map.

Preferably, the calculation method of the normalized normalized water index NDWI is:

Among them, NIR is the near-infrared band, and GREEN is the green band.

Preferably, the method comprises:

The remote sensing image of the target area includes a first remote sensing image of the target area in a first period and a second remote sensing image of the target area in a second period after the first period;

Obtaining a first classification label map based on the first remote sensing image;

Obtaining a second classification label map based on the second remote sensing image;

Construct the first classification model and the second classification model;

Collecting feature data of woodland areas and non-woodland areas from the first classification label map to obtain a first training sample; using the first training sample to train the first classification model to obtain a third classification model;

Collecting feature data of woodland areas and non-woodland areas from the second classification label map to obtain second training samples; using the second training samples to train the second classification model to obtain a fourth classification model.

Among them, this method trains two models respectively, namely the third classification model and the fourth classification model, the third classification model is used to process the remote sensing images corresponding to the first period, and the fourth classification model is used to process the remote sensing images corresponding to the second period Image, the purpose of this design is that the preset area has different data in different periods, that is, the data before and after the forest land change, and the corresponding model parameters are also different. Therefore, two models were trained respectively, and the two models were used to process the remote sensing images before and after the change, namely One model is used to process the remote sensing image before the forest land change, and the other model is used to process the remote sensing image after the forest land change, so that the forest area and non-forest area corresponding to the remote sensing image can be obtained more accurately.

Preferably, the method comprises:

Obtaining the third classification model and the fourth classification model by using the model training method;

Obtaining the remote sensing image x of the area to be detected in the A period and the remote sensing image y of the area to be detected in the B period, wherein the A period is before the B period;

Input the remote sensing image x into the third classification model, and output the woodland area K in the remote sensing image x;

Input the remote sensing image y into the fourth classification model, and output the forest area P in the remote sensing image y;

Based on the difference between the forest area K and the forest area P, a forest change detection result map of the area to be detected is obtained.

Among them, this method can obtain the woodland area in the remote sensing images of the area to be detected in different periods, and the change of the woodland area in the actual section of the area can be obtained through the change results of the woodland area.

Preferably, the method further includes: converting the forestland change detection result map shown into a vector to obtain forestland change patterns. The change of forest land can be displayed more intuitively and vividly through the map of forest land change.

The present invention also provides a model training system, said system comprising:

The first obtaining unit is used to obtain the remote sensing image of the target area;

a classification label map obtaining unit, configured to extract a vegetation coverage area from the remote sensing image, divide the vegetation coverage area into a woodland area and a non-woodland area, and classify and mark the woodland area and the non-woodland area, Obtain the classification label map;

A classification model construction unit, configured to construct a classification model, the input of the classification model is the input remote sensing image of the preset area, and the output of the classification model is the woodland area and the non-woodland area in the input remote sensing image;

A first training unit, configured to collect spectral features of the woodland area and the non-woodland area from the classification label map, obtain training samples based on the spectral features; use the training samples to train the classification model.

The present invention also provides a model training device, comprising a memory, a processor, and a computer program stored in the memory and operable on the processor, and the model is realized when the processor executes the computer program. The steps of the training method.

The present invention also provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the model training method are implemented.

The present invention also provides a forest land change detection system, said system comprising:

A second training unit, configured to train and obtain the third classification model and the fourth classification model by using the model training method;

The second obtaining unit is used to obtain the remote sensing image x of the area to be detected in the A period and the remote sensing image y of the area to be detected in the B period, wherein the A period is before the B period;

A first processing unit, configured to input the remote sensing image x into the third classification model, and output the woodland area K in the remote sensing image x;

The second processing unit is configured to input the remote sensing image y into the fourth classification model, and output the forest area P in the remote sensing image y;

A comparing unit, configured to obtain a forest change detection result map of the area to be detected based on the difference between the forest area K and the forest area P.

The present invention also provides a forest land change detection device, which includes a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the computer program, the Steps of forest land change detection method.

The present invention also provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the forestland change detection method are implemented.

One or more technical solutions provided by the present invention have at least the following technical effects or advantages:

The classification model trained by the invention can quickly and accurately obtain the woodland area and non-woodland area in the remote sensing image.

The present invention preprocesses the two phases of remote sensing images of the target area and extracts the woodland area, and then automatically extracts the forest land change area, intelligently realizes the long-term change detection of large-scale forest resources, and solves the low efficiency and inaccuracy of manual visual interpretation High issues, to provide support for the decision-making of forest land supervision.

Description of drawings

The drawings described here are used to provide a further understanding of the embodiments of the present invention, constitute a part of the present invention, and do not constitute a limitation to the embodiments of the present invention;

Fig. 1 is the schematic flow chart of model training method;

Figure 2 is a schematic diagram of the composition of the model training system.

Detailed ways

In order to understand the above-mentioned purpose, features and advantages of the present invention more clearly, the present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and the features in the embodiments may be combined with each other without conflicting with each other.

In the following description, many specific details are set forth in order to fully understand the present invention. However, the present invention can also be implemented in other ways different from the scope of this description. Therefore, the protection scope of the present invention is not limited by the following disclosure. limitations of specific examples.

It should be understood that "system", "device", "unit" and/or "module" as used herein is a method for distinguishing different components, elements, components, parts or assemblies of different levels. However, the words may be replaced by other expressions if other words can achieve the same purpose.

As indicated in the specification and claims, the terms "a", "an", "an" and/or "the" are not specific to the singular and may include the plural unless the context clearly indicates an exception. Generally speaking, the terms "comprising" and "comprising" only suggest the inclusion of clearly identified steps and elements, and these steps and elements do not constitute an exclusive list, and the method or device may also contain other steps or elements.

The flowchart is used in this specification to illustrate the operations performed by the system according to the embodiment of this specification. It should be understood that the preceding or following operations are not necessarily performed in the exact order. Instead, various steps may be processed in reverse order or simultaneously. At the same time, other operations can be added to these procedures, or a certain step or steps can be removed from these procedures.

Embodiment one

Please refer to FIG. 1. FIG. 1 is a schematic flow chart of a model training method. Embodiment 1 of the present invention provides a model training method. The method includes:

Obtain remote sensing images of the target area;

Wherein, the target area may be any area, such as mountains, plains, cities, villages, etc., and the present invention does not specifically limit the specific type and location of the target area.

Wherein, the remote sensing image acquisition method in this embodiment can be any one of various methods, such as obtaining through a satellite system, or obtaining through a network, or obtaining through a database, etc. For limitation, for example, the remote sensing image of this method can be obtained through Sentinel-2, and can also be obtained through other methods.

Wherein, in Embodiment 1 of the present invention, the feature data includes one or more of the following types of data: spectral feature data, normalized difference vegetation index NDVI data, normalized difference water index NDWI data and terrain feature data.

Among them, the spectral characteristics of ground objects are that any ground objects in nature have their own electromagnetic radiation laws, such as reflection and absorption of certain bands of ultraviolet, visible light, infrared and microwave, and they all have the ability to emit certain infrared rays. , Microwave characteristics; a small number of surface objects also have the characteristics of transmitting electromagnetic waves, which is called the spectral characteristics of surface objects.

Among them, the normalized difference vegetation index is one of the important parameters reflecting the growth and nutritional information of crops. The application of NDVI: detection of vegetation growth status, vegetation coverage and elimination of partial radiation errors, etc., this embodiment uses it to detect vegetation coverage.

Among them, NDWI (Normalized Difference Water Index, normalized water index), uses a specific band of remote sensing images to perform normalized difference processing to highlight the water body information in the image, and the vegetation moisture index NDWI is based on mid-infrared and near-infrared bands The normalized ratio index of .

Among them, the topographic feature data include topographic factors, such as slope, slope aspect, slope change rate and so on.

Wherein, in the embodiment of the present invention, after the step of obtaining the remote sensing image of the target area, and before the step of extracting the vegetation coverage area from the remote sensing image, the method further includes the step of: performing the remote sensing image of the target area Atmospheric correction processing. Atmospheric correction means that the total radiance of the ground target finally measured by the sensor is not a reflection of the real reflectance of the ground surface, which includes the radiation amount error caused by atmospheric absorption, especially scattering. Atmospheric correction is the process of eliminating these radiation errors caused by atmospheric influences and retrieving the true surface reflectance of ground objects.

Wherein, in the embodiment of the present invention, the sen2cor tool is used to perform atmospheric correction on the Sentinel-2 remote sensing image in two periods (if the processing level of the Sentinel-2 remote sensing image is L2A, this step is skipped); this method can also use other The tools or methods to perform atmospheric correction on remote sensing images, this method is not specifically limited.

Wherein, in the embodiment of the present invention, after the step of obtaining the remote sensing image of the target area and before the step of extracting the vegetation coverage area from the remote sensing image, the method further includes the step of removing the remote sensing image of the target area clouds and cloud shadows.

Wherein, in the embodiment of the present invention, the method adopts a mask method to remove clouds and cloud shadows in the remote sensing image of the target area.

Wherein, in the embodiment of the present invention, the removal of clouds and cloud shadows in the remote sensing image of the target area by using a mask method specifically includes:

Wherein, in the embodiment of the present invention, relative radiation correction processing is performed on remote sensing images of other phases in the same period as the remote sensing images before filling the base map.

This method performs cloud removal processing on the obtained remote sensing images, and the cloud removal processing uses multi-temporal remote sensing images to replace corresponding regions. First, use fmask to detect clouds and cloud shadows on the atmospherically corrected multi-temporal images (cloud shadows are cloud shadows) Shadows produced on the ground) to obtain clouds and cloud shadow masks, and then select remote sensing images with the least coverage of clouds and cloud shadows in each period as the base map, and use remote sensing images of other phases in the same period (according to cloud coverage from small To sort) to fill the cloud and cloud shadow area corresponding to the mask in the base map. Before filling the base map, the remote sensing images of other time phases need to be corrected for relative radiation, so as to reduce the radiation difference between remote sensing images of different time phases.

Wherein, in the embodiment of the present invention, the method includes: calculating the normalized normalized vegetation index NDVI of the remote sensing image, determining a threshold based on the normalized normalized vegetation index NDVI, and performing binarization based on the threshold to obtain vegetation A mask, using the vegetation mask to perform mask processing on the remote sensing image to obtain the vegetation coverage area.

According to the red band (4 bands) and near-infrared band (8 bands), calculate the normalized difference vegetation index NDVI of the two phases of remote sensing images, determine the threshold value and binarize to obtain the vegetation mask, and use the vegetation mask to mask the two phases of images For processing, only the vegetation coverage area is retained, and the non-vegetation area is set to no data (nodata). The calculation method of the normalized difference vegetation index NDVI of the remote sensing image is:

Among them, NIR is the near-infrared band, and RED is the red band.

Wherein, in the embodiment of the present invention, the method includes: performing multi-scale segmentation of the vegetation coverage area to obtain a segmentation map, converting the segmentation map into a vector to obtain a segmentation vector; creating a new classification in the segmentation vector Field, select forest land and non-forest map spots based on the segmentation vector, and assign different values to the forest map spots and non-forest map spots according to the category, convert the segmentation vector into a raster map according to the classification field, and obtain Classification label map.

Among them, multi-scale segmentation is performed on the multi-spectral remote sensing image after the two-stage mask to obtain a segmentation map, and then the segmentation map is converted into a shp vector to obtain a segmentation vector; a new classification field is created in the segmentation vector, and a uniform distribution is selected based on the segmentation vector Forest land and non-forest map spots, and assign values of 1 and 2 for forest land and non-forest land respectively according to the category. After the selection is completed, convert the vector into a raster map according to the classification field to obtain a classification label map.

Wherein, in the embodiment of the present invention, the calculation method of the normalized normalized water index NDWI is:

Among them, NIR is the near-infrared band, and GREEN is the green band.

Wherein, in the embodiment of the present invention, the method includes:

Construct the first classification model and the second classification model;

Use the green band (3 bands) and near infrared band (8 bands) of the two phases of multispectral images to calculate the normalized water index NDWI, and sample the digital elevation model DEM resolution to 10m to facilitate the collection of terrain features and use classification labels The spectral features of forest land and non-forest land, NDVI, NDWI, and DEM terrain features are collected to construct sample sets, and then the random forest algorithm is used to train two classification models based on the sample sets.

Among them, the Digital Elevation Model (Digital Elevation Model), referred to as DEM, is to realize the digital simulation of the ground terrain through limited terrain elevation data (that is, the digital expression of the terrain surface shape). A solid ground model of elevation is a branch of the Digital Terrain Model (DTM), from which various other terrain feature values can be derived. It is generally believed that DTM describes the spatial distribution of various geomorphic factors including elevation, such as slope, slope aspect, slope change rate and other factors including linear and nonlinear combinations, and DEM is a zero-order simple single-item digital geomorphic model , other landform characteristics such as slope, aspect and slope change rate can be derived on the basis of DEM.

Embodiment two

Embodiment 2 of the present invention provides a forest land change detection method, the method comprising:

Using the model training method described in Embodiment 1 to train and obtain the third classification model and the fourth classification model;

Wherein, in the second embodiment of the present invention, the method further includes: converting the shown forestland change detection result map into a vector to obtain forestland change map spots.

Use the trained classification model to classify and predict the two phases of multispectral images, and obtain forest land and non-forest land classification maps;

Use the difference between the previous classification map and the next classification map to obtain the forest land change detection result map. The positive value is the forest land reduction area, and the negative value is the forest land increase area. Convert the forest land change detection result map into a vector to obtain the forest land change map.

Embodiment Three

Please refer to FIG. 2. FIG. 2 is a schematic diagram of the composition of the model training system. Embodiment 3 of the present invention provides a model training system. The system includes:

Embodiment four

Embodiment 4 of the present invention provides a model training device, including a memory, a processor, and a computer program stored in the memory and operable on the processor, and the processor implements the computer program when executing the computer program. The steps of the model training method are described.

Embodiment five

Embodiment 5 of the present invention provides a computer-readable storage medium, where the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the model training method are implemented.

Embodiment six

Embodiment 6 of the present invention provides a forest land change detection system, the system comprising:

The second training unit is used for training the model training method described in Embodiment 1 to obtain the third classification model and the fourth classification model;

Embodiment seven

Embodiment 7 of the present invention provides a forest land change detection device, including a memory, a processor, and a computer program stored in the memory and operable on the processor. When the processor executes the computer program, it realizes The steps of the forest land change detection method described in the second embodiment.

Embodiment eight

Embodiment 8 of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the forestland change detection method described in Embodiment 2 are realized.

Wherein, the processor can be a central processing unit (CPU, Central Processing Unit), and can also be other general-purpose processors, digital signal processors (digital signal processors), application specific integrated circuits (Application Specific Integrated Circuits), off-the-shelf programmable Gate array (Field programmable gate array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor, or the processor may be any conventional processor, or the like.

The memory can be used to store the computer program and/or module, and the processor realizes various functions of the forest land change detection device or the model training device in the invention by running or executing the data stored in the memory. The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application required by a function (such as a sound playback function, an image playback function, etc.) and the like. In addition, the memory can include high-speed random access memory, and can also include non-volatile memory, such as hard disk, internal memory, plug-in hard disk, smart memory card, secure digital card, flash memory card, at least one magnetic disk storage device, flash memory device, or other volatile solid-state memory devices.

If the forest change detection device or the model training device is realized in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention realizes all or part of the processes in the methods of the above-mentioned embodiments, and can also be stored in a computer-readable storage medium through a computer program. When the computer program is executed by a processor, it can realize the implementation of each of the above-mentioned methods. example steps. Wherein, the computer program includes computer program code, object code form, executable file or some intermediate form and the like. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, U disk, removable hard disk, magnetic disk, optical disk, computer memory, read-only memory, random access memory, point carrier signal , telecommunication signals, and software distribution media. It should be noted that the content contained in the computer-readable medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction.

The present invention has described the basic concepts. Obviously, for those skilled in the art, the above detailed disclosure is only an example, and does not constitute a limitation to this specification. Although not expressly stated here, various modifications, improvements, and corrections to this description may occur to those skilled in the art. Such modifications, improvements and corrections are suggested in this specification, so such modifications, improvements and corrections still belong to the spirit and scope of the exemplary embodiments of this specification.

Meanwhile, this specification uses specific words to describe the embodiments of this specification. For example, "one embodiment", "an embodiment", and/or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of this specification. Therefore, it should be emphasized and noted that two or more references to "an embodiment" or "an embodiment" or "an alternative embodiment" in different places in this specification do not necessarily refer to the same embodiment . In addition, certain features, structures or characteristics in one or more embodiments of this specification may be properly combined.

In addition, those skilled in the art will understand that various aspects of this specification can be illustrated and described by several patentable categories or situations, including any new and useful process, machine, product or combination of substances, or any combination of them Any new and useful improvements. Correspondingly, various aspects of this specification may be entirely executed by hardware, may be entirely executed by software (including firmware, resident software, microcode, etc.), or may be executed by a combination of hardware and software. The above hardware or software may be referred to as "block", "module", "engine", "unit", "component" or "system". Additionally, aspects of this specification may be embodied as a computer product comprising computer readable program code on one or more computer readable media.

A computer storage medium may contain a propagated data signal embodying a computer program code, for example, in baseband or as part of a carrier wave. The propagated signal may have various manifestations, including electromagnetic form, optical form, etc., or a suitable combination. A computer storage medium may be any computer-readable medium, other than a computer-readable storage medium, that can be used to communicate, propagate, or transfer a program for use by being coupled to an instruction execution system, apparatus, or device. Program code residing on a computer storage medium may be transmitted over any suitable medium, including radio, electrical cable, fiber optic cable, RF, or the like, or combinations of any of the foregoing.

The computer program codes required for the operation of each part of this manual can be written in any one or more programming languages, including object-oriented programming languages such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python etc., conventional procedural programming languages such as C language, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages such as Python, Ruby and Groovy, or other programming languages. The program code may run entirely on the user's computer, or as a stand-alone software package, or run partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the latter case, the remote computer can be connected to the user computer through any form of network, such as a local area network (LAN) or wide area network (WAN), or to an external computer (such as through the Internet), or in a cloud computing environment, or as a service Use software as a service (SaaS).

In addition, unless explicitly stated in the claims, the order of processing elements and sequences described in this specification, the use of numbers and letters, or the use of other names are not used to limit the sequence of processes and methods in this specification. While the foregoing disclosure has discussed by way of various examples some embodiments of the invention that are presently believed to be useful, it should be understood that such detail is for illustrative purposes only and that the appended claims are not limited to the disclosed embodiments, but rather, the claims The claims are intended to cover all modifications and equivalent combinations that fall within the spirit and scope of the embodiments of this specification. For example, although the system components described above may be implemented by hardware devices, they may also be implemented by a software-only solution, such as installing the described system on an existing server or mobile device.

In the same way, it should be noted that in order to simplify the expression disclosed in this specification and help the understanding of one or more embodiments of the invention, in the foregoing description of the embodiments of this specification, sometimes multiple features are combined into one embodiment, drawings or descriptions thereof. This method of disclosure does not, however, imply that the subject matter of the specification requires more features than are recited in the claims. Indeed, embodiment features are less than all features of a single foregoing disclosed embodiment.

Each patent, patent application, patent application publication, and other material, such as article, book, specification, publication, document, etc., cited in this specification is hereby incorporated by reference in its entirety. Application history documents that are inconsistent with or conflict with the content of this specification are excluded, and documents (currently or later appended to this specification) that limit the broadest scope of the claims of this specification are also excluded. It should be noted that if there is any inconsistency or conflict between the descriptions, definitions, and/or terms used in the accompanying materials of this manual and the contents of this manual, the descriptions, definitions and/or terms used in this manual shall prevail .

While preferred embodiments of the invention have been described, additional changes and modifications to these embodiments can be made by those skilled in the art once the basic inventive concept is appreciated. Therefore, it is intended that the appended claims be construed to cover the preferred embodiment as well as all changes and modifications which fall within the scope of the invention.

Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

Model training method, is characterized in that, described method comprises:

Obtain remote sensing images of the target area;

Extracting a vegetation coverage area from the remote sensing image, dividing the vegetation coverage area into a woodland area and a non-woodland area, classifying and marking the woodland area and the non-woodland area to obtain a classification label map;

Constructing a classification model, the input of the classification model is the remote sensing image of the preset area, and the output of the classification model is the woodland area and the non-woodland area in the remote sensing image of the preset area;

A number of feature data of the woodland area and the non-woodland area are collected from the classification label map, and the classification model is trained based on the feature data.
The classification model training method according to claim 1, wherein the feature data includes one or more of the following types of data: spectral feature data, normalized normalized vegetation index NDVI data, normalized normalized water index NDWI data and terrain feature data.
The model training method according to claim 1, characterized in that, after the step obtains the remote sensing image of the target area, and before the step extracts the vegetation coverage area from the remote sensing image, the method further includes the step of: The remote sensing images of the target area are subjected to atmospheric correction processing.
The model training method according to claim 1, characterized in that, after the remote sensing image of the target area is obtained in the step, and before the vegetation coverage area is extracted from the remote sensing image in the step, the method further includes the step of removing the Clouds and cloud shadows in remote sensing images of target areas.
The model training method according to claim 4, characterized in that the method adopts a mask method to remove clouds and cloud shadows in the remote sensing image of the target area.
The model training method according to claim 5, wherein said removing clouds and cloud shadows in the remote sensing image of the target area by using a mask method specifically includes:

Performing cloud and cloud shadow detection on the remote sensing image to obtain a cloud and cloud shadow mask;

From the historical remote sensing image data corresponding to the target area, select the historical remote sensing image with the lowest cloud and cloud shadow coverage as the base map;

Use standard remote sensing images of other phases in the same period as the remote sensing image to fill the cloud and cloud shadow area corresponding to the cloud and cloud shadow mask in the base map, and the standard remote sensing image is the same as the remote sensing image The remote sensing image with the lowest cloud and cloud shadow coverage among the remote sensing images of other time phases in the same period.
The model training method according to claim 6, characterized in that before filling the base map, relative radiation correction processing is performed on remote sensing images of other phases in the same period as the remote sensing images.
The model training method according to claim 1, characterized in that, the method comprises: calculating the normalized normalized vegetation index (NDVI) of the remote sensing image, determining a threshold based on the normalized normalized vegetation index (NDVI), and performing based on the threshold A vegetation mask is obtained through binarization, and the remote sensing image is masked using the vegetation mask to obtain the vegetation coverage area.
The model training method according to claim 8, wherein the calculation method of the normalized difference vegetation index NDVI of the remote sensing image is:

Among them, NIR is the near-infrared band, and RED is the red band.
The model training method according to claim 1, wherein the method comprises:

Carry out multi-scale segmentation of the vegetation coverage area to obtain a segmentation map, convert the segmentation map into a vector to obtain a segmentation vector; create a new classification field in the segmentation vector, and select forest land and non-forest map spots based on the segmentation vector , and assign different values to the forest map spots and non-forest map spots according to the category, convert the segmentation vector into a raster map according to the classification field, and obtain a classification label map.
The model training method according to claim 10, wherein the calculation method of the normalized normalized water index NDWI is:

Among them, NIR is the near-infrared band, and GREEN is the green band.
The model training method according to claim 1, wherein the method comprises:

The remote sensing image of the target area includes a first remote sensing image of the target area in a first period and a second remote sensing image of the target area in a second period after the first period;

Obtaining a first classification label map based on the first remote sensing image;

Obtaining a second classification label map based on the second remote sensing image;

Construct the first classification model and the second classification model;

Collecting feature data of woodland areas and non-woodland areas from the first classification label map to obtain a first training sample; using the first training sample to train the first classification model to obtain a third classification model;

Collecting feature data of woodland areas and non-woodland areas from the second classification label map to obtain second training samples; using the second training samples to train the second classification model to obtain a fourth classification model.
Forest land change detection method, is characterized in that, described method comprises:

Adopting the model training method described in claim 12 to train and obtain the third classification model and the fourth classification model;

Obtaining the remote sensing image x of the area to be detected in the period A and the remote sensing image y of the area to be detected in the period B, wherein the period A is before the period B;

Input the remote sensing image x into the third classification model, and output the woodland area K in the remote sensing image x;

Input the remote sensing image y into the fourth classification model, and output the forest area P in the remote sensing image y;

Based on the difference between the forest area K and the forest area P, a forest change detection result map of the area to be detected is obtained.
The forest land change detection method according to claim 13, further comprising: converting the forest land change detection result map shown into a vector to obtain forest land change map spots.
A model training system, characterized in that the system includes:

The first obtaining unit is used to obtain the remote sensing image of the target area;

a classification label map obtaining unit, configured to extract a vegetation coverage area from the remote sensing image, divide the vegetation coverage area into a woodland area and a non-woodland area, and classify and mark the woodland area and the non-woodland area, Obtain the classification label map;

A classification model construction unit, configured to construct a classification model, the input of the classification model is the input remote sensing image of the preset area, and the output of the classification model is the woodland area and the non-woodland area in the input remote sensing image;

A first training unit, configured to collect spectral features of the woodland area and the non-woodland area from the classification label map, obtain training samples based on the spectral features; use the training samples to train the classification model.
A model training device, comprising a memory, a processor, and a computer program stored in the memory and operable on the processor, characterized in that, when the processor executes the computer program, the computer program according to claim 1 is realized. The steps of any one of the model training methods in -12.
A computer-readable storage medium, the computer-readable storage medium stores a computer program, characterized in that, when the computer program is executed by a processor, it implements the steps of the model training method according to any one of claims 1-12 .
The woodland change detection system is characterized in that the system includes:

A second training unit, configured to train and obtain the third classification model and the fourth classification model by using the model training method according to claim 12;

The second obtaining unit is used to obtain the remote sensing image x of the area to be detected in the A period and the remote sensing image y of the area to be detected in the B period, wherein the A period is before the B period;

A first processing unit, configured to input the remote sensing image x into the third classification model, and output the woodland area K in the remote sensing image x;

The second processing unit is configured to input the remote sensing image y into the fourth classification model, and output the forest area P in the remote sensing image y;

A comparing unit, configured to obtain a forest change detection result map of the area to be detected based on the difference between the forest area K and the forest area P.
A forest land change detection device, comprising a memory, a processor, and a computer program stored in the memory and operable on the processor, characterized in that, when the processor executes the computer program, the computer program according to the claims is realized. 13. Steps of the forestland change detection method.
A computer-readable storage medium, the computer-readable storage medium storing a computer program, characterized in that, when the computer program is executed by a processor, the steps of the forestland change detection method according to claim 13 are realized.