CN115468917A - Method and system for extracting crop information of farmland plot based on high-resolution remote sensing - Google Patents

Abstract

The invention belongs to the technical field of farmland information extraction, and discloses a method and a system for extracting farmland plot crop information based on high-resolution remote sensing, wherein a partitioned and classified image interpretation method is adopted, and a research area is divided into a plurality of sub-areas based on the differences of different landform units in the research area in the aspects of remote sensing image spectrum, surface vegetation density, residential distribution and the like; and precision evaluation and verification, namely calculating the overall classification precision and kappa coefficient of various land features by establishing a confusion matrix to judge the precision of the classification result. According to the method, the traditional Landsat remote sensing image rough classification and the Google Earth high-precision image fine classification technology are organically combined, so that the extraction precision of the type and the space distribution information of the farmland land crops can be greatly improved; by adopting the image interpretation method of partition classification, the interference of the same object and different spectrum or the same spectrum of foreign objects of the remote sensing image in the region can be effectively reduced, and the classification precision is improved.

Description

Method and system for extracting crop information of farmland plot based on high-resolution remote sensing

Technical Field

The invention belongs to the technical field of farmland information extraction, and particularly relates to a method and a system for extracting farmland plot crop information based on high-resolution remote sensing.

Background

At present, high-precision agricultural land utilization and farmland plot crop information are important prerequisites for developing modern agricultural development space planning and farmland resource protection in China. However, the field crop information data used in the existing agricultural production is mainly obtained by manual investigation or based on some relatively low-resolution digital image manner. The information acquisition mode is time-consuming and labor-consuming, and the precision is difficult to guarantee. Although some large-scale remote sensing technologies are applied to farmland information extraction in recent years, related technical means are often low in precision, and it is difficult to quickly and accurately acquire refined planting information such as crop types, distribution areas and the like of dimensions of farmland plots. In addition, the traditional remote sensing image classification method generally selects training samples by using a uniform standard in the whole research area range, and although the method is simple and convenient, the problems of same-object different spectrums, same-object same spectrums and the like are not easy to solve, and the phenomena of wrong division and missing division often occur.

Through the above analysis, the problems and defects of the prior art are as follows:

(1) The land utilization data used in the existing agricultural production is mainly acquired through manual investigation or based on a digital image mode with low resolution, the traditional acquisition mode not only wastes time and labor, but also is difficult to guarantee the precision, and the crop planting information of the dimensions of the farmland plots is difficult to acquire quickly and accurately.

(2) The traditional remote sensing image classification method is not easy to solve the problems of same-object different spectrums, same-object spectrums and the like, and the phenomena of wrong division and missing division often occur, so that the error between the image classification result and the actual situation is larger.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides a method and a system for extracting crop information of a farmland plot based on high-resolution remote sensing.

The invention is realized in such a way that a method for extracting farmland plot crop information based on high-resolution remote sensing comprises the following steps:

step one, classifying remote sensing images in a partitioning manner: adopting a subarea classification image interpretation method, dividing a research area into a plurality of sub-areas based on the difference of different landform units in the research area in the aspects of remote sensing image spectrum, surface vegetation density, residential point distribution and the like;

step two, precision evaluation and verification: and calculating the overall classification precision and kappa coefficient of various ground features by establishing a confusion matrix to judge the precision of the classification result.

The remote sensing image partition classification in the first step specifically comprises:

(1) Pretreatment before classification: preprocessing image data;

(2) And (4) partition classification: adopting a maximum likelihood method to carry out supervision and classification;

(3) And (4) classification post-treatment: and splicing the partitioning results.

The image data preprocessing in the step (1) comprises geometric correction, projection transformation, image mosaic and research area cutting extraction.

And (2) the remote sensing image of the research area after correction and registration processing in the step (1) is superposed with the administrative boundary, and then the administrative partition boundary of the typical area is used for cutting the image of the whole area into a plurality of sub-areas.

And (3) selecting the remote sensing images with the most abundant information content in the step (2) to combine the wave bands of 5, 4 and 3, and generating a false color composite image by matching three colors of red, green and blue.

In the step (2), during classification, the latest Google Earth high-definition image of a research area is used as a contrast, and a land utilization planning map is used as a reference.

And (3) when the training samples are collected in the areas in the step (2), the sample points are uniformly distributed in the area range, and the pixels are obtained by selecting the pure spectrum.

When the partition results are spliced in the step (3), according to land utilization classification results obtained by different partition interpretation, a decision tree is established in remote sensing image processing software to unify land type codes and color matching of each partition, image splicing and boundary fusion processing are carried out, and the partition interpretation results are synthesized into a complete whole-partition land utilization and crop type distribution map; some of the fine patches are merged and smoothed using a class clustering method.

Further, in the second precision evaluation and verification, a certain number of verification points are randomly generated in the research range, the real types of the verification points are judged according to Google earth high-resolution images or ground actual survey data, the real types are compared with remote sensing classification results, and the overall precision is counted.

Another object of the present invention is to provide a technical system for extracting crop information of farmland plots based on high resolution remote sensing, comprising:

the image preprocessing module is used for preprocessing the image data before classification;

the partition classification module is used for supervising and classifying the remote sensing image by adopting a maximum likelihood method;

the classification post-processing module is used for splicing the partitioning results;

and the precision evaluation and verification module is used for calculating the overall classification precision and kappa coefficient of various ground features by establishing a confusion matrix to judge the precision of the classification result.

In combination with the technical solutions and the technical problems to be solved, please analyze the advantages and positive effects of the technical solutions to be protected in the present invention from the following aspects:

first, aiming at the technical problems existing in the prior art and the difficulty in solving the problems, the technical problems to be solved by the technical scheme of the present invention are closely combined with the technical scheme to be protected and the results and data in the research and development process, and some creative technical effects brought after the problems are solved are analyzed in detail and deeply. The specific description is as follows:

(1) According to the method, the traditional Landsat remote sensing image rough classification and the Google Earth high-precision image fine classification technology are organically combined, so that the extraction precision of the farmland plot crop type and space distribution information can be greatly improved.

(2) The invention adopts the image interpretation method of partition classification, can effectively reduce the interference of the same thing and different spectrum or the same spectrum of foreign matters of the remote sensing image in the region, improve the classification precision, and effectively solve the technical difficult problem in the remote sensing monitoring of the farmland crops in a large range.

(3) The invention applies the high-precision remote sensing technology to agricultural land utilization, crop information extraction and precision verification, thereby not only saving the workload, but also effectively improving the precision of the classification result. Therefore, the method realizes the rapid and accurate extraction of the large-range farmland land utilization and plot scale crop planting information, and greatly saves the time and economic cost which only depend on manual investigation in the past. Meanwhile, a set of simple, convenient and efficient technical method system which can be used by industrial personnel facing agricultural scientific research, agricultural regionalists and the like and non-remote sensing professionals is constructed, and a certain promotion effect is played for developing a new round of land resource general survey analysis, farmland crop information acquisition and related subsequent scientific research and research work.

Secondly, considering the technical scheme as a whole or from the perspective of products, the technical effect and advantages of the technical scheme to be protected by the invention are specifically described as follows:

(1) The invention explores and forms a technical scheme for realizing the fine classification and the quick extraction of the crop information of the farmland plots based on the high-precision remote sensing image and the partition classification thought; a set of monitoring model of high-resolution remote sensing multi-source heterogeneous data analysis and crop partition classification facing to farmland plots is constructed, modularization and systematization operation of farmland crop information remote sensing extraction are achieved, and working efficiency is greatly improved.

(2) Based on the technical scheme, a set of mature, convenient and large-area remote sensing product system convenient to popularize and use is formed preliminarily, and the remote sensing product system is popularized and applied and subjected to precision verification in typical areas of three major food producing areas in China preliminarily. And next, relevant technical parameters and methods are integrated, a multi-source heterogeneous spatial database and a remote sensing result open service platform suitable for different areas in China are constructed, and the productization and marketization application of the technical result is realized.

Third, as an inventive supplementary proof of the claims of the present invention, there are also presented several important aspects:

(1) The expected income and commercial value after the technical scheme of the invention is converted are as follows:

the technical scheme of the invention is adopted to verify the typical regional precision of the county scale and the farmland plot scale of three major grain producing areas in China, and the typical regional precision proves that the land utilization classification and crop information extraction accuracy of different regions can reach more than 85 percent, and the accuracy of some regions is even close to 95 percent. The technical achievement is in the leading level in the industry in the aspects of fine classification and overall precision by testing according to the national standard of 'quality inspection and acceptance of digital surveying and mapping achievement'.

(2) The technical scheme of the invention effectively solves the technical problems to be solved urgently in the field of agricultural remote sensing:

the method is used for rapidly and accurately acquiring farmland plot crop information (including farmland crop types, distribution areas and the like in different seasons) in a large range, and is always a technical difficult problem to be solved urgently in the field of agricultural remote sensing. The main innovation and advantages of the invention are embodied in two aspects: 1) In terms of data sources: according to the invention, a set of data fusion thinking combining traditional Landsat remote sensing images and Google Earth high-precision images is established by trying to take multi-source and multi-temporal images as data sources, so that advantage complementation of multi-source data is realized; 2) In the classification method, the innovative method for remote sensing partition classification provided by the invention is assisted by image splicing and boundary fusion processing technologies, so that the misclassification and misclassification errors caused by 'same-object different spectrum, same-foreign-object spectrum' and the like in the past are effectively improved, and the classification precision of the remote sensing image is remarkably improved.

In conclusion, the technical idea and the system scheme provided by the invention effectively solve the technical difficult problem in the field of agricultural remote sensing at home and abroad at present. The method is in the leading level in the industry in the aspects of rapid and accurate extraction and refined classification of farmland crop information, and the related technical method has a wide application prospect.

Drawings

FIG. 1 is a flow chart of a method for extracting crop information of a farmland plot based on high-resolution remote sensing, which is provided by the embodiment of the invention.

Fig. 2 is a schematic diagram of a partition classification method according to an embodiment of the present invention.

Fig. 3 is a flowchart of a regional land use partition classification method based on remote sensing images according to an embodiment of the present invention.

FIG. 4 is a sectional view of elevation remote sensing images of typical areas of agriculture in China according to an embodiment of the present invention.

Fig. 5 is a technical flowchart effect diagram of extracting crop distribution information by high-precision remote sensing provided by the embodiment of the invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is further described in detail with reference to the following embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.

1. Illustrative embodiments are explained. This section is an explanatory embodiment expanding on the claims so as to fully understand how the present invention is embodied by those skilled in the art.

As shown in fig. 1, the method for extracting crop information of a farmland plot based on high-resolution remote sensing provided by the embodiment of the invention comprises the following steps:

s101, remote sensing image partition classification: adopting an image interpretation method of partition classification, and dividing a research area into a plurality of sub-areas based on the differences of different landform units in the research area in the aspects of remote sensing image spectrum, surface vegetation density, residential distribution and the like;

s102, precision evaluation and verification: and calculating the overall classification precision and kappa coefficient of various ground features by establishing a confusion matrix to judge the precision of the classification result.

In the embodiment of the present invention, the remote sensing image partition classification in step S101 specifically includes:

(1) Pretreatment before classification: preprocessing image data;

(2) And (4) partition classification: adopting a maximum likelihood method to supervise and classify;

(3) And (4) classification post-treatment: and splicing the partitioning results.

In the embodiment of the present invention, the image data preprocessing in step (1) includes geometric correction, projective transformation, image mosaicking, and region-of-interest cropping extraction.

In the embodiment of the invention, the remote sensing image of the research area after correction and registration processing in step (1) is superposed with the administrative boundary, and then the administrative partition boundary of the typical area is used for cutting the image of the whole area into a plurality of sub-areas.

In the step (2) of the embodiment of the invention, the remote sensing images 5, 4 and 3 with the most abundant information content are selected to be combined, and three colors of red, green and blue are matched to generate a false color composite image.

In the step (2) of the embodiment of the invention, during classification, the latest Google Earth high-definition image in the research area is used as a contrast, and a land utilization planning map is used as a reference.

When the training samples are collected in the area in the step (2), sample points are uniformly distributed in the area range, and the pixels are obtained by selecting the pure spectrum.

When the partition results are spliced in the step (3), according to land use classification results obtained by different partition interpretation, a decision tree is established in remote sensing image processing software to unify land type codes and color matching of each partition, image splicing and boundary fusion processing are carried out, and the partition interpretation results are synthesized into a complete whole-partition land use and crop type distribution map; some of the fine patches are merged and smoothed using a class clustering method.

In the precision evaluation and verification of the step S102 in the embodiment of the invention, a certain number of verification points are randomly generated in a research range, the real type of the verification points is judged according to Google earth high-resolution images or ground actual survey data, the real type is compared with remote sensing classification results, and the overall precision is counted.

2. Application examples. In order to prove the creativity and the technical value of the technical scheme of the invention, the part is the application example of the technical scheme of the claims on specific products or related technologies.

The technical scheme provided by the invention is applied to three large main grain production areas, the distribution pattern and sowing area data of main crops such as wheat, corn, rice, soybean and rape are obtained, the monitoring result precision is averagely more than 90%, and compared with other large-scale crop monitoring, the method provided by the invention is in the leading level in the aspects of crop type refinement degree and extraction precision.

By applying the technical scheme of the invention, the classification results in the three typical agricultural production areas have higher precision. The total precision of a certain county is 85.96%, the total precision of a certain county is 86.17%, the total precision of a certain district is 93.74%, and the precision of some typical areas is even close to 95%, which shows that the technology has good application and popularization effects.

3. Evidence of the relevant effects of the examples. The embodiment of the invention achieves some positive effects in the process of research and development or use, and has great advantages compared with the prior art, and the following contents are described by combining data, diagrams and the like in the test process.

1. The remote sensing image partition classification technical process comprises the following steps: the traditional remote sensing image classification method generally selects training samples by using a unified standard in the whole research area range, is simple and convenient, but is not easy to solve the problems of same object, different spectrums, same foreign object spectrums and the like, and often has the phenomena of wrong division and missing division. In order to reduce interference of the same object and different spectrum or the same spectrum of the foreign object in the remote sensing image in the region and improve the classification precision, the embodiment of the invention adopts an image interpretation method of partition classification (as shown in fig. 2). The basic idea of the method is as follows: based on the differences of different landform units in the research area in the aspects of remote sensing image spectrum, surface vegetation density, residential distribution and the like, the research area is divided into a plurality of sub-areas.

Through the subarea acquisition training sample and the subarea development supervision classification, the training sample selection error caused by the same object different spectrum or the same foreign object spectrum in different landform units can be effectively avoided, and the precision of the classification result is improved.

On the basis, a technical process of agricultural land utilization and crop information partition classification based on high-precision remote sensing images is constructed (as shown in figure 3).

During specific operation, firstly, dividing a complete remote sensing image covering an area into a plurality of small units (or called sub-areas) according to the difference of landform and landform, as shown in fig. 4, a) in fig. 4 is an inner Mongolian and county subarea; b) Dividing the ancient city into areas in Hunan province; c) In a region from the chenchen station county, north river. Then, a classification system is established in a partition mode, training samples are collected in the partition mode, and classification is carried out by applying a supervision classification method through preprocessing steps such as image correction and partition cutting. When the training sample is selected, a Google earth high-definition image map is adopted, so that the accuracy of sample selection is ensured as much as possible. And after the subareas are classified, splicing and fusing the result graphs of all the subareas to obtain the land utilization and crop information classification results of the whole subarea.

(1) Establishing a classification system and a technical method flow:

the establishment of the classification system is the basis for developing remote sensing image classification. In the embodiment of the invention, when a classification system of a typical region is determined, the original 12 major categories are adjusted into 9 major categories (table 1) by mainly referring to the current state of land utilization classification (GB/T21010-2007) standard and combining classification targets and land utilization type characteristics of the typical region. As the research of the embodiment of the invention focuses on agricultural land utilization types such as agriculture, forestry, pasturing and the like, the cultivated land, the woodland and the garden land are refined, and other types (such as industrial and mining storage land, residential land, public service land, special land, transportation land and the like) are appropriately merged and simplified. When the planting information of the farmland plot crops is extracted, selecting characteristic parameters such as normalized vegetation index NDVI and the like to construct a high-precision remote sensing extraction technical scheme, wherein (1) in fig. 5 is a multi-temporal image (extracting the growth cycle of the crops) as shown in fig. 5; (2) For image processing (radiation correction, vector clipping, NDVI conversion); (3) For graded segmentation (different color rendering) with NDVI value size; (4) Performing superposition and interactive interpretation, and adjusting crop identification parameters; (5) determining a threshold model, and performing decision extraction; and (6) is an extraction effect graph.

The target area is subdivided into more than 100 monitoring subareas according to factors such as crop phenology, image date and the like, and a monitoring model is constructed region by region through sampling analysis for monitoring and extraction, so that the distribution data of main crops such as wheat, corn, rice, soybean and the like in a typical grain main production area are obtained.

TABLE 1 Classification system for land utilization

(2) Image classification and processing flow

Pretreatment before classification: the image data preprocessing comprises geometric correction, projective transformation, image mosaic, research area cutting extraction and the like. Images and administrative charts are commonly used with the UTM projection (Universal transform Mercator, UTM), WGS1984 ellipsoid coordinates. The remote sensing image of the research area after correction and registration processing can be well overlapped with the administrative boundary. And then the administrative partition boundary of the typical area is used for respectively cutting the whole area image into a plurality of sub-areas.

And (4) partition classification: among the conventional supervised classification methods, the most commonly used supervised classification method is the Maximum Likelihood Method (MLC). The maximum likelihood method has a strict theoretical basis, and particularly for data in normal distribution, a discriminant function is easy to establish and has good statistical characteristics. And the method can fully utilize various prior knowledge, and the man-machine interaction operation is visual and simple, so that the embodiment of the invention adopts a maximum likelihood method to supervise and classify.

The selection of the training samples is the key of classification, and the embodiment of the invention selects the combination of the remote sensing images 5, 4 and 3 with the most abundant information quantity and generates the false color synthetic image by matching three colors of red, green and blue. The combined synthetic image is similar to natural color, relatively accords with the visual habits of people, is rich in information quantity, can fully display the difference of various ground feature image characteristics, and is convenient for selecting training samples. In classification, the latest Google Earth high-definition image in the research area is used as a contrast, and a land utilization planning map is used as a reference. When the training samples are collected in the subareas, the sample points are distributed as uniformly as possible in the area range, and spectral purity pixels are selected as much as possible, so that the classification precision is improved.

And (4) classification post-treatment: first, the splicing process for the partitioned results. And according to land utilization classification results obtained by interpretation of different partitions, establishing a decision tree in remote sensing image processing software to unify land type codes and color matching of each partition, performing image splicing and boundary fusion processing, and synthesizing the partition interpretation results into a complete whole-partition land utilization and crop type distribution map. There are usually many fine patches in the initial classification result graph, and it needs to be post-classified. The technical scheme mainly uses a class clustering method (Clump Classes) to merge and smooth some small plaques.

2. Precision evaluation and verification method

The most important step after remote sensing classification is precision evaluation, which generally needs to select real land type sample points on the ground as reference standards and adopt some quantitative indexes and parameters to check the classification precision. In the past, the image classification precision is usually checked through on-site acquisition and check, time and labor are wasted, the cost is high, and the check work of a large-range classification result is not practical.

The embodiment of the invention provides great convenience for checking the classification result by adopting the free and high-precision images provided by the Google earth, and the used remote sensing image analysis software can be smoothly linked with the Google earth, so that the accurate positioning and identification of the ground features of the remote sensing image are realized. Google Earth provides free meter-level high-precision satellite images, and provides possibility for land use change detection in land parcel scale. The embodiment of the invention calculates the overall classification precision and kappa coefficient of various land features to judge the precision of the classification result by establishing a confusion matrix (table 2). Randomly generating a certain number of verification points in a research range, judging the real type of the verification points by taking Google earth high-resolution images or ground actual survey data as the basis, comparing the real type with remote sensing classification results, and counting the overall accuracy. The typical regional precision verification of the county scale and the farmland plot scale of the three major grain producing areas in China by adopting the embodiment of the invention shows that the accuracy rate of land utilization classification and crop information extraction reaches more than 85 percent, and some regions even approach 95 percent. The overall accuracy meets the requirements of the relevant national standards.

TABLE 2 remote sensing classification precision confusion matrix

And Pm and n represent the number of verification points with the extraction type of m and the actual type of n.

By adopting the technical method provided by the invention, the precision test of the classification result graphs of three typical regions (inner Mongolian Xinhe county, hebei province, the Chenchen platform county and Hunan province, changde city and ancient city district) of agricultural production in China shows that: the overall accuracy of the soil utilization map in the year 2010 of inner mongolian xing and county was 85.96%. The overall accuracy in the year 2010 of the chenchenchen station county, north, river was 86.17%. The overall accuracy in 2010 in the ancient city of Changde city of Hunan province is 93.74%. It can be seen that the precision of the classification results of different typical areas in the south and the north of China is over 85%, and the precision of the classification results of some areas is even close to 95%. The technical achievement is in the leading level in the industry in the aspects of fine classification and overall precision by inspection according to the national standard of 'quality inspection and acceptance of digital surveying and mapping achievement'.

The above description is only for the purpose of illustrating the embodiments of the present invention, and the scope of the present invention should not be limited thereto, and any modifications, equivalents and improvements made by those skilled in the art within the technical scope of the present invention as disclosed in the present invention should be covered by the scope of the present invention.

Claims (10)

1. A method for extracting crop information of farmland plots based on high-resolution remote sensing is characterized by comprising the following steps:

step one, remote sensing image partition classification: adopting an image interpretation method of partition classification, and dividing a research area into a plurality of sub-areas based on the differences of different landform units in the research area in the aspects of remote sensing image spectrum, surface vegetation density, residential distribution and the like;

step two, precision evaluation and verification: and calculating the overall classification precision and kappa coefficient of various ground features by establishing a confusion matrix to judge the precision of the classification result.

2. The method for extracting crop information of farmland plots based on high resolution remote sensing as claimed in claim 1, wherein the remote sensing image partition classification in the first step specifically comprises:

(1) Pretreatment before classification: preprocessing image data;

(2) And (4) partition classification: adopting a maximum likelihood method to supervise and classify;

(3) And (4) classification post-treatment: and splicing the partition results.

3. The method for extracting farmland land crop information based on high resolution remote sensing as claimed in claim 2, wherein the image data preprocessing in the step (1) comprises geometric correction, projective transformation, image mosaicing and study region cropping extraction.

4. The method for extracting crop information of farmland plots based on high resolution remote sensing as set forth in claim 2, wherein the remote sensing images of the research area after the correction and registration processing in step (1) coincide with the administrative boundaries, and then the administrative division boundaries of the typical area are used to cut the whole area image into several sub-areas.

5. The method for extracting crop information of farmland plots based on high resolution remote sensing as claimed in claim 2, wherein said step (2) selects the combination of 5, 4, 3 bands of remote sensing images with the most abundant information content, and generates a pseudo-color composite image by using three colors of red, green and blue.

6. The method for extracting crop information of farmland plots based on high resolution remote sensing as claimed in claim 2, wherein in the step (2), the latest Google Earth high-definition image of the research area is used as a contrast in classification, and a land use planning map is used as a reference.

7. The method for extracting crop information of farmland plots based on high resolution remote sensing as claimed in claim 2, wherein when training samples are collected in a partitioned manner in step (2), sample points are uniformly distributed in a regional range, and spectral purity pixels are selected.

8. The method for extracting crop information of farmland plots based on high resolution remote sensing as claimed in claim 2, wherein when the segmentation results are subjected to splicing processing in step (3), according to land utilization classification results obtained by different segmentation interpretations, a decision tree is established in remote sensing image processing software to unify land type codes and color matching of each region, image splicing and boundary fusion processing are performed, and the segmentation interpretation results are synthesized into a complete whole region land utilization and crop type distribution map; some of the fine patches are merged and smoothed using a class clustering method.

9. The method for extracting farmland land crop information based on high-resolution remote sensing as claimed in claim 1, wherein in the second precision evaluation and verification step, a certain number of verification points are randomly generated in the research range, the real types of the verification points are judged according to Google earth high-resolution images or ground actual survey data, and the real types are compared with the remote sensing classification results to count the overall precision.

10. A system for extracting information of farmland crops based on high resolution remote sensing for implementing the method for extracting information of farmland crops based on high resolution remote sensing according to any one of claims 1 to 9, which is characterized in that the system for extracting information of farmland crops based on high resolution remote sensing comprises:

the image preprocessing module is used for preprocessing the image data before classification;

the partition classification module is used for supervising and classifying the remote sensing images by adopting a maximum likelihood method;

the classification post-processing module is used for splicing the partitioning results;

and the precision evaluation and verification module is used for calculating the overall classification precision and kappa coefficient of various ground features by establishing a confusion matrix to judge the precision of the classification result.

Images