Natural grassland state identification method and electronic equipment

Abstract

The invention discloses a natural grassland state identification method, and the method comprises the steps: obtaining a remote sensing image of a natural grassland, and obtaining the spatial distribution data of the grassland through a trained grassland state identification model. And constructing a grass mowing field state recognition model based on the grass mowing characteristic indexes. The grass trimming characteristic indexes comprise a topographic characteristic index, a spectral characteristic index, a texture characteristic index and a user-defined characteristic index. And for each grass trimming characteristic index, generating corresponding characteristic index raster data through calculation, and constructing grass trimming field characteristic index data through a wave band synthesis tool.

Description

Technical Field

[0001] The present invention belongs to the technical field of animal husbandry, and particularly relates to a method for identifying the state of a natural grassland hayfield and an electronic device. Background Art

[0002] The utilization methods of natural grasslands are mainly divided into grazing and haymaking. Among them, the distribution and area of hayfields are very important for the reserve of winter livestock forage in pastoral areas. At the same time, due to continuous haymaking in hayfields, the number and types of seeds in the soil seed bank may decrease, leading to a series of grassland problems such as grassland degradation. Therefore, the identification of natural grassland hayfields plays a very important role in regulating the reserve of winter forage, restoring the soil seed bank, and preventing the decline of grassland soil fertility and other grassland problems. Summary of the Invention

[0003] One embodiment of the present disclosure is a method for automatically extracting natural grassland hayfield information based on object-oriented. The method includes obtaining a remote sensing image of a natural grassland, and obtaining spatial distribution data of the hayfield through a trained hayfield state recognition model. Here, object-oriented refers to a method in remote sensing interpretation and analysis. Using object-oriented classification technology, adjacent pixels are grouped into objects to identify spectral elements of interest, and the spatial, texture, and spectral information of the panchromatic and multispectral data of remote sensing satellite images are fully utilized for segmentation, and then the land cover types in the target area are interpreted.

[0004] The hayfield state recognition model is constructed based on haymaking characteristic indices, and the haymaking characteristic indices include topographic characteristic indices, spectral characteristic indices, texture characteristic indices, and custom characteristic indices.

[0005] After obtaining each haymaking characteristic index, corresponding characteristic index raster data is generated through calculation, and hayfield characteristic index data is constructed through a band synthesis tool. The hayfield state recognition model inputs the hayfield characteristic index data, and obtains hayfield spatial distribution data through the CART classification tree algorithm. Brief Description of the Drawings

[0006] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features, and advantages of the exemplary embodiments of the present invention will become readily understood. In the drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, wherein: Figure 1 Flowchart of the method for identifying the state of a natural grassland hayfield according to one embodiment of the present invention. Detailed Embodiments

[0007] In the existing research on the identification method of mowing grasslands, usually only the change in the Normalized Difference Vegetation Index (NDVI) data of the grassland before and after mowing is considered, without considering other mowing characteristics. Therefore, if by introducing the mowing feature index of natural grasslands (topographic feature index, spectral feature index, texture feature index, custom feature index), combining the training sample library of image features before and after mowing of natural mowing grasslands, and using the classification decision tree classification method to construct a method for automatic extraction of natural grassland mowing grassland information based on object-oriented, the workload of manual visual interpretation and ground investigation verification can be greatly reduced, and the distribution and area of natural grassland mowing grasslands in the target area can be obtained quickly and accurately. Here, interpretation refers to the process of obtaining information from images. The interpretation process of remote sensing images is the reverse process of obtaining remote sensing images, that is, the process of extracting remote sensing information from the simulated images of ground truth by remote sensing and inversely inferring the ground prototype. Remote sensing image interpretation includes visual interpretation, man-machine interactive interpretation, and computer digital image processing.

[0008] According to one or more embodiments, the specific invention process is as follows: The following separately explains the mowing feature index of natural grasslands.

[0009] (1) Topographic feature index.

[0010] Natural grassland mowing grasslands are mainly distributed in areas with flat terrain and small slopes. Therefore, the slope can be used to describe the topographic features. The slope data is calculated from the DEM (Digital Elevation Model), and the DEM is the data collected by the phased array L-band synthetic aperture radar (PALSAR) sensor with a spatial resolution of 12.5 m. Resample the DEM data to be consistent with the spatial resolution of the high-resolution optical remote sensing vegetation index. Calculate the slope factor using the DEM data, and the specific calculation formula is: (1) Among them, Slope is the slope of the point (x, y), is the relative height in the horizontal direction of the point (x, y), is the relative height in the vertical direction of the point (x, y).

[0011] (2) Spectral feature index.

[0012] The spectral features of the remote sensing image of the grassland after mowing are an important feature basis for the extraction and identification of mowing grassland information, and also one of the important bases for selecting training samples of mowing grasslands. Spectral features include spectral brightness (Brightness) and the mean value of each band (Mean). In the embodiments of the present disclosure, based on high-resolution optical remote sensing image data, the bands include the near-infrared band (Band NearIR), the red band (Band Red), the green band (Band Green), and the blue band (Band Blue). The specific calculation formula is: (2) In formula (2), Mean (g(i, j)) is the average gray value in the image row direction, N is the number of rows of the image, (i, j) is the row and column position of a unit pixel in the image, g(i, j) is the gray value of the pixel in the image.

[0013] Due to mowing, there is a large difference in the gray values of adjacent pixels between mowed and unmowed areas in the mowing field. Therefore, the spectral brightness feature uses a weighted average to calculate the brightness mean, and the calculation formula is as follows: (3) In formula (3), BRI (g(i, j)) is the spectral brightness in the image row direction, that is, the weighted average gray value in the image row direction, Px is the weight of the pixel in the image in the row direction.

[0014] (3) Texture feature index.

[0015] The texture feature after mowing is the most important feature for identifying the mowing field. The artificial texture feature formed after mowing is the main basis for selecting the training samples of the mowing field. The texture feature index is obtained by calculating the gray-level co-occurrence matrix. The gray-level co-occurrence matrix quantitatively extracts and describes the statistical attributes of the texture feature based on the gray direction, change amplitude, and interval of the image pixels, and is used to represent the probability that a pixel with gray level j appears at a distance d and direction θ from a pixel with gray level i.

[0016] There are many types of gray-level co-occurrence matrix texture feature indexes. To reduce data redundancy and improve the recognition calculation speed of the mowing field, in the embodiments of the present disclosure, the texture feature indexes of the mowing field are selected as Contrast, Correlation, and Homogeneity. The specific calculation formulas are as follows: Contrast reflects the contrast of the gray value of a certain pixel in the remote sensing image with the pixel values of its adjacent pixels. If the pixels deviating from the diagonal of this pixel have large gray values, that is, the brightness values of the remote sensing image have large differences, then the value of CON (Contrast) is large. Contrast reflects the clarity of the remote sensing image and the depth of the texture grooves. The deeper the texture grooves, the greater the contrast, the brighter the remote sensing image, and the clearer the visual effect; on the contrary, the shallower the texture grooves, the smaller the contrast, the darker the remote sensing image, and the more blurred the visual effect.

[0017] (4) In formula (4) is the contrast, , are respectively the gray levels of and The pixel is the total number of pixels, is the probability that a pixel with gray level j appears at a distance d and direction θ from a pixel with gray level i.

[0018] The correlation degree represents the similarity degree of pixels in a remote sensing image in the row or column direction. Therefore, the magnitude of the correlation degree value reflects the correlation relationship of pixel gray level values in the image. When the gray level values of matrix pixels in a remote sensing image are uniform and approximately equal, the larger the correlation degree value, the greater the similarity degree of the pixels in the image; on the contrary, when the difference in gray level values of matrix pixels in a remote sensing image is larger, the smaller the correlation degree value, the smaller the similarity degree of the pixels in the image.

[0019] (5) In formula (5), is the correlation degree, , are respectively the pixels with gray levels and respectively, is the total number of pixels, is the mean value of the gray level gray level, is the probability that a pixel with gray level j appears at a distance d and direction θ from a pixel with gray level i, is the variance of the gray level variance, is the variance of the gray level variance.

[0020] The homogeneity is a measure unit for measuring the uniformity of pixel gray level values in a remote sensing image, indicating whether the distribution of pixel gray level values in the image is uniform. If the pixels in the image have the same or nearly similar gray level values, the homogeneity of the image is larger; on the contrary, if the distribution of pixel gray level values is significantly non-uniform, the homogeneity of the image is smaller.

[0021] (6) In formula (6), is the energy, , are respectively the pixels with gray levels and respectively, is the total number of pixels, is the probability that a pixel with gray level j appears at a distance d and direction θ from a pixel with gray level i.

[0022] (4) Custom feature index The custom feature index consists of the Normalized Difference Vegetation Index (NDVI) and the rate of change of the Normalized Difference Vegetation Index (R NDVI)(Composition. The Normalized Difference Vegetation Index (NDVI) is one of the vegetation indices used to reflect the health status of vegetation, based on the high reflectance value of vegetation in the near-infrared band and the strong absorption characteristics of chlorophyll in the red band. The normalized vegetation index data is derived from high-resolution optical remote sensing satellite data products, and the image time resolution is monthly multi-spectral remote sensing data (including 4 bands: blue, green, red, and near-infrared) during the growing season of natural grasslands.)

[0023] The calculation formula of NDVI is: (7) In formula (7), is the normalized vegetation index of the i-th month during the growing season of natural grassland vegetation, is the reflectance value in the near-infrared band; is the reflectance value in the red band.

[0024] R NDVI The calculation formula is: (8) In formula 8, is the change rate of the normalized vegetation index from the i-th month to the j-th month during the growing season of natural grassland, is the normalized vegetation index of the i-th month during the growing season of natural grassland, that is, the normalized vegetation index of natural grassland before mowing; is the normalized vegetation index of the j-th month of natural grassland, that is, the normalized vegetation index of natural grassland after mowing.

[0025] Method for constructing mowing feature index data. Assume that the mowing feature index includes terrain feature index, spectral feature index, texture feature index, and custom feature index. According to the calculation methods of each mowing field feature index, the corresponding feature index raster data is produced, and the mowing field feature index data is constructed through a band composition tool. Here, the band composition tool refers to the process of outputting the bands in the raster dataset in the order of the bands in the input control box in the analysis environment of images and raster datasets.)

[0026] Furthermore, a mowing field identification model based on the mowing feature index. Based on the training samples and the mowing feature index data (terrain feature index, spectral feature index, texture feature index, and custom feature index), the spatial distribution data of the mowing fields is obtained through the CART (Classification and Regression Tree) classification tree algorithm. The CART classification regression tree can be used for both classification and regression during the classification process. The output result of the classification is the category of the classification samples, and the output result of the regression tree is a real number.)

[0027] The CART classification tree algorithm uses the Gini coefficient to select features. The Gini coefficient represents the impurity of the model. The smaller the Gini coefficient, the lower the impurity and the better the feature.

[0028] Among them, the calculation formula of the Gini value is:

[0029] (9) In formula (9), is the probability of the classification appearing, is the number of classification categories.

[0030] reflects the probability that two randomly selected samples in the dataset D have inconsistent labels. Therefore, the smaller it is, the higher the purity of the two samples. According to the statistical data of local hayfields, the recognition and interpretation results are optimized through interactive recognition and interpretation. Based on the optimization results, the haying distribution and area are statistically analyzed. Specifically, according to the statistical survey data of local hayfields, the sample feature space of natural grassland hayfields is evaluated, and the recognition and interpretation results automatically recognized based on this sample feature space are optimized through human-computer interaction recognition and interpretation, and then the hayfield distribution and area are obtained.

[0031] As Figure 1 shown, the process of the method for identifying the status of natural grassland hayfields includes:

[0032] Obtain high-resolution remote sensing image data through dedicated channels for various remote sensing images.

[0033] Use remote sensing image processing tools to perform radiometric correction, geometric correction, atmospheric correction, image cropping, etc. on the original remote sensing image data.

[0034] Use the scale set theory to perform multi-scale segmentation on the high-resolution remote sensing image, obtain the morphological contours and attribute features of the segmentation objects, and extract the optimal segmentation scale for natural grassland hayfields.

[0035] Identify and train natural grassland hayfields by constructing terrain features, spectral features, texture features, and custom features of natural grassland hayfields.

[0036] Based on the regions of interest of natural grassland hayfields in the field survey, use machine learning algorithms to construct terrain features, spectral features, texture features, and custom feature recognition sample feature space training of natural grassland hayfields.

[0037] Based on the sample feature space training, select the K-nearest neighbor method, decision tree, support vector machine, and random forest algorithms to construct an algorithm set.

[0038] Automatically interpret the natural grassland hayfields using the constructed algorithm set for each algorithm, evaluate the interpretation results, and screen out the optimal algorithm.

[0039] Evaluate the interpretation accuracy using the producer accuracy, overall accuracy, and Kappa coefficient in the confusion matrix.

[0040] Evaluate the sample feature space of the natural grassland hayfields according to the local hayfield statistical survey data, and optimize the human-computer interaction recognition and interpretation for the interpretation results automatically identified based on this sample feature space, so as to obtain the distribution and area of the hayfields.

[0041] According to the interactive interpretation results, perform processing such as fusion and elimination on the obtained final result data, and the obtained result data is the distribution and area data of the natural grassland hayfields.

[0042] The beneficial effects of the present disclosure are as follows. Since the distribution and area of the natural grassland hayfields are very important for the winter forage reserves of livestock in pastoral areas, and at the same time, due to continuous mowing of the hayfields, the quantity and variety of seeds in the soil seed bank decrease, leading to a series of grassland problems such as grassland degradation.

[0043] Therefore, the identification of natural grassland hayfields plays a very important role in regulating winter forage reserves, restoring the soil seed bank, and preventing grassland soil fertility decline and other grassland problems. At the same time, it is also the basis for accurately grasping the grassland-livestock balance of the natural grassland grazing land. Based on the object-oriented method for automatic extraction of natural grassland hayfield information, the present disclosure greatly reduces the workload of manual visual interpretation and on-site investigation and verification, and can quickly and accurately obtain the distribution and area of the natural grassland hayfields in the target area.

[0044] It should be understood that in the embodiments of the present invention, the term "and / or" is merely a description of the association relationship between associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this article generally represents an "or" relationship between the preceding and following associated objects.

[0045] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0046] In addition, in each embodiment of the present invention, each functional unit may be integrated into a processing unit, or each unit may exist physically alone, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional unit.

[0047] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, mobile hard disks, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), magnetic disks, or optical discs.

[0048] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A method for identifying the status of a natural grassland mowing field, characterized in that, obtain the remote sensing image of the natural grassland, and through the trained mowing field status recognition model, obtain the spatial distribution data of the mowing field.

2. The method according to claim 1, characterized in that, construct the mowing field status recognition model based on the mowing feature index.

3. The method according to claim 2, characterized in that, the mowing feature index includes a terrain feature index, a spectral feature index, a texture feature index, and a custom feature index.

4. The method according to claim 3, characterized in that, for each mowing feature index, calculate the corresponding feature index raster data, and construct the mowing field feature index data through the band synthesis tool.

5. The method according to claim 4, characterized in that, for the mowing field status recognition model, input the mowing field feature index data, and obtain the spatial distribution data of the mowing field through the CART classification tree algorithm.

6. The method according to claim 3, characterized in that, the terrain feature index uses the slope to describe the terrain feature of the mowing field.

7. The method according to claim 3, characterized in that, the spectral feature index includes spectral brightness and the average brightness of each spectral band.

8. The method according to claim 3, characterized in that, the texture feature index is the artificial texture feature formed after mowing the mowing field.

9. The method according to claim 3, characterized in that, the custom feature index is composed of the normalized difference vegetation index and the change rate of the normalized difference vegetation index.

10. An electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, the processor runs the computer program to implement the method according to any one of claims 1-9.

Images