Remote sensing image vegetation extraction method and system based on NDVI and HSV
Patent Information
- Application Number
- CN202610785570.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-06-02
- Publication Date
- 2026-10-20
AI Technical Summary
发明人发现，指数提取法多依赖单一光谱特征，对特定植被提取的适配性不足，提取精度有限；基于像元的分类法难以兼顾植被细微特征，易产生信息冗余，提取精度难以保障；面向对象分类法需反复调试分割参数，耗时费力，时间成本较高；深度学习法虽具备高效识别优势，但需大量标签数据支撑，数据获取成本高昂
1）本发明的基于NDVI与HSV的遥感影像植被提取方法，通过融合多时序NDVI特征与HSV色彩特征，构建互补性强的双维度特征体系，充分发挥时序特征反映植被生长规律、色彩特征凸显地物视觉差异的优势，突破传统单一光谱特征的提取局限，有效解决复杂场景下不同植被光谱混淆导致的识别精度低的问题，提升遥感影像植被分类提取的准确性与可靠性。
Smart Images
Figure FT_1 
Figure FT_2 
Figure FT_3
Abstract
Description
Technical Field
[0001] This invention belongs to the field of satellite remote sensing image monitoring technology, specifically relating to a method and system for extracting vegetation from remote sensing images based on NDVI and HSV. Background Technology
[0002] The statements herein provide only background information in relation to this invention and do not necessarily constitute prior art.
[0003] Vegetation is widely distributed in ecosystems such as lake nature reserves, and its spatial distribution and dynamic changes directly reflect the health of the ecosystem, making it one of the core indicators for ecological monitoring and protection. Traditional vegetation extraction methods rely on manual means such as field surveys and quadrat monitoring. While these methods can obtain high-precision vegetation information, they suffer from drawbacks such as high labor intensity, long monitoring cycles, high labor costs, and limited coverage, making it difficult to meet the actual needs of dynamic ecological management. With the rapid development of remote sensing technology, satellite remote sensing, with its unique advantages of large-scale, rapid, and periodic monitoring, has become the mainstream technical means for vegetation remote sensing extraction.
[0004] Existing remote sensing vegetation extraction methods mainly include index extraction, pixel-based classification, object-oriented classification, and deep learning. The inventors found that index extraction relies heavily on single spectral features, resulting in insufficient adaptability to specific vegetation types and limited extraction accuracy; pixel-based classification struggles to capture subtle vegetation features, easily leading to information redundancy and inconsistent extraction accuracy; object-oriented classification requires repeated debugging of segmentation parameters, which is time-consuming and labor-intensive; and while deep learning offers high efficiency, it requires a large amount of labeled data, resulting in high data acquisition costs.
[0005] In addition, existing methods often neglect the integrated application of temporal and color features, failing to fully combine the temporal patterns of vegetation growth with color differences to achieve accurate extraction. This results in insufficient extraction accuracy for single vegetation types, making it difficult to meet the actual needs of ecological monitoring and vegetation management. Summary of the Invention
[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method and system for extracting vegetation from remote sensing images based on NDVI and HSV.
[0007] To achieve the above objectives, the present invention is implemented through the following technical solution: In a first aspect, the technical solution of the present invention provides a method for vegetation extraction from remote sensing images based on NDVI and HSV, including: Acquire satellite remote sensing images of the target vegetation in the first and second periods; The normalized vegetation index was calculated for the first and second period satellite remote sensing images, respectively. HSV color transformation was performed on satellite remote sensing images from the first and second periods to obtain color feature components from the first and second periods, including hue, saturation and brightness components. A decision tree model is constructed based on the normalized vegetation index and color feature components. Interfering features are sequentially removed and the target vegetation is extracted according to preset decision nodes. The preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
[0008] In at least one implementation, the first period is specifically the germination period of the target vegetation; the second period is specifically the vigorous growth period of the target vegetation.
[0009] In at least one implementation, the first type of disturbing features includes bulrushes and / or water bodies; the second type of disturbing features includes cultivated land and / or forest land; and the third type of disturbing features includes grassland.
[0010] In at least one implementation, the first type of target vegetation is vegetation with significant color saturation characteristics; the second type of target vegetation is vegetation with prominent hue characteristics and distributed in blocky patches.
[0011] In at least one implementation, the second type of target vegetation is extracted using the hue components of the first period, specifically including: performing window mean operation on the hue components of the first period to obtain optimized hue features, and extracting the second type of target vegetation based on the optimized hue features of the first period.
[0012] In at least one implementation, the window averaging operation uses a 5×5 pixel window.
[0013] In at least one implementation, the method further includes: post-processing the extracted vector information of the first and second types of target vegetation, specifically including: calculating the perimeter to area ratio of the target vegetation vector patch, setting a threshold to remove irregular strip-shaped patches; using mathematical morphology methods to eliminate holes in the vector patch; setting an area threshold for the target vegetation vector patch to remove small, fragmented patches.
[0014] Secondly, the technical solution of the present invention also provides a vegetation extraction system based on NDVI and HSV remote sensing images, comprising: The image acquisition module is configured to acquire satellite remote sensing images of the target vegetation in the first and second periods. The NDVI calculation module is configured to calculate the normalized vegetation index of satellite remote sensing images from the first and second periods, respectively. The HSV transformation module is configured to perform HSV color transformation on satellite remote sensing images from the first and second periods to obtain color feature components from the first and second periods, including hue, saturation and brightness components. The classification module is configured to: construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to preset decision nodes; wherein, the preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
[0015] Thirdly, the technical solution of the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the NDVI and HSV-based remote sensing image vegetation extraction method described in the first aspect.
[0016] Fourthly, the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the NDVI and HSV-based remote sensing image vegetation extraction method described in the first aspect.
[0017] The beneficial effects of the above-described technical solution of the present invention are as follows: 1) The vegetation extraction method of remote sensing images based on NDVI and HSV of the present invention integrates multi-temporal NDVI features and HSV color features to construct a highly complementary two-dimensional feature system. It gives full play to the advantages of temporal features reflecting vegetation growth patterns and color features highlighting visual differences of ground objects. It breaks through the limitations of traditional single spectral feature extraction, effectively solves the problem of low recognition accuracy caused by spectral confusion of different vegetation in complex scenes, and improves the accuracy and reliability of vegetation classification and extraction in remote sensing images.
[0018] 2) This invention addresses the impact of various interfering features in complex scenarios by designing a hierarchical decision model. Through differentiated combinations of multi-temporal NDVI and HSV color features, various interfering features are eliminated sequentially, avoiding misclassification and omission caused by direct classification, and significantly improving the anti-interference ability of vegetation classification and extraction in complex scenarios.
[0019] 3) This invention combines the growth time patterns and color feature differences of different vegetation types to optimize feature application strategies: For the characteristic of some vegetation having bright colors during its vigorous growth period, the HSV saturation component of the corresponding time period is used to achieve accurate identification, solving the problem of distinguishing it from other vegetation; For the characteristic of some vegetation having prominent hue features and concentrated spatial distribution at specific growth stages, the HSV hue component of the corresponding time period is used in combination with window mean operation optimization to improve the accuracy and spatial continuity of this type of vegetation extraction, ultimately achieving accurate and efficient classification and extraction of various target vegetation types. Attached Figure Description
[0020] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.
[0021] Figure 1 This is a schematic diagram of the vegetation extraction method based on NDVI and HSV from remote sensing images disclosed in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the NDVI changes of typical land features in the test area in April and June, as disclosed in Embodiment 1 of the present invention; Figure 3 This is a schematic diagram comparing the separability of vegetation based on HSV color features as disclosed in Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the vegetation classification and extraction model for the experimental area disclosed in Embodiment 1 of the present invention; Figure 5 This is a typical vegetation extraction result image of the experimental area disclosed in Embodiment 1 of the present invention. Detailed Implementation
[0022] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.
[0023] Example 1 As introduced in the background section, existing remote sensing vegetation extraction methods often rely on single spectral features, resulting in insufficient adaptability to specific vegetation types and limited extraction accuracy. Pixel-based classification methods struggle to capture subtle vegetation features, leading to information redundancy and compromised extraction accuracy. Object-oriented classification methods require repeated debugging of segmentation parameters, which is time-consuming and labor-intensive. While deep learning methods offer high efficiency, they require substantial amounts of labeled data, resulting in high data acquisition costs. Furthermore, existing methods often neglect the fusion of temporal and color features, failing to fully integrate the temporal patterns of vegetation growth with color differences for accurate extraction. This leads to insufficient extraction accuracy for single vegetation types, making it difficult to meet the practical needs of ecological monitoring and vegetation management.
[0024] To overcome the shortcomings of the prior art, in a typical embodiment of the present invention, such as... Figures 1 to 5 As shown, this embodiment discloses a method for vegetation extraction from remote sensing images based on NDVI and HSV. This method combines the spectral characteristics and growth patterns of the target vegetation, and extracts target vegetation from remote sensing images based on temporal NDVI and HSV (Hue-Saturation-Value) color features, achieving accurate and efficient classification and extraction of the target vegetation. Figures 1 to 5 As shown, the vegetation extraction method based on NDVI and HSV remote sensing images specifically includes the following steps: S1. Acquire satellite remote sensing images of the target vegetation in the first and second periods; S2. Calculate the normalized vegetation index for the first-period satellite remote sensing image and the second-period satellite remote sensing image respectively; S3. Perform HSV color transformation on the satellite remote sensing images of the first period and the satellite remote sensing images of the second period to obtain the color feature components of the first period and the second period, including the hue component, saturation component and brightness component. S4. Construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to the preset decision nodes; wherein, the preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
[0025] The following detailed description of the above-mentioned vegetation extraction method based on NDVI and HSV remote sensing images, with reference to specific implementation methods, provides a detailed explanation.
[0026] S1. Acquire satellite remote sensing images of the target vegetation in the first and second periods.
[0027] In this step, Sentinel L2A-level remote sensing images of the target vegetation from the first and second periods of the same year are selected, and the acquired satellite remote sensing images are preprocessed to obtain preprocessed satellite remote sensing images. Specifically, the first period refers to the budding stage of the target vegetation, and the second period refers to the vigorous growth stage of the target vegetation.
[0028] As a further implementation method, the acquired satellite remote sensing image is preprocessed, specifically including: selecting band data with a set resolution, and sequentially performing band synthesis, image mosaicking, and image cropping to obtain the preprocessed satellite remote sensing image. As a further implementation method, the preferred band resolution is 10m.
[0029] S2. Calculate the normalized vegetation index for the first-period satellite remote sensing image and the second-period satellite remote sensing image, respectively.
[0030] In this step, based on the preprocessed first-period satellite remote sensing images and the second-period satellite remote sensing images, the normalized vegetation index for the first period is calculated respectively. Normalized Difference Vegetation Index in the Second Period Specifically, it is expressed as:
[0031] In the formula, For the first Normalized Difference Vegetation Index for the period ; For the first Near-infrared pixel values of satellite remote sensing images during a given period; For the first Red band pixel values of satellite remote sensing images of a given period.
[0032] S3. Perform HSV color transformation on the satellite remote sensing images of the first and second periods to obtain the color feature components of the first and second periods, including hue component, saturation component and brightness component.
[0033] In this step, HSV color transformation is performed on the red, green, and blue bands of the preprocessed first-period satellite remote sensing image to obtain the color feature components of the first period, including the hue component. saturation component and brightness components The red, green, and blue bands of the preprocessed second-period satellite remote sensing imagery are subjected to HSV color transformation to obtain the color feature components of the second period, including the hue component. saturation component and brightness components The HSV color transformation can be specifically represented as:
[0034]
[0035]
[0036] In the formula, , , The first The hue, saturation, and brightness components of the period ; , , The first Normalized values for the red, green, and blue bands of the period; For the first The maximum normalized values of the red, green, and blue bands during the period; For the first The minimum normalized values of the red, green, and blue bands during the period; For the first period and The difference.
[0037] S4. Construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to the preset decision nodes.
[0038] In this step, a decision tree model is constructed based on the synergistic effect of the time-series normalized vegetation index (NZVI) and color feature components. The model sequentially completes the removal of interfering features and the extraction of target vegetation classification according to preset decision nodes. Specifically, based on the time-series NZVI and color feature components, the decision nodes are preset as follows: First, a first threshold is set using the NZVI during the vigorous growth period to remove the first type of interfering features, including *Potamogeton crispus* and water bodies. Second, a second threshold is set using the NZVI during the germination period to remove the second type of interfering features, including non-target vegetation such as cultivated land and forest land. Finally, the saturation component during the vigorous growth period is used to further refine the decision tree. A third threshold is set to extract the first type of target vegetation, which is vegetation with significant color saturation characteristics; subsequently, the brightness component at the budding stage is used. A fourth threshold is set to remove the third type of disturbing features, such as grasslands that still exist after being removed by the second threshold; finally, the hue components of the germination stage are... Window averaging is performed to obtain optimized hue features. A fifth threshold is set based on the optimized hue features at the germination stage to extract the second type of target vegetation, which is vegetation with prominent hue features and a blocky, contiguous distribution. As a further implementation, the window averaging operation uses a 5×5 pixel window to reduce interference from small ground features.
[0039] In this embodiment, the NDVI and HSV-based remote sensing image vegetation extraction method further includes step S5, which is used to post-process the extracted vector information of the first and second types of target vegetation. Specifically, the perimeter-to-area ratio of the extracted target vegetation vector patches is calculated, a sixth threshold is set to remove irregular strip-shaped patches; mathematical morphology methods are used to eliminate holes in the vector patches; an area threshold is set on the target vegetation vector patches to remove small, fragmented patches, ultimately obtaining accurate remote sensing image target vegetation extraction results.
[0040] Using a nature reserve with rich vegetation types and complex ground disturbances as the test area, aquatic vegetation such as reeds and lotus were selected as target vegetation to verify the effectiveness and applicability of the vegetation extraction method based on NDVI and HSV remote sensing images disclosed in this embodiment.
[0041] First, Sentinel L2A-level remote sensing images of aquatic plants such as reeds and lotus during their budding period (April) and peak growth period (June) in the same year were acquired. 10m band data were selected, and band synthesis, image mosaicking, and image cropping were performed sequentially to complete the remote sensing image preprocessing, resulting in preprocessed satellite remote sensing images of the budding period and the peak growth period.
[0042] Then, based on the preprocessed satellite remote sensing images of the budding stage and the preprocessed satellite remote sensing images of the vigorous growth stage, the normalized vegetation index for the first period was calculated. Normalized Difference Vegetation Index in the Second Period .
[0043] Next, HSV color transformation was performed on the red, green, and blue bands of the preprocessed budding satellite remote sensing image to obtain the color feature components of the budding stage, including the hue component. saturation component and brightness components The red, green, and blue bands of the preprocessed satellite remote sensing image during the vigorous growth period were subjected to HSV color transformation to obtain the color feature components of the vigorous growth period, including the hue component. saturation component and brightness components .
[0044] To explore the phenological differences of typical land cover in the study area and accurately distinguish between target vegetation and various disturbance land cover, based on sampling point data, the mean NDVI values of various land cover in April (germination period) and June (vigorous growth period) of the experimental area from 2018 to 2025 were calculated, and NDVI variation curves were plotted (e.g., Figure 2 (As shown). In April, aquatic vegetation, represented by reeds and lotus, is in its budding stage. The NDVI of reeds is similar to that of duckweed and grassland. Lotus often appears as a water background in the image, with a low NDVI value. Cultivated land and forest land generally have higher NDVI values, which can be used to remove these interfering features. In June, the NDVI of reeds and lotus increases significantly, but they are easily confused with the spectrum of forest land and grassland. The NDVI of duckweed and water bodies is generally low, which can be used to remove these interfering features, thereby improving the extraction accuracy of typical aquatic vegetation.
[0045] In the experimental area corresponding to this embodiment, in April, one type of vegetation, represented by reeds, exhibited prominent color characteristics, with low chlorophyll content and an overall yellowish-brown hue. In June, another type of vegetation, represented by lotus, showed significant color saturation, mostly appearing as light green, clearly different from other dark green vegetation. HSV transformation was performed on the April and June images to obtain hue, saturation, and brightness components. Based on the sampled data, the Jeffries-Matusita (JM) distances of the target vegetation and other interfering features were calculated for each component. The results are as follows: Figure 3 As shown. By Figure 3 It can be seen that, in the hue component in April, the JM distance between the vegetation type of reeds and lotus, pondweed, woodland, and water is greater than 1.8, indicating that the various land features are well separable under this feature; in the brightness component in April, the JM distance between this vegetation type and grassland is close to 1.4, indicating that they are separable but there is some overlap; in the saturation component in June, the JM distance between the vegetation type of lotus and reeds, pondweed, woodland, grassland, and water is greater than 1.8, indicating that the various land features are well separable under this feature, which can provide a reliable basis for accurate classification.
[0046] Based on the above analysis, a decision tree model was constructed. Leveraging the synergistic effect of temporal NDVI and HSV color features, interfering features were removed and target vegetation was extracted within the experimental area. Figure 4 As shown. Specifically, firstly, a first threshold is set using NDVI data from the peak growth period (June) to remove interfering features such as duckweed and water bodies; secondly, a second threshold is set using NDVI data from the germination period (April) to remove non-target vegetation interference such as cultivated land and forest land; then, based on the saturation component of the peak growth period (June)... A third threshold was set to extract brightly colored lotus-like vegetation; the brightness component during the budding period (April) was used as a reference. A fourth threshold was set to further eliminate disturbing features such as grassland; finally, the hue components during the budding period (April) were analyzed. A 5×5 pixel window was used to perform mean calculation to obtain optimized tone features. Based on the optimized tone features during the budding period (April), a fifth threshold was set to extract reed vegetation with prominent tone features and a blocky, continuous distribution.
[0047] Based on the target vegetation vector information obtained from the decision tree model, with reeds and lotus as typical representatives, the perimeter-to-area ratio of the vector patches was calculated. An appropriate threshold was set to remove irregular strip-shaped patches, and then mathematical morphology methods were used to eliminate holes in the vector patches. An area threshold was set for the above typical target vegetation vector patches to remove small, fragmented patches, resulting in the final classification and extraction results of typical target vegetation. The results are as follows: Figure 5 As shown, this demonstrates that the method proposed in this embodiment has good applicability in the application of vegetation extraction from large-scale, multi-temporal remote sensing images.
[0048] Example 2 In a typical embodiment of the present invention, this embodiment discloses a vegetation extraction system based on NDVI and HSV remote sensing images, comprising: The image acquisition module is configured to acquire satellite remote sensing images of the target vegetation in the first and second periods. The NDVI calculation module is configured to calculate the normalized vegetation index of satellite remote sensing images from the first and second periods, respectively. The HSV transformation module is configured to perform HSV color transformation on satellite remote sensing images from the first and second periods to obtain color feature components from the first and second periods, including hue, saturation and brightness components. The classification module is configured to: construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to preset decision nodes; wherein, the preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
[0049] Example 3 In a typical embodiment of the present invention, this embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the NDVI and HSV-based remote sensing image vegetation extraction method described in Embodiment 1. These steps include: S1. Acquire satellite remote sensing images of the target vegetation in the first and second periods; S2. Calculate the normalized vegetation index for the first-period satellite remote sensing image and the second-period satellite remote sensing image respectively; S3. Perform HSV color transformation on the satellite remote sensing images of the first period and the satellite remote sensing images of the second period to obtain the color feature components of the first period and the second period, including the hue component, saturation component and brightness component. S4. Construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to the preset decision nodes; wherein, the preset decision nodes include: The first type of interfering land cover was removed using the normalized vegetation index of the second period; the second type of interfering land cover was removed using the normalized vegetation index of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering land cover was removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period. S5. Post-process the extracted vector information of the first and second types of target vegetation. Specifically, calculate the perimeter to area ratio of the extracted target vegetation vector patches, set a sixth threshold to remove irregular strip-shaped patches; use mathematical morphology methods to eliminate holes in the vector patches; set an area threshold for the target vegetation vector patches to remove small, fragmented patches, and finally obtain accurate remote sensing image target vegetation extraction results.
[0050] Example 4 In a typical embodiment of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the NDVI and HSV-based remote sensing image vegetation extraction method described in Embodiment 1. These steps include: S1. Acquire satellite remote sensing images of the target vegetation in the first and second periods; S2. Calculate the normalized vegetation index for satellite remote sensing images of the first period and satellite remote sensing images of the second period, respectively; S3. Perform HSV color transformation on the satellite remote sensing images of the first period and the satellite remote sensing images of the second period to obtain the color feature components of the first period and the second period, including the hue component, saturation component and brightness component. S4. Construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to the preset decision nodes; wherein, the preset decision nodes include: The first type of interfering land cover was removed using the normalized vegetation index of the second period; the second type of interfering land cover was removed using the normalized vegetation index of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering land cover was removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period. S5. Post-process the extracted vector information of the first and second types of target vegetation. Specifically, calculate the perimeter to area ratio of the extracted target vegetation vector patches, set a sixth threshold to remove irregular strip-shaped patches; use mathematical morphology methods to eliminate holes in the vector patches; set an area threshold for the target vegetation vector patches to remove small, fragmented patches, and finally obtain accurate remote sensing image target vegetation extraction results.
[0051] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for vegetation extraction from remote sensing images based on NDVI and HSV, characterized in that, include: Acquire satellite remote sensing images of the target vegetation in the first and second periods; The normalized vegetation index was calculated for the first and second period satellite remote sensing images, respectively. HSV color transformation was performed on satellite remote sensing images from the first and second periods to obtain color feature components from the first and second periods, including hue, saturation and brightness components. A decision tree model is constructed based on the normalized vegetation index and color feature components. Interfering features are sequentially removed and the target vegetation is extracted according to preset decision nodes. The preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
2. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 1, characterized in that, The first period is specifically the budding stage of the target vegetation; the second period is specifically the vigorous growth stage of the target vegetation.
3. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 1, characterized in that, Category I disturbance features include pickled mustard greens and / or water bodies; Category II disturbance features include cultivated land and / or forest land; Category III disturbance features include grassland.
4. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 1, characterized in that, The first category of target vegetation consists of vegetation with significant color saturation characteristics; the second category of target vegetation consists of vegetation with prominent color characteristics and distributed in blocky patches.
5. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 1, characterized in that, The second type of target vegetation is extracted using the hue components of the first period. Specifically, this includes: performing window mean operation on the hue components of the first period to obtain optimized hue features, and extracting the second type of target vegetation based on the optimized hue features of the first period.
6. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 5, characterized in that, The window mean calculation uses a 5×5 pixel window.
7. The method for vegetation extraction from remote sensing images based on NDVI and HSV as described in claim 1, characterized in that, It also includes: post-processing the extracted vector information of the first and second types of target vegetation, specifically including: calculating the perimeter to area ratio of the target vegetation vector patch, setting a threshold to remove irregular strip-shaped patches; using mathematical morphology methods to eliminate holes in the vector patch; setting an area threshold for the target vegetation vector patch to remove small, fragmented patches.
8. A vegetation extraction system based on NDVI and HSV remote sensing images, characterized in that, include: The image acquisition module is configured to acquire satellite remote sensing images of the target vegetation in the first and second periods. The NDVI calculation module is configured to calculate the normalized vegetation index of satellite remote sensing images from the first and second periods, respectively. The HSV transformation module is configured to perform HSV color transformation on satellite remote sensing images from the first and second periods to obtain color feature components from the first and second periods, including hue, saturation and brightness components. The classification module is configured to: construct a decision tree model based on the normalized vegetation index and color feature components, and sequentially remove interfering features and extract target vegetation according to preset decision nodes; wherein, the preset decision nodes include: The first type of interfering features were removed using the normalized vegetation index (NDI) of the second period; the second type of interfering features were removed using the NDI of the first period; the first type of target vegetation was extracted using the saturation component of the second period; the third type of interfering features were removed using the brightness component of the first period; and the second type of target vegetation was extracted using the hue component of the first period.
9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the method for extracting vegetation from remote sensing images based on NDVI and HSV as described in any one of claims 1-7.
10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the method for extracting vegetation from remote sensing images based on NDVI and HSV as described in any one of claims 1-7.