Construction method of remote sensing comprehensive vegetation index based on local window adaptive entropy weight method
Patent Information
- Application Number
- CN202611029716.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-10
- Publication Date
- 2026-10-13
AI Technical Summary
[0006]针对现有技术存在的问题，本发明提供了基于局部窗口自适应熵权法的遥感综合植被指数构建方法，具备空间自适应、高低植被覆盖区间数值稳定、多指数均衡融合、精细反映地表异质性的优点，解决了现有技术中传统植被指数表征片面、融合权重僵化、监测精度不足的问题
提高植被信息表达的适应性与准确性：针对传统植被指数（如NDVI、EVI、MSAVI等）在不同地表环境下采用固定计算模型、权重恒定、适应性不足的问题；本发明通过构建局部空间窗口，引入基于局部统计特征的评估机制，使不同植被指数在不同空间位置能够依据其实际信息贡献自适应参与综合计算，从而提高植被信息在沙漠、森林、农田等不同环境中的表达准确性；
Smart Images
Figure CN122906548A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of remote sensing ecological monitoring and vegetation information extraction technology, and particularly relates to a method for constructing a comprehensive vegetation index based on the local window adaptive entropy weight method. Background Technology
[0002] Vegetation is an important component of terrestrial ecosystems. Its cover and structural characteristics directly reflect the regional ecological environment quality and ecological security level. Extracting vegetation information based on remote sensing data has become a core technical means for large-scale, long-term ecological monitoring. Among them, traditional vegetation indices such as the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) are widely used in vegetation cover estimation, vegetation change monitoring, and ecological assessment research due to their simple calculation methods and convenient data acquisition.
[0003] Existing traditional vegetation indices still have significant limitations in practical applications. Indices such as NDVI and EVI mostly rely on single or a small number of spectral bands for construction, and are essentially empirical expressions of the relationship between near-infrared and red light reflectance, making it difficult to comprehensively characterize the structural complexity of vegetation and its coupling relationship with environmental factors. In high-coverage areas such as forests and shrublands, NDVI is prone to saturation, resulting in a significant decrease in the index's sensitivity to changes in high-biomass vegetation; while in deserts, bare land, or sparsely vegetated areas, the index values are easily affected by soil background, resulting in poor stability.
[0004] Traditional vegetation indices exhibit significant mechanical and singular adaptability to different ecological environments. Existing methods typically employ uniform index models and fixed parameters, failing to adaptively adjust for ecosystem types with significantly varying vegetation cover, such as forests, grasslands, and deserts. This results in insufficient comparability and accuracy when applied across regions and climate zones. Furthermore, traditional indices often exist as single indicators, neglecting the comprehensive expressive power of multidimensional remote sensing information and failing to reflect the comprehensive characteristics of vegetation ecosystems. This single-indicator-driven evaluation approach can easily lead to one-sided ecological status identification in complex surface environments. In addition, regarding statistical analysis at the research scale, existing studies often focus on simple statistics such as means or extreme values, ignoring the spatial distribution structure of vegetation indices and failing to reveal the heterogeneity and hierarchical characteristics of vegetation conditions within a region.
[0005] Therefore, there is a need for a CRSVI construction method that can integrate multiple remote sensing indicators, overcome the mechanical defects of traditional vegetation indices, and possess efficient computing power and multidimensional statistical analysis capabilities, so as to more accurately and comprehensively characterize the vegetation status under different ecological environments and provide reliable technical support for regional ecological monitoring and management. Summary of the Invention
[0006] To address the problems of existing technologies, this invention provides a method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method. This method has the advantages of spatial adaptability, numerical stability between high and low vegetation cover intervals, balanced fusion of multiple indices, and precise reflection of surface heterogeneity. It solves the problems of one-sided representation, rigid fusion weights, and insufficient monitoring accuracy of traditional vegetation indices in existing technologies.
[0007] This invention is implemented as follows: a remote sensing integrated vegetation index construction method based on the local window adaptive entropy weight method includes the following steps: S1. Remote Sensing Data Acquisition and Preprocessing This is used to acquire multispectral remote sensing image data of the study area. It can be multi-temporal or non-multi-temporal. The remote sensing image data should contain red, near-infrared and blue light bands, etc. It supports the calculation of a variety of commonly used remote sensing vegetation indices. The data is preprocessed with cloud removal, atmospheric correction, geometric correction, mask cropping and other preprocessing. Commonly recommended remote sensing data sources include Landsat, Sentinel-2, MODIS and other remote sensing images. S2, Calculation and temporal synthesis of multiple vegetation indices This technical method is mainly based on the integration of the following three remote sensing vegetation indices: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Improved Soil-Regulated Vegetation Index (MSAVI) were used. The time dimension of remote sensing images acquired in the study area within the preset time range was statistically processed. If it was multi-temporal data, the monthly remote sensing vegetation index should be generated by median or mean synthesis according to the target requirements to reduce the impact of outliers on subsequent calculations. S3, Standardized treatment of vegetation index Since the entropy weight method calculates weights based on "fluctuation" rather than "absolute value," to avoid problems such as entropy weight distortion, incomparability within the window, and abnormal comprehensive index caused by different vegetation index value ranges, this method standardizes each vegetation index involved in the calculation before adopting the entropy weight method. The standardization method is to obtain the 1st and 99th quantiles of the corresponding index in the study area as the upper and lower limits. For indices above 99% and below 1%, they are assigned values equivalent to 99% and 1%, respectively. Then, vegetation indices with different dimensions and value ranges are normalized to make each vegetation index fall within a uniform value range, providing a basis for subsequent weight calculation. S4, Local Space Window Construction To overcome the problem that traditional vegetation index analysis methods rely solely on global statistical features and fail to reflect spatial heterogeneity, this invention introduces the construction of local spatial windows at the pixel scale: a local spatial window of a preset size is constructed around the target pixel in the remote sensing image to define the spatial range for statistical analysis. Through this local spatial window, the vegetation information of a single pixel is expanded to a comprehensive expression within a certain spatial scale around it, thereby depicting the spatial distribution characteristics and structural differences of vegetation in a local area. By defining the statistical range through local windows, the variation characteristics of vegetation patch boundaries, transition zones, and heterogeneous landscape units can be effectively reflected, avoiding the obscuring of local information by global statistical methods. In the test of this method, a 9×9 window size is a window size that can balance computational load and entropy weight determination, and is the recommended window setting parameter. S5, Local Adaptive Entropy Weight Calculation This method assesses the local spatial information content of each vegetation index based on its coefficient of variation within a local window, and assigns weights to each vegetation index based on the concept of entropy weight. The larger the coefficient of variation, the stronger the discriminative power of the index in the local space, the smaller the corresponding entropy value, and the higher the assigned weight. By assigning corresponding weight values to different vegetation indices within a window based on their information entropy, and by adjusting the entropy weight calculation within each window as the window moves, the weight allocation of each index can be dynamically adjusted according to changes in spatial location—this is the adaptive entropy weight allocation method. Through this method, the advantages of each vegetation index can be leveraged in different surface environments such as forests, grasslands, and bare land to enhance their respective weight values in the calculation of CRSVI, thereby amplifying the information expression capabilities of vegetation indices in various surface environments. Overall, this method achieves adaptive weight adjustment of different vegetation indices under different ecological conditions, avoiding mechanically fixed index weights, thereby significantly improving the stability and applicability of CRSVI in complex ecological environments. S6. Weight Constraints and Normalization Correction To avoid excessive concentration or extreme imbalance in the weight allocation results in local areas, this invention further sets up a weight constraint and normalization module to constrain and standardize the local weights: the weight values obtained by each vegetation index within the local spatial window are subject to amplitude constraints. By setting the upper and lower limits of the weights, the weight values exceeding the reasonable range are constrained and adjusted to prevent a certain vegetation index from obtaining too high a weight due to local abnormal fluctuations in a specific area, or from being completely suppressed due to insufficient information expression. Through this amplitude constraint, it can be ensured that different vegetation indices can participate in the result construction during the comprehensive calculation process, thereby maintaining the stability and balance of multi-index fusion. According to testing, the weight amplitude currently recommended by this method is 0.2-0.5. After completing the weight limiting process, the adjusted weights are normalized, which can effectively suppress the weight polarization phenomenon that may occur in highly heterogeneous regions by the local entropy weight method, avoid the excessive dependence of CRSVI on a single index, and significantly improve the spatial continuity and numerical stability of the fusion results. S7, Comprehensive Index Fusion Calculation After completing the local weight constraint and normalization process, based on the normalized local weights corresponding to each vegetation index, entropy weight calculation based on the moving window is performed on multiple vegetation indices to generate CRSVI images. S8, Image Optimization Processing Edge processing: The progressive window method is used to process the edge raster units (black borders and areas with no value) of the CRSVI results. This method uses 3×3 mean filtering to smooth and fill the image edges, with 2 iterations to ensure that the final image is complete and without missing edges. Denoising: A Gaussian smoothing filter (radius: 1, sigma: 0.5) is used to lightly modify the final image, removing noise and imperfections, and enhancing image quality and effect; The results of CRSVI can form a rasterized index image with spatial continuity. This image can be used for applications such as vegetation cover classification, vegetation spatial pattern analysis, temporal change monitoring, and ecological environment assessment, providing reliable data support for regional ecological status assessment and related decision-making.
[0008] Compared with the prior art, the beneficial effects of the present invention are as follows: Improving the adaptability and accuracy of vegetation information representation: Addressing the shortcomings of traditional vegetation indices (such as NDVI, EVI, MSAVI, etc.) which employ fixed calculation models, constant weights, and insufficient adaptability in different land surface environments, this invention constructs a local spatial window and introduces an evaluation mechanism based on local statistical characteristics. This allows different vegetation indices to adaptively participate in comprehensive calculations based on their actual information contributions at different spatial locations, thereby improving the accuracy of vegetation information representation in different environments such as deserts, forests, and farmlands. Effectively alleviate the saturation problem of single vegetation indices: In traditional techniques, single vegetation indices are prone to saturation in high-coverage areas and are easily affected by soil background interference in low-coverage or bare land areas, leading to information distortion. This invention utilizes the complementary characteristics of different indices under different cover conditions through a local weighted fusion method of multiple vegetation indices, thereby reducing the impact of single index failure on the results and improving the stability of CRSVI over a wide cover gradient range. Attached Figure Description
[0009] Figure 1 This is a schematic diagram of the method flow provided in an embodiment of the present invention; Figure 2This is a schematic diagram of a high vegetation cover area (Guizhou forest area) provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of a low vegetation cover area (Xinjiang desert region) provided in an embodiment of the present invention. Detailed Implementation
[0010] To further understand the invention's content, features, and effects, the following embodiments are provided, and detailed descriptions are given in conjunction with the accompanying drawings.
[0011] The structure of the present invention will now be described in detail with reference to the accompanying drawings.
[0012] refer to Figures 1 to 3 As shown, the remote sensing integrated vegetation index construction method based on the local window adaptive entropy weight method provided in this embodiment of the invention includes the following steps: S1. Remote Sensing Data Acquisition and Preprocessing This is used to acquire multispectral remote sensing image data of the study area. It can be multi-temporal or non-multi-temporal. The remote sensing image data should contain red, near-infrared and blue light bands, etc. It supports the calculation of a variety of commonly used remote sensing vegetation indices. The data is preprocessed with cloud removal, atmospheric correction, geometric correction, mask cropping and other preprocessing. Commonly recommended remote sensing data sources include Landsat, Sentinel-2, MODIS and other remote sensing images. S2, Calculation and temporal synthesis of multiple vegetation indices This technical method is mainly based on the integration of the following three remote sensing vegetation indices: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Improved Soil-Regulated Vegetation Index (MSAVI) were used. The time dimension of remote sensing images acquired in the study area within the preset time range was statistically processed. If it was multi-temporal data, the monthly remote sensing vegetation index should be generated by median or mean synthesis according to the target requirements to reduce the impact of outliers on subsequent calculations. S3, Standardized treatment of vegetation index Since the entropy weight method calculates weights based on "fluctuation" rather than "absolute value," to avoid problems such as entropy weight distortion, incomparability within the window, and abnormal comprehensive index caused by different vegetation index value ranges, this method standardizes each vegetation index involved in the calculation before adopting the entropy weight method. The standardization method is to obtain the 1st and 99th quantiles of the corresponding index in the study area as the upper and lower limits. For indices above 99% and below 1%, they are assigned values equivalent to 99% and 1%, respectively. Then, vegetation indices with different dimensions and value ranges are normalized to make each vegetation index fall within a uniform value range, providing a basis for subsequent weight calculation. S4, Local Space Window Construction To overcome the problem that traditional vegetation index analysis methods rely solely on global statistical features and fail to reflect spatial heterogeneity, this invention introduces the construction of local spatial windows at the pixel scale: a local spatial window of a preset size is constructed around the target pixel in the remote sensing image to define the spatial range for statistical analysis. Through this local spatial window, the vegetation information of a single pixel is expanded to a comprehensive expression within a certain spatial scale around it, thereby depicting the spatial distribution characteristics and structural differences of vegetation in a local area. By defining the statistical range through local windows, the variation characteristics of vegetation patch boundaries, transition zones, and heterogeneous landscape units can be effectively reflected, avoiding the obscuring of local information by global statistical methods. In the test of this method, a 9×9 window size is a window size that can balance computational load and entropy weight determination, and is the recommended window setting parameter. S5, Local Adaptive Entropy Weight Calculation This method assesses the local spatial information content of each vegetation index based on its coefficient of variation within a local window, and assigns weights to each vegetation index based on the concept of entropy weight. The larger the coefficient of variation, the stronger the discriminative power of the index in the local space, the smaller the corresponding entropy value, and the higher the assigned weight. By assigning corresponding weight values to different vegetation indices within a window based on their information entropy, and by adjusting the entropy weight calculation within each window as the window moves, the weight allocation of each index can be dynamically adjusted according to changes in spatial location—this is the adaptive entropy weight allocation method. Through this method, the advantages of each vegetation index can be leveraged in different surface environments such as forests, grasslands, and bare land to enhance their respective weight values in the calculation of CRSVI, thereby amplifying the information expression capabilities of vegetation indices in various surface environments. Overall, this method achieves adaptive weight adjustment of different vegetation indices under different ecological conditions, avoiding mechanically fixed index weights, thereby significantly improving the stability and applicability of CRSVI in complex ecological environments. S6. Weight Constraints and Normalization Correction To avoid excessive concentration or extreme imbalance in the weight allocation results in local areas, this invention further sets up a weight constraint and normalization module to constrain and standardize the local weights: the weight values obtained by each vegetation index within the local spatial window are subject to amplitude constraints. By setting the upper and lower limits of the weights, the weight values exceeding the reasonable range are constrained and adjusted to prevent a certain vegetation index from obtaining too high a weight due to local abnormal fluctuations in a specific area, or from being completely suppressed due to insufficient information expression. Through this amplitude constraint, it can be ensured that different vegetation indices can participate in the result construction during the comprehensive calculation process, thereby maintaining the stability and balance of multi-index fusion. According to testing, the weight amplitude currently recommended by this method is 0.2-0.5. After completing the weight limiting process, the adjusted weights are normalized, which can effectively suppress the weight polarization phenomenon that may occur in highly heterogeneous regions by the local entropy weight method, avoid the excessive dependence of CRSVI on a single index, and significantly improve the spatial continuity and numerical stability of the fusion results. S7, Comprehensive Index Fusion Calculation After completing the local weight constraint and normalization process, based on the normalized local weights corresponding to each vegetation index, entropy weight calculation based on the moving window is performed on multiple vegetation indices to generate CRSVI images. S8, Image Optimization Processing Edge processing: The progressive window method is used to process the edge raster units (black borders and areas with no value) of the CRSVI results. This method uses 3×3 mean filtering to smooth and fill the image edges, with 2 iterations to ensure that the final image is complete and without missing edges. Denoising: A Gaussian smoothing filter (radius: 1, sigma: 0.5) is used to lightly modify the final image, removing noise and imperfections, and enhancing image quality and effect; The results of CRSVI can form a rasterized index image with spatial continuity. This image can be used for applications such as vegetation cover classification, vegetation spatial pattern analysis, temporal change monitoring, and ecological environment assessment, providing reliable data support for regional ecological status assessment and related decision-making.
[0013] This invention utilizes a local window adaptive entropy weighting method to dynamically fuse multiple vegetation indices to construct a comprehensive remote sensing vegetation index. It overcomes the limitations of traditional vegetation index models, which suffer from fixed parameters and weights and poor scene adaptability. This method adaptively allocates index weights based on the variation characteristics of vegetation information in different local land surfaces, effectively overcoming the shortcomings of traditional indices, such as easy saturation in high-coverage vegetation areas and susceptibility to soil background interference in low-coverage and bare land areas. Furthermore, through weight constraints and normalization, it avoids weight polarization and ensures the stability and spatial continuity of the fusion results. This significantly improves the accuracy, reliability, and universality of vegetation status representation under different ecological environments, providing high-precision and highly stable technical support for large-scale, long-term vegetation dynamic monitoring and regional ecological environment assessment.
[0014] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.
[0015] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method, characterized in that, Includes the following steps: S1. Remote sensing data acquisition and preprocessing: Acquire multispectral remote sensing images of the study area containing red, near-red and blue light bands, and perform cloud removal, atmospheric correction, geometric correction and study area masking preprocessing on the remote sensing images in sequence to obtain preprocessed image data. S2. Calculation and temporal synthesis of multiple vegetation indices: Based on preprocessed image data, the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Modified Soil-Regulated Vegetation Index (MSAVI) are calculated respectively. For multi-temporal remote sensing data, the median or mean synthesis method is used to generate monthly vegetation index datasets, resulting in three phases of single vegetation index images. S3. Vegetation index standardization: The 1% and 99th percentiles of each vegetation index are extracted as the upper and lower limits of the values. Abnormal values that exceed the quantile range are truncated and constrained. The three vegetation indices after truncation are normalized to eliminate the differences in dimensions and value ranges, and the standardized vegetation index image is obtained. S4. Local spatial window construction: Using each target pixel in the image as the center, a local window of a preset size is constructed to limit the spatial range of local statistical analysis. Local vegetation spatial feature information is collected by sliding the window across the entire area. S5. Local Adaptive Entropy Weight Calculation: Calculate the coefficient of variation of each standardized vegetation index window by window. Based on the correlation between the coefficient of variation and information entropy, dynamically calculate the local initial weight of each vegetation index to achieve adaptive adjustment of the weight as the spatial surface scene changes. S6. Weight Constraint and Normalization Correction: The initial local weights are constrained to limit the weight values to a preset reasonable range. Then, the weights after the constraint are normalized to obtain the balanced adaptive fusion weights. S7. Comprehensive index fusion calculation: Based on the balanced adaptive fusion weight, the three-phase standardized vegetation indices are weighted and fused to generate the initial comprehensive remote sensing vegetation index (CRSVI) image. S8. Image optimization processing: Edge completion and noise reduction are performed on the initial CRSVI image to obtain the final high-precision CRSVI result image.
2. The method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method as described in claim 1, characterized in that: In step S1, the multispectral remote sensing image is any one or more combinations of Landsat, Sentinel-2, and MODIS remote sensing images.
3. The method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method as described in claim 1, characterized in that: In step S4, the preset size of the local movable window is 9×9 pixels.
4. The method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method as described in claim 1, characterized in that: In step S6, the preset reasonable range for the weight limit is 0.2-0.
5.
5. The method for constructing a comprehensive vegetation index based on the local window adaptive entropy weight method as described in claim 1, characterized in that: In step S8, the edge completion adopts the 3×3 mean filtering progressive window method, with the number of iterations set to 2, to fill the image edge areas with no value or black borders.
6. The method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method as described in claim 1, characterized in that: In step S8, the noise reduction process uses Gaussian smoothing filtering with the filter parameters set to radius 1 and sigma=0.
5.
7. The method for constructing a remote sensing integrated vegetation index based on the local window adaptive entropy weight method as described in claim 6, characterized in that: In step S5, the coefficient of variation is negatively correlated with information entropy.