Intelligent inversion system and method for vegetation coverage based on remote sensing time series analysis
By constructing a time-series normalized vegetation index sequence and adaptive growing season segmentation, combined with phenological type and spatiotemporal consistency correction, the stability and continuity issues of vegetation cover inversion were solved, and high-precision vegetation cover inversion was achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG HONGSEN ECOLOGICAL TECHNOLOGY CO LTD
- Filing Date
- 2026-07-01
- Publication Date
- 2026-07-31
AI Technical Summary
Existing vegetation cover inversion technologies cannot fully depict the temporal changes of vegetation throughout the entire growing season, cannot adaptively match the growth fluctuation characteristics of different regions and different vegetation, and are easily affected by cloud interference and geometric registration bias, resulting in insufficient stability and poor spatial continuity of the inversion results.
A time-series normalized vegetation index sequence for the target region is constructed. The start and end points of the growing season are identified through an adaptive growing season segmentation module. Combined with phenological type and spatiotemporal consistency correction, an accurate vegetation cover inversion map is generated.
It achieves high-precision and dynamic vegetation cover inversion, generating accurate, continuous, and stable vegetation cover inversion maps, which are suitable for vegetation ecological monitoring and resource management.
Smart Images
Figure CN122493300A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of image processing technology, and in particular to an intelligent vegetation cover retrieval system and method based on remote sensing time-series analysis. Background Technology
[0002] Existing vegetation cover inversion technologies mostly rely on single-temporal or limited-temporal remote sensing images, which cannot fully depict the temporal changes of vegetation throughout the entire growing season. The identification of the start and end points of the growing season adopts a fixed threshold judgment mode, which cannot adaptively match the growth fluctuation characteristics of different regions and different vegetation. This results in low segmentation accuracy of growing season segments, and the inversion results are difficult to match the actual growth status of vegetation.
[0003] Traditional inversion methods do not establish differentiated coverage calculation rules based on vegetation phenology. During the processing of time-series remote sensing data, they are easily affected by factors such as cloud interference and geometric registration deviation. At the same time, they lack a spatiotemporal consistency correction step, resulting in insufficient stability and poor spatial continuity of vegetation coverage inversion results, making it impossible to achieve high-precision and dynamic vegetation coverage inversion and monitoring. Summary of the Invention
[0004] This invention provides an intelligent vegetation cover retrieval system and method based on remote sensing time series analysis to solve the problems mentioned in the background art.
[0005] To achieve the above objectives, the present invention provides an intelligent vegetation cover retrieval system based on remote sensing time-series analysis, characterized in that the system includes a time-series construction module, an adaptive growing season segmentation module, and a vegetation cover retrieval module, wherein: The temporal construction module is used to sort the collected multi-temporal remote sensing image sequences pixel by pixel under the complete vegetation growing season cycle in the target area and construct the temporal normalized vegetation index sequence of the target area. The adaptive growing season segmentation module is used to perform fluctuation characteristic analysis on the time-series normalized vegetation index sequence. It adaptively identifies the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, and uses the start and end points of the growing season as the segmentation boundary to divide the time-series normalized vegetation index sequence into multiple growing season segments. The vegetation cover inversion module is used to estimate the vegetation cover of the target area based on the time series feature vectors and phenological types in the growing season segments, and convert the vegetation cover corrected by spatiotemporal consistency into a vegetation cover inversion map of the target area.
[0006] In a preferred embodiment, when the time-series construction module performs pixel-by-pixel sorting of the acquired multi-temporal remote sensing image sequences during the complete vegetation growing season cycle in the target area, it is specifically used for: Obtain the imaging date and corresponding cloud cover detection results for each image in a multi-temporal remote sensing image sequence; For each pixel location in the target area, within the complete vegetation growing season, images with cloud coverage areas below a preset tolerance range at independent pixel locations are selected as candidate images according to the order of imaging dates. The images whose imaging dates are closest to the equal time points within the growing season cycle are selected from the candidate images as the valid images for the pixel locations. Invalid images with cloud coverage exceeding the allowable range are removed, and spatial geometric registration is performed on the retained temporal images to ensure that all images have a consistent pixel spatial correspondence within the target area.
[0007] In a preferred embodiment, when the time-series construction module executes the construction of a time-series normalized vegetation index sequence for the target region, it is specifically used for: In the images of each time phase after geometric registration, the near-infrared band values and red band values at the coordinate positions of each pixel in the target area are extracted and arranged in chronological order to form a time series of band values at the pixel coordinate positions. Calculate the normalized vegetation index based on the band numerical time series; The normalized vegetation index is statistically arranged to obtain the time-series normalized vegetation index sequence of the target area.
[0008] In a preferred embodiment, the normalized vegetation index is calculated using the following formula: ; In the formula, Normalized Difference Vegetation Index (NDVI) The near-infrared reflectance value at the time of target acquisition. The red band reflectance values at the same time. The preset regularization strength coefficient, This represents the sample standard deviation of the red band reflectance values of a pixel at all valid acquisition times within a complete vegetation growing season.
[0009] In a preferred embodiment, when the adaptive growing season segmentation module performs fluctuation characteristic analysis on the time-series normalized vegetation index sequence and adaptively identifies the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, it is specifically used for: The rate of change of the index between adjacent time points is determined based on the difference in the normalized vegetation index for each pair of adjacent time points in the time series normalized vegetation index sequence. Traversing backward from the starting point of the time-series normalized vegetation index sequence, the first time point in which the rate of change of the index turns from negative to positive and the positive value appears consecutively for multiple time steps is identified as the candidate starting point of the growing season. Continuing to iterate backward from the start of the growing season, the first time point in time where the rate of change turns from positive to negative and appears consecutively for multiple time steps is identified as the candidate end of the growing season.
[0010] In a preferred embodiment, the adaptive growing season segmentation module, when adaptively identifying the start and end points of the growing season based on the exponential rate of change of adjacent time points in the sequence, further includes: Obtain the peak value of the index after the start of the candidate growing season, and obtain the trough value of the index before the end of the candidate growing season. If the peak of the index is later than the candidate growth season start point and the index increase between them is greater than the preset increase threshold, then the candidate growth season start point is confirmed as a valid growth season start point. If the index trough occurs earlier than the candidate growth season endpoint and the index decline between the two exceeds a preset decline threshold, then the candidate growth season endpoint is confirmed as a valid growth season endpoint.
[0011] In a preferred embodiment, the adaptive growing season segmentation module, when performing the division of the time-normalized vegetation index sequence into multiple growing season segments using the start and end points of the growing season as segmentation boundaries, is specifically used for: Using the start and end of the growing season as the boundaries of the current growing season segment, a subsequence from the start to the end of the growing season is extracted from the time-normalized vegetation index sequence as the current growing season segment. Using the next time point after the end of the growing season as the new sequence start point, the operation of identifying the start and end points of the next growing season is repeated until the complete time-series normalized vegetation index sequence is traversed to obtain multiple growing season segments of the target area. If the duration of one of the multiple growing season segments is less than the predetermined minimum growing season duration threshold, the segment will be merged into the adjacent previous growing season segment, and the merged segment will be used as a new growing season segment for subsequent processing.
[0012] In a preferred embodiment, the vegetation cover inversion module, when estimating the vegetation cover of a target area based on time-series feature vectors and phenological types in growing season segments, is specifically used for: The peak value, mean value and rate of increase of the normalized vegetation index sequence within multiple growing season segments were extracted as time series feature vectors. The phenological type of the growing season segment is determined based on the rate of increase. Phenological types include evergreen, deciduous, and grassland. The evergreen type corresponds to a pre-defined coverage mapping rule that sets the vegetation coverage to a high-value stable range. The summer green type corresponds to a coverage mapping rule that sets the coverage in segments according to the temporal position of the peak value. The grassland type corresponds to a mapping rule that linearly correlates the coverage with the peak value of the normalized vegetation index. By inputting the time series feature vector into the selected coverage mapping rule, the vegetation coverage estimate corresponding to the growing season segment is obtained; By combining the estimated values of all growing season segments within the same target area at the same time point according to pixel location, a full-time vegetation cover estimation map of the target area is obtained.
[0013] In a preferred embodiment, when the vegetation cover inversion module performs the conversion of vegetation cover corrected for spatiotemporal consistency into a vegetation cover inversion map of the target area, it is specifically used for: Based on the full-time vegetation cover estimation map, the change in cover between two adjacent time phases is examined pixel by pixel on the time axis; If the change exceeds the normal fluctuation range of the change of pixels in the same period of the previous complete growing season, the coverage estimate of the time phase is replaced with the average of the coverage values of the two consecutive time phases. On a unified spatial axis, the vegetation cover map for each time phase is smoothed by neighborhood smoothing, and the cover value of each cell is updated to the weighted average of the cover values of the surrounding neighboring cells. The vegetation cover map after time correction and spatial smoothing is used as the final vegetation cover inversion map of the target area.
[0014] To address the aforementioned problems, this invention also provides an intelligent vegetation cover retrieval method based on remote sensing time-series analysis, the method comprising: Under the complete vegetation growing season in the target area, the collected multi-temporal remote sensing image sequences are sorted pixel by pixel to construct the temporal normalized vegetation index sequence of the target area. Fluctuation characteristics analysis was performed on the time-series normalized vegetation index (NZV) sequence. The start and end points of the growing season were adaptively identified based on the rate of change of the index at adjacent time points in the sequence. The NZV sequence was then divided into multiple growing season segments using the start and end points of the growing season as dividing boundaries. Based on the time series feature vectors and phenological types in the growing season segments, the vegetation cover of the target area is estimated, and the vegetation cover corrected by spatiotemporal consistency is converted into a vegetation cover inversion map of the target area.
[0015] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention constructs a time-series normalized vegetation index sequence of the complete vegetation growing season cycle in the target area, and adaptively completes the segmentation of growing season segments by combining the index fluctuation characteristics. It can accurately match the natural laws of vegetation growth, stably process multi-temporal remote sensing image data, and efficiently capture the temporal change characteristics of vegetation growth. It has reliable technical feasibility and data processing stability.
[0016] 2. This invention achieves differentiated vegetation coverage estimation based on vegetation phenology types, and optimizes the inversion results through spatiotemporal consistency correction, generating accurate, continuous, and stable vegetation coverage inversion maps. It can provide efficient and feasible technical solutions for vegetation ecological monitoring, resource management, and other scenarios, and has significant practical value and broad application prospects. Attached Figure Description
[0017] Figure 1 This is a system architecture diagram of an intelligent vegetation cover retrieval system based on remote sensing time-series analysis provided in an embodiment of the present invention. Figure 2 This is a flowchart illustrating an intelligent vegetation cover inversion method based on remote sensing time-series analysis provided in an embodiment of the present invention.
[0018] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation
[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments belong to some, but not all, embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0020] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “said” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms, and “multiple” generally includes at least two unless the context clearly indicates otherwise.
[0021] Depending on the context, the word "if" or "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."
[0022] Furthermore, the timing of the steps in the following method embodiments is merely an example and not a strict limitation.
[0023] In practice, the server-side equipment deployed in the intelligent vegetation cover inversion system based on remote sensing time-series analysis may consist of one or more devices. This system can be implemented as a business instance, a virtual machine, or hardware devices. For example, it can be implemented as a business instance deployed on one or more devices in a cloud node. Simply put, this system can be understood as software deployed on a cloud node, providing intelligent vegetation cover inversion based on remote sensing time-series analysis to various user terminals. Alternatively, it can also be implemented as a virtual machine deployed on one or more devices in a cloud node, with application software installed to manage each user terminal. Alternatively, the intelligent vegetation cover inversion system based on remote sensing time series analysis can also be implemented as a server consisting of many identical or different types of hardware devices, with one or more hardware devices set up to provide the intelligent vegetation cover inversion system based on remote sensing time series analysis to each user terminal.
[0024] In terms of implementation, the intelligent vegetation cover retrieval system based on remote sensing time-series analysis and the user terminal are mutually adaptable. That is, if the intelligent vegetation cover retrieval system based on remote sensing time-series analysis is implemented as an application installed on a cloud service platform, then the user terminal is implemented as a client that establishes a communication connection with the application; or if the intelligent vegetation cover retrieval system based on remote sensing time-series analysis is implemented as a website, then the user terminal is implemented as a webpage; or if the intelligent vegetation cover retrieval system based on remote sensing time-series analysis is implemented as a cloud service platform, then the user terminal is implemented as a mini-program in an instant messaging application.
[0025] like Figure 1 The figure shown is a system architecture diagram of an intelligent vegetation cover inversion system based on remote sensing time series analysis provided in an embodiment of the present invention.
[0026] The intelligent vegetation cover inversion system based on remote sensing time-series analysis described in this invention can be set up in a cloud server. In terms of implementation, it can be implemented as one or more service devices, or as an application installed in the cloud (e.g., a mobile service operator's server, server cluster, etc.), or it can be developed as a website. Depending on the functions implemented, the intelligent vegetation cover inversion system based on remote sensing time-series analysis may include a time-series construction module, an adaptive growing season segmentation module, and a vegetation cover inversion module. The module described in this invention can also be called a unit, referring to a series of computer program segments that can be executed by an electronic device's processor and perform a fixed function, stored in the electronic device's memory.
[0027] In this embodiment of the invention, in the intelligent vegetation cover inversion system based on remote sensing time-series analysis, each of the above modules can be implemented independently and can call other modules. Here, "calling" can be understood as one module connecting to multiple modules of another type and providing corresponding services to those connected modules. In the intelligent vegetation cover inversion system based on remote sensing time-series analysis provided by this embodiment of the invention, the applicable scope of the system architecture can be adjusted by adding modules and directly calling them without modifying the program code, achieving cluster-based horizontal expansion to quickly and flexibly expand the system. In practical applications, the above modules can be set in the same device or different devices, or they can be set in virtual devices, such as service instances in a cloud server.
[0028] The following describes, with reference to specific embodiments, each component and its specific workflow of the intelligent vegetation cover retrieval system based on remote sensing time-series analysis: The temporal construction module is used to sort the collected multi-temporal remote sensing image sequences pixel by pixel under the complete vegetation growing season cycle in the target area and construct the temporal normalized vegetation index sequence of the target area. In this embodiment of the invention, when the time-series construction module performs pixel-by-pixel sorting of the acquired multi-temporal remote sensing image sequences during the execution of a complete vegetation growing season cycle in the target area, it is specifically used for: Obtain the imaging date and corresponding cloud cover detection results for each image in a multi-temporal remote sensing image sequence; For each pixel location in the target area, within the complete vegetation growing season, images with cloud coverage areas below a preset tolerance range at independent pixel locations are selected as candidate images according to the order of imaging dates. The images whose imaging dates are closest to the equal time points within the growing season cycle are selected from the candidate images as the valid images for the pixel locations. Invalid images with cloud coverage exceeding the allowable range are removed, and spatial geometric registration is performed on the retained temporal images to ensure that all images have a consistent pixel spatial correspondence within the target area.
[0029] When the time-series construction module executes the construction of the time-series normalized vegetation index sequence for the target region, it is specifically used for: In the images of each time phase after geometric registration, the near-infrared band values and red band values at the coordinate positions of each pixel in the target area are extracted and arranged in chronological order to form a time series of band values at the pixel coordinate positions. Calculate the normalized vegetation index based on the band numerical time series; The normalized vegetation index is statistically arranged to obtain the time-series normalized vegetation index sequence of the target area.
[0030] The formula for calculating the normalized vegetation index is as follows: ; In the formula, Normalized Difference Vegetation Index (NDVI) The near-infrared reflectance value at the time of target acquisition. The red band reflectance values at the same time. The preset regularization strength coefficient, This represents the sample standard deviation of the red band reflectance values of a pixel at all valid acquisition times within a complete vegetation growing season.
[0031] The near-infrared reflectance value of the target at the acquisition time is taken from the remote sensing images of each time phase that have completed spatial geometric registration. It is the surface near-infrared reflectance observation data directly acquired by the remote sensing sensor at the acquisition time of the corresponding pixel coordinate position of the target area. The data directly reflects the true near-infrared reflectance state of the pixel at the corresponding time.
[0032] The red band reflectance values at the same time are taken from remote sensing images of various time phases that have completed spatial geometric registration. They are the surface red band reflectance observation data obtained by remote sensing sensors and near-infrared band synchronously at the same acquisition time for the corresponding pixel coordinates of the target area. They are the matching observation data of the same time phase as the near-infrared band reflectance values.
[0033] The preset regularization intensity coefficient is determined based on industry standards for vegetation remote sensing index calculation, combined with the noise distribution characteristics of remote sensing data within a complete vegetation growing season. It was determined through multiple sets of vegetation cover inversion verification experiments and remains fixed as a regularization constraint for index calculation. The sample standard deviation of the red band reflectance values of a pixel at all valid acquisition times within a complete vegetation growing season is calculated by first acquiring the red band reflectance values of the pixel at all valid times throughout the growing season, calculating the arithmetic mean of all values, then calculating the difference between each reflectance value and the mean, squaring the result, summing all squared results, dividing by the total number of valid acquisition times, and finally taking the square root of the result to obtain the final sample standard deviation.
[0034] This calculation formula constructs the core calculation logic of the index by using the difference and relationship between the reflectance of the near-infrared band and the red band. It completes the numerical correction by combining the regularization intensity coefficient and the standard deviation of the red band reflectance sample. It can stably calculate the normalized vegetation index at the target collection time, providing accurate single-phase index data for constructing a time-series normalized vegetation index sequence, and supporting the orderly development of subsequent growing season segmentation and vegetation cover inversion processes.
[0035] The imaging date of each image is extracted from the metadata information of the multi-temporal remote sensing image sequence. By performing cloud coverage identification on the full-domain pixels of each remote sensing image, the cloud coverage detection result corresponding to each image is determined. This result directly reflects the cloud coverage distribution status of the image.
[0036] Using individual pixels within the target area as processing units, the multi-temporal remote sensing image sequence is arranged in chronological order of imaging dates within the time range of a complete vegetation growing season. For each independent pixel location, the cloud coverage area ratio of that pixel in the corresponding image is detected, and images with a cloud coverage area ratio value less than a preset allowable range value are identified as candidate images for that pixel location.
[0037] The complete vegetation growing season cycle is divided into several equally divided time points. For all candidate images at each pixel location, the matching degree between the imaging date of each candidate image and the time points of each equally divided time point is compared. The candidate image with the highest matching degree between the imaging date and the equally divided time point is selected as the valid image for that pixel location.
[0038] Images in a multi-temporal remote sensing image sequence whose cloud coverage area ratio exceeds a preset allowable range are identified as invalid images and directly removed. For all retained valid temporal images, the spatial location parameters of the images are adjusted by aligning the pixel coordinates of the same surface features in the images, so that every pixel in all images within the target area corresponds to the same surface location, forming a consistent pixel spatial correspondence.
[0039] On remote sensing images of each time phase that have completed spatial geometric registration and have consistent pixel spatial correspondence, the coordinate position of each fixed pixel in the target area is accurately located. The near-infrared band value and red band value of the pixel coordinate position are read in the current time phase image. According to the chronological order of the imaging dates of each time phase image, the near-infrared band value and red band value of the same pixel coordinate position are arranged in sequence to form the time series of band values corresponding to the pixel coordinate position.
[0040] Based on the near-infrared and red band values at the same time phase in the band numerical time series, the difference between the two types of band values is first calculated, and then the sum of the two types of band values is calculated. The difference and the sum are processed according to the vegetation index characterization rules to obtain the normalized vegetation index of the pixel coordinate position at the current time phase.
[0041] The normalized vegetation index (NDI) calculated from the coordinates of all pixels in the target area at various time phases is statistically analyzed and arranged in an orderly manner according to the chronological order of the imaging date, ultimately yielding a time-series NDI sequence that fully covers the target area.
[0042] The beneficial effects include the ability to accurately acquire the imaging date and cloud cover detection results of multi-temporal remote sensing images, complete the screening of candidate and valid images pixel by pixel, eliminate invalid images and achieve spatial geometric registration of images of each temporal phase, form a unified pixel spatial correspondence, orderly construct band numerical time series, accurately calculate the normalized vegetation index, and finally generate a complete temporal normalized vegetation index sequence of the target area, providing accurate and stable data support for subsequent vegetation cover inversion.
[0043] The adaptive growing season segmentation module is used to perform fluctuation characteristic analysis on the time-series normalized vegetation index sequence. It adaptively identifies the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, and uses the start and end points of the growing season as the segmentation boundary to divide the time-series normalized vegetation index sequence into multiple growing season segments. In this embodiment of the invention, when the adaptive growing season segmentation module performs fluctuation feature analysis on the time-series normalized vegetation index sequence and adaptively identifies the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, it is specifically used for: The rate of change of the index between adjacent time points is determined based on the difference in the normalized vegetation index for each pair of adjacent time points in the time series normalized vegetation index sequence. Traversing backward from the starting point of the time-series normalized vegetation index sequence, the first time point in which the rate of change of the index turns from negative to positive and the positive value appears consecutively for multiple time steps is identified as the candidate starting point of the growing season. Continuing to iterate backward from the start of the growing season, the first time point in time where the rate of change turns from positive to negative and appears consecutively for multiple time steps is identified as the candidate end of the growing season.
[0044] The adaptive growing season segmentation module, when performing adaptive identification of the start and end points of the growing season based on the exponential rate of change of adjacent time points in the sequence, further includes: Obtain the peak value of the index after the start of the candidate growing season, and obtain the trough value of the index before the end of the candidate growing season. If the peak of the index is later than the candidate growth season start point and the index increase between them is greater than the preset increase threshold, then the candidate growth season start point is confirmed as a valid growth season start point. If the index trough occurs earlier than the candidate growth season endpoint and the index decline between the two exceeds a preset decline threshold, then the candidate growth season endpoint is confirmed as a valid growth season endpoint.
[0045] The adaptive growing season segmentation module, when dividing the time-normalized vegetation index sequence into multiple growing season segments using the start and end points of the growing season as segmentation boundaries, is specifically used for: Using the start and end of the growing season as the boundaries of the current growing season segment, a subsequence from the start to the end of the growing season is extracted from the time-normalized vegetation index sequence as the current growing season segment. Using the next time point after the end of the growing season as the new sequence start point, the operation of identifying the start and end points of the next growing season is repeated until the complete time-series normalized vegetation index sequence is traversed to obtain multiple growing season segments of the target area. If the duration of one of the multiple growing season segments is less than the predetermined minimum growing season duration threshold, the segment will be merged into the adjacent previous growing season segment, and the merged segment will be used as a new growing season segment for subsequent processing.
[0046] Each adjacent time point in the time-series normalized vegetation index (NZVI) sequence is processed one by one. The NZVI value of the next adjacent time point is subtracted from the NZVI value of the previous adjacent time point to obtain the NZVI difference for that group of adjacent time points. This NZVI difference is directly used as the rate of change of the index between adjacent time points, thus completing the determination of the rate of change of the index for all adjacent time points in the time-series normalized vegetation index sequence.
[0047] Starting from the beginning of the time-series normalized vegetation index sequence, the index change rate corresponding to each time point is traversed sequentially in the direction of time progression. When the index change rate is detected to change from a negative state to a positive state, and the positive state after the change is maintained within a preset fixed number of time steps, the first time point in which the state change of the index change rate occurs is marked and determined as the candidate starting point of the growing season.
[0048] Starting from the determined candidate growing season starting point, continue to iterate backwards in the direction of time to the corresponding exponential rate of change at each time point. When the exponential rate of change is detected to change from a positive state to a negative state, and the negative state after the change is continuously maintained within a preset fixed number of time steps, mark the first time point at which the exponential rate of change changes is changed and determine it as the candidate growing season end point.
[0049] The normalized vegetation index (NDI) values are traversed for all time points after the start of the candidate growing season. The time point with the largest value is identified by comparing the values point by point. This time point is the peak value of the NDI after the start of the candidate growing season.
[0050] The normalized vegetation index values corresponding to all time points before the end of the candidate growing season are traversed. The time point with the smallest value is identified by comparing the values point by point. This time point is the trough of the index before the end of the candidate growing season.
[0051] The time point corresponding to the peak of the index is determined to be after the time point corresponding to the start of the candidate growing season. The difference between the normalized vegetation index value of the candidate growing season start point and the normalized vegetation index value at the peak of the index is calculated. This difference is the index increase rate between the two. When the index increase rate is greater than the preset increase rate threshold value, the candidate growing season start point is confirmed as a valid growing season start point.
[0052] The time point corresponding to the index trough is determined to be earlier than the time point corresponding to the candidate growing season end. The difference between the normalized vegetation index value of the candidate growing season end and the normalized vegetation index value at the index trough is calculated. This difference is the index decrease rate between the two. When the index decrease rate is greater than the preset decrease rate threshold, the candidate growing season end is confirmed as a valid growing season end.
[0053] The confirmed effective start and end points of the growing season are set as the dividing boundaries between the current growing season segment. The time points corresponding to these two boundaries are accurately located in the time series normalized vegetation index (NDVI) sequence. All normalized vegetation index data between the two boundary time points are completely extracted in the original time order, and the extracted subsequence is directly used as the current growing season segment.
[0054] Set the next time point corresponding to the currently determined end of the growing season as the new traversal start position of the time-normalized vegetation index sequence. Repeat the complete operation process of identifying the effective start and end of the growing season, and continue to complete the operation of identifying the start of the growing season, identifying the end of the growing season, and extracting the growing season segments until all time points of the time-normalized vegetation index sequence have been traversed and processed, and finally obtain multiple growing season segments corresponding to the target area.
[0055] Calculate the total number of time points covered by each growing season segment, and compare this total number of time points with the number of time points corresponding to the predetermined minimum growing season duration threshold. When the total number of time points of a growing season segment is less than the number of time points corresponding to the minimum growing season duration threshold, all the data of that growing season segment are concatenated in chronological order at the end of the adjacent previous growing season segment. The new data sequence formed after merging is used as a new growing season segment for subsequent vegetation cover inversion related processing.
[0056] The beneficial effects include the ability to accurately determine the rate of change of the index at adjacent time points in the time-series normalized vegetation index sequence, accurately identify candidate growing season start and end points, verify the effective growing season start and end points through the index peak and trough times, use these as boundaries to complete the growing season segmentation of the time-series sequence, merge segments that do not meet the duration criteria, adaptively match the fluctuation characteristics of vegetation growth, improve the accuracy and rationality of growing season segment segmentation, and provide stable and reliable growing season segment data for subsequent vegetation cover inversion.
[0057] The vegetation cover inversion module is used to estimate the vegetation cover of the target area based on the time series feature vectors and phenological types in the growing season segments, and convert the vegetation cover corrected by spatiotemporal consistency into a vegetation cover inversion map of the target area.
[0058] In this embodiment of the invention, when the vegetation cover inversion module estimates the vegetation cover of the target area based on the time series feature vector and phenological type in the growing season segment, it is specifically used for: The peak value, mean value and rate of increase of the normalized vegetation index sequence within multiple growing season segments were extracted as time series feature vectors. The phenological type of the growing season segment is determined based on the rate of increase. Phenological types include evergreen, deciduous, and grassland. The evergreen type corresponds to a pre-defined coverage mapping rule that sets the vegetation coverage to a high-value stable range. The summer green type corresponds to a coverage mapping rule that sets the coverage in segments according to the temporal position of the peak value. The grassland type corresponds to a mapping rule that linearly correlates the coverage with the peak value of the normalized vegetation index. By inputting the time series feature vector into the selected coverage mapping rule, the vegetation coverage estimate corresponding to the growing season segment is obtained; By combining the estimated values of all growing season segments within the same target area at the same time point according to pixel location, a full-time vegetation cover estimation map of the target area is obtained.
[0059] When the vegetation cover inversion module converts the vegetation cover corrected for spatiotemporal consistency into a vegetation cover inversion map of the target area, it is specifically used for: Based on the full-time vegetation cover estimation map, the change in cover between two adjacent time phases is examined pixel by pixel on the time axis; If the change exceeds the normal fluctuation range of the change of pixels in the same period of the previous complete growing season, the coverage estimate of the time phase is replaced with the average of the coverage values of the two consecutive time phases. On a unified spatial axis, the vegetation cover map for each time phase is smoothed by neighborhood smoothing, and the cover value of each cell is updated to the weighted average of the cover values of the surrounding neighboring cells. The vegetation cover map after time correction and spatial smoothing is used as the final vegetation cover inversion map of the target area.
[0060] The normalized vegetation index (NVI) sequence within each growing season segment is compared point by point to determine the maximum value in the sequence as the peak value. The mean value is obtained by summing all the NVI values in the sequence and dividing by the total number of time points in the sequence. The rate of increase is obtained by calculating the rate of change of the values from the start of the growing season to the peak value in the sequence according to the time order. The peak value, mean value, and rate of increase are combined to form the time series feature vector corresponding to the growing season segment.
[0061] The rising rate of the growing season segment is compared one by one with the preset threshold values for the three phenological types. If the rising rate is within the threshold range corresponding to the evergreen type, the growing season segment is identified as evergreen; if the rising rate is within the threshold range corresponding to the summer green type, the growing season segment is identified as summer green; and if the rising rate is within the threshold range corresponding to the grassland type, the growing season segment is identified as grassland.
[0062] For growing season segments identified as evergreen, the vegetation cover is set within a fixed high-value stable range by directly applying the preset coverage mapping rules. For growing season segments identified as deciduous, different time periods are divided according to the position of the index peak in the time series, and the vegetation cover is set according to the corresponding rules for each time period. For growing season segments identified as grassland, the vegetation cover is directly correlated with the peak value of the normalized vegetation index according to the preset linear correspondence.
[0063] The time-series feature vector corresponding to each growing season segment is substituted into the coverage mapping rule selected for the phenological type to which the segment belongs. Numerical matching and transformation are then performed according to the rule to obtain the vegetation cover estimate for each growing season segment at each time point. For all growing season segments within the target area, the same time point is selected to extract the vegetation cover estimate for each segment. All estimates are then systematically stitched together according to the spatial coordinates of pixels within the target area to form a complete temporal vegetation cover estimate map covering the entire target area.
[0064] Using the generated full-time vegetation cover estimation map as the processing basis, each independent pixel in the target area is traversed along the direction of time extension. The vegetation cover value of the pixel in two adjacent imaging phases is extracted in turn. The change in coverage of the pixel between the two phases is obtained by subtracting the value of the previous phase from the value of the later phase. The check of the change in coverage of all pixels in adjacent phases is completed.
[0065] Retrieve the normal fluctuation range of vegetation cover change for each pixel in the target area during the same period of the previous complete vegetation growing season. Compare the currently calculated vegetation cover change with the upper and lower limits of the normal fluctuation range. When the vegetation cover change is greater than the upper limit of the normal fluctuation range or less than the lower limit of the normal fluctuation range, replace the vegetation cover estimate for that time phase with the sum of the vegetation cover value of the previous time phase and the vegetation cover value of the next time phase, and divide by two.
[0066] Under a unified spatial coordinate system, for each vegetation cover map corresponding to each imaging time, a predetermined number of neighboring pixels are determined around each pixel. The vegetation cover values of the central pixel and its surrounding neighboring pixels are calculated according to a predetermined fixed weight. The resulting weighted average value is then used to replace the original vegetation cover value of the central pixel, completing the neighborhood smoothing process for the entire map. The vegetation cover map after completing the coverage change correction processing on the time axis and the neighborhood smoothing processing on the spatial axis is directly determined as the final vegetation cover inversion map of the target area.
[0067] The beneficial effects are that it can extract the temporal features of growing season segments and accurately determine the vegetation phenology type, obtain accurate vegetation coverage estimates by matching the corresponding coverage mapping rules according to different phenology types, integrate and generate a full-time vegetation coverage estimation map, correct abnormal coverage changes through the time axis and carry out neighborhood smoothing processing on the spatial axis, and finally obtain an accurate, stable and spatially continuous vegetation coverage inversion map of the target area.
[0068] Reference Figure 2The diagram shown is a flowchart illustrating an intelligent vegetation cover retrieval method based on remote sensing time-series analysis according to an embodiment of the present invention. In this embodiment, the intelligent vegetation cover retrieval method based on remote sensing time-series analysis includes: Under the complete vegetation growing season in the target area, the collected multi-temporal remote sensing image sequences are sorted pixel by pixel to construct the temporal normalized vegetation index sequence of the target area. Fluctuation characteristics analysis was performed on the time-series normalized vegetation index (NZV) sequence. The start and end points of the growing season were adaptively identified based on the rate of change of the index at adjacent time points in the sequence. The NZV sequence was then divided into multiple growing season segments using the start and end points of the growing season as dividing boundaries. Based on the time series feature vectors and phenological types in the growing season segments, the vegetation cover of the target area is estimated, and the vegetation cover corrected by spatiotemporal consistency is converted into a vegetation cover inversion map of the target area.
[0069] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.
[0070] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.
[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims
1. A smart vegetation cover retrieval system based on remote sensing time-series analysis, characterized in that, The system includes a time-series construction module, an adaptive growing season segmentation module, and a vegetation cover inversion module, wherein: The temporal construction module is used to sort the collected multi-temporal remote sensing image sequences pixel by pixel under the complete vegetation growing season cycle in the target area and construct the temporal normalized vegetation index sequence of the target area. The adaptive growing season segmentation module is used to perform fluctuation characteristic analysis on the time-series normalized vegetation index sequence. It adaptively identifies the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, and uses the start and end points of the growing season as the segmentation boundary to divide the time-series normalized vegetation index sequence into multiple growing season segments. The vegetation cover inversion module is used to estimate the vegetation cover of the target area based on the time series feature vectors and phenological types in the growing season segments, and convert the vegetation cover corrected by spatiotemporal consistency into a vegetation cover inversion map of the target area.
2. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 1, characterized in that, When the time-series construction module performs pixel-by-pixel sorting of the acquired multi-temporal remote sensing image sequences during the execution of a complete vegetation growing season cycle in the target area, it is specifically used for: Obtain the imaging date and corresponding cloud cover detection results for each image in a multi-temporal remote sensing image sequence; For each pixel location in the target area, within the complete vegetation growing season, images with cloud coverage areas below a preset allowable range at independent pixel locations are selected as candidate images according to the order of imaging dates. The images whose imaging dates are closest to the equal time points within the growing season cycle are selected from the candidate images as the valid images for the pixel locations. Invalid images with cloud coverage exceeding the allowable range are removed, and spatial geometric registration is performed on the retained temporal images to ensure that all images have a consistent pixel spatial correspondence within the target area.
3. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 2, characterized in that, When the time-series construction module executes the construction of the time-series normalized vegetation index sequence for the target region, it is specifically used for: In the images of each time phase after geometric registration, the near-infrared band values and red band values at the coordinate positions of each pixel in the target area are extracted and arranged in chronological order to form a time series of band values at the pixel coordinate positions. Calculate the normalized vegetation index based on the band numerical time series; The normalized vegetation index is statistically arranged to obtain the time-series normalized vegetation index sequence of the target area.
4. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 3, characterized in that, The formula for calculating the normalized vegetation index is as follows: ； In the formula, Normalized Difference Vegetation Index (NDVI) The near-infrared reflectance value at the time of target acquisition. The red band reflectance values at the same time. The preset regularization strength coefficient, This represents the sample standard deviation of the red band reflectance values of a pixel at all valid acquisition times within a complete vegetation growing season.
5. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 1, characterized in that, The adaptive growing season segmentation module, when performing fluctuation characteristic analysis on the time-series normalized vegetation index sequence and adaptively identifying the start and end points of the growing season based on the rate of change of the index at adjacent time points in the sequence, is specifically used for: The rate of change of the index between adjacent time points is determined based on the difference in the normalized vegetation index for each pair of adjacent time points in the time series normalized vegetation index sequence. Traversing backward from the starting point of the time-series normalized vegetation index sequence, the first time point in which the rate of change of the index turns from negative to positive and the positive value appears consecutively for multiple time steps is identified as the candidate starting point of the growing season. Continuing to iterate backward from the start of the growing season, the first time point in time where the rate of change turns from positive to negative and appears consecutively for multiple time steps is identified as the candidate end of the growing season.
6. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 5, characterized in that, The adaptive growing season segmentation module, when performing adaptive identification of the start and end points of the growing season based on the exponential rate of change of adjacent time points in the sequence, further includes: Obtain the peak value of the index after the start of the candidate growing season, and obtain the trough value of the index before the end of the candidate growing season. If the peak of the index is later than the candidate growth season start point and the index increase between them is greater than the preset increase threshold, then the candidate growth season start point is confirmed as a valid growth season start point. If the index trough occurs earlier than the candidate growth season endpoint and the index decline between the two exceeds a preset decline threshold, then the candidate growth season endpoint is confirmed as a valid growth season endpoint.
7. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 1, characterized in that, The adaptive growing season segmentation module, when dividing the time-normalized vegetation index sequence into multiple growing season segments using the start and end points of the growing season as segmentation boundaries, is specifically used for: Using the start and end of the growing season as the boundaries of the current growing season segment, a subsequence from the start to the end of the growing season is extracted from the time-normalized vegetation index sequence as the current growing season segment. Using the next time point after the end of the growing season as the new sequence start point, the operation of identifying the start and end points of the next growing season is repeated until the complete time-series normalized vegetation index sequence is traversed to obtain multiple growing season segments of the target area. If the duration of one of the multiple growing season segments is less than the predetermined minimum growing season duration threshold, the segment will be merged into the adjacent previous growing season segment, and the merged segment will be used as a new growing season segment for subsequent processing.
8. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 1, characterized in that, When the vegetation cover inversion module estimates the vegetation cover of a target area based on the time-series feature vectors and phenological types in the growing season segments, it is specifically used for: The peak value, mean value and rate of increase of the normalized vegetation index sequence within multiple growing season segments were extracted as time series feature vectors. The phenological type of the growing season segment is determined based on the rate of increase. Phenological types include evergreen, deciduous, and grassland. The evergreen type corresponds to a pre-defined coverage mapping rule that sets the vegetation coverage to a high-value stable range. The summer green type corresponds to a coverage mapping rule that sets the coverage in segments according to the temporal position of the peak value. The grassland type corresponds to a mapping rule that linearly correlates the coverage with the peak value of the normalized vegetation index. By inputting the time series feature vector into the selected coverage mapping rule, the vegetation coverage estimate corresponding to the growing season segment is obtained; By combining the estimated values of all growing season segments within the same target area at the same time point according to pixel location, a full-time vegetation cover estimation map of the target area is obtained.
9. The intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 8, characterized in that, When the vegetation cover inversion module converts the vegetation cover corrected for spatiotemporal consistency into a vegetation cover inversion map of the target area, it is specifically used for: Based on the full-time vegetation cover estimation map, the change in cover between two adjacent time phases is examined pixel by pixel on the time axis; If the change exceeds the normal fluctuation range of the change of pixels in the same period of the previous complete growing season, the coverage estimate of the time phase is replaced with the average of the coverage values of the two consecutive time phases. On a unified spatial axis, the vegetation cover map for each time phase is smoothed by neighborhood smoothing, and the cover value of each cell is updated to the weighted average of the cover values of the surrounding neighboring cells. The vegetation cover map after time correction and spatial smoothing is used as the final vegetation cover inversion map of the target area.
10. A method for intelligent inversion of vegetation cover based on remote sensing time-series analysis, characterized in that, The method for using the intelligent vegetation cover retrieval system based on remote sensing time-series analysis as described in claim 1: Under the complete vegetation growing season in the target area, the collected multi-temporal remote sensing image sequences are sorted pixel by pixel to construct the temporal normalized vegetation index sequence of the target area. Fluctuation characteristics analysis was performed on the time-series normalized vegetation index (NZV) sequence. The start and end points of the growing season were adaptively identified based on the rate of change of the index at adjacent time points in the sequence. The NZV sequence was then divided into multiple growing season segments using the start and end points of the growing season as dividing boundaries. Based on the time series feature vectors and phenological types in the growing season segments, the vegetation cover of the target area is estimated, and the vegetation cover corrected by spatiotemporal consistency is converted into a vegetation cover inversion map of the target area.