Remote sensing image shooting area optimization method based on historical income analysis

Abstract

The invention discloses a remote sensing image shooting area optimization method based on historical income analysis, and belongs to the technical field of remote sensing satellite task planning and resource scheduling. The method comprises the following steps: carrying out local grid division and income distribution on historical shooting task data, and constructing a standardized historical block database; establishing an industry income area; respectively calculating a long-term income index and a short-term income index for each industry income area with the time window containing the current date; normalizing the two, calculating a comprehensive income index, projecting the comprehensive income index to a block unit covered by an industry income area, and generating a block-level income estimation value; screening according to the estimated values to obtain a prediction point set; and carrying out visibility screening and executable shooting task optimization on the prediction point set to obtain a local optimal task set. According to the method, the potential high-value area is automatically excavated in the idle time period of the satellite, the autonomous imaging task is generated, and the satellite resource utilization rate and the imaging income efficiency are improved.

Description

Technical Field

[0001] This invention relates to the field of remote sensing satellite mission planning and resource scheduling technology, and in particular to a method for optimizing remote sensing image acquisition areas based on historical benefit analysis. Background Technology

[0002] With the rapid development of the commercial remote sensing market, the deployment of large-scale, low-cost satellite constellations (30-50 or more satellites) is becoming increasingly common. Their high-frequency coverage and wide-area imaging capabilities have been widely applied in various industries such as agriculture, forestry, water conservancy, and urban management. However, in actual operation, existing mission planning and scheduling systems still have many limitations, restricting the utilization efficiency of satellite resources and the commercial value conversion capability of remote sensing data.

[0003] Current remote sensing systems largely rely on an "order-driven" task execution model, where the system matches tasks and schedules orbits only after the customer submits a specific request. This passive response mechanism can meet routine needs when there are sufficient orders, but during periods of sparse or no orders, a large amount of idle time for satellites is wasted, and orbital resources, attitude adjustment opportunities, and image storage capacity are not effectively utilized, resulting in a serious problem of resource idleness.

[0004] Meanwhile, although commercial remote sensing platforms have accumulated a large number of image capture and sales records, existing systems are usually only used for static statistical analysis, lacking mechanisms to extract spatial revenue patterns and industry cycle regularities. The systems cannot identify potentially high-value areas from historical data, let alone proactively predict or generate imaging tasks without customer orders, leading to a continuous loss of potential business opportunities.

[0005] Furthermore, the demand for remote sensing images from different industries exhibits distinct temporal and spatial periodic characteristics. For example, the water conservancy industry experiences concentrated demand during the flood season, the forestry industry sees frequent fire monitoring during the dry season, and the agricultural industry experiences imaging peaks around the sowing and harvesting periods. However, existing systems lack readily available "industry-specific revenue region models" or demand evolution mechanisms, making it difficult to model and predict these trends.

[0006] More importantly, existing scheduling systems still primarily rely on physical constraints such as orbital attitude, angle cost, and storage capacity when prioritizing tasks, neglecting the potential differences in commercial benefits that tasks may bring, and thus failing to achieve revenue-driven resource allocation.

[0007] Therefore, there is an urgent need in this field for a new method that can proactively predict high-value imaging areas and optimize remote sensing image capture areas in a profit-oriented manner. Summary of the Invention

[0008] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for optimizing remote sensing image acquisition areas based on historical revenue analysis. This method is particularly suitable for multi-satellite systems for commercial operation, and can automatically discover potential high-value areas and generate autonomous imaging tasks during satellite idle periods to improve satellite resource utilization and imaging revenue efficiency.

[0009] To achieve the above objectives, the present invention adopts the following technical solution:

[0010] A remote sensing image acquisition area optimization method based on historical benefit analysis includes the following steps:

[0011] Step 1: Obtain historical shooting task data, and perform local grid division and revenue allocation on the historical shooting task data in sequence to generate several block units corresponding to each historical shooting task. Each block unit includes a unique task number, coordinates, date tag, industry tag, local metadata and revenue value. All the block units corresponding to the historical shooting tasks form a standardized historical block database.

[0012] Step 2: Establish industry revenue regions with independent spatial ranges and time windows. Each industry revenue region includes a region number, spatial range, time window, and industry code.

[0013] Step 3: For each industry return region within the time window that includes the current date, calculate the long-term return index and short-term return index for that industry return region.

[0014] When calculating the long-term return index, the block units corresponding to the industry's return area within the same time window over multiple past years are retrieved from the standardized historical block database. Based on the return values ​​included in the retrieved block units, the historical average return and average growth rate of the industry's return area are calculated, thereby obtaining the long-term return index of the industry's return area.

[0015] When calculating the short-term return index, the block unit corresponding to the industry's return area within the current period is retrieved from the standardized historical block database. The total return and return growth rate of the current period are calculated based on the return values ​​included in the retrieved block unit, thereby obtaining the short-term return index of the industry's return area. The current period refers to a fixed-length window that ends on the current date and traces back a preset number of days. When the start date of the window is earlier than the start time of the corresponding industry's return area, the start time of the corresponding industry's return area is used as the start date of the window.

[0016] Step 4: Normalize the long-term return index and the short-term return index respectively. Calculate the comprehensive return index for each industry's return region based on the normalized long-term return index and short-term return index. Project the comprehensive return index onto the block cell covered by the corresponding industry's return region to generate a block-level return estimate.

[0017] Step 5: For all block units covered by the industry revenue region within the time window that includes the current date, sort them by their block-level revenue estimates and filter out the top block units with the highest block-level revenue estimates. Each block unit generates a set of prediction points for task planning;

[0018] Step 6: Based on the available satellites in the constellation and their imaging parameters when performing imaging tasks, perform visibility screening and optimize the executable imaging tasks of the predicted point set to obtain a locally optimal task set, wherein each task includes the executing satellite number, execution time interval, the set of predicted point numbers contained therein, and total task benefits.

[0019] The remote sensing image acquisition area optimization method proposed in this invention, through data-driven and intelligent prediction methods, significantly improves the utilization efficiency of satellite resources, the commercial value density of task allocation, and the initiative and foresight of the operation and scheduling system. Compared with existing technologies, this invention has the following beneficial effects:

[0020] (1) This invention has autonomous prediction capabilities and no longer relies on external manual input:

[0021] This invention achieves the following by constructing a unified return index system (long-term return index + short-term return index):

[0022] Automatically learn the spatial benefit distribution patterns behind historical imaging tasks;

[0023] Automatically extrapolates historical return trends to future timeframes;

[0024] It can generate predictions of high-value areas in the future without relying on manual planning or external industry tips;

[0025] It possesses "spontaneous and autonomous" imaging planning capabilities, and can independently form a set of high-yield shooting areas, i.e., a set of locally optimal tasks, without external task input, thereby improving the automation and foresight of imaging task generation.

[0026] (2) This invention spatially standardizes historical shooting tasks, solving the problems of complex image orientation and difficulty in aligning regions:

[0027] This invention generates blocks with fixed orientations in the WGS84 coordinate system (which are not affected by image tilt) by dividing the local grid, and represents remote sensing images with arbitrary poses and tilt angles as unified spatial units. This enables the mapping of a large number of historical shooting tasks with inconsistent formats and poses onto a unified spatial structure, allowing historical shooting data from different times, different satellites, and different imaging angles to be directly compared under the same spatial reference, eliminating geometric differences between historical shooting data and improving the model's efficiency in utilizing historical information.

[0028] (3) This invention establishes a cross-regional and cross-industry benefit evaluation system on a unified grid (block):

[0029] This invention calculates a long-term return index, a short-term return index, and a comprehensive return index obtained by combining the two after normalization. It uses blocks as the basic granularity to achieve high-resolution return projection. Based on a unified spatial unit, it constructs a return evaluation system that is consistent in location and scale and can be compared horizontally, realizing unified evaluation across industries and time. This makes the subsequent prediction point selection, return calculation and task generation comparable and stable.

[0030] (4) This invention brings a “high-resolution revenue field” through revenue projection and spatial decay model:

[0031] This invention projects the comprehensive return index onto the block cell covered by the corresponding industry return area and calculates the block-level return estimate through a spatial decay function. This maps the industry return area (usually large in scale) onto a fine 1km × 1km grid, making the return field continuous, smooth and with high spatial resolution. This is beneficial for more accurately locating high-return points in a small area and improving the accuracy of prediction points.

[0032] (5) The task generation mechanism of this invention based on visual filtering and optimized combination has real-time performance and executability:

[0033] This invention optimizes the set of predicted points by performing visibility screening and executable imaging tasks. Based on the benefit evaluation at the predicted point level, it further introduces constraints such as satellite orbit, attitude, and imaging conditions, ultimately forming a set of executable locally optimal tasks. This achieves closed-loop automatic generation from "predicted benefit points" to "executable imaging tasks," and is engineering-practical. Attached Figure Description

[0034] Figure 1 This is a flowchart of the remote sensing image acquisition area optimization method based on historical benefit analysis according to the present invention.

[0035] Figure 2 A schematic diagram illustrating the division of a local mesh;

[0036] Figure 3 A schematic diagram of the revenue areas for the water conservancy industry;

[0037] Figure 4 This is a schematic diagram of the shooting tasks within the locally optimal task set. Detailed Implementation

[0038] The technical solution of the present invention will now be described in detail with reference to the accompanying drawings and preferred embodiments.

[0039] like Figure 1 As shown, this invention provides a method for optimizing remote sensing image acquisition areas based on historical benefit analysis, specifically including the following steps:

[0040] Step 1: Local grid division and revenue allocation, and establishment of a standardized historical block database.

[0041] First, acquire historical shooting task data. This data includes several historical shooting tasks, each denoted as […]. It contains the following information:

[0042] The image coverage area (i.e., the remote sensing image) has a spatial range of approximately 20km × 20km.

[0043] The mission center point, i.e., the area covered by the image. The geometric centroid;

[0044] The date of the photo, accurate to the day;

[0045] : Industry label, which is one of the following: water conservancy, agriculture, forestry, urban management, environmental protection and emergency response;

[0046] Task metadata, including resolution, cloud cover, and observation angle;

[0047] : The total reward for this task.

[0048] Next, the historical shooting task data is divided into local grids and the revenue is allocated in sequence to generate several block units corresponding to each historical shooting task. All the block units corresponding to the historical shooting tasks form a standardized historical block database.

[0049] Specifically, when performing local grid division, within the image coverage area... Within the smallest bounding rectangle, based on historical shooting tasks center point Using the origin as the coordinate system, generate coordinates parallel to the WGS84 coordinate system in all directions. A grid set is obtained by dividing the area into 1km × 1km blocks. , .

[0050] Local mesh generation specifically includes the following steps:

[0051] Step 1.1.1: For historical shooting tasks At a preset fixed interval (e.g., 1 km) in the image coverage area Within the smallest bounding rectangle, a regular grid is directly divided in the WGS84 coordinate system according to the latitude and longitude directions, where the longitude direction is the east-west axis and the latitude direction is the north-south axis. All the resulting blocks are strictly parallel to the latitude and longitude directions of the WGS84 coordinate system and do not rotate with the image's imaging pose or tilt angle. Figure 2 As shown, Figure 2 The area formed by the blue lines is the image coverage area. The area formed by the red lines is a regular grid.

[0052] Step 1.1.2: For the first grid Calculate its relationship with the image coverage area Intersection area The calculation formula is as follows:

[0053] ;

[0054] in, Indicates calculation and The area of ​​their intersection.

[0055] Step 1.1.3: Determine the intersection area Is it greater than the area threshold (e.g.) When the intersection area satisfy At that time, the grid It is added to the grid set as an effective block unit.

[0056] Step 1.1.4: Let Then repeat steps 1.1.2-1.1.3, traversing all the meshes generated in step 1.1.1, to obtain the historical shooting task. Corresponding grid set .

[0057] It should be noted that in this step, the spatial position of each grid is independent of other tasks, and no cross-task alignment is performed.

[0058] After completing the local grid division, the revenue is allocated based on historical shooting mission data, which includes the following steps:

[0059] Step 1.2.1: For historical shooting tasks Corresponding grid set The first in grid Computational grid The center of mass and the center point Distance between The distance decay function is obtained. ,in The preset industry attenuation coefficients are shown in Table 1. The industry attenuation coefficient reflects the sensitivity of industry returns to attenuation with distance from the center. For example, the water conservancy industry has a broad scope of concern and a slower attenuation rate, therefore its industry attenuation coefficient is relatively small; the environmental protection and emergency response industries have a strong central focus and the most significant attenuation, therefore their industry attenuation coefficient is relatively large.

[0060] Table 1. Industry Decline Coefficient Comparison Table

[0061]

[0062] Step 1.2.2: The distribution of profits follows the principle of "central decay + area weighting," therefore, the following formula is used to calculate the... grid weight base :

[0063] ;

[0064] in:

[0065] : grid From the center of mass to the center point The distance, in km;

[0066] : grid The effective coverage area, i.e., the area covered by the image. The area of ​​the intersection, in km²;

[0067] : Industry-specific depreciation coefficient;

[0068] Distance decay function.

[0069] Step 1.2.3: Calculate the normalization factor The calculation formula is as follows:

[0070] ;

[0071] Step 1.2.4: Calculate the mesh Benefits The calculation formula is as follows:

[0072] ;

[0073] Step 1.2.5: Let Then repeat steps 1.2.1-1.2.4 to traverse the mesh set. After completing all the grids in the image, the historical shooting task is finished. The distribution of profits, and historical filming missions. Corresponding total revenue It must be fully allocated to its divided set of grids. In this process, the conservation of returns is guaranteed, that is, the following conditions are met:

[0074] ;

[0075] in, To be assigned to historical filming tasks The division of the first grid The benefits.

[0076] It should be noted that revenue distribution is completed independently within the task level, and the date tags are uniformly set as follows. .

[0077] Record grid set Information for each grid cell, including unique spatial polygon coordinates, date tags accurate to the day, industry tags, metadata, and allocated revenue, is used to generate historical shooting tasks. The corresponding number of block units.

[0078] Ultimately, the historical filming mission Transformed into Each block unit forms a block data set, which is denoted as: The standardized historical block database consists of the block data sets corresponding to all historical shooting missions, where each block unit includes the following information:

[0079] block_id: A unique identifier within the task;

[0080] wkt: The coordinates of the 1km×1km grid of this block unit, with the coordinate system being WGS84.

[0081] time: Date tag, which is the same as the shooting date of the corresponding historical shooting task;

[0082] field_code: Industry tag, which is the same as the industry tag of the corresponding historical shooting task;

[0083] meta: Local metadata, inherited or extracted from the task metadata of the corresponding historical shooting task;

[0084] w: The allocated income, calculated from steps 1.2.1 to 1.2.4.

[0085] Below is an example of profit distribution.

[0086] Historical filming mission Water conservancy industry, industry attenuation coefficient Total revenue (yuan), generate 3 valid meshes, with parameters as shown in Table 2.

[0087] Table 2 Parameters of the effective mesh

[0088]

[0089] The standardized historical block database obtained in step 1 aggregates historical returns within the corresponding time period and spatial range, providing a basis for subsequent short-term and long-term predictions.

[0090] Step 2: Establish industry revenue regions with independent spatial scope and time windows. Each industry revenue region includes a region number, spatial scope, time window, and industry code.

[0091] This step aims to construct the driving input set for long-term and short-term revenue forecasts. In this step, multiple independent industry revenue regions are defined based on industry characteristics and geographical differences. Each industry revenue region includes a pre-defined spatial range and time window, and is associated with an industry type (water conservancy, agriculture, forestry, urban management, environmental protection, and emergencies). During forecasting, the system automatically determines which industry revenue regions are in their active periods based on the current date, selectively retrieving relevant historical block data to form the forecast input. For example... Figure 3 The image shows the revenue area for the water conservancy industry in the downstream area of ​​a certain river. The green polygon area in the image represents the spatial range of this water conservancy industry revenue area, with the time label being June 1st to August 1st.

[0092] The term "industry revenue region" is used to describe "a spatial unit within a specific time period and spatial range where a certain industry generates revenue." Different regions, even those belonging to the same industry, have completely independent time windows and geographical boundaries, and there is no logical coupling between them.

[0093] No. The revenue regions of an industry can be defined in the following structured form:

[0094] ;

[0095] in:

[0096] : No. Each industry's revenue region has a unique region number;

[0097] The spatial extent of the region boundary is represented by a polygon in the WGS84 coordinate system. This spatial extent can be a polygon of any shape, and the spatial coordinates of the polygon boundary are defined and stored in the WGS84 coordinate system. It should be noted that each industry revenue area is defined based on actual business or geographical boundaries (e.g., watershed, agricultural belt, urban agglomeration, forest belt, etc.). Different industry revenue areas can overlap, but they are processed independently in the forecasting stage without logical conflicts.

[0098] : Time window, in days, formatted as ,in The start date of the time window. This is the end date of the time window;

[0099] : Industry code, used for subsequent weighting and normalization of prediction results.

[0100] In step 2, the time window for each industry revenue region is determined by the region's climate conditions, geographical features, and industry work cycle, and the time windows of different industry revenue regions within the same industry are not shared or merged.

[0101] For example, the revenue windows for different sectors within the water conservancy industry may differ:

[0102] Middle and lower reaches of the Yangtze River: June 1 – August 15;

[0103] Songhua River Basin Area: July 10 – September 20;

[0104] Pearl River Basin region: May 20 – July 30.

[0105] When establishing industry revenue regions, the specific time range, or time window, for each industry revenue region is recorded independently. The time window is accurate to the day and can be stored in the format [YYYY-MM-DD_start, YYYY-MM-DD_end]. If it is a cross-year cycle (such as September to February of the following year), it can be processed logically continuously.

[0106] When calculating the long-term and short-term yield indices in step 3, obtain the current date. ;

[0107] like Then the profit area of ​​this industry Activated and participates in the calculation;

[0108] like Then the profit area of ​​this industry It remains inactive and does not participate in calculations.

[0109] Therefore, even if multiple regions exist in the same industry, only the region whose time window matches the current date will be retrieved during the calculation.

[0110] Each industry revenue region is bound to an industry code field_code, as shown in Table 3. This code is used for weighting and normalization calculations in the prediction output stage, so that different types of revenue can be compared under a unified indicator.

[0111] Table 3 Industry Codes

[0112]

[0113] The following is an example of an industry revenue region, as shown in Table 4.

[0114] Table 4 Industry Revenue Regions

[0115]

[0116] The storage and indexing mechanism adopted in step 2 is as follows:

[0117] (1) All industry revenue regions are stored in structured record format, including:

[0118] Region ID (region_id), industry code (field_code), spatial range (wkt), and time window (time_window);

[0119] (2) Index creation:

[0120] Spatial index: Used to quickly determine whether a block intersects with a revenue region of a certain industry;

[0121] Time index: Used to match the current date within a day-level range.

[0122] When forecasting, the system automatically filters and activates industry revenue regions that match the current date and time window.

[0123] Step 2 defines the core input structure for Step 3. When calculating the long-term and short-term return indices, the system activates only the industry return regions that match the current date and retrieves historical block units through spatial and temporal joint matching. This ensures differentiated accuracy in both space and time and avoids distortion caused by a uniform industry window.

[0124] Step 3: Calculate the long-term return index and the short-term return index.

[0125] The purpose of this step is to target time windows that include the current date. For each industry's revenue region, calculate the long-term and short-term return indices for that industry's revenue region.

[0126] The core of the long-term return index is to calculate the trend of industry return changes over the same historical time period, in order to reflect cyclical patterns and long-term return potential.

[0127] When calculating the long-term return index, the current date is used. Using the industry revenue region established in step 2 as input, activate all regions that meet the requirements. The industry profit areas form a set of activated industry profit areas. Each industry's revenue region includes:

[0128] Spatial range ( );

[0129] Time window ( );

[0130] Industry Code ( ).

[0131] Next, retrieve past data from the standardized historical block database. Within the same time window of each year, meet the requirements The search identifies the block units corresponding to the industry's revenue regions. The search uses "region + time" as the query criteria to retrieve historical block units from the database that meet the spatial and temporal overlap criteria, which are then used for subsequent revenue calculations.

[0132] Based on the revenue values ​​included in the retrieved block units, the historical average revenue and average growth rate of the corresponding industry revenue region are calculated. Then, based on the historical average revenue and average growth rate, the Long-Term Revenue Index (LTRI) for that industry revenue region is calculated. The LTRI quantifies the average revenue level and growth trend of an industry revenue region over multiple years, reflecting the long-term performance of the industry revenue region in seasonal or annual cycles. This step, through statistical analysis and trend extraction of historical data, forms a forward-predictable revenue index, providing input for subsequent multi-scale revenue normalization and task optimization.

[0133] For industry revenue regions The activated first Profitable regions for each industry, historical average returns The calculation formula is as follows:

[0134] ;

[0135] average growth rate The calculation formula is as follows:

[0136] ;

[0137] Long-term return index The calculation formula is as follows:

[0138] ;

[0139] in, For the first The industry's profit region is in the first The total annual revenue is equal to the sum of the revenue values ​​included in all retrieved block units; For the current date The year in which it occurred; The global reference return used for normalization is a value from the past. The first year The maximum historical average return for each industry's revenue region; The quantity for the previous year; This is the balance coefficient, with a typical value of 0.7.

[0140] Long-term return index It comprehensively reflects the historical return level and growth potential of the industry's revenue region and is a core quantitative indicator for long-term forecasting.

[0141] The process of calculating the long-term return index in this step has the following characteristics:

[0142] Rolling annual search with window alignment: Ensuring periodic consistency by shifting the time window annually;

[0143] The yield index is concise and unified: it reflects both the scale and growth trend of yields in a weighted sum format.

[0144] Strong spatial independence: Revenue for each industry is calculated independently within its designated region, eliminating cross-regional interference;

[0145] It has good scalability: the time period (such as 3 years, 5 years, 10 years) can be adjusted according to industry characteristics.

[0146] The following example retrieves data from the past 5 years in a standardized historical block database (i.e., ...). Within the same time window, satisfying Taking the block unit corresponding to the industry's return region as an example, the calculation process of the long-term return index is illustrated.

[0147] When calculating the long-term return index, a rolling search is performed along the annual dimension using the time window of the industry's return region as the benchmark to statistically analyze historical returns over the past 5 years. The search interval for each year is a shifted version of the time window of the industry's return region in the corresponding year.

[0148] First, determine the range of years to search:

[0149] Let the current year be Going back 5 years, traversing:

[0150] ;

[0151] This involves sequentially retrieving revenue data for the same time window across the past five years.

[0152] For each year Generate the corresponding time interval for that year:

[0153] ;

[0154] in, This means mapping the "month and day" of the original time window to the target year while maintaining a consistent time period length.

[0155] For example, if the time window is from June 1 to August 31, then the corresponding time interval for 2023 is [2023-06-01, 2023-08-31].

[0156] Taking the "water conservancy industry - the middle and lower reaches of the Yangtze River" as an example:

[0157] The profit window for this industry is from June 1st to August 31st.

[0158] The current date is July 1, 2025.

[0159] The five annual windows were searched sequentially in the standardized historical block database, as shown in Table 5.

[0160] Table 5 Annual Window

[0161]

[0162] Within each annual window, all historical block units within the industry's revenue region are extracted. The sum of the revenue values ​​included in all block units retrieved under each annual window is calculated to obtain the total revenue of the industry's revenue region for each year. These total returns will be used to calculate historical average returns and average growth rates.

[0163] Specifically, based on the total returns of the industry's revenue region over the past 5 years, the historical average return and average growth rate of the industry's revenue region are calculated, and a long-term return index is generated. The calculation formula is as follows:

[0164] Historical average returns:

[0165] ;

[0166] Average growth rate:

[0167] ;

[0168] Long-term return index:

[0169] ;

[0170] in:

[0171] : Average return over the past 5 years;

[0172] Average growth rate;

[0173] Global reference return, used for normalization calculation. Its value is the maximum value of the historical average return of the industry's return range over the past 5 years, ensuring that the normalization range fluctuates between 0 and 1.

[0174] : Balance coefficient, used to balance the weight ratio of the two parts of the formula; its default value is 0.5. When The larger the value, the more the long-term return index focuses on historical average returns (reflecting scale); when... The smaller the value, the more the long-term return index focuses on the average growth rate (reflecting the trend).

[0175] In step 3, while calculating the long-term return index, the time window also includes the current date. Calculate the short-term return index for each industry's return region.

[0176] The purpose of calculating the short-term yield index is to use the current date... Using July 1st as a benchmark, the returns and change rates of each industry's return region are calculated within the most recent preset number of days (e.g., 30 days), thereby identifying industry return regions with significant short-term increases or rapid changes in returns within the current year. The short-term return index reflects the immediate volatility of returns and serves as a dynamic supplement to the long-term return index.

[0177] When calculating the short-term return index, the block unit corresponding to the industry's return area within the current period is retrieved from the standardized historical block database. The total return and return growth rate of the current period are calculated based on the return values ​​included in the retrieved block unit, thereby obtaining the short-term return index of the industry's return area. The current period refers to a fixed-length window that starts from the current date and goes back a preset number of days. When the start date of the window is earlier than the start time of the corresponding industry's return area, the start time of the corresponding industry's return area is used as the start date of the window.

[0178] First, the time window is trimmed to determine the current period: the short-term calculation period for each activated industry revenue zone is set to a maximum of a preset number of days, such as 30 days, and cannot be earlier than the start time of that industry revenue zone. Therefore, the current period is defined. for:

[0179] ;

[0180] in:

[0181] This indicates a date 30 days prior to the current date;

[0182] It is the start date of the time window for the industry's profit zone;

[0183] Indicates taking and The later of the two ensures that the start date of the current cycle is no earlier than the start date of the time window defined for the industry's revenue zone.

[0184] For example, suppose the current date If the industry's profit zone begins on June 15, the current period is [June 15, July 1], without going back to before June 15.

[0185] The determination of the current cycle mentioned above is only based on a preset number of days of 30 days. Those skilled in the art can choose other reasonable number of days according to actual needs, which will not be elaborated here.

[0186] After determining the current period, traverse all records in the standardized historical block database and retrieve block units that meet the following conditions:

[0187] Spatially, it intersects with the polygonal range (wkt) of the corresponding industry's revenue region;

[0188] In terms of time, it falls within the current cycle. Within the range.

[0189] The revenue value included in the retrieved block units that meet the above conditions is denoted as... The revenue values ​​contained in all the block units retrieved corresponding to the revenue region of this industry. By summing the results, the total revenue of the industry's revenue region in the current period can be obtained. .

[0190] The above process maps the revenue from image tasks from different time and space sources to a unified revenue statistics for the same industry revenue region.

[0191] Next, we calculate the revenue growth rate. Using a preset period of 30 days, and assuming the current period's start date is no earlier than the start date of its corresponding industry revenue region, within the same 30-day window, we divide the revenue values ​​contained in the block unit according to time sequence (days) and perform linear regression to calculate the revenue growth rate of that industry revenue region. The calculation formula is as follows:

[0192] ;

[0193] in:

[0194] This represents the average return of the block unit over the first 7 days of the current period (using 7 days as an example only).

[0195] This represents the average return of the block unit over the last 7 days of the current period (using 7 days as an example only).

[0196] To prevent division by zero of extremely small constants.

[0197] Earnings growth rate This reflects the trend of earnings over the past month:

[0198] like This indicates that recent returns are rising;

[0199] like This indicates that recent returns are declining;

[0200] If a region has only a small amount of data, the default growth rate is 0 (neutral processing).

[0201] This calculation method avoids cross-month bias while maintaining sensitivity to short-term changes.

[0202] Finally, based on total revenue and revenue growth rate Calculate the short-term return index , The specific calculation formula, which combines the level of profitability and the growth rate, is as follows:

[0203] ;

[0204] in:

[0205] The average return of all industry revenue regions within the time window that include the current date in the current period is the average of the total returns of all industry revenue regions within the time window that include the current date in the current period.

[0206] This is the weighting coefficient, with a typical value of 0.6;

[0207] For the first The revenue growth rate of each industry's revenue region over the past 30 days.

[0208] In the aforementioned short-term return index In the calculation formula, the first term contributes more when the total revenue is high; the second term contributes more when the revenue increases rapidly. The combination of the two reflects the regional characteristic of "both making money and growing".

[0209] Step 4: Normalization of return index and allocation of return projection.

[0210] In this step, to achieve comparability across different time scales, the long-term return index and the short-term return index are first normalized to unify the return indicators at different time scales onto the same dimension.

[0211] The formulas for normalizing the long-term return index and the short-term return index in this step are as follows:

[0212] (1) Normalization of long-term return index:

[0213] ;

[0214] in:

[0215] For the first Long-term return index of each industry after regional normalization of returns;

[0216] The maximum value among the long-term return indices of all activated industry return regions;

[0217] It is the minimum value among all the long-term return indices of the activated industry return regions.

[0218] (2) Normalization of short-term return index:

[0219] ;

[0220] in:

[0221] For the first Short-term return index after regional normalization of industry returns;

[0222] The maximum value among the short-term return indices of all activated industry return regions;

[0223] It is the minimum value among the short-term return indices of all activated industry return regions.

[0224] (3) Calculation of comprehensive benefit index:

[0225] For the For each industry's revenue region, its comprehensive return index is calculated based on the normalized long-term return index and short-term return index. The calculation formula is:

[0226] ;

[0227] in:

[0228] For the first Comprehensive revenue indicators for each industry's revenue region;

[0229] This is the weighting factor for long-term returns (typically 0.6–0.8).

[0230] Finally, the comprehensive benefit index Project to the Generate block-level revenue estimates for each industry's revenue region, covering all block cells within that region. Its block-level profit estimate is defined as:

[0231] ;

[0232] in:

[0233] For block unit The estimated block-level revenue, where the block unit number is... Same as the corresponding grid;

[0234] It is a spatial decay function (such as an exponential function or a Gaussian function);

[0235] This is the distance from the center of the block cell's grid to the center of the industry revenue area.

[0236] The above block-level revenue estimates It can ensure that the returns decline smoothly within the region and maintain the total return.

[0237] Step 5: For all block units covered by the industry revenue region within the time window that includes the current date, sort them by their block-level revenue estimates and filter out the top block units with the highest block-level revenue estimates. Each block unit generates a set of prediction points for task planning.

[0238] In this step, all block units covered by the active industry revenue regions (i.e., all industry revenue regions within the time window that include the current date) are aggregated into a candidate set, and then ranked according to block-level revenue estimates. Sort all block units in the candidate set in descending order and select the top ones. Block-level revenue estimate The corresponding block unit generates a set of prediction points, or Result points, for task planning in step 6. The Result points set consists of a group of high-yield prediction points, each representing a target point with the highest predicted yield in both space and time. In step 6, based on these prediction points and combined with the available satellites in the constellation and their imaging parameters during imaging missions, a set of actually executable imaging missions will be generated. Each prediction point contains the following complete attribute information:

[0239] result_id: Predicted point number;

[0240] block_id: The source historical block number;

[0241] field_code: Industry code;

[0242] R_block: Overall return metric;

[0243] time_window: The time window during which the photo can be taken;

[0244] region_id: The region code of the revenue region of the industry;

[0245] Location: The coordinate center in the WGS84 coordinate system, i.e., the spatial location;

[0246] priority_rank: Ranks the block-level revenue estimates.

[0247] Step 6: Local Optimal Scheduling and Shooting Task Generation. Through multiple rounds of combination and adjustment, the set of task solutions with the highest relative benefit is found to achieve local optimal scheduling and generate actual executable shooting tasks.

[0248] The purpose of this step is to generate a set of actionable shooting tasks (denoted as ) based on the Result point set generated in step 5, under satellite visibility conditions and time constraints. ), which is the set of locally optimal tasks.

[0249] In this step, based on the available satellites in the constellation and their imaging parameters (i.e., visual information, referring to the shooting parameters that satisfy satellite resource constraints) when performing imaging tasks, the set of predicted points is filtered for visibility and the executable imaging tasks are optimized to obtain a locally optimal set of tasks. Each task includes the executing satellite number, execution time interval, set of predicted point numbers, and total task benefit. The imaging parameters of the satellites when performing imaging tasks include the orbits, visible range, imaging bandwidth, attitude control parameters, and shooting schedule of each satellite.

[0250] Specifically, the visibility filtering of the predicted point set includes the following steps:

[0251] Iterate through the prediction points in the prediction point set and determine whether each prediction point meets the imaging conditions of a certain satellite:

[0252] The predicted point is located within the satellite's visible band.

[0253] The shooting time fell within the satellite's available shooting window;

[0254] The satellite's imaging angle and attitude control parameters meet the allowed preset range;

[0255] Predicted points that meet the above conditions are marked as executable points.

[0256] This step checks whether each predicted point can be actually photographed by a satellite, which is equivalent to performing a physical feasibility screening.

[0257] After identifying executable points, an executable shooting task is constructed, which includes the following steps:

[0258] Adjacent executable points that can be captured by the same satellite within the same time period will be automatically merged into a single candidate task.

[0259] For each candidate task, its task revenue is calculated. This task revenue is the sum of the estimated block-level revenues of all executable points contained in the candidate task, expressed by the following formula:

[0260] ;

[0261] in, Indicates the first Candidate tasks The rewards for completing the task.

[0262] By merging multiple high-yield executable points in the same orbit and time period into a single candidate mission, satellite imaging can achieve higher yields in a single operation.

[0263] Since multiple candidate tasks may overlap in time, task combination and local optimum search are required. That is, among all feasible candidate tasks, a set of tasks that do not conflict in time and meet the shooting conditions are found and combined to maximize the total task benefit of the combination.

[0264] Specifically, when performing task combination and local optimum search, an iterative local search method is used to obtain the local optimum result. The search process is as follows:

[0265] ① Initialization:

[0266] According to task rewards Sort all candidate tasks from highest to lowest order;

[0267] Select the first few candidate tasks to form the initial solution .

[0268] ② Combination attempts:

[0269] Select a subset of candidate tasks randomly or heuristically;

[0270] Check whether these candidate tasks conflict in time and whether they meet the shooting conditions (i.e., the same satellite cannot perform them simultaneously).

[0271] ③ Comparison of benefits:

[0272] If the total task reward of the new combination is higher, then update the combination to the current solution;

[0273] If the returns of the new portfolio are lower than those of the original portfolio, then the original portfolio remains unchanged.

[0274] ④ Iterative loop:

[0275] Repeat steps ② and ③ until the total task reward change for several consecutive rounds is less than a threshold (e.g., 1%), then stop the search.

[0276] The final result is the locally optimal combination of tasks, which maximizes the overall task benefit. This combination is then designated as the locally optimal task set. This can be expressed by the formula:

[0277] .

[0278] The locally optimal task set obtained by the above iterative local search method satisfies all shooting conditions and has the highest relative benefit. Figure 4 The diagram illustrates the effect of local optimal scheduling and image acquisition task generation in step 6. After the optimization calculation in step 6, the dispersed multiple grids are optimized and combined into three image acquisition tasks (Task1, Task2, Task3) that do not conflict with time and resources. Each task specifies the satellite to be executed (Task1 is Gaofen 03D11, Task2 is Gaofen 03D03, and Task3 is Gaofen 03D27) and the precise imaging time (Task1 is 2025-07-28, Task2 is 2025-08-01, and Task3 is 2025-07-29), demonstrating the ability of this invention to output an optimized remote sensing image acquisition area scheme.

[0279] In the locally optimal task set, each task contains:

[0280] satellite_id: The satellite number to be executed;

[0281] start_time / end_time: The execution time window;

[0282] points: The set of predicted point numbers contained within;

[0283] R_task: Total task reward.

[0284] In addition, each task may also include:

[0285] task_id: Shooting task number;

[0286] Priority: Execution priority (in descending order of task benefits).

[0287] It should be noted that the above explanation of the task combination optimization process using iterative local search illustrates the same principle. It is understandable that other classic local optimization algorithms, such as hill climbing or greedy local search, can achieve similar results. Their commonality lies in repeatedly adjusting the combination within the feasible range until the reward no longer increases, causing the total task reward of the task combination to converge to a local optimum.

[0288] Step 6, through the above specific implementation methods, achieves the following technical effects:

[0289] Prioritize task scheduling based on local optimum search: automatically select the task combination with the highest yield;

[0290] Mission feasibility assurance: All missions meet the satellite imaging requirements and there are no conflicts;

[0291] Fast solution: Local optimization is performed only within the feasible task scope to avoid high computational overhead caused by global traversal;

[0292] Prediction and execution closed loop: realize the automatic transformation from the revenue prediction point to the actual task scheduling.

[0293] Thus, through steps 1-6 above, the present invention has completed the final transformation from "regional value prediction" to "executable task generation". The resulting "locally optimal task set" is the final solution output by the present invention after optimizing the remote sensing image acquisition area, which can directly guide satellite imaging.

[0294] The remote sensing image acquisition area optimization method proposed in this invention is particularly suitable for commercially operational multi-satellite systems, such as large-scale, low-cost satellite constellations (30-50 satellites, orbiting at an altitude of approximately 500 km). This method, through data-driven and intelligent prediction techniques, significantly improves the utilization efficiency of satellite resources, the commercial value density of task allocation, and the initiative and foresight of the operation and scheduling system. Compared with existing technologies, this invention has the following beneficial effects:

[0295] (1) By constructing a standardized historical block database and industry revenue regions, and designing long-term and short-term revenue indices, this invention can proactively identify and generate high-value imaging tasks during periods without customer orders, significantly reducing the idleness of satellite resources and improving the utilization efficiency of satellite resources.

[0296] (2) By defining industry revenue regions with independent spatiotemporal attributes, this invention integrates the periodic prior knowledge of different industries (such as flood season, agricultural conditions, and fire prevention period) into the optimization method in a structured manner, so that the method can make differentiated and accurate predictions of revenue potential in different regions and at different times, overcoming the drawback of traditional scheduling systems that are difficult to model and predict the trends of different industries.

[0297] (3) This invention establishes a scheduling orientation centered on benefits. Through a comprehensive benefit scoring mechanism, long-term and short-term benefit indices are normalized and integrated into a unified comprehensive benefit index. By combining the available satellites in the constellation and their imaging parameters, visibility screening and executable shooting tasks are optimized. Finally, a set of executable locally optimal tasks with priority on benefits is generated, realizing an intelligent scheduling strategy that shifts from "technology-oriented" to "benefit-oriented".

[0298] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0299] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. A remote sensing image shooting area optimization method based on historical income analysis, characterized in that, The method comprises the following steps: Step 1: Obtain historical shooting task data, and sequentially perform local grid division and income distribution on the historical shooting task data to generate a plurality of block units corresponding to each historical shooting task, each block unit comprising a unique task number, coordinates, a date label, an industry label, local metadata and an income value, and the block units corresponding to all the historical shooting tasks forming a standardized historical block database; Step 2: Establish an industry income region with an independent spatial range and a time window, each industry income region comprising a region number, a spatial range, a time window and an industry code; Step 3: For each industry income region containing the current date in the time window, calculate the long-term income index and the short-term income index of the industry income region, respectively; When calculating the long-term income index, search for the block units corresponding to the industry income region in the same time window in the past several years in the standardized historical block database, and calculate the historical average income and the average growth rate of the industry income region according to the income values included in the searched block units, and then obtain the long-term income index of the industry income region; When calculating the short-term income index, search for the block units corresponding to the industry income region in the current period in the standardized historical block database, and calculate the total income and the income growth rate of the current period according to the income values included in the searched block units, and then obtain the short-term income index of the industry income region, wherein the current period refers to a fixed-length window with the current date as the end date and tracing back a predetermined number of days, and when the start date of the window is earlier than the start time of the corresponding industry income region, the start time of the corresponding industry income region is taken as the start date of the window; Step 4: Normalize the long-term income index and the short-term income index, respectively, calculate the comprehensive income index of each industry income region according to the normalized long-term income index and the short-term income index, and project the comprehensive income index onto the block units covered by the corresponding industry income region to generate block-level income estimation values; Step 5: Sort all block units covered by all industry return area containing current date in time window by their block level return estimate value, select the top block units with the highest block level return estimate value, generate the prediction point set for task planning; Step 6: According to the available satellites in the constellation and the imaging parameters when they perform imaging tasks, perform visibility filtering and executable shooting task optimization on the set of prediction points to obtain a locally optimal task set, wherein each task comprises a satellite number, an execution time interval, a set of prediction point numbers and a total task income. 2.The remote-sensing image shooting area optimization method based on historical revenue analysis according to claim 1, wherein, The process of local grid division according to the historical shooting task data comprises the following steps: Step 1.1.1: Generating a grid within the minimum bounding rectangle of the included imagery coverage area at a pre-set fixed spacing in historical shooting tasks the included imagery coverage area at a pre-set fixed spacing in historical shooting tasks Step 1.1.2: For any one of the grid cells, calculate its relationship with the image coverage area. Intersection area ; Step 1.1.3: judging intersection area whether it is greater than an area threshold, if yes, the grid is added to the grid set as a valid block unit. Step 1.1.4: After traversing all the grids generated in step 1.1.1, the historical shooting tasks are obtained corresponding grid set. 3.The remote-sensing image shooting area optimization method based on historical revenue analysis according to claim 2, characterized in that, The process of income distribution according to the historical shooting task data comprises the following steps: Step 1.2.1: For historical shooting tasks The distance between the center of the corresponding grid set and the center point The distance between the center of the corresponding grid set and the center point , and the distance attenuation function is obtained, where is a preset industry attenuation coefficient;​ Step 1.2.2: Calculate weight base ; Step 1.2.3: Calculate normalization factor ; Step 1.2.4: Calculate the first... Benefits per grid ,and ,in Historical filming mission Total revenue; Step 1.2.5: After traversing all grids in the grid set, complete the historical shooting task of the revenue distribution. 4.The method of claim 3, wherein, The values of the industry attenuation coefficients are as follows: 0.15 for the water conservancy industry, 0.1 for the agricultural industry, 0.25 for the forestry industry, 0.05 for the urban management industry, and 0.4 for environmental protection and emergencies.

5. The remote-sensing image shooting area optimization method based on historical revenue analysis according to any one of claims 1 to 4, characterized in that, In step 3, the historical average return is calculated as follows: ； Average growth rate The formula for calculating the average growth rate is as follows: ； Long-term return index The formula for calculating the long-term return index is as follows: ； wherein, is the total revenue of the industry revenue area in the year, is the global reference revenue, is the balancing factor, is the number of past years.

6. The remote-sensing image shooting area optimization method based on historical revenue analysis according to any one of claims 1 to 4, characterized in that, In step 3, the total earnings of the current period is equal to the sum of the earnings values included in the retrieved block element; The revenue growth rate The formula for calculating the revenue growth rate is as follows: ； Short-term earnings index The formula for the calculation is as follows: ； where, is the average return of the block units in the previous 7 days in the current period, is the average return of the block units in the last 7 days in the current period, is a small constant to prevent division by zero, is the average return of all industry return areas in the current period, is the weight coefficient.

7. The remote-sensing image shooting area optimization method based on historical revenue analysis according to any one of claims 1 to 4, characterized in that, In step 4, the comprehensive revenue index of the industry revenue region The calculation formula is as follows: ； wherein, is the normalized long-term return index, is the normalized short-term return index, is the long-term return weight coefficient; block level revenue estimate The formula for calculating the block level revenue estimate is as follows: ； wherein, is a spatial decay function, is the distance from the grid center of the block unit to the center of the industry revenue area.

8. The remote-sensing image shooting area optimization method based on historical revenue analysis according to any one of claims 1 to 4, characterized in that, In step 6, when performing the visibility screening on the set of prediction points, the set of prediction points is traversed to screen executable points satisfying the following satellite shooting conditions: the spatial position of the prediction point is located in the visibility band of the satellite; the shooting time falls within the shootable time window of the satellite; and the imaging angle and attitude control parameter of the satellite satisfy a preset range. 9.The remote-sensing image shooting area optimization method based on historical revenue analysis of claim 8, wherein, In step 6, after the executable points are screened, adjacent executable points shot by the same satellite in the same time period are automatically merged into a candidate task, and the task benefits of each candidate task are calculated. An iterative local search method is used to find a combination of tasks from all candidate tasks, which is time conflict-free and satisfies the shooting conditions, so that the total task benefit of the combination is maximum, and the combination is taken as the local optimal task set.

10. The remote-sensing image shooting area optimization method based on historical revenue analysis according to any one of claims 1 to 4, characterized in that, The spatial range of the industry benefit area is an arbitrary polygon, and the spatial coordinates of the polygon boundary are defined and stored in the WGS84 coordinate system; and the preset number of days is 30 days.

Images