Method for evaluating and pricing remote sensing data asset value by using AI

Abstract

The invention relates to the field of commercial and financial data processing, and discloses a method for evaluating and pricing remote sensing data asset value by using AI, which comprises the following steps of: establishing a set of computer data processing regulations which represent information novelty task scene matching degree and three utility components of data structure definition through combination; according to the invention, black box type price prediction is abandoned, and data asset value judgment is converted into objective examination of a traceable and understood utility profile, so that pricing decisions in the fields of finance and asset management are made to be more accurate. And a transparent and reliable technical basis with internal logic self-checking capability is provided.

Description

Technical Field

[0001] This invention relates to a method for assessing and pricing remote sensing data assets using AI, belonging to the field of commercial and financial data processing technology. Background Technology

[0002] A current mainstream technical approach is to build complex prediction models based on deep learning. This approach attempts to directly regress a price or value level by inputting multidimensional technical features of remote sensing data and historical transaction data, which to some extent achieves the automated preliminary quantification of the value of data assets.

[0003] However, when this approach is applied to professional fields such as finance and insurance or key asset management, which have stringent requirements for the timeliness and reliability of decision-making results, its inherent limitations begin to emerge. In these scenarios, the value of data assets is not an isolated, objective scalar that can be accurately predicted, but rather a reflection of the utility of data in providing key decision-making information under specific business tasks. Existing methods, by focusing excessively on fitting a dynamic and subjective price variable endogenously to the market, generally fall into a systematic neglect of the utility of data tasks themselves. This neglect often leads to situations where, in actual business, data purchased at high prices becomes unusable due to a mismatch between information dimensions and task requirements, resulting in decision-making errors and economic losses.

[0004] Faced with the dilemma of a disconnect between evaluation results and actual utility, a direct approach to improvement is to further increase the complexity of the model or introduce more market-related factors, hoping to improve the accuracy of predictions. However, this approach does not address the root of the problem and instead exacerbates the black-box nature of the evaluation process and computational costs. This makes it even more difficult for professional users in fields such as finance to base their core business decisions on such an algorithmic model whose logic is untraceable and whose results are unexplainable. Specifically, existing technologies have the following shortcomings: 1. They simplify the core issue of data's task utility to a problem of predicting market prices, leading to a disconnect between evaluation objectives and commercial substance; 2. The evaluation process heavily relies on complex black-box models whose internal logic is unexplainable, making it difficult for professional decision-makers to trust the evaluation results; 3. There is a lack of a standardized data utility metric that can span different data sources and different task scenarios, resulting in a lack of objective and reliable basis for data procurement and risk management. Therefore, how to break away from the mindset of predicting a single price and establish an evaluation and pricing method that can transparently, standardizedly, and at low cost reveal the core utility of remote sensing data for specific commercial tasks becomes the technical problem that this invention aims to solve. Summary of the Invention

[0005] This invention provides a method for assessing and pricing remote sensing data assets using AI. Its main purpose is to solve the problem that existing technologies mistakenly simplify data utility assessment into price prediction, resulting in a black box assessment process, unreliable results, and a lack of standardized utility metrics.

[0006] To achieve the above objectives, this invention provides a method for assessing and pricing remote sensing data assets using AI. This method establishes a set of computer data processing procedures, which include the following steps: Step a, generating a differential utility fingerprint characterizing the decision-making utility of remote sensing data under a specific task, specifically includes: after acquiring the remote sensing data to be evaluated and a known benchmark remote sensing data, performing a benchmark data credibility gating review: applying signal processing methods to the known benchmark remote sensing data to obtain its own structural clarity component, and determining whether the structural clarity component meets the credibility threshold; if the judgment result is not satisfied, another known benchmark remote sensing data is selected and this review is repeated until a valid benchmark data that meets the credibility threshold is found; after the valid benchmark data is determined, by calculating the difference in information distribution between the data to be evaluated and the valid benchmark data, a first utility component characterizing the information novelty of the data to be evaluated is obtained; based on a task template that defines specific task requirements, the metadata of the data to be evaluated is logically matched with the task template to obtain a second utility component characterizing the scenario matching degree of the data to be evaluated; applying signal processing methods to the data to be evaluated to calculate the ratio of its internal structured information to unstructured information, to obtain a third utility component characterizing the structural clarity of the data to be evaluated itself. Step b: Combine the first utility component, the second utility component, and the third utility component to form a differential utility fingerprint of the remote sensing data to be evaluated. Step c involves outputting the differential utility fingerprint as a standardized technical data structure to support the valuation and pricing decisions of remote sensing data assets.

[0007] Preferably, the acquisition methods for the first utility component, the second utility component, and the third utility component in step a are further defined as follows: The first utility component is acquired by calculating the gray-level histograms of the data to be evaluated and the effective reference data for their respective predetermined bands, and using KL divergence to measure the difference between the two gray-level histograms; the second utility component is acquired by performing a step-by-step logical comparison and weighted summation of the band information, spatial resolution, and imaging time information contained in the metadata of the data to be evaluated with the key band requirements, optimal resolution range requirements, and phase requirements defined for a specific task in the task template; the third utility component is acquired by applying the Laplacian operator to filter the data to be evaluated and calculating the variance ratio between the filtered image and the original image.

[0008] Preferably, in step a, the structural sharpness component The calculation method is precisely defined as follows: ,in, To determine the variance of pixel value distribution in an image after applying the Laplacian operator to remote sensing data. This represents the variance of the pixel value distribution in the original remote sensing image.

[0009] Preferably, the confidence threshold is a value determined based on statistical analysis of historical remote sensing data to identify low-quality data with large-area cloud and fog obscuration or severe sensor noise.

[0010] Preferably, in step b, the differential utility fingerprint further includes a fourth utility component. The method for generating the fourth utility component includes: obtaining the spatial resolution of the remote sensing data to be evaluated and the effective reference remote sensing data; calculating the ratio of their spatial resolutions and performing a nonlinear mapping of the ratio through an arctangent function to generate a fourth utility component that characterizes the degree of scale inconsistency between the two, serving as a confidence index for the first utility component.

[0011] Preferably, when performing step a, the method also includes the following steps in parallel: recording the mismatch information between the metadata of the data to be evaluated and each requirement in the task template; and combining all non-zero mismatch information to generate a structured diagnostic mismatch fingerprint that characterizes the capability shortcomings of the data to be evaluated under specific task requirements.

[0012] Preferably, the method further includes: based on the content of the diagnostic mismatch fingerprint, invoking a data search instruction to search in the data catalog for other remote sensing data products whose metadata matches one or more mismatch information recorded in the diagnostic mismatch fingerprint, and generating a data combination suggestion containing information of other remote sensing data products.

[0013] Preferably, the method further includes a dynamic optimization mechanism for task templates, which includes the following steps: recording differential utility fingerprints of transacted remote sensing data associated with a specific task; periodically performing statistical analysis on a set of recorded differential utility fingerprints associated with the same specific task, determining a consensus utility pattern reflecting market preferences by calculating the statistical median of each utility component dimension; and comparing the consensus utility pattern with the task template of the specific task, and generating a set of optimized data for optimizing the task template based on the comparison results.

[0014] Preferably, the task template is stored in JSON format, which allows the task template to be updated or supplemented through text editing.

[0015] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention, through parallel processing, simultaneously generates a first utility component representing information increment, a second utility component representing task fit, and a third utility component representing signal stability for a single remote sensing data set. The information increment is obtained based on the difference in information distribution between the data to be evaluated and a known benchmark. The task fit is determined by a logical comparison between the data metadata and a preset task template, while the signal stability is quantified by the internal ratio of structured to unstructured information within the data itself. This method unifies three independent analytical results from different dimensions into a standardized differential utility fingerprint. This transforms the judgment of data asset value from relying on isolated interpretations of single technical indicators or predictions of market prices into an examination of a traceable and understandable utility profile composed of information novelty, scene matching, and structural clarity.

[0016] 2. Before assessing the novelty of information, this invention first invokes a signal processing step for evaluating the structural clarity of the data itself. This step processes the known data used as the benchmark and determines the validity of the benchmark data based on its own structural clarity component. This method of reusing the internal functional modules of the assessment system in advance to examine the reliability of the assessment premise itself establishes an inherent logical self-checking loop for the entire assessment method. It avoids the situation where the assessment results are distorted due to the introduction of contaminated benchmark data, and enables the assessment system to resist input source contamination and maintain the reliability of its conclusions when facing uncertain data quality in the real world.

[0017] 3. The method of the present invention also includes recording differential utility fingerprints of transacted remote sensing data associated with a specific task, and periodically performing statistical analysis on these fingerprint data derived from real market behavior to determine a consensus utility pattern that reflects market preferences; it compares this consensus utility pattern with a preset task template and provides the differences as optimization suggestions to the managers. This mechanism constructs a feedback path from discrete market transaction results to structured technical rules, enabling the evaluation system, which was originally based on static expert knowledge, to absorb and reflect dynamically changing market cognition, thereby possessing the ability to self-improve and continuously evolve.

[0018] 4. When performing scenario matching evaluation, this invention also records the mismatch information between the metadata of the data to be evaluated and the various requirements in the task template in parallel, and combines this information into an independent diagnostic mismatch fingerprint that represents the shortcomings in capability. This transforms the intermediate information discarded during the matching process into a structured new information product that can be used for subsequent business logic. It not only reveals why there is a mismatch, but also points out the direction for capability improvement, so that the evaluation method is no longer limited to obtaining a single matching score, but extends to a function that can provide data combination suggestions. Attached Figure Description

[0019] Figure 1 This is a schematic diagram illustrating the process of generating diagnostic mismatch fingerprints and data combination suggestions in this invention. Figure 2 This is a diagram showing the overall data flow and mechanism of this invention. Figure 3 This is a use case diagram of the remote sensing data asset valuation system of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below. However, it should be understood that the following embodiments are intended for explanation and illustration, and not for limiting the scope of protection of the present invention.

[0021] This invention provides a method for evaluating and pricing remote sensing data assets using AI. It establishes a computer data processing procedure designed to eliminate unreliable price predictions and possesses inherent logical self-checking and utility transparency. The core of this procedure lies in generating a standardized differential utility fingerprint. This fingerprint is formed by combining three utility components representing information novelty, task scenario matching, and the clarity of the data's structure. In specific application scenarios, such as when financial institutions need to monitor asset changes in a region, existing technologies often suffer from opaque evaluation processes or over-reliance on single indicators, leading to a disconnect between purchased data and actual task requirements, thus affecting the reliability of decisions. To address this utility evaluation dilemma in similar business management and supervision scenarios, this invention provides a solution... The method is configured to perform the following steps; the method first performs step a, which generates a differential utility fingerprint characterizing the decision-making utility of remote sensing data under a specific task. Specifically, when processing a set of remote sensing data to be evaluated, in order to objectively measure the information increment it contains, the system needs to introduce a known benchmark remote sensing data as a reference. Considering that the quality of known data may be uncertain in the real world due to cloud cover or sensor noise, in order to avoid distortion of the evaluation results due to the introduction of a contaminated benchmark, the system performs a benchmark data credibility gating review: before calculating the information novelty, the procedure first calls the signal processing method used to evaluate the data's own sharpness to process the known benchmark remote sensing data to obtain its own structural sharpness component. The calculation is limited to ,in, To determine the variance of pixel value distribution in an image after applying the Laplacian operator to remote sensing data. The system calculates the variance of the pixel value distribution of the original remote sensing image. Then, it compares the calculated structural sharpness component of the reference data with a confidence threshold. This confidence threshold is a value predetermined based on statistical analysis of historical remote sensing data to identify low-quality data. If the component value fails to meet this threshold, it indicates that the reference data itself has quality problems. The system will select another known reference remote sensing data and repeat this review until a valid reference data that meets the confidence threshold is found.

[0022] After determining a valid benchmark, the system calculates the difference in information distribution between the data to be evaluated and the valid benchmark to obtain a first utility component characterizing the novelty of the data to be evaluated. To transform the concept of information distribution difference into a computable engineering implementation, this step is limited to: calculating the gray-level histograms of the data to be evaluated and the valid benchmark for their respective predetermined bands, and using KL divergence to measure the difference between the two gray-level histograms. KL divergence, as an information theory metric, can effectively measure the amount of information lost when approximating one probability distribution to another. Its calculation result numerically reflects the degree of deviation of the information distribution of the data to be evaluated from the known information benchmark, thus characterizing its inherent information novelty. Furthermore, to determine the applicability of the data to be evaluated in a specific business task, the system logically matches the metadata of the data to be evaluated with a task template that defines specific task requirements, thereby obtaining a second utility component characterizing the scenario matching degree of the data to be evaluated. To ensure the transparency and scalability of this matching process, the task... The template is stored in JSON format, allowing for updates and additions via text editing. It defines the key bands required for a specific task, the optimal resolution range, and, if necessary, equivalence information. The matching process involves logically comparing and weighting the band information, spatial resolution, and imaging time information contained in the metadata of the data to be evaluated against these requirements in the task template. This design transforms the judgment of task fit into a series of traceable logical operations and arithmetic summations. Simultaneously, to quantify the signal quality of the data itself, the system applies signal processing methods to the data to be evaluated. By calculating the ratio of its internal structured information to unstructured information, a third utility component characterizing the structural clarity of the data is obtained. Specifically, this process involves applying the Laplacian operator to the data to be evaluated and calculating the variance ratio between the filtered image and the original image. The technical logic is that the Laplacian operator, as a second-order differential operator, is sensitive to high-frequency noise signals in images but weakly responds to smooth structural information. Therefore, by calculating... This ratio quantifies the proportion of structural parts in an image that carry effective information, providing an objective technical indicator for the stability and operability of the data itself.

[0023] After independently calculating the three utility components, the method executes step b, combining the first, second, and third utility components to form a differential utility fingerprint of the remote sensing data to be evaluated. To address the reliability issue of the information novelty component caused by potential spatial resolution differences between the data to be evaluated and known benchmark data, the differential utility fingerprint can be configured to include a fourth utility component. This fourth utility component is generated by: obtaining the spatial resolution of the remote sensing data to be evaluated and the valid benchmark remote sensing data; calculating the ratio of their spatial resolutions; and performing a nonlinear mapping of this ratio using an arctangent function to generate a fourth utility component characterizing the degree of scale inconsistency between the two. This fourth utility component serves as... A confidence index for the first utility component; correspondingly, in order to transform the intermediate information generated during the evaluation process into a commercially valuable data product, when performing step a, the method also records the mismatch information between the metadata of the data to be evaluated and each requirement in the task template in parallel, and combines all non-zero mismatch information to generate a structured diagnostic mismatch fingerprint that characterizes the capability shortcomings of the data to be evaluated under specific task requirements. This diagnostic mismatch fingerprint reveals the specific reasons for the data mismatch and indicates the direction for capability supplementation. It can be further used to invoke data search instructions to search the data catalog for other remote sensing data products that can supplement capability shortcomings, and generate a data combination suggestion containing information on other remote sensing data products.

[0024] Finally, in step c, the differential utility fingerprint is output as a standardized technical data structure to support the valuation and pricing decisions of remote sensing data assets. To enhance its usability in commercial scenarios, this step may also include visualizing the differential utility fingerprint on a data trading platform, for example, presenting the first, second, and third utility components as a radar chart as an objective technical reference. Furthermore, to enable the entire evaluation system to have self-evolution capabilities, the method may also include a dynamic optimization mechanism for task templates. This mechanism records the differential utility fingerprints of transacted remote sensing data associated with specific tasks and periodically performs statistical analysis on a set of fingerprints under the same task. By calculating the statistical median of each utility component dimension, a consensus utility pattern reflecting market preferences is determined. The system compares this consensus utility pattern with a preset task template and generates a set of optimized data for optimizing the task template based on the comparison results. In this way, a feedback path from market transaction results to technical rules is constructed, enabling the evaluation system to absorb and reflect dynamically changing market perceptions.

[0025] Example 1: In a business decision-making scenario, an agricultural supply chain finance company, facing a sudden regional drought, needs to conduct a risk assessment of the future output of multiple plantations in the region that have already granted credit, in order to adjust its financial derivatives positions. At this time, there are several remote sensing data options available in the data market: one is an expensive optical image with 0.5-meter spatial resolution; another is a moderately priced multispectral image with 10-meter spatial resolution, but including the red-edge band; and a publicly known benchmark remote sensing data, also with 10-meter resolution, that is freely available. The typical data procurement decision... A common pitfall is focusing solely on spatial resolution, which may lead to the underestimation of risk due to the inability of expensive optical images to reflect the crop's intrinsic physiological stress state before wilting. When the method of this invention is applied to this scenario, the system does not directly predict the price or grade of the three datasets. Instead, it generates a differential utility fingerprint for each dataset to be evaluated. When evaluating a moderately priced multispectral image, the system acquires this data along with a known benchmark remote sensing data set. It first performs a benchmark data credibility gate check on the benchmark data set, i.e., it calls the structural sharpness component calculation method to calculate the benchmark data set's own... The system proceeds to the next step only after confirming that the value meets the preset confidence threshold, indicating that the data quality can be used for benchmark comparison. This collaborative mechanism uses the calculation method of the third utility component, which is used to evaluate the stability of the data signal to be evaluated, to pre-examine the reliability of the evaluation premise itself, thereby ensuring that the acquisition of the first utility component, information novelty, is based on a reliable benchmark.

[0026] When calculating the first utility component, the system compares the gray-level histograms of the multispectral image and the effective reference data in the red-edge bands and uses KL divergence for measurement. Because changes in crop chlorophyll content under drought stress are reflected in the red-edge band, causing a deviation in information distribution from known reference data under normal conditions, the system calculates a higher information novelty component. When evaluating scene matching, the red-edge band is listed as a necessary band requirement in the JSON-formatted task template for this drought stress monitoring task, and the optimal resolution range is set between 5 and 15 meters. Therefore, the metadata of this multispectral image is logically integrated with the task template. After comparison and weighted summation, a higher second utility component was obtained. Conversely, although the optical image with a resolution of 0.5 meters far exceeds the requirement in terms of resolution, its second utility component is lower because it does not contain the red edge band. This method resolves the contradiction between the superiority of technical indicators and the actual utility of the task that often occurs in data procurement. During the evaluation process, the system also records the mismatch information between the metadata of the data to be evaluated and the task template in parallel. For example, if the task template contains a requirement for necessity within 72 hours after a disaster, and the imaging time of the multispectral image exceeds this range, the system will record this mismatch information and combine it to generate an independent diagnostic mismatch fingerprint. This approach transforms the intermediate information discarded during the matching process into a structured information product that can be used for subsequent business logic. It not only reveals the data's timeliness shortcomings to decision-makers but also drives the generation of data combination suggestions, such as invoking data search commands to find other data products that can compensate for this timeliness shortcoming. Ultimately, the financial company's data analysis department no longer faces a set of isolated and difficult-to-compare prices or technical parameters, but receives a differential utility fingerprint of the multispectral image, composed of three utility components: high information novelty, high scene matching, and high structural clarity, as well as a diagnostic mismatch fingerprint indicating its timeliness shortcomings. The basis for decision-making shifts from predicting a single price to examining a traceable and understandable utility profile composed of information novelty, scene matching, and structural clarity. This gives data asset risk management and procurement decisions logical self-checking capabilities and transparent technical evidence.

[0027] Example 2: To objectively verify the effectiveness of the method of the present invention in distinguishing the utility of different remote sensing data for a specific task compared with conventional methods, this experiment was designed. The purpose of the experiment is to quantitatively compare the differential utility fingerprint evaluation system used in the present invention with a conventional evaluation method that relies on a single technical indicator, and to determine the degree of conformity between the evaluation results and the preset true utility levels when faced with a complex dataset containing high-quality, low-quality, and pseudo-high-value data. To this end, a computer-generated remote sensing data simulation test platform was constructed. This platform includes a test dataset containing 10 remote sensing data with different characteristics, and a software environment for performing data utility evaluation. In the test dataset used in the experiment, different true utility levels were preset for different data, which represent their application in the specific management task of monitoring the addition of illegal buildings in a monitored ecological protection zone. The application value of this method is evident. The dataset contains various types of interference data, such as high-resolution pseudo-color data with high spatial resolution but whose imaging timeframe does not meet the task template requirements, and high-resolution, low-quality data with large areas of cloud cover. The experiment sets up a control group and an experimental group. The control group uses an evaluation method with spatial resolution as the primary weighting indicator, while the experimental group uses the complete evaluation method based on differential utility fingerprinting, as described in this invention. For the baseline data reliability gating review stage in the experimental group method, the reliability threshold is set to 0.25. This parameter aims to balance the rigor and usability of the review; values ​​below this typically indicate noise or occlusion in the data that may affect the evaluation. This value is based on statistical analysis of historical public datasets and is an engineering example determined for image data that can effectively filter out cloud cover exceeding 30%.

[0028] After the experiment began, the control group and the experimental group each processed all 10 remote sensing data points in the test dataset using their respective evaluation methods, and each generated a utility ranking. The results showed that the experimental group's ranking was highly correlated with the actual utility level, while the control group's ranking showed several discrepancies with the actual utility level, resulting in a utility level of 5. The experimental group's method, based on its scene matching component, assigned a ranking of 5 to D03, consistent with the actual situation. For data point D07 (high score, low quality - cloud), its actual utility level was 8, while the control group assigned a ranking of 6. The experimental group's method, based on its structural clarity component... The calculations showed that the data quality did not meet the requirements, and a ranking of 8, consistent with the actual utility level, was given. Furthermore, the experimental group correctly identified data number D02 (median multispectral - high quality) as having the highest utility ranking, while the control group ranked it third. This difference in ranking results is due to the differential utility fingerprint mechanism of this invention, which, through the collaborative evaluation of three independent dimensions—information novelty, scene matching, and structural clarity—and supplemented by benchmark data credibility gating review, avoids the one-sidedness of a single technical indicator and can more comprehensively reflect the decision-making utility of data under specific tasks. The experimental results confirm that, compared with conventional methods relying on a single technical indicator, the method of this invention can identify and distinguish the intrinsic quality and task applicability of different remote sensing data. Its evaluation results are highly consistent with the application value of the data, providing a data processing procedure for the valuation of remote sensing data assets in fields such as administrative management and supervision.

[0029] Example 3: This example combines Figures 1 to 3 This paper describes the implementation of a method for assessing and pricing remote sensing data assets using AI, such as... Figure 1 As shown in the diagram, the process begins with the user submitting data to the evaluation system. The system then initiates a scene matching assessment, and the scene matcher reads the task template requirements and compares the band information, spatial resolution, and imaging time item by item. During this process, the mismatch recorder simultaneously records various mismatch information and combines this information after the comparison is completed to generate a diagnostic mismatch fingerprint. When the system determines that there is a capability deficiency, it triggers an alternative process, which sends a data search command to the data search engine. The search engine then searches for supplementary data in the data product catalog that can fill the gap and generates data combination suggestions based on the returned matching product list. Finally, the evaluation system outputs the diagnostic mismatch fingerprint and combination suggestions to the user.

[0030] like Figure 2 As shown in the figure, the data to be evaluated, the benchmark data, and the task template in JSON format are used as core inputs to enter the differential utility fingerprint evaluation core module. This module comprehensively evaluates the novelty of information, the matching degree of the scenario, and the clarity of the structure. On the one hand, it generates a differential utility fingerprint as a standardized objective utility profile to support value assessment and pricing decisions. On the other hand, it drives the generation of diagnostic mismatch fingerprints, that is, it records the mismatch information between metadata and task templates to reveal capability shortcomings, thereby forming structured capability shortcoming information, and can generate data combination suggestions based on this. At the same time, the figure also reveals a dynamic optimization mechanism for task templates. By analyzing the differential utility fingerprints of transacted remote sensing data, it calculates the utility pattern that reflects market consensus, and then proposes optimization suggestions for task templates, realizing the self-evolution of the system.

[0031] like Figure 3As shown in the figure, this diagram illustrates the use case relationships of the remote sensing data asset valuation system of this invention from the perspective of system participants. Data analysts or decision-makers, as core users, interact with the system to perform two main functions: valuing data assets and obtaining diagnostic and enhancement suggestions. System administrators are responsible for managing and optimizing task templates to ensure the accuracy and timeliness of the valuation system. In addition, the figure also shows the data providers that serve as the source of the system's valuation objects.

[0032] Example 4: When applying the method of this invention to a new business task, namely, to conduct pre-emptive financial risk monitoring of suspected illegal logging activities in a specific tropical rainforest area, a calibrated task template and evaluation procedure need to be constructed for this task. This example describes the systematic calibration process for completing this task configuration. Before the calibration process is initiated, a reference dataset for the tropical rainforest illegal logging task is first obtained. This dataset contains 20 to 30 positive sample images, which have been labeled and confirmed by analysts. These images clearly show recent logging patches. The dataset also contains a considerable number of negative samples. The imagery contains non-target changes such as seasonal leaf fall or crop rotation. The first step in the process is to calibrate the internal weights of the scene matching component in the task template. The system iteratively adjusts the relative weights of three items: key band requirements, optimal resolution range requirements, and phase requirements when necessary. Using these weights, the system calculates the second utility component for all samples in the reference dataset. The objective function of this iterative process is to find a set of weights that maximizes the mean score of all positive samples and minimizes the mean score of all negative samples. This weight combination is then embedded into a task template in JSON format specific to this task.

[0033] The second step in the process is to calibrate the credibility threshold used in the benchmark data credibility gating review. Given the differences in atmospheric and noise characteristics across different geographical regions and from commonly used sensors, a matching threshold needs to be set. The calibration procedure involves randomly selecting 100 images from historical public data archives associated with the target area, and having analysts or auxiliary tools categorize them into three types: clear, lightly cloudy or hazy, and heavily cloudy or hazy. Subsequently, the system calculates the structural sharpness component for each of these 100 images. And analyze the three types of images The score distribution and the value of the confidence threshold were used to determine the image as clear. The 10th percentile of the score distribution is set to ensure the threshold has discriminative power while avoiding the exclusion of most usable data due to overly stringent standards. The third step in the process is to refine the calculation method of the fourth utility component, which characterizes the degree of scale inconsistency between the data to be evaluated and the benchmark data. This is done by nonlinearly mapping the ratio of their spatial resolutions through an arctangent function to generate a scale inconsistency factor with a value range in the [0,1] interval. Its calculation formula has been determined as follows: ,in The ratio of the larger to the smaller value of the spatial resolution of the data to be evaluated and the valid reference data, i.e. This function will hour The value is mapped to 0, and as... The value increases, The value of approaches 1 non-linearly. By executing the above calibration process, the system finally generates a complete evaluation procedure for the task of illegal logging in tropical rainforests, which includes precise internal weights, localized confidence thresholds, and deterministic calculation formulas. This process transforms the setting of all key parameters in the evaluation method from dependence on experience into a set of traceable and reproducible engineering calibration steps, providing a standardized deployment path for the application of this invention to new commercial tasks.

[0034] Example 5: The task template dynamic optimization mechanism of the present invention operates as follows in a continuously running commercial environment: When the system records a statistically valid number of differential utility fingerprints of transacted remote sensing data for a specific task of drought stress monitoring within a predetermined period, an optimization analysis procedure is triggered. This procedure calculates the statistical median of this fingerprint set in three dimensions: information novelty, scene matching degree, and structural clarity, to form a consensus utility pattern reflecting market preferences. Subsequently, the system compares this consensus utility pattern with the currently preset task template. If a persistent deviation exceeding a preset statistical threshold is found between the two in any dimension, a template optimization suggestion report is generated. This report does not automatically modify system parameters but is submitted to a designated system administrator, who reviews the data comparison results and optimization suggestions presented in the report and makes a decision on whether to update the task template. All decision-making processes are recorded by the system log.

[0035] In the routine operation of the method, when a remote sensing data to be evaluated generates a diagnostic mismatch fingerprint characterizing its capability shortcomings because its metadata fails to fully meet the requirements of the specified task template, the system will initiate a business opportunity discovery mechanism. If the diagnostic mismatch fingerprint indicates that the current data has a shortcoming in meeting the spatial resolution requirement, the system will automatically transform the mismatch information into a structured data search instruction. This instruction searches the data product catalog for other remote sensing data products that can make up for this capability shortcoming and generates a data combination recommendation. This recommendation includes the differential utility fingerprints of the original data and the recommended supplementary data, as well as an expected combined utility fingerprint that can meet all the requirements of the initial task template after recalculating the combination of the two features. This provides data users with a quantifiable combination scheme basis for procurement decisions.

[0036] Example 6: When the method of the present invention faces the boundary condition of no valid publicly known benchmark remote sensing data, the system is configured to execute a set of preset fault-tolerant procedures. The procedure is defined as follows: when the system is a piece of data to be evaluated, after a preset number of consecutive searches in all configured public data archives, it is still unable to obtain a valid benchmark data that meets the correlation requirements in time and space and has passed the benchmark data credibility gating review, the system determines that it is currently in a benchmark unavailable state. In this state, in order to ensure the continuity of the evaluation process, the system automatically sets the first utility component of the data to be evaluated, i.e., information novelty, to a preset neutral non-zero value of 0.5, and at the same time, adds a status flag indicating that information novelty has not been effectively calculated to the differential utility fingerprint generated in this evaluation.

[0037] To determine the internal weights of the scene matching component in the task template, so that its utility evaluation has a reproducible quantitative basis, the system is configured to execute an offline optimization parameter search process. The objective function of this process is defined as finding a set of weight combinations that maximizes the mean scene matching score calculated from a set of positive sample reference data and minimizes the mean score calculated from a set of negative sample reference data. To solve this objective function, the system adopts a coordinate ascending iterative optimization algorithm. The initial state of this algorithm is to set the internal weights of the three items under the scene matching component—key band requirement, optimal resolution range requirement, and necessary relative requirement—to the same initial value. In each iteration, the algorithm fixes all weights except one weight and traverses the weight in the interval [0,1] with a preset step size to find the weight value that optimizes the objective function value. This process is applied to all weights in turn until a fixed number of iterations is completed, or the gain of the objective function value is lower than a preset threshold in two consecutive iterations, and then the final weight combination is output.

[0038] 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.

[0039] 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 method for assessing and pricing remote sensing data assets using AI, characterized in that, This method establishes a set of computer data processing procedures, which include the following steps: Step a, generating a differential utility fingerprint characterizing the decision-making utility of remote sensing data under a specific task, specifically includes: after acquiring the remote sensing data to be evaluated and a known benchmark remote sensing data, performing a benchmark data credibility gating review: applying signal processing methods to the known benchmark remote sensing data to obtain its own structural clarity component, and determining whether the structural clarity component meets the credibility threshold; if the judgment result is not satisfied, another known benchmark remote sensing data is selected and this review is repeated until a valid benchmark data that meets the credibility threshold is found; after the valid benchmark data is determined, by calculating the difference in information distribution between the data to be evaluated and the valid benchmark data, a first utility component characterizing the information novelty of the data to be evaluated is obtained; based on a task template that defines specific task requirements, the metadata of the data to be evaluated is logically matched with the task template to obtain a second utility component characterizing the scenario matching degree of the data to be evaluated; applying signal processing methods to the data to be evaluated to calculate the ratio of its internal structured information to unstructured information, to obtain a third utility component characterizing the structural clarity of the data to be evaluated itself. Step b: Combine the first utility component, the second utility component, and the third utility component to form a differential utility fingerprint of the remote sensing data to be evaluated. Step c involves outputting the differential utility fingerprint as a standardized technical data structure to support the valuation and pricing decisions of remote sensing data assets.

2. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, In step a, the acquisition methods for the first utility component, the second utility component, and the third utility component are further defined as follows: The first utility component is acquired by calculating the gray-level histograms of the data to be evaluated and the effective reference data for their respective predetermined bands, and using KL divergence to measure the difference between the two gray-level histograms; the second utility component is acquired by logically comparing and weighting the band information, spatial resolution, and imaging time information contained in the metadata of the data to be evaluated with the key band requirements, optimal resolution range requirements, and phase requirements defined for the specific task in the task template; the third utility component is acquired by applying the Laplacian operator to filter the data to be evaluated and calculating the variance ratio between the filtered image and the original image.

3. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, In step a, the structural sharpness component The calculation method is precisely defined as follows: ,in, To determine the variance of pixel value distribution in an image after applying the Laplacian operator to remote sensing data. This represents the variance of the pixel value distribution in the original remote sensing image.

4. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, The confidence threshold is a value determined based on statistical analysis of historical remote sensing data to identify low-quality data with large-area cloud and fog obscuration or severe sensor noise.

5. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, In step b, the differential utility fingerprint also includes a fourth utility component. The method for generating the fourth utility component includes: obtaining the spatial resolution of the remote sensing data to be evaluated and the effective reference remote sensing data; calculating the ratio of their spatial resolutions and performing a nonlinear mapping of the ratio through an arctangent function to generate a fourth utility component that characterizes the degree of scale inconsistency between the two, serving as a confidence index for the first utility component.

6. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, When performing step a, the method also includes the following steps in parallel: recording the mismatch information between the metadata of the data to be evaluated and each requirement in the task template; and combining all non-zero mismatch information to generate a structured diagnostic mismatch fingerprint that characterizes the capability shortcomings of the data to be evaluated under specific task requirements.

7. The method for evaluating and pricing remote sensing data assets using AI according to claim 6, characterized in that, The method also includes: based on the content of the diagnostic mismatch fingerprint, invoking a data search command to search the data catalog for other remote sensing data products whose metadata matches one or more mismatch information recorded in the diagnostic mismatch fingerprint, and generating a data combination suggestion that includes information from other remote sensing data products.

8. The method for evaluating and pricing remote sensing data assets using AI according to claim 1, characterized in that, The method also includes a dynamic optimization mechanism for task templates, which includes the following steps: recording differential utility fingerprints of transacted remote sensing data associated with a specific task; periodically performing statistical analysis on a set of recorded differential utility fingerprints associated with the same specific task, and determining a consensus utility pattern reflecting market preferences by calculating the statistical median of each utility component dimension; and comparing the consensus utility pattern with the task template of the specific task, and generating a set of optimized data for optimizing the task template based on the comparison results.

9. A method for assessing and pricing remote sensing data assets using AI according to claim 2, characterized in that, The task templates are stored in JSON format, which allows for updating or adding to the task templates via text editing.

Images