Landscape seasonal change prediction and four-season landscape effect intelligent generation system

Abstract

This invention relates to the field of landscape seasonal image processing technology, specifically a system for predicting landscape seasonal changes and intelligently generating seasonal landscape effects. The system includes acquiring basic landscape image data and corresponding seasonal annotation data; generating landscape seasonal image data based on display scene parameters; acquiring observer identification information corresponding to the generated landscape seasonal image data; modeling perceptual differences among different observers to generate deviation modeling results characterizing seasonal perception biases; and adjusting the generated landscape seasonal image data accordingly before outputting the display results. By introducing an observer differentiation mechanism into the landscape image generation and processing process, seasonal changes are made more consistent with the perceptual characteristics of different observers in terms of visibility, expression of change types, and seasonal understanding. This reduces the deviation between perceptual results and design expectations in public participatory landscape scheme presentations, improving the consistency and reliability of the display effect.

Description

Technical Field

[0001] This invention relates to the field of landscape seasonal image processing, specifically a system for predicting landscape seasonal changes and intelligently generating seasonal landscape effects. Background Technology

[0002] With the development of digital landscape design, 3D visualization, and participatory planning technologies, seasonal landscape effects are typically presented to different groups through image generation and visualization to aid in scheme evaluation and optimization. These technologies are widely used in urban public space design, landscape planning, and ecological landscape display, simulating changes in vegetation color, morphology, and spatial hierarchy across different seasons to achieve a visual representation of landscape schemes over time.

[0003] In existing technologies, the generation of landscape seasonal effects is usually based on historical landscape images or design models. This involves analyzing the patterns of landscape change under different seasonal conditions and generating corresponding seasonal landscape image data using image generation algorithms or seasonal prediction models. Related technologies generally focus on the overall modeling and image rendering of seasonal change trends. For example, they may adjust the landscape's color, texture, or morphological features based on seasonal tags to obtain display results for different seasonal phases such as spring, summer, autumn, and winter.

[0004] The inventors of this application have discovered that the existing landscape seasonal change prediction and display technologies have at least the following technical problems in practical applications: Existing public participatory landscape design display technologies typically assume that observers can objectively and accurately see seasonal changes and distinguish their content. However, in practice, different observers differ even at the basic level of being able to see changes; for example, some elderly people are insensitive to color changes, and children lack attention to subtle morphological changes. Secondly, even if changes can be seen, different groups still show significant differences in their ability to distinguish the specific content of the changes; for example, they may not be able to clearly distinguish whether it is a color change, a structural change, or a spatial hierarchy change. Furthermore, even if the content of the changes can be distinguished, different observers have varying degrees of understanding of the seasonal characteristics and landscape intentions conveyed by the changes, leading to a deviation between the final perceived results and the design expectations. These problems are further amplified under different display times, lighting conditions, viewing distances, and other environmental changes. Existing technologies typically only treat differences in observer ability and environmental differences as horizontally superimposed a priori conditions, making it difficult to identify the reasons and degree of discrepancy between the observed effectiveness and the generated results. Summary of the Invention

[0005] This invention proposes a landscape seasonal change prediction and intelligent generation system for four-season landscape effects. It aims to solve the problem that existing landscape display schemes can only present seasonal effects based on fixed rules or static materials, which makes it difficult to accurately reflect the dynamic changes of the real landscape under different seasonal conditions, and there is a deviation between the generated results and human perception.

[0006] To this end, this invention introduces a closed-loop modulation mechanism for seasonal change prediction and perception feedback, which links the objective characteristics of landscape change with the subjective characteristics of the observer's perception. This allows for the generation of seasonal landscape effects that are more in line with human visual cognition and understanding habits, while ensuring the rationality of seasonal change patterns.

[0007] To achieve the above objectives, the embodiments of this application disclose the following technical solutions:

[0008] This solution discloses a landscape seasonal change prediction and intelligent generation system for four-season landscape effects, including a processor and a memory; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory to achieve the following steps:

[0009] The system acquires basic landscape image data, seasonal annotation data, and display scene parameter data for the target landscape area. Basic landscape image data characterizes the spatial structure, color distribution, and landscape element composition of the target landscape area at an objective level, serving as the foundational data source for seasonal change prediction and landscape effect generation. Seasonal annotation data indicates the season or seasonal phase state of the corresponding landscape image at the time of acquisition, enabling the system to establish a correspondence between landscape features and seasonal states. Display scene parameter data reflects the differences in landscape effect presentation under different display times, lighting conditions, and viewing distances, thus providing a basis for scene differentiation in subsequent perception modeling.

[0010] Based on basic landscape image data and seasonal annotation data, seasonal feature information is extracted to characterize the changing patterns of the target landscape area under different seasonal states. Seasonal change trend prediction processing is performed on the target landscape area to generate an initial seasonal change prediction result containing at least one type of seasonal feature weight information.

[0011] By jointly analyzing basic landscape image data and its corresponding seasonal annotation data, seasonal characteristic information reflecting the changing trends of the target landscape area under different seasonal states can be extracted. This seasonal characteristic information is used to describe the overall changing pattern of landscape elements with seasonal changes.

[0012] Based on this, performing seasonal change trend prediction processing on the target landscape area helps to avoid the randomness and discontinuity caused by generating landscape effects based solely on images at a single point in time, thus ensuring that the generated results have the rationality and consistency of evolution in the time dimension.

[0013] Based on the initial seasonal change prediction results, image feature expression parameters for characterizing different seasonal states are determined. Based on the image feature expression parameters, seasonal landscape effect generation processing is performed on the target landscape area to generate at least one set of landscape seasonal generation image data.

[0014] By transforming the initial seasonal change prediction results into image feature expression parameters that characterize different seasonal states, the abstract seasonal change patterns can be mapped into a visualized landscape representation, thereby generating the four-season landscape effect of the target landscape area under different seasonal conditions.

[0015] This generation process ensures that the landscape effect not only reflects the objective characteristics of seasonal changes, but also maintains stylistic consistency and continuity of change between different seasonal phases.

[0016] While outputting landscape seasonality generated image data, the system acquires observer's perceptual feedback data on the landscape seasonality generated image data.

[0017] Since the evaluation of landscape effects depends not only on the objective quality of the generated images but also on the observer's visual perception, comprehension, and cognitive habits, incorporating observer perceptual feedback data alongside the generated landscape seasonal image data allows us to obtain subjective perceptual information about the generated results in real-world usage scenarios. This perceptual feedback data provides a foundation for subsequently identifying discrepancies between the generated results and human perception.

[0018] Data alignment and feature extraction are performed on the perception feedback data to generate perception result feature data;

[0019] By performing data alignment and feature extraction on the perceived feedback data, raw, discrete subjective feedback information can be transformed into structured perceived result feature data. This allows the perceived information to be correlated and modeled within the same analytical framework as the landscape seasonal image data. This process helps reduce the interference of differences in feedback formats from different observers on subsequent analysis and improves the stability of the perceived information data.

[0020] Based on the perceptual result feature data and the landscape seasonal generation image data, a correspondence is established between the perceptual result feature data and various landscape features extracted from the landscape seasonal generation image data according to the different feature categories of the perceptual result feature data. A perceptual bias model is constructed, and perceptual bias parameters corresponding to various landscape features are output.

[0021] By establishing a correspondence between the perceptual result feature data and various landscape features in the landscape seasonal image data, a perceptual bias model can be constructed to characterize the difference between the generated result and human perception.

[0022] Perception bias models are used to quantify the degree of bias generated by different landscape features in the actual perception process, thereby providing a basis for subsequent adjustments to the generation process.

[0023] Based on the perception bias parameter, according to the landscape feature type corresponding to the perception bias parameter, the corresponding seasonal feature weights and image feature expression parameters in the initial seasonal change prediction results are reverse-modulated to generate updated landscape seasonal generation image data.

[0024] By inversely modulating the seasonal feature weights and image feature representation parameters based on perceptual bias parameters, human perceptual preferences can be introduced into the process of seasonal change prediction and landscape effect generation, enabling the system to have adaptive adjustment capabilities. This inverse modulation mechanism forms a closed-loop optimization process based on perceptual feedback, allowing the generated seasonal landscape effects to gradually approach the presentation results that conform to human visual perception and understanding habits through multiple iterations.

[0025] This invention constructs a closed-loop technical architecture of "seasonal change prediction - landscape effect generation - perception feedback modeling - reverse modulation optimization", which realizes the collaborative modeling of the seasonal change law of landscape and human perception characteristics, and effectively improves the rationality, realism and perceptual consistency of the generated landscape effect in all four seasons.

[0026] This invention further proposes that, in order to establish a stable correlation between landscape and seasonal changes, seasonal phase annotation data corresponding to the basic landscape image data is introduced to indicate the seasonal type or seasonal phase state of the landscape image at the time of acquisition. By establishing a correspondence between the seasonal phase annotation data and the basic landscape image data, interference caused by the overlapping of landscape features from different seasons on the modeling of change patterns can be avoided, thereby improving the accuracy and consistency of seasonal change prediction results.

[0027] This invention further proposes that when generating and analyzing the effects of seasonal landscapes, the changes in landscapes under different seasonal phases are usually reflected in multiple dimensions, including but not limited to changes in color distribution, changes in landscape structure and form, and changes in the overall spatial hierarchy.

[0028] Correspondingly, an observer's perception of seasonal changes in a landscape can also be characterized from different dimensions, such as the perception of the visibility of the landscape as easy to identify, the perception of the distinguishing of the forms of change, and the perception of the understanding of seasonal characteristics.

[0029] By modeling different types of landscape features in relation to different perceptual dimensions, we can more precisely characterize the sources of difference between generated landscape effects and human perception, thereby avoiding the problem of perceptual bias masked by using only a single evaluation index.

[0030] The present invention further proposes that, since perceptual feedback data usually originates from the observer's subjective judgment, its expression may vary. In order to facilitate correlation analysis with landscape features, perceptual feedback data is processed in a structured manner.

[0031] By uniformly encoding, characterizing, and time-aligning different types of sensory feedback information, sensory feedback data can participate in subsequent sensory bias modeling processes in a computable form, thereby improving the utilization efficiency and stability of sensory information in the overall technical solution.

[0032] This invention further proposes that, in order to accurately reflect the difference between the generated landscape effect and the actual human perception under different perceptual dimensions, this invention introduces a multi-dimensional perceptual bias model mechanism.

[0033] By establishing corresponding deviation modeling units for different perceptual dimensions—namely, visibility deviation sub-models, discrimination deviation sub-models, and comprehensibility deviation sub-models—interference between different perceptual factors can be avoided. This allows various perceptual deviation parameters to more accurately reflect the degree of deviation of corresponding landscape features at the perceptual level. This hierarchical modeling approach helps improve the interpretability of perceptual deviation analysis and provides a clear modulation direction for subsequent perceptual deviation-based adjustments.

[0034] The present invention further proposes that, based on the perception deviation parameters output by the perception deviation model, a reverse modulation mechanism is introduced to adjust the seasonal change prediction process and the landscape effect generation process.

[0035] By feeding back the perceptual bias parameter to the adjustment process of seasonal feature weights and image feature representation parameters, the system can gradually correct the prediction results and generation effects during multiple generation and feedback processes, making them more consistent with the observer's perceptual characteristics and cognitive habits. This inverse modulation mechanism constitutes a closed-loop optimization process based on perceptual feedback, enabling the system to have continuous learning and adaptive optimization capabilities.

[0036] The present invention further proposes that, in practical application scenarios, the perceived results of landscape effects are not only related to the generated image itself, but are also affected by factors such as display time, lighting conditions, viewing distance, and individual differences of observers.

[0037] Therefore, introducing display scene parameters and observer identification information, namely display time, lighting conditions and viewing distance parameters, into the perception modeling process helps to distinguish and analyze the perception feedback under different usage scenarios and different observer groups, thereby improving the adaptability of the perception bias model to complex real-world application environments.

[0038] In summary, this invention, through the collaborative design of seasonal phase annotation, landscape feature types, perception dimension division, perception bias hierarchical modeling, and reverse modulation closed-loop optimization mechanism, enables the prediction of landscape seasonal changes and the generation of seasonal landscape effects to establish an effective mapping relationship between objective change laws and subjective perception characteristics, thereby significantly improving the rationality and perceptual consistency of the generated results.

[0039] This invention processes basic landscape image data of the target landscape area and combines it with seasonal annotation data to predict the changing patterns of the landscape under different seasonal states. Based on this, it generates corresponding seasonal landscape effects, thereby avoiding the problem of distorted seasonal expression caused by relying solely on static rules or single display results.

[0040] Unlike existing technologies, this invention incorporates observer feedback data on the generated results during the construction of landscape seasonality image data. This feedback data undergoes data alignment and feature extraction processing, enabling the system to distinguish perceptual differences among observers at different levels, such as "whether they can see changes," "whether they can distinguish the type of change," and "their understanding of seasonal characteristics." By constructing a perceptual bias model and outputting perceptual bias parameters corresponding to various landscape features, the system can identify the sources of deviation between the generated landscape seasonality image data and the actual perceptual results.

[0041] Based on the aforementioned perception bias parameter, this invention reverse-modulates the seasonal feature weights in the seasonal change prediction results and the image feature expression parameters in the generation and processing of seasonal landscape effects. This enables the subsequent generation of landscape seasonal image data to be specifically adjusted in terms of color change features, structural change features, and spatial hierarchy change features, thereby improving the visibility, discernibility, and consistency of understanding of seasonal changes among different observer groups.

[0042] Therefore, by processing and generating landscape images and implementing closed-loop adjustment based on perceptual feedback, this invention enables the seasonal landscape effect to not only visually present reasonable seasonal changes, but also better adapt to the differences in perceptual abilities of different observers, reduce the deviation between perceptual results and design expectations, and improve the effectiveness and reliability of public participatory landscape scheme presentation. Attached Figure Description

[0043] Figure 1 This is the overall flowchart of the present invention;

[0044] Figure 2 This is a flowchart of the four-season landscape effect generation process in an embodiment of the present invention;

[0045] Figure 3 This is a structural diagram of the perception bias model according to an embodiment of the present invention;

[0046] Figure 4 This is a graph showing the relationship between the sensing deviation parameter and the reverse modulation in an embodiment of the present invention. Detailed Implementation

[0047] Specific embodiments of the invention will now be described in detail. Although the invention is described in conjunction with these specific embodiments, it should be understood that the invention is not intended to be limited to these specific embodiments. Rather, these embodiments are intended to cover alternative, modified, or equivalent embodiments that may be included within the spirit and scope of the invention as defined by the claims. In the following description, numerous specific details are set forth in order to provide a thorough understanding of the invention. The invention may be practiced without some or all of these specific details. In other instances, well-known processes have not been described in detail so as not to unnecessarily obscure the invention.

[0048] When used in conjunction with the terms "comprising," "method comprising," or similar language in this specification and appended claims, the singular forms "a," "some," and "the" include plural references unless the context clearly indicates otherwise. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0049] Example

[0050] A landscape seasonal change prediction and intelligent generation system for four-season landscape effects includes a processor and a memory; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory to achieve the following steps:

[0051] Acquire basic landscape image data, seasonal annotation data, and display scene parameter data for the target landscape area. The seasonal annotation data is annotation information that corresponds one-to-one with the basic landscape image data, used to indicate the seasonal type or seasonal state of the basic landscape image at the time of acquisition. The display scene parameter data includes display time, lighting conditions, and viewing distance parameters. The display scene parameter data participates in the calculation as an independent input parameter in the process of constructing the perception deviation model, and is used to distinguish and model the perception feedback data under different display times, lighting conditions, and viewing distances.

[0052] In this embodiment, basic landscape image data can be acquired within the target landscape area using fixed shooting equipment, drone aerial photography equipment, or mobile acquisition terminals. The acquired images objectively reflect the spatial morphology, vegetation distribution, and overall composition of the target landscape in the actual environment. Seasonal annotation data, as auxiliary information corresponding one-to-one with the basic landscape image data, can be manually annotated or automatically generated based on time information. It is used to clarify the season type or specific seasonal state at the time of image acquisition, thereby providing basic semantic constraints for subsequent seasonal change analysis. Display scene parameter data, in this embodiment, is used to describe the display environment of the landscape effect. It is introduced as an independent input parameter into the subsequent perception bias model calculation, enabling the system to distinguish the perceptual differences generated under different display times, lighting conditions, and viewing distances, avoiding misjudging perceptual changes caused by environmental factors as seasonal change characteristics of the landscape itself.

[0053] Based on basic landscape image data and seasonal annotation data, seasonal feature information is extracted to characterize the changing patterns of the target landscape area under different seasonal states. Seasonal change trend prediction processing is performed on the target landscape area to generate an initial seasonal change prediction result containing at least one type of seasonal feature weight information.

[0054] In this embodiment, seasonal feature information includes abstract features related to vegetation status, color distribution, and spatial structure changes, used to reflect the overall trend of landscape changes with the seasons. Based on the seasonal feature information, seasonal change trend prediction processing is performed on the target landscape area, enabling the system to learn the relationship between different seasonal phases from historical or existing images, and assign corresponding seasonal feature weights to different seasonal features during the prediction process, thereby generating an initial seasonal change prediction result containing at least one type of seasonal feature weight information. This result is used to describe the relative influence of different seasonal states in the overall prediction.

[0055] In this embodiment, based on feature extraction and change pattern analysis of basic landscape image data and seasonal phase annotation data, various seasonal phase feature information are correlated and modeled in chronological order. By fitting and extrapolating the landscape feature change trends under different seasonal conditions, a seasonal change prediction mechanism for characterizing the seasonal change pattern of the target landscape area is constructed.

[0056] The seasonal change prediction mechanism uses seasonal feature weights and their variation parameters as the core calculation basis. Through comprehensive analysis of the relationship between landscape feature changes under different seasonal conditions, it outputs the initial seasonal change prediction results to describe the visual performance of the target landscape under future or target seasonal conditions.

[0057] In the above-mentioned seasonal change prediction mechanism, in order to quantify the influence of different seasonal features in the prediction process, this embodiment represents the extracted seasonal feature information as feature vectors and models the pattern of their change over time based on the seasonal labeling data.

[0058] Specifically, the seasonal feature information of the target landscape area acquired at different times is represented as a seasonal feature vector, and a corresponding seasonal feature weight is assigned to each seasonal feature. By weighting the seasonal feature vector and the seasonal feature weight, the initial seasonal change prediction result used to characterize the overall visual state under the target seasonal conditions is obtained, and then the following seasonal change prediction model formula is obtained:

[0059] ;

[0060] in, The initial seasonal change prediction results are used to describe the overall visual appearance of the target landscape area at the target time point or under the target seasonal conditions.

[0061] The amount of seasonal feature information is determined by the number of seasonal feature types extracted from the basic landscape image data and the seasonal annotation data;

[0062] For the first Seasonal characteristic information in time The corresponding feature values ​​can be represented as numerical features or feature vectors calculated from the corresponding image features, which are used to participate in the seasonal change prediction calculation. The seasonal feature information comes from the vegetation status features, color distribution features and spatial structure change features extracted from the basic landscape image data. When the seasonal feature information is represented in the form of feature vectors, the weighted calculation means that the corresponding feature vectors are weighted and combined to obtain a comprehensive feature representation that characterizes the overall seasonal change trend.

[0063] To and The corresponding seasonal feature weights are based on the landscape feature change patterns under different seasonal conditions in historical seasonal labeling data. They are obtained by statistical analysis or trend fitting of the relationship between the corresponding seasonal features under historical seasonal conditions and time, and are used to characterize the relative influence of the seasonal feature in the prediction results.

[0064] In this embodiment, the above-mentioned modeling and extrapolation calculation process for generating the initial seasonal change prediction results constitutes a seasonal change prediction model for landscape seasonal change prediction. This seasonal change prediction model provides a basic prediction basis for subsequent landscape seasonal effect generation processing and perception bias correction.

[0065] In this embodiment, seasonal feature information is used to characterize the changing trend of the target landscape under different seasonal conditions, and belongs to the abstract features used for prediction modeling; landscape features are used to characterize the specific perceptible visual elements in the landscape seasonal generated image, and belong to the objective features at the generated image level.

[0066] Based on the initial seasonal change prediction results, image feature expression parameters for characterizing different seasonal states are determined. Based on the image feature expression parameters, seasonal landscape effect generation processing is performed on the target landscape area to generate at least one set of landscape seasonal effect generation image data. The seasonal landscape effect generation processing includes: determining image feature expression configuration parameters for characterizing different seasonal states based on the initial seasonal change prediction results, and generating landscape seasonal effect generation image data with corresponding seasonal features based on the image feature expression configuration parameters. The various landscape features extracted from the landscape seasonal effect generation image data include at least color change features, structural change features, and spatial hierarchy change features.

[0067] In practice, image feature representation parameters are used to constrain the presentation of different visual elements during subsequent image generation, ensuring that the generated seasonal landscape images visually reflect the predicted seasonal change trends. By applying these image feature representation parameters to the target landscape area, the process of generating seasonal landscape effects can be achieved through comprehensive control of color representation, structural morphology, and spatial hierarchy. This ensures that the generated seasonal landscape images conform to the visual characteristics of different seasons in terms of overall style and detail variations, thus providing a stable display object for subsequent sensory feedback collection.

[0068] While outputting landscape seasonality generated image data, the system acquires observer perception feedback data on the generated image data. This perception feedback data includes at least visibility feedback data, discrimination feedback data, and interpretability feedback data. The acquired perception feedback data is structured data collected through an interactive terminal. This structured data includes visibility indicators for the generated landscape seasonality image data, change type selection information, and seasonal interpretation annotation information. Furthermore, the perception feedback data is associated with the corresponding display scene parameter data through a unified time identifier. During the perception feedback data acquisition process, observer identification information is simultaneously acquired. The perception feedback data is categorized and stored based on this observer identification information, which is used to distinguish observer groups with different perceptual ability characteristics.

[0069] Perceptual feedback data is collected in a structured format to clearly record the observer's direct judgments when viewing landscape seasonal images. By setting visibility indicators, change type selection information, and seasonal understanding annotations, observers can provide clear feedback on "whether they perceived the change," "what type of change it was," and "whether they understood its corresponding seasonal meaning." Simultaneously, a unified time stamp links the perceptual feedback data with corresponding display scene parameter data, ensuring that the perceptual results match the specific display environment conditions.

[0070] The perceived feedback data undergoes data alignment and feature extraction to eliminate format differences between different feedback sources, generating perceptual result feature data suitable for model calculation. This perceptual result feature data includes visibility feature data, change type feature data, and understanding feature data. The data alignment and feature extraction process includes: binarizing or hierarchically encoding the visibility feedback data to transform it into visibility feature data indicating whether a change was perceived; performing category mapping on the discriminative feedback data to generate change type feature data, reflecting the observer's differentiation of change content; and semantically vectorizing the understanding feedback data to generate understanding feature data, characterizing the observer's level of understanding of the seasonal meaning. Each of these perceptual result feature data types serves as an independent feature dimension, providing structured input for subsequent perceptual bias modeling.

[0071] Based on the perceptual result feature data and the landscape seasonal image data, a perceptual bias model is constructed by establishing correspondences between the different feature categories of the perceptual result feature data and various landscape features extracted from the landscape seasonal image data, and outputting perceptual bias parameters corresponding to each type of landscape feature. The constructed perceptual bias model is a multi-layer perceptual bias model, which includes a visibility bias sub-model, a resolution bias sub-model, and a comprehension bias sub-model. Each sub-model establishes a mapping relationship between its corresponding perceptual result feature data and the landscape seasonal image data. The perceptual bias parameters include sub-bias parameters corresponding to the outputs of the visibility bias sub-model, the resolution bias sub-model, and the comprehension bias sub-model, and each sub-bias parameter establishes a correspondence with color change features, structural change features, and spatial hierarchy change features, respectively. Visibility feature data establishes a correspondence with color change features, change type feature data establishes a correspondence with structural change features, and comprehension feature data establishes a correspondence with spatial hierarchy change features. During the construction of the perceptual bias model, corresponding observer identification information is introduced as one of the input features.

[0072] In this step, the perception bias model adopts a multi-layered structure, including a visibility bias sub-model, a discrimination bias sub-model, and a comprehension bias sub-model. Each sub-model is used to characterize the bias at different perception levels. By inputting visibility feature data, change type feature data, and comprehension feature data into the corresponding sub-models, and combining them with color change features, structural change features, and spatial hierarchy change features for mapping calculation, the output is the perception bias parameters corresponding to various landscape features.

[0073] In this embodiment, in order to quantitatively characterize the degree of deviation between the observer's actual perception results and the objective landscape features in the landscape seasonal generation image, the feature data of various perception results are compared and calculated with the corresponding landscape features.

[0074] Specifically, by modeling the difference between the objective feature values ​​of the same landscape feature in the generated image and the feature values ​​perceived by the observer, a perception bias parameter is obtained to characterize the degree of perception bias, and thus the following formula for calculating the perception bias parameter is derived:

[0075] ;

[0076] in, For the first The perceptual deviation parameter corresponding to the landscape feature is used to characterize the degree of deviation between the observer's perception result and the landscape feature in the landscape seasonal image data. Before calculating the perceptual deviation parameter, the perceptual result feature data and the corresponding landscape feature are aligned in feature space to make them in a comparable feature representation space.

[0077] To extract the first from landscape seasonal image data The feature values ​​of the landscape features include color change features, structural change features, and spatial hierarchy change features; they are obtained by performing image feature extraction processing on the landscape seasonal image data.

[0078] In order to be with the first The perception result feature data corresponding to the landscape features are derived from the visibility feature data, change type feature data, and understanding feature data obtained after data alignment and feature extraction processing of the perception feedback data.

[0079] Observer identification information is introduced into the model as an additional input feature, enabling the model to distinguish perceptual differences among different observer groups.

[0080] Based on the perception bias parameter, according to the landscape feature type corresponding to the perception bias parameter, the seasonal feature weights and image feature expression parameters corresponding to the initial seasonal change prediction results are back-modulated to generate updated landscape seasonal generation image data. The back-modulation includes: by identifying the sub-bias parameters corresponding to the seasonal feature weights in the perception bias parameter, adjusting the seasonal feature weights involved in the calculation in the seasonal change prediction process, and generating updated seasonal change prediction results based on the adjusted seasonal feature weights.

[0081] In this embodiment, in order to enable the seasonal change prediction model to adaptively correct itself based on the perception bias, a perception bias parameter is introduced to reverse-modulate the original seasonal feature weights.

[0082] Specifically, based on the magnitude and direction of the perception bias parameter, the corresponding seasonal feature weights are adjusted by increasing or decreasing them to suppress prediction deviations caused by perception bias, thereby obtaining the updated seasonal feature weights. The calculation process can be expressed as the following weight update formula:

[0083] ;

[0084] in, The updated version Seasonal feature weights are used to recalculate subsequent seasonal change prediction results;

[0085] The first method used in the initial seasonal change prediction process Seasonal feature weights;

[0086] In order to be with the first The perceptual deviation parameters corresponding to the seasonal characteristics are output by the perceptual deviation model;

[0087] The weight modulation coefficient is used to control the degree of influence of the perception bias parameter on the adjustment range of the seasonal feature weights, and to limit the adjustment range of the seasonal feature weights in a single adjustment, so as to avoid over-correction of the seasonal change prediction results. The weight modulation coefficient is a preset parameter or obtained through statistical analysis of historical generated data.

[0088] In the actual update process, the updated seasonal feature weights can be subject to validity constraints to ensure that they meet the requirements of the prediction model for weight stability.

[0089] The reverse modulation also includes: modulating the image feature expression parameters in the generation and processing of seasonal landscape effects based on the sub-bias parameters corresponding to the seasonal feature weights in the perception bias parameters, so as to correct the expression of color, structure and spatial hierarchy in the seasonal landscape generation images. The image feature expression parameters include at least color expression parameters, structure expression parameters and spatial hierarchy expression parameters. The updated seasonal change prediction results are stored as historical generation data and participate in the parameter update processing of the perception bias model together with the newly acquired perception feedback data, thereby forming a closed-loop generation mechanism for continuous iterative optimization.

[0090] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation methods of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A system for predicting seasonal landscape changes and intelligently generating seasonal landscape effects, characterized in that: It includes a processor and a memory; the memory is used to store computer programs; the processor is used to execute the computer programs stored in the memory to perform the following steps: Acquire basic landscape image data, seasonal annotation data, and display scene parameter data for the target landscape area; Based on basic landscape image data and seasonal annotation data, seasonal feature information is extracted to characterize the changing patterns of the target landscape area under different seasonal states. Seasonal change trend prediction processing is performed on the target landscape area to generate an initial seasonal change prediction result containing at least one type of seasonal feature weight information. Based on the initial seasonal change prediction results, image feature expression parameters for characterizing different seasonal states are determined. Based on the image feature expression parameters, seasonal landscape effect generation processing is performed on the target landscape area to generate at least one set of landscape seasonal generation image data. While outputting landscape seasonality generated image data, the system acquires observer's perceptual feedback data on the landscape seasonality generated image data. Data alignment and feature extraction are performed on the perception feedback data to generate perception result feature data; Based on the perceptual result feature data and the landscape seasonal generation image data, a correspondence is established between the perceptual result feature data and various landscape features extracted from the landscape seasonal generation image data according to the different feature categories of the perceptual result feature data. A perceptual bias model is constructed, and perceptual bias parameters corresponding to various landscape features are output. Based on the perception bias parameter, according to the landscape feature type corresponding to the perception bias parameter, the corresponding seasonal feature weights and image feature expression parameters in the initial seasonal change prediction results are reverse-modulated to generate updated landscape seasonal generation image data.

2. The landscape seasonal change prediction and intelligent generation system for four-season landscape effects according to claim 1, characterized in that, Seasonal annotation data are annotation information that corresponds one-to-one with the basic landscape image data, used to indicate the seasonal type or seasonal state of the basic landscape image at the time of acquisition.

3. The landscape seasonal change prediction and intelligent generation system for four-season landscape effects according to claim 2, characterized in that, The generation and processing of the four seasons landscape effect includes: Based on the initial seasonal change prediction results, the image feature representation configuration parameters for characterizing different seasonal states are determined, and landscape seasonal generation image data with corresponding seasonal features are generated based on the image feature representation configuration parameters. The various landscape features extracted from landscape seasonal image data include at least color change features, structural change features, and spatial hierarchy change features; Perceptual feedback data includes at least visual feedback data, discriminative feedback data, and comprehension feedback data; The perception result feature data includes visibility feature data, change type feature data, and understanding feature data; the data alignment and feature extraction processing of the perception feedback data includes: performing binarization or hierarchical encoding processing on the visibility feedback data to generate visibility feature data; performing category mapping processing on the discriminative feedback data to generate change type feature data; and performing semantic vectorization processing on the understanding feedback data to generate understanding feature data. Visibility feature data, change type feature data, and understanding feature data are each input as an independent feature dimension into the perception bias model.

4. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 3, characterized in that, The constructed perception bias model is a multi-layer perception bias model, which includes a visibility bias sub-model, a resolution bias sub-model, and a comprehension bias sub-model. Each sub-model establishes a mapping relationship between the corresponding perception result feature data and the landscape seasonal generation image data.

5. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 4, characterized in that, The perceptual bias parameters include sub-bias parameters corresponding to the outputs of the visibility bias sub-model, the discrimination bias sub-model, and the comprehension bias sub-model, respectively, and each sub-bias parameter establishes a correspondence with color change features, structural change features, and spatial hierarchy change features, respectively; the visibility feature data establishes a correspondence with the color change features, the change type feature data establishes a correspondence with the structural change features, and the comprehension feature data establishes a correspondence with the spatial hierarchy change features.

6. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 2, characterized in that, The reverse modulation includes: adjusting the seasonal feature weights involved in the calculation of seasonal change prediction based on the sub-bias parameters corresponding to the seasonal feature weights in the sensing bias parameters, and generating updated seasonal change prediction results based on the adjusted seasonal feature weights.

7. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 3, characterized in that, The reverse modulation also includes: modulating the image feature expression parameters in the generation and processing of the four seasons landscape effect based on the sub-bias parameters corresponding to the seasonal feature weights in the perception bias parameters. The image feature expression parameters include at least color expression parameters, structural expression parameters, and spatial hierarchy expression parameters.

8. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 1, characterized in that, The display scene parameter data includes display time, lighting conditions, and viewing distance parameters. These parameters are used as independent input parameters in the construction of the perception deviation model to differentiate and model the perception feedback data under different display times, lighting conditions, and viewing distances.

9. The intelligent system for predicting landscape seasonal changes and generating four-season landscape effects according to claim 1, characterized in that, The updated seasonal change prediction results are stored as historical generated data and, together with the newly acquired perception feedback data, participate in the parameter update processing of the perception bias model.

10. The intelligent system for predicting seasonal landscape changes and generating four-season landscape effects according to claim 3, characterized in that, The perceptual feedback data is structured data collected through an interactive terminal. The structured data includes visibility information, change type selection information, and seasonal interpretation annotation information for the landscape seasonal image data. The perceptual feedback data is associated with the corresponding display scene parameter data through a unified time identifier. During the perceptual feedback data collection process, observer identification information is acquired simultaneously. The perceptual feedback data is classified and stored according to the observer identification information. The observer identification information is used to distinguish observer groups with different perceptual ability characteristics. The perceptual bias model incorporates the corresponding observer identification information as one of the input features during its construction.

Images