Urban landscape semantic segmentation method and system based on large model

Abstract

The invention relates to the field of smart cities, in particular to an urban landscape semantic segmentation method and system based on a large model, and the method comprises the steps: obtaining street view image data of a target geographic region; inputting the streetscape image data into a preset image semantic segmentation model, analyzing the streetscape image data through the image semantic segmentation model, and identifying various predefined environmental element categories contained in each image; calculating an urban landscape quantitative index based on the recognition result of the environmental element category; associating the calculated value of the urban landscape quantitative index to a corresponding road section or node in the road network spatial data corresponding to the target geographic area; generating a road network space visualization layer representing the value of the index value in a color grading manner based on the result after association; and according to a preset index threshold value, identifying the spatial road sections or nodes meeting the early warning conditions in the visual layer and generating prompt information. And the scientificity and comparability of an analysis result are ensured.

Description

Technical Field

[0001] This application relates to the field of smart city technology, and in particular to a method and system for semantic segmentation of urban landscapes based on a large model. Background Technology

[0002] Large-scale urban landscape semantic segmentation is a technique that uses a deep learning model pre-trained on massive amounts of data and possessing powerful general visual understanding capabilities to perform pixel-level analysis of urban street scene images and automatically classify the visual elements into a series of predefined semantic categories.

[0003] In the quantitative assessment of urban landscapes using current technologies, mainstream computer vision analysis methods can extract color, texture, or identify specific objects from street view images and calculate basic indicators such as vegetation cover. However, the results of these calculations are often abstract numerical values ​​or statistical charts detached from specific geographic spaces. It is difficult to determine crucial questions for planning decisions, such as "where exactly the problems are located in the city" and "how the severity of problems is spatially correlated across different locations." This disconnect between data calculation results and geospatial information makes it difficult to directly apply the analysis findings to planning, design, and management actions that require precise location. Decision-makers still need to perform a significant amount of additional, manual spatial interpretation and transformation work. Existing methods largely focus on extracting general visual features, and the indicators they calculate differ from the specialized and multifaceted evaluation dimensions focused on in fields such as urban planning and landscape design. For example, simple object recognition or color statistics are insufficient for effectively quantifying professional concepts such as "street enclosure" and "interface diversity," which require a comprehensive understanding of spatial relationships and element combinations. This results in insufficient professional depth in the analysis results.

[0004] Secondly, urban spaces are multifunctional complexes, and the core functions of different locations (such as traffic efficiency, commercial vitality, and historical preservation) determine their primary landscape optimization goals. Existing methods, after calculating multiple indicators, typically present them side-by-side or use simple weighted averages, making it difficult to identify and interpret the interrelationships or mutual reinforcement among different indicators. When multiple indicators perform poorly simultaneously, the system cannot provide decision-makers with rational analytical clues about "which contradiction to prioritize" based on the core functional attributes of the location, potentially outputting confusing or even contradictory optimization directions. Summary of the Invention

[0005] To address one or more problems in the existing technology, the main objective of this application is to provide a method and system for semantic segmentation of urban landscapes based on a large model.

[0006] To achieve the aforementioned objectives, this application proposes a semantic segmentation method for urban landscapes based on a large model, the method comprising: Acquire street view image data of the target geographic area; The street view image data is input into a preset image semantic segmentation model. The image semantic segmentation model is used to parse the street view image data and identify the various predefined environmental element categories contained in each image. Based on the identification results of the environmental element categories, urban landscape quantitative indicators are calculated; The calculated values ​​of the urban landscape quantitative indicators are associated with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. Based on the correlation results, a road network spatial visualization layer is generated, which uses color grading to represent the high and low values ​​of the index. Based on preset indicator thresholds, spatial road segments or nodes in the visualization layer that meet the warning conditions are identified and prompt information is generated.

[0007] This application also provides a large-model-based urban landscape semantic segmentation system, including: The street view image acquisition module is used to acquire street view image data of the target geographic area; The semantic segmentation and parsing module is used to input the street view image data into a preset image semantic segmentation model, and to parse the street view image data through the image semantic segmentation model to identify multiple predefined environmental element categories contained in each image. The landscape index calculation module is used to calculate urban landscape quantitative indicators based on the identification results of the environmental element categories. The spatial data association module is used to associate the calculated values ​​of urban landscape quantitative indicators with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. The visualization layer generation module is used to generate a road network spatial visualization layer based on the associated results, using color grading to represent the level of index values. The early warning information generation module is used to identify spatial road segments or nodes in the visualization layer that meet the early warning conditions based on preset indicator thresholds and generate prompt information.

[0008] This application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of any of the methods described above.

[0009] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the methods described above.

[0010] The urban landscape semantic segmentation method and system based on a large model, as described in this application, realizes a paradigm shift in urban landscape assessment from relying on subjective human interpretation to objective, batch-automated analysis. By introducing a pre-trained large model for semantic segmentation, it can efficiently and accurately identify various environmental elements from massive street view images, solving the core pain points of traditional methods such as low efficiency and inconsistent standards. Subsequently, the identification results are automatically calculated into professional quantitative indicators, establishing a reliable bridge from visual information to professional knowledge, ensuring the scientific nature and comparability of the analysis results. In terms of decision support depth, the indicator values ​​are precisely correlated with urban road network spatial data. This allows abstract assessment data to acquire specific geographical location attributes, thereby generating intuitive color-coded visualization maps. Decision-makers can go beyond traditional tables and charts, directly observing the spatial distribution patterns of landscape indicators in geographic space, identifying high-value areas and weak areas, greatly improving spatial cognition efficiency and the intuitiveness of analysis results. Based on preset thresholds, it realizes automatic identification and early warning of abnormal spaces. The system can proactively discover problem road sections or nodes and generate targeted prompts, thereby upgrading the function from "post-event description" to "pre-event early warning" and "in-event monitoring". This provides clear and identifiable decision-making targets for refined urban governance. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a large-model-based urban landscape semantic segmentation method according to an embodiment of this application. Figure 2 This is a flowchart illustrating a large-model-based urban landscape semantic segmentation method according to an embodiment of this application. Figure 3 This is a schematic block diagram of a large-model-based urban landscape semantic segmentation system according to an embodiment of this application; Figure 4 This is a schematic block diagram of the structure of a computer device according to an embodiment of this application.

[0012] The realization of the purpose, functional features and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0014] Reference Figure 1 This application provides a method for semantic segmentation of urban landscapes based on a large model, the method comprising: S1. Obtain street view image data of the target geographic area; S2. Input the street view image data into a preset image semantic segmentation model, and use the image semantic segmentation model to parse the street view image data and identify the multiple predefined environmental element categories contained in each image; S3. Based on the identification results of the environmental element categories, calculate the quantitative indicators of urban landscape; S4. Associate the calculated values ​​of the urban landscape quantitative indicators with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. S5. Based on the results of association, generate a road network spatial visualization layer that uses color grading to represent the high and low values ​​of the index. S6. Based on the preset indicator thresholds, identify the spatial road segments or nodes in the visualization layer that meet the warning conditions and generate prompt information.

[0015] As described in steps S1-S3 above, step one is the data input stage of the method. The target geographic area refers to the specific urban area to be analyzed, such as a block, an administrative district, or a planning area. Street view image data can be obtained in batches by calling the application programming interface of a commercial map platform, or it can be collected in a customized manner using a professional data collection vehicle. These images are usually panoramic views or perspective images taken from multiple directions, accompanied by precise geographic coordinates and timestamps. Step two is the core perception stage of the method. The preset image semantic segmentation model is a computer vision model based on deep learning and pre-trained on massive image data. When street view images are input, the model classifies each pixel and assigns it to a predefined semantic category related to the urban environment. These categories are specifically designed to cover key physical elements that affect the perception of the urban landscape. This achieves an automated, refined, and structured understanding of the content of street view images. It transforms unstructured image pixels into a set of machine-readable semantic tags for "environmental elements," converting human visual cognition of images into computable data, laying a solid foundation for subsequent quantitative analysis. Step three is the quantitative analysis stage of the method. The calculation process is based on mathematical formulas established using knowledge from fields such as landscape architecture and urban planning. Each quantitative indicator is directly related to one or more specific environmental element categories. For example, a typical calculation method is based on the proportion of pixels of a specific element category within the entire image or region of interest. The visual element identification results obtained in the previous step are transformed into a series of measurable and comparable numerical indicators with clear professional meaning. This process elevates qualitative visual perception to quantitative scientific evaluation, enabling the assessment of urban spatial quality to move beyond a model entirely reliant on subjective experience and achieve objectivity and standardization.

[0016] As described in steps S4-S6 above, step four is the spatial integration stage of the method. Road network spatial data is the basic vector layer in the urban geographic information system, containing linear elements such as roads and paths, as well as their geographical location information. The association operation relies on the coordinates of the collection points carried by the street view images themselves. Through spatial location matching algorithms, the index values ​​calculated for each collection point are assigned to the road segment or node in the road network that is closest to it or corresponds to it through projection relationships. This is a key step in realizing the leap "from data to space". It allows the analysis results, which were originally isolated and belonged only to a certain image collection point, to be effectively "embedded" into the digital skeleton of the urban space. As a result, the abstract index values ​​have gained clear spatial ownership and geographical context, generating spatial data rich in semantic information. Step five is the visualization stage of the method's results. Based on the index values ​​carried by each road segment or node after association, rendering is performed according to the preset color mapping rules. Usually, a continuous gradient color band is used, for example, using green to red to represent index values ​​from good to bad. Using a map rendering engine, the rendering results are generated into an independent visualization layer that can be overlaid on the map base map. This transforms complex tabular data into an intuitive "map language". Decision-makers no longer need to interpret complex reports; a single map provides a comprehensive and intuitive overview of the spatial distribution patterns, elevation differences, and problem clusters of various landscape indicators within the target area, significantly improving information acquisition efficiency and spatial insight. Step six presets threshold values ​​based on professional standards or management objectives. The system automatically traverses all spatial elements in the visualization layer, comparing their indicator values ​​with the thresholds to identify non-compliant road sections or nodes. For these non-compliant elements, structured prompts are automatically generated, such as pop-ups, tags, or summary reports.

[0017] As described above, this method realizes a paradigm shift in urban landscape assessment from relying on subjective human interpretation to objective, batch-automated analysis. By introducing a pre-trained large-scale model for semantic segmentation, it can efficiently and accurately identify various environmental elements from massive street view images, solving the core pain points of traditional methods such as low efficiency and inconsistent standards. Subsequently, the identification results are automatically calculated into professional quantitative indicators, establishing a reliable bridge from visual information to professional knowledge, ensuring the scientific nature and comparability of the analysis results. In terms of decision support depth, the indicator values ​​are precisely correlated with urban road network spatial data. This gives the abstract assessment data specific geographical location attributes, thereby generating intuitive color-coded visualization maps. Decision-makers can go beyond traditional tables and charts, directly observing the spatial distribution patterns of landscape indicators in geographic space, identifying high-value areas and weak areas, greatly improving spatial cognitive efficiency and the intuitiveness of analysis results. Based on preset thresholds, it realizes automatic identification and early warning of abnormal spaces. It can proactively discover problem road sections or nodes and generate targeted prompts, thereby upgrading the function from "post-event description" to "pre-event early warning" and "in-event monitoring". This provides clear and identifiable decision-making targets for refined urban governance.

[0018] In one embodiment, the environmental element categories include vegetation elements, sky elements, building interface elements, pedestrian facility elements, motorized transportation elements, and diversity representation elements; the step of calculating urban landscape quantitative indicators based on the identification results of the environmental element categories includes: Based on the identification results of the vegetation elements, the street green view rate is calculated; Calculate the sky openness based on the recognition results of the aforementioned sky-related elements; Calculate the street enclosure degree based on the identification results of the building interface elements; Based on the identification results of the pedestrian facility elements, the proportion of pedestrian walkways is calculated; Based on the identification results of the aforementioned motorized traffic elements, the degree of road motorization is calculated; Based on the identification results of the diversity characterization elements, the index values ​​used to characterize interface diversity are determined.

[0019] As mentioned above, based on the professional theoretical system of urban landscape perception and evaluation, each category precisely corresponds to a key physical dimension affecting people's street space experience. Specifically: Vegetation elements refer to all naturally or artificially planted green plants in the street space, such as trees, shrubs, lawns, and climbing plants. Their proportion in the field of vision is directly related to people's ecological feeling and psychological comfort. Sky elements refer to the visible portion of the sky not obstructed by buildings, structures, or dense tree canopies. Their area directly reflects the sense of openness or oppression of the street. Building facade elements mainly refer to the visual boundaries formed by the vertical facades of buildings and walls on both sides of the street, which are the main factors shaping the sense of enclosure and place of the street. Pedestrian facility elements refer to paved areas specifically designed for pedestrian passage and activities, including sidewalks, pedestrian plazas, and pedestrian safety islands. Motor vehicle elements refer to road areas serving motor vehicle traffic and the vehicles themselves appearing in the image. Diverse representational elements constitute a composite category, typically referring to detailed elements on the building facade such as shop signs, windows, balconies, decorative components, and diverse vegetation forms. Their richness collectively determines the visual interest and vitality of the facade. By reducing and aggregating the general recognition results of the large model across these six professional dimensions, we effectively mapped basic visual semantics such as "trees," "grass," "sky," and "walls" into the professional semantic framework of urban planning and design, thus providing a clearly structured and targeted input for subsequent quantitative calculations. Regarding the calculation and value of landscape quantitative indicators, we constructed a systematic method for calculating landscape quantitative indicators: Street green view rate, obtained by statistically analyzing the total number of pixels identified as vegetation elements in a single street view image and calculating their proportion of the total number of pixels in the effective analysis area of ​​the image. This indicator provides an objective and repeatable quantitative standard, replacing the subjective description of "good or bad greening," and achieving a standardized measurement of the visual supply capacity of street greening. Sky openness is obtained by calculating the proportion of sky element pixels in a specific region of interest at the top of the image. This indicator quantifies the visible range of the sky for humans and is an important visual basis for assessing the sense of confinement and ventilation and lighting potential of street spaces. Street enclosure is assessed by comprehensively calculating the pixel proportion of building interface elements within the visual areas on both sides of the image. Generally, a higher proportion indicates a stronger sense of spatial enclosure and a more defined street interface. Pedestrian walkway proportion is measured by calculating the proportion of pedestrian facility element pixels to the total pixels of the entire road cross-section. It directly reflects the degree of "people-first" in road resource allocation and provides key data for slow-traffic system planning and human-centered street renovation. Road motorization level is characterized by the proportion of motor vehicle element pixels to the total pixels of the entire road cross-section. This indicator directly reflects the functional intensity of street space in serving vehicular traffic.The interface diversity index, while relatively complex to calculate, is typically implemented by first counting the number of unique subcategories based on the identification results of diversity-representing elements, and then combining this with the evenness of the distribution of these subcategories' pixels on the building interface, for example, through information entropy calculation, to obtain a comprehensive score. This index overcomes the limitation of merely measuring physical proportions, enabling a quantitative evaluation of the visual richness and commercial vitality of street facades, which is particularly important for commercial street quality assessment and historical district landscape management. By systematically proposing and calculating these six complementary indicators, our solution can create a three-dimensional portrait of the streetscape from multiple core dimensions such as ecology, space, function, and perception. This not only overcomes the limitations of existing technologies that often focus on only a single indicator, but also makes the final analysis conclusions more comprehensive and reliable.

[0020] Reference Figure 2 In one embodiment, the step of associating the calculated urban landscape quantitative index value with the corresponding road segment or node in the road network spatial data corresponding to the target geographical area includes: S41. Obtain the spatial location data of the collection points that are bound to each street view image; S42. Obtain the road network centerline data of the target geographical area; S43. Spatial matching is performed between the spatial location data of each collection point and the road network centerline data to determine the target line element corresponding to each collection point on the road network centerline data. The target line element is used to generate a segmented road network with attached landscape indicators. S44. The calculated values ​​of all urban landscape quantitative indicators corresponding to each street view image are used as a set of attribute information and assigned to the target line element corresponding to the street view image collection point. S45. Based on the attribute assignment results, generate a new segmented road network spatial dataset containing urban landscape quantitative indicator attribute fields. The geometric shape of each road segment element is defined by the road network centerline data, and the attribute table of each road segment element in the segmented road network spatial dataset records the landscape indicator values ​​of the associated collection points.

[0021] As described above, in Step One, during the street view data acquisition process, the positioning system on the acquisition vehicle or equipment synchronously records the precise geographic coordinates at the moment each image is captured. These coordinates are uniquely bound to the image file, collectively forming the "collection point." The data format can include coordinate pairs of longitude, latitude, and elevation, stored in an independent positioning file or image metadata. This assigns a precise "geographic ID card" to each landscape analysis result. This ensures that all subsequent calculations based on street view images are no longer numerical values ​​floating in abstract statistical space, but rather geospatial information uniquely anchored to specific locations in the real world, serving as the logical starting point for establishing spatial relationships. Step Two: Road network centerline data is a core vector data type in urban geographic information systems. It expresses the direction, length, and topological relationships of roads in the form of abstract line elements through a series of continuous coordinate points. This data originates from urban basic surveying or open map data and includes attributes such as road geometry, grade, and name. It provides a standardized spatial framework representing the urban traffic network and public space skeleton. The road network centerline, as a digital expression of urban fabric, is a key geographical basis for integrating discrete, point-like landscape observations into a continuous, linear urban spatial structure for understanding and analysis. Step 3: Geospatial Calculation. Using spatial analysis algorithms in the Geographic Information System (GIS), such as "nearest neighbor analysis" or "linear reference," the vertical distance from the coordinates of each street view acquisition point to the road network centerline element is calculated. The system finds the spatially closest road network centerline for each acquisition point, within a set tolerance range, and designates it as the "target line element" for that acquisition point. This step establishes the spatial attribution relationship between "image acquisition points" and "road segments." Isolated street view measurement points representing instantaneous perspectives are logically "assigned" to the urban streets they represent. This solves the core problem of "which road's landscape does this image represent," establishing the mathematical and spatial foundation for aggregating image-dimensional analysis results to the road dimension. Step 4: After determining the attribution relationship, attribute transfer operations are performed. It takes all the previously calculated urban landscape quantitative indicators derived from the street view image and writes them as a new set of attribute fields into the attribute table of the target line element. For example, the attribute table of a road network centerline element will add fields such as "green view rate" and "sky openness" and record the specific values ​​associated with this point. This realizes the fusion of information from the "image domain" to the "geospatial domain". Road elements are thus endowed with "genetic information" describing their landscape quality, transforming from a simple line with only geometric and basic attributes into a rich and feature-rich intelligent spatial object. Step five, based on the original, continuous and complete road network centerline data, and according to the correlation of all collection points, logically segments the original centerline at spatial matching locations or creates corresponding elements to generate a new spatial dataset.Each element in this dataset represents a "segment unit" carrying specific street view survey results. Its geometry is inherited from the original road network, while its extended attribute table records detailed landscape indicator values ​​for the associated collection points. This dataset can be stored using standard geospatial data formats such as GeoJSON and Shapefile. This embodiment addresses the shortcomings of existing technologies: the "disconnect between computation and space," that is, the separation of abstract data calculation results from operable geospatial information. Through spatial matching algorithms, it solves the problem of "nowhere to place" the massive, discrete street view image analysis results in the macro-urban space, accurately and systematically "embedding" them into the continuous spatial framework of the urban road network, giving each data point a clear spatial affiliation and context. A qualitative leap from "statistical reports" to "spatial data products": This embodiment goes beyond simply establishing associations; it further defines how to generate a standardized new spatial dataset rich in professional attributes. This makes the final output no longer the charts or tables in traditional research reports, but a structured geographic data layer that can be directly imported into urban planning information systems for overlay analysis, buffer analysis, and spatial queries. This has completely changed the form and application mode of analytical results.

[0022] In one embodiment, prior to the step of generating a road network spatial visualization layer that represents the level of index values ​​using a color-grading method, the method further includes: Read the new segmented road network spatial dataset; Determine whether the new segmented road network spatial dataset meets the multi-indicator anomaly conditions; When the attributes of a new segmented road network spatial dataset contain records of at least two urban landscape quantitative indicators whose values ​​exceed their respective preset thresholds, it is determined that the multi-indicator abnormality condition is met. Based on the judgment result, a preset indicator interaction relationship rule is invoked, and the interaction relationship between the at least two abnormal indicators is determined according to the indicator interaction relationship rule, wherein the interaction relationship includes conflict or synergy; Based on the interaction between the at least two abnormal indicators, an optimization strategy is determined.

[0023] As described above, step one calls the data access interface to load the segmented road network spatial dataset generated above, which contains attribute fields of urban landscape quantitative indicators. This dataset exists in the form of a standard geospatial data file (such as GeoJSON or Shapefile), and the reading process includes parsing its geometric information and extended attribute tables. This ensures that subsequent advanced analysis is based on a structured data source that has achieved "data-space" integration. It provides input for examining regional issues from a macro perspective and prepares the data for moving from single-point assessment to area diagnosis. In steps two and three, the attribute tables of each road segment element in the dataset are traversed. For each road segment, the system compares the values ​​of multiple urban landscape quantitative indicators recorded in the attribute table, such as green view rate and road motorization level, with professional thresholds set independently for each indicator in advance. These thresholds are critical values ​​set based on planning standards, specifications, or empirical values. Only when at least two or more indicator values ​​in the attributes of the same road segment element simultaneously exceed their respective preset thresholds is the road segment determined to meet the "multi-indicator anomaly condition". Instead of viewing the merits or demerits of individual indicators in isolation, this approach identifies "complex problem sections" where multiple dimensions simultaneously exhibit deficiencies. This helps decision-makers prioritize key spaces with more complex problems requiring systematic intervention from massive amounts of data, improving analytical efficiency and problem targeting. Step four, after determining the existence of multiple indicator anomalies, accesses a pre-built "indicator interaction rule base." This rule base is built based on professional knowledge in urban planning and landscape design, encoding common interaction patterns between different indicators in the form of logical rules. For example, the rule base might define that "green view rate" and "road motorization level" often have a "conflicting" relationship in spatial allocation; while "green view rate" and "sky openness" may also exhibit a "conflicting" relationship in some cases due to canopy shading, or form a "synergistic" relationship with "interface diversity" in specific commercial street scenarios. Based on the specific combination of indicators currently exhibiting anomalies, the corresponding rules are queried and applied to qualitatively determine whether the interaction relationship between indicators is "conflicting" or "synergistic." Step five, based on the "conflicting" or "synergistic" relationship determined in the previous step, proceeds to different strategy generation paths. If the relationship is determined to be "synergistic," the strategy focuses on finding comprehensive solutions that can simultaneously improve multiple indicators. If the relationship is determined to be "conflicting," the strategy needs to consider how to make trade-offs and choices within limited space or resource constraints, providing a clear framework for subsequent, more refined prioritization.

[0024] In this embodiment, multiple indicators are presented in a remarkably simple, juxtaposed manner. By introducing "indicator interaction rules," it forces the terminal to consider the correlation between issues and make a professional judgment on whether these correlations constitute "conflict" or "synergy." In urban planning, simultaneously optimizing all objectives is often impractical; the core lies in trade-offs. This solution directly addresses this core challenge by identifying "conflicting" relationships and determining the strategic direction for "trade-offs." It clearly informs decision-makers that there are inherent contradictions among these objectives in this road segment, and subsequent decisions must be based on conscious trade-offs.

[0025] In one specific embodiment, suppose a road segment is identified during the analysis, with a "green view rate" of 8% (lower than the preset 15% threshold) and a "road motorization level" of 85% (higher than the preset 70% threshold) in its attribute table. Traditional methods would output two independent warnings in parallel: "green view rate too low" and "road motorization level too high". After reading the data for this road segment, it is confirmed that it meets the multi-indicator anomaly condition of "at least two indicators exceeding the threshold". A pre-set rule base for indicator interaction relationships is invoked. Based on professional knowledge, the rule base defines that in typical urban road spaces, there is often a spatial resource competition relationship between the two indicators, "green view rate" and "road motorization level," meaning that increasing greenery often requires compressing vehicle space or affecting traffic efficiency, and vice versa. Therefore, the interaction relationship between these two abnormal indicators is determined to be "conflict". Strategy orientation: Based on the determination of the "conflict" relationship, the optimization strategy for this road segment is determined to be "requiring trade-offs," rather than simply suggesting simultaneous improvement of both indicators. Improvement point one: From 'listing up' to 'revealing contradictions'. When faced with the above scenario, existing technologies can only mechanically list two isolated problems side by side, like pointing out that a piece of land simultaneously "lacks parks" and "lacks parking lots." The "relationship determination" step introduced in this application, however, is like further analyzing the direct competition between "building parks" and "building parking lots" in terms of spatial use. It reveals that behind the surface demands of "insufficient greening" and "excessive traffic load," the core issue is the fundamental contradiction in the allocation of limited road space resources between ecological and traffic functions. This intelligent identification of the inherent conflict in the problem is a prerequisite for any reasonable planning and prioritization. Improvement point two: It constructs a crucial "problem framework" for professional decision-making. After determining a "conflicting" relationship, the determined strategic direction essentially constructs a correct problem framework for decision-makers. It clearly states that "the best of both worlds" cannot be expected here; the trade-off of "which comes first" must be directly addressed. This directly elevates the decision focus from "whether to do greening" or "whether to maintain traffic flow" to a more professional and practical level of decision-making: "how to optimize greening as much as possible under the constraint of prioritizing traffic efficiency." This fundamentally changes the decision-making confusion caused by traditional parallel early warning systems. The third improvement is the "value leap" achieved in analytical conclusions. Through this process, the output is no longer a "problem list" that might cause internal controversy, but an "analysis briefing" containing in-depth diagnosis and clear decision-making pathways. For the aforementioned traffic bottleneck roads, the conclusion is essentially: "The core contradiction is a conflict in spatial allocation; it is recommended to seek greening improvement solutions within the framework of prioritizing traffic congestion relief." This seamlessly connects technical analysis results with the real decision-making pressure and professional knowledge faced by managers, enhancing the operability and decision-support value of the analysis results.

[0026] In one embodiment, the step of determining the optimization strategy based on the interaction relationship between the at least two abnormal indicators includes: If the interaction relationship is determined to be conflicting, the spatial functional attributes of the target road segment that meet the multi-index anomaly conditions are identified, and priority ranking is performed among the at least two anomaly indicators. If the interaction is determined to be synergistic, an optimization strategy is generated to simultaneously improve the at least two abnormal indicators.

[0027] As mentioned above, this step is initiated after step one determines that there is a conflict between the abnormal indicators. Its primary operation is to identify the spatial functional attributes of the target road segment. This attribute is predefined and assigned to the road network's metadata, such as traffic-dominated, commercial / service-oriented, historically protected, or residential. Identification can be done by reading preset fields in the road segment attribute table or by analyzing surrounding land use, road grade, etc. Subsequently, based on this functional attribute, the associated priority ranking logic is invoked. This logic is essentially a decision rule that clearly specifies which of several conflicting indicators should be given higher priority for improvement on such functional road segments. For example, for traffic-dominated road segments, the rule might stipulate that when road mobility conflicts with green visibility, ensuring or improving traffic efficiency is the priority objective. This transforms the general "conflict" judgment into a "priority" conclusion with clear action guidance. It ensures that the given strategic recommendations are not vague but closely linked to the fundamental positioning of the road segment and the primary goals of urban management. This ensures that the strategy output is based on professional planning logic from the outset, filtering out invalid options that do not align with the functional positioning for decision-makers. Step two proceeds to this path once a synergistic relationship is determined between the abnormal indicators. A synergistic relationship means that measures to improve one indicator may also have a positive impact on the other, or at least not a negative impact. Based on a pre-built knowledge base or algorithm model, comprehensive strategy recommendations aimed at simultaneously improving all abnormal indicators will be generated. This strategy may point to a composite spatial design or management scheme. For example, if green view rate and sky visibility are determined to be synergistically improveable, it might be recommended to "plant street trees with moderate crown width and high branching points," which can increase green coverage while reducing excessive obstruction of the sky view.

[0028] In this embodiment, by "identifying spatial functional attributes" and applying differentiated processing accordingly, the fundamental principle of "classification guidance and precise policy implementation" in urban planning is encoded into the algorithm logic. Instead of providing uniform advice for all "green view rate-motorization conflict" issues, it differentiates between prioritizing smooth traffic flow on main roads and prioritizing environmental improvement in residential streets and alleys. This reflects a profound respect for the complexity and functional heterogeneity of urban space. Facing urban problems sometimes requires difficult trade-offs and choices, while other times it involves clever collaboration and win-win outcomes. This distinction not only makes the strategy generation logic clear and reliable but also educates and guides decision-makers to view multi-indicator evaluation results with a more systematic and dialectical mindset.

[0029] In one embodiment, if the interaction relationship is determined to be conflicting, the spatial functional attributes of the target road segment that meet the multi-index anomaly conditions are identified, and priority is ranked among the at least two anomaly indicators. The steps include: Based on the type of the spatial function attribute of the target road segment, invoke the preset priority rule bound to that type; If the type is traffic-driven, and the at least two abnormal indicators include road motorization level and green view rate, then the preset priority rule specifies that the road motorization level indicator should be prioritized as the dominant optimization target. If the type is historical protection type, and the at least two abnormal indicators include green view rate and street enclosure degree, then the preset priority rule specifies that the street enclosure degree indicator should be prioritized as the main optimization target.

[0030] As described above, step one is the priority ranking decision engine. First, based on the "spatial function attribute" field recorded or analyzed in the target road segment attribute table, its type is determined, such as traffic-oriented or historical preservation type. Then, a structured "priority rule base" is accessed. This rule base is a pre-built knowledge base that associates "road segment function type" with "priority instructions under specific indicator conflict combinations." The retrieval process involves matching the current road segment's type with specific abnormal indicator combinations to retrieve the unique or most applicable priority rule from the base. This process transforms the ambiguous "ranking" requirement into a deterministic rule query and execution, achieving configurability, transparency, and professionalization of the decision logic. It solidifies expert knowledge in urban planning, such as "prioritizing smooth traffic flow on arterial roads," into clear rules within the system, ensuring that every priority ranking is traceable and systematic, rather than based on a black-box model or ad-hoc calculations. Steps two and three involve the specific application and output of these rules. For traffic-dominated road sections: When the query conditions in the rule base are traffic-dominated and the abnormal indicators include road mobility and green view rate, and these conditions are met, the executed rule instruction is clear: prioritizing the improvement of road mobility indicators is designated as the primary optimization objective. This does not mean completely abandoning greening, but rather establishing that when resources are limited or space conflicts exist, solving traffic efficiency is the top priority, and greening improvements must be sought within this constraint. For historically protected road sections: When the type is historically protected and the abnormal indicators include green view rate and street enclosure, the executed rule instruction is: prioritizing the improvement of street enclosure indicators is designated as the primary optimization objective. This rule is based on the core principle of historical preservation—maintaining the spatial scale and interface integrity of streets is the cornerstone of protecting their historical value. Compared to adding greenery that may disrupt the original spatial proportions, protecting the sense of enclosure has a higher priority. It provides clear, unambiguous, and highly contextualized action guidance. It transforms complex professional trade-offs into a single primary objective instruction that both machines and users can directly understand. For example, when faced with the aforementioned traffic bottlenecks, instead of simply stating "there are two problems here," it is clearly pointed out that "at this location, priority should be given to ensuring and improving traffic capacity." This provides designers and managers with an unshakeable decision-making foothold and a starting point for all subsequent work, effectively ending potential disputes about objectives between different departments or professions.

[0031] This embodiment addresses the practical problem of "how to implement priority ranking in a system," a combination of engineering and professional expertise. By constructing a mapping rule base of "functional type - indicator combination - priority objective," it transforms ranking decisions that rely on expert experience and value judgments into logical queries that can be executed automatically, accurately, and consistently by a computer. In classic urban planning dilemmas such as transportation versus ecology, and protection versus renewal, this embodiment provides pre-defined "standard answers" in the form of rules. For example, it explicitly stipulates that "efficiency takes precedence over greening" in key transportation corridors, and "preservation of historical character is more important than increasing green space" in historic districts. This is not rigid, but rather pre-encodes long-term, practically validated, and consensus-based professional value judgments into the system, ensuring that when facing these high-value or highly sensitive areas, technical analysis can automatically align with the highest-level strategic objectives, avoiding the output of misleading recommendations that contradict core values.

[0032] In one embodiment, before calculating the street green view rate based on the identification results of the vegetation type elements, the method further includes: For the same street view collection point, acquire multiple street view images of the same street view collection point within a preset time window; For each of the street view images, vegetation feature recognition is performed to distinguish ecologically functional vegetation areas from the identified vegetation features. The percentage of pixels representing ecologically functional vegetation identified at the collection point within the time window was statistically analyzed. Based on the pixel percentage at each time point, the statistical median or the average value after removing outliers is calculated as the green visibility rate of the collection point.

[0033] As described above, Step One retrieves all valid historical images taken at the same location within a preset time period, such as the past 24 months, from the street view image database, based on the spatial coordinates of a single street view acquisition point. This time window is designed to cover seasonal cycles and include a sufficient number of sampling points. This expands the analytical basis from a "snapshot" of a single point in time to an "observation sequence" that reflects temporal changes. This provides the necessary data foundation for subsequent filtering out random factors and capturing stable states. Step Two, after identifying "vegetation-related element" pixels through a large model, performs a more refined secondary classification or filtering. This can be achieved through machine learning models based on high-level features in computer vision, such as specific texture patterns and near-infrared spectral reflectance characteristics, or through feature matching and comparison with known ecologically functional vegetation sample libraries, such as typical morphologies of different tree species and grasslands. Its core is to exclude man-made or temporary objects that, while visually green, do not actually possess ecological functions or belong to stable greening. This directly corrects the "misjudgments" of general semantic segmentation models in professional applications. This ensures that subsequent calculations are based on truly ecologically valuable and relatively stable green entities, rather than green billboards, painted walls, or temporarily placed potted plants, thus fundamentally enhancing the professional purity and practical significance of the "green view rate" indicator. Step three, for each historical image within the time window, after the fine differentiation in step two, calculates the percentage of pixels identified as ecologically functional vegetation within the effective analysis area of ​​that image. This step generates an independent, professionally defined temporary green view rate value for each historical image. This transforms multi-temporal image data into a numerical sequence suitable for time-series analysis. This sequence visually demonstrates the specific fluctuations in vegetation cover at the collection point over time, providing direct input for the next step of data integration and denoising. Step four statistically integrates the series of all temporary green view rate values ​​generated in step three. By employing a statistical median or first removing outliers (e.g., using box plots to identify and exclude abnormally high or low values ​​caused by extreme weather, temporary obstruction, or identification errors), and then calculating the average, a single value representing the stable greening level of the sampling point is synthesized. This robust statistical method effectively resists the interference of random factors in a single image. Whether it's a temporary drop in green coverage due to winter leaf fall or temporary obstruction by a large truck, the extreme impact is significantly mitigated by the median or denoised mean algorithm. The resulting "calculated green visibility rate" more reliably reflects the inherent greening level of the location under normal conditions, significantly improving the stability, reliability, and comparability of the assessment results.

[0034] In this embodiment, by distinguishing between ecologically functional vegetation, it ensures strict alignment between the measured object and concepts such as "ecological benefits" and "long-term greening" that are of concern in the planning field. This solves the fundamental problem of indicator distortion caused by semantic generalization in existing technologies, making "green view rate" a visually descriptive indicator. By introducing multi-temporal analysis and combining robust statistics, this embodiment installs a "stabilizer" for the evaluation system. It acknowledges and actively handles the dynamism and uncertainty in the real world, rather than pretending they don't exist as in existing technologies. This allows the final output green view rate value to represent a location's "typical state" or "baseline level" over time, rather than an instantaneous reading influenced by random factors. This greatly enhances the scientific rigor and credibility of the indicator for long-term trend monitoring and fair comparisons between different periods or regions.

[0035] For example, during the long cycle of large-scale urban renewal or infrastructure construction, which may last 2-5 years, an area may be occupied by construction fences, temporary roads, construction machinery, and exposed soil for an extended period. At this time, a fundamental dilemma arises: Distortion in identification: Semantic segmentation models may identify fences as "building interfaces" or unknown categories, significantly distorting the "street enclosure degree"; temporary roads and muddy ground affect the calculation of "pedestrian walkway ratio" and "road motorization level." The paradox of stability value calculation: During the years-long construction period, all historical images are in an "abnormal" state. The calculated "stable green view rate" may be close to 0%, but this reflects the stability of the "construction state," not the stability of the "streetscape." The terminal will continuously mark this potentially transformed area as a long-term, serious "problem area," potentially triggering optimization strategies incorrectly, when in fact the largest "optimization" (the construction itself) is being implemented here.

[0036] Based on the existing problems, a solution may include: after the step of acquiring street view image data of the target geographic area, the method further includes: Acquire spatial range data within the target geographic area that is in a special state, including construction status; Based on the spatial range data, determine whether the spatial location of each street view image acquisition point falls within the spatial range of the special state. If it is determined that the data falls within the spatial range of the special state, then the street view image data originating from that collection point is marked differently from the normal state.

[0037] As mentioned above, at the start of the analysis process, geospatial data reflecting the temporary or special states of urban space are acquired in parallel. The most typical examples are "Construction Red Line Data for Projects Under Construction" or "Urban Renewal Project Scope Data" released by urban planning and management departments. This data exists in the form of polygonal vector layers, with clear start and end time attributes, authoritatively defining which areas are under construction during a specific time period. After acquiring the coordinates of street view collection points, these coordinates are first compared with the aforementioned "Special State Spatial Data" to determine their spatial location. If a collection point is located within any valid special state area, it is marked as a "Special State Point." For street view data marked as "Special State Points," a different processing strategy is adopted than for regular points. For example, Option A: Suspend Evaluation: Directly skip or postpone the calculation of indicators and all subsequent analyses of the street view image at that point to avoid producing invalid or misleading results. Option B output: The conventional calculation is still performed, but when generating the final visualization layer and warning information, the point or the road segment to which it belongs is specially marked, such as with a semi-transparent or dashed border, and a note is added stating "This area is under construction and the data is for reference only", thereby isolating it from interference with the overall assessment conclusion.

[0038] Reference Figure 3 This application also provides a large-model-based urban landscape semantic segmentation system, including: Street view image acquisition module 1 is used to acquire street view image data of the target geographic area; The semantic segmentation and parsing module 2 is used to input the street view image data into a preset image semantic segmentation model, and to parse the street view image data through the image semantic segmentation model to identify multiple predefined environmental element categories contained in each image; Landscape index calculation module 3 is used to calculate urban landscape quantitative indicators based on the identification results of the environmental element categories. Spatial data association module 4 is used to associate the calculated values ​​of urban landscape quantitative indicators with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. Visualization layer generation module 5 is used to generate a road network spatial visualization layer based on the associated results, using color grading to represent the high and low values ​​of the index. The early warning information generation module 6 is used to identify spatial road segments or nodes in the visualization layer that meet the early warning conditions and generate prompt information based on preset indicator thresholds.

[0039] As described above, it is understood that each component of the large-model-based urban landscape semantic segmentation system proposed in this application can realize the function of any of the large-model-based urban landscape semantic segmentation methods described above, and the specific structure will not be repeated.

[0040] Reference Figure 4 This application also provides a computer device, which may be a server, and its internal structure may be as follows: Figure 4 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores monitoring data and other data. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a semantic segmentation method for urban landscapes based on a large model.

[0041] The processor described above executes the urban landscape semantic segmentation method based on a large model, including: acquiring street view image data of the target geographic area; inputting the street view image data into a preset image semantic segmentation model, parsing the street view image data through the image semantic segmentation model, and identifying multiple predefined environmental element categories contained in each image; calculating urban landscape quantitative indicators based on the identification results of the environmental element categories; associating the calculated urban landscape quantitative indicator values ​​with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographic area; generating a road network spatial visualization layer that represents the level of indicator values ​​in a color-coded manner based on the association results; and identifying spatial road segments or nodes in the visualization layer that meet the warning conditions and generating prompt information according to preset indicator thresholds.

[0042] One embodiment of this application also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements a large-model-based urban landscape semantic segmentation method, including the following steps: acquiring street view image data of a target geographic area; inputting the street view image data into a preset image semantic segmentation model, parsing the street view image data through the image semantic segmentation model, and identifying multiple predefined environmental element categories contained in each image; calculating urban landscape quantitative indicators based on the identification results of the environmental element categories; associating the calculated urban landscape quantitative indicator values ​​with corresponding road segments or nodes in the road network spatial data corresponding to the target geographic area; generating a road network spatial visualization layer that represents the level of indicator values ​​using a color grading method based on the association results; and identifying spatial road segments or nodes in the visualization layer that meet the warning conditions and generating prompt information according to preset indicator thresholds.

[0043] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0044] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0045] The above description is only a preferred embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural changes made based on the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A semantic segmentation method for urban landscapes based on a large model, characterized in that, The method includes: Acquire street view image data of the target geographic area; The street view image data is input into a preset image semantic segmentation model. The image semantic segmentation model is used to parse the street view image data and identify the various predefined environmental element categories contained in each image. Based on the identification results of the environmental element categories, urban landscape quantitative indicators are calculated; The calculated values ​​of the urban landscape quantitative indicators are associated with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. Based on the correlation results, a road network spatial visualization layer is generated, which uses color grading to represent the high and low values ​​of the index. Based on preset indicator thresholds, spatial road segments or nodes in the visualization layer that meet the warning conditions are identified and prompt information is generated.

2. The urban landscape semantic segmentation method based on a large model according to claim 1, characterized in that, The environmental element categories include vegetation elements, sky elements, building interface elements, pedestrian facility elements, motorized transportation elements, and diversity representation elements; the step of calculating urban landscape quantitative indicators based on the identification results of the environmental element categories includes: Based on the identification results of the vegetation elements, the street green view rate is calculated; Calculate the sky openness based on the recognition results of the aforementioned sky-related elements; Calculate the street enclosure degree based on the identification results of the building interface elements; Based on the identification results of the pedestrian facility elements, the proportion of pedestrian walkways is calculated; Based on the identification results of the aforementioned motorized traffic elements, the degree of road motorization is calculated; Based on the identification results of the diversity characterization elements, the index values ​​used to characterize interface diversity are determined.

3. The urban landscape semantic segmentation method based on a large model according to claim 2, characterized in that, The step of associating the calculated urban landscape quantitative index values ​​with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area includes: Acquire spatial location data of collection points that are bound to each street view image; Obtain the road network centerline data for the target geographical area; Spatial matching is performed between the spatial location data of each collection point and the road network centerline data to determine the target line element corresponding to each collection point on the road network centerline data. The target line element is used to generate a segmented road network with attached landscape indicators. The calculated values ​​of all urban landscape quantitative indicators corresponding to each street view image are used as a set of attribute information and assigned to the target line element corresponding to the street view image collection point. Based on the attribute assignment results, a new segmented road network spatial dataset is generated, which includes attribute fields of urban landscape quantitative indicators. The geometric shape of each road segment element is defined by the road network centerline data, and the attribute table of each road segment element in the segmented road network spatial dataset records the landscape indicator values ​​of the associated collection points.

4. The urban landscape semantic segmentation method based on a large model according to claim 3, characterized in that, Prior to the step of generating a road network spatial visualization layer that represents the level of index values ​​using a color-grading method, the method further includes: Read the new segmented road network spatial dataset; Determine whether the new segmented road network spatial dataset meets the multi-indicator anomaly conditions; When the attributes of a new segmented road network spatial dataset contain records of at least two urban landscape quantitative indicators whose values ​​exceed their respective preset thresholds, it is determined that the multi-indicator abnormality condition is met. Based on the judgment result, a preset indicator interaction relationship rule is invoked, and the interaction relationship between the at least two abnormal indicators is determined according to the indicator interaction relationship rule, wherein the interaction relationship includes conflict or synergy; Based on the interaction between the at least two abnormal indicators, an optimization strategy is determined.

5. The urban landscape semantic segmentation method based on a large model according to claim 4, characterized in that, The step of determining the optimization strategy based on the interaction relationship between the at least two abnormal indicators includes: If the interaction relationship is determined to be conflicting, the spatial functional attributes of the target road segment that meet the multi-index anomaly conditions are identified, and priority ranking is performed among the at least two anomaly indicators. If the interaction is determined to be synergistic, an optimization strategy is generated to simultaneously improve the at least two abnormal indicators.

6. The urban landscape semantic segmentation method based on a large model according to claim 5, characterized in that, If the interaction relationship is determined to be conflicting, then the spatial functional attributes of the target road segment that meet the multi-index anomaly conditions are identified, and priority is ranked among the at least two anomaly indicators. The steps include: Based on the type of the spatial function attribute of the target road segment, invoke the preset priority rule bound to that type; If the type is traffic-driven, and the at least two abnormal indicators include road motorization level and green view rate, then the preset priority rule specifies that the road motorization level indicator should be prioritized as the dominant optimization target. If the type is historical protection type, and the at least two abnormal indicators include green view rate and street enclosure degree, then the preset priority rule specifies that the street enclosure degree indicator should be prioritized as the main optimization target.

7. The urban landscape semantic segmentation method based on a large model according to claim 2, characterized in that, Before calculating the street green view rate based on the identification results of the vegetation type elements, the method further includes: For the same street view collection point, acquire multiple street view images of the same street view collection point within a preset time window; For each of the street view images, vegetation feature recognition is performed to distinguish ecologically functional vegetation areas from the identified vegetation features. The percentage of pixels representing ecologically functional vegetation identified at the collection point within the time window was statistically analyzed. Based on the pixel percentage at each time point, the statistical median of the set of pixel percentage values ​​or the average value after removing outliers is calculated as the green visibility rate of that collection point.

8. A semantic segmentation system for urban landscapes based on a large model, characterized in that, include: The street view image acquisition module is used to acquire street view image data of the target geographic area; The semantic segmentation and parsing module is used to input the street view image data into a preset image semantic segmentation model, and to parse the street view image data through the image semantic segmentation model to identify multiple predefined environmental element categories contained in each image. The landscape index calculation module is used to calculate urban landscape quantitative indicators based on the identification results of the environmental element categories. The spatial data association module is used to associate the calculated values ​​of urban landscape quantitative indicators with the corresponding road segments or nodes in the road network spatial data corresponding to the target geographical area. The visualization layer generation module is used to generate a road network spatial visualization layer based on the associated results, using color grading to represent the level of index values. The early warning information generation module is used to identify spatial road segments or nodes in the visualization layer that meet the early warning conditions based on preset indicator thresholds and generate prompt information.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images