Intelligent evaluation method and system for explainable street view environment quality

By generating a temporal sequence of street view images and dividing them into stable environment, periodic management, and temporary interference elements, the problem of insufficient stability and interpretability in street view image evaluation is solved, and more accurate environmental quality assessment and governance guidance are achieved.

CN123071920APending Publication Date: 2026-11-10XIAN UNIV OF SCI & TECH
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611232137.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing environmental quality assessment methods based on street view images are easily affected by temporary factors, resulting in insufficient stability and interpretability of the assessment results, and making it difficult to accurately distinguish between long-term environmental quality and temporary disturbances.

Method used

By acquiring current and historical street view images of the target sampling points, a co-located street view image group is formed. Spatial alignment and street view element recognition are then performed to form a temporal sequence of street view elements. Based on the temporal stability state, the street view elements are divided into stable environmental elements, periodic management elements, and temporary interference elements. Evaluation results are generated and the impact of temporary interference is suppressed.

Benefits of technology

It improves the stability and objectivity of street view environmental quality assessment, can accurately distinguish between long-term environmental quality and temporary disturbances, and outputs interpretation results to facilitate subsequent governance.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123071920A_ABST
    Figure CN123071920A_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of data processing, in particular to an intelligent evaluation method and system for explainable street environment quality, the method comprising: obtaining a current street image and a corresponding historical street image of a target sampling point, to form a same-position street image group; after spatial alignment and street element identification, associating street elements at the same spatial position in time to form a street element time sequence; dividing stable environment elements, periodic management elements and temporary interference elements according to time stability, and generating a street environment quality evaluation result and an explanation result accordingly; and the system is used to execute the above method. The present application divides street elements, thereby distinguishing long-term existing street environment content, repeatedly appearing management order content and occasional temporary visual content, reducing the interference of single collection state on the street environment quality evaluation result, and improving the stability and objectivity of the evaluation result.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to an intelligent evaluation method and system for interpreting street view environmental quality. Background Technology

[0002] With the advancement of refined urban governance and the improvement of street space quality, street environment quality assessment based on street view images is increasingly being applied to scenarios such as road environment inspection, pedestrian space assessment, street renewal analysis, and urban public space governance. Street view images can record a variety of visual information such as roads, pedestrian spaces, green spaces, building facades, traffic facilities, and temporary occupants. Compared with manual on-site inspections, they have the advantages of wide coverage, high data collection efficiency, and ease of generating spatial evaluation results.

[0003] Existing environmental quality assessment methods based on street view images typically use a single street view image as the assessment object. They extract street view elements through image recognition, semantic segmentation, object detection, or machine learning models, and generate street view environmental quality assessment results based on the category, area proportion, spatial location, or visual characteristics of the street view elements. Although this can improve the automation level of street view environmental assessment to some extent, street view images themselves reflect the visible state at a specific acquisition time. The images may simultaneously contain long-term road spatial structures, recurring management order states, and occasional temporary visual content.

[0004] In real-world data collection environments, the same street space may be affected by factors such as temporarily parked vehicles, pedestrian obstructions, short-term construction barriers, temporary objects, obstructions from the camera, and weather changes at different collection times. Some of these factors do not necessarily reflect the long-term environmental quality of the street space itself. If all visible elements in a single street view image are directly used as the evaluation basis, the evaluation results are easily affected by occasional events, making it difficult to accurately attribute low scores to long-term spatial environmental problems, periodic management problems, or temporary visual interference problems. This affects the stability, interpretability, and remediation direction of the street view environmental quality evaluation results.

[0005] To address this, a smart evaluation method and system for interpreting street view environmental quality is proposed. Summary of the Invention

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent evaluation method for interpretable street view environmental quality includes: S1, acquiring the current street view image of the target sampling point, and acquiring historical street view images according to the street spatial location corresponding to the target sampling point, and forming a co-located street view image group by combining the current street view image and the historical street view image; S2, performing spatial alignment and street view element recognition on the street view images in the co-located street view image group to obtain street view element information at multiple acquisition times; S3, performing temporal correlation on street view elements located at the same street spatial location at multiple acquisition times according to the street view element information to form a street view element temporal sequence; S4, determining the temporal stable state of the street view elements according to the street view element temporal sequence, and classifying the street view elements into stable environmental elements, periodic management elements, and temporary interference elements according to the temporal stable state; S5, generating street view environmental quality evaluation results based on stable environmental elements and periodic management elements, and suppressing the evaluation contribution of temporary interference elements to the street view environmental quality evaluation results; S6, outputting the street view environmental quality evaluation results and the corresponding interpretation results.

[0007] A system for executing the intelligent evaluation method for interpretable street view environmental quality as described in claim 1 includes an image acquisition module, an alignment recognition module, a temporal correlation module, a state classification module, an evaluation processing module, and an interpretation output module. The image acquisition module is used to acquire street view images and compose co-located street view image groups. The alignment recognition module is used to perform spatial alignment of street view images and street view element recognition. The temporal correlation module is used to perform temporal correlation of street view elements to form a temporal sequence of street view elements. The state classification module is used to determine the temporal stable state of street view elements and classify street view elements according to the temporal stable state. The evaluation processing module is used to generate street view environmental quality evaluation results. The interpretation output module is used to output interpretation results.

[0008] As a preferred technical solution of the present invention, the acquisition of historical street view images includes mapping the target sampling point to road space data to obtain the road number, road mileage location, road side and shooting direction corresponding to the target sampling point, and then selecting historical street view images from the historical street view image library that are in the same road space range as the target sampling point, have similar road mileage locations, consistent road sides and similar shooting directions, so that the historical street view images and the current street view images jointly reflect the same street space area; before forming a group of identical street view images, the selected historical street view images are deduplicated, including identifying street view images that were repeatedly collected in a short period of time based on the collection time, collection location, road mileage location and image similarity of the historical street view images, and retaining the street view images with higher image clarity among the repeatedly collected street view images.

[0009] As a preferred technical solution of the present invention, the spatial alignment includes using the current street view image as a reference image, extracting static spatial features from the current street view image and historical street view images, and establishing a spatial mapping relationship from historical street view images to the current street view image based on the static spatial features, transforming the road area, pedestrian space area, building interface area, and green area in the historical street view image to a unified image coordinate system; the street view element recognition includes performing pixel-level category recognition on the spatially aligned street view image to obtain a category label map, extracting connected components from the pixel areas of the same category and spatially connected in the category label map to obtain candidate street view element areas, deleting candidate street view element areas with an area lower than the noise area condition, and merging candidate street view element areas of the same category and spatially adjacent to obtain street view element information; the street view element information includes street view element category, street view element area, street view element area ratio, center position of street view element area, and road functional area to which the street view element belongs, wherein the road functional area to which the street view element belongs is determined based on the positional relationship between the street view element area and the pedestrian space, vehicle space, road edge space, building interface space, or green space in the road spatial data.

[0010] As a preferred technical solution of the present invention, the time correlation includes taking street scene elements of the same category under different collection times as candidate correlation objects, matching them according to the degree of regional overlap, regional center distance and consistency of the functional area of ​​the road to which they belong, and adding the successfully matched candidate correlation objects to the time phase sequence of the same street scene element.

[0011] As a preferred technical solution of the present invention, determining the temporal stable state of street view elements includes: determining the occurrence and coverage state based on the frequency of occurrence of street view elements at multiple acquisition times; determining the location stable state based on the changes in the regional center position of street view elements at multiple acquisition times; determining the area stable state based on the changes in the area proportion of street view elements at multiple acquisition times; and determining the discrete repetition state based on the intermittent repetition of street view elements at multiple acquisition times. Stable environmental elements refer to street view elements that continuously appear at multiple acquisition times, are mainly concentrated in the same street space, and whose area proportion changes are within the allowable range. Periodic management elements refer to street view elements that do not meet the classification conditions of stable environmental elements but repeatedly appear at multiple acquisition times and are mainly concentrated in the same street functional area. Temporary interference elements refer to street view elements that have a low frequency of occurrence and whose location or area is unstable.

[0012] As a preferred technical solution of the present invention, the step of generating street view environmental quality evaluation results and outputting corresponding interpretation results includes: determining the evaluation contribution method of street view elements to the evaluation dimensions according to the classification type of street view elements, enabling stable environmental elements to participate in the evaluation as the basis for long-term environmental quality evaluation, enabling periodic management elements to participate in the evaluation as the basis for management order evaluation, and enabling temporary interference elements not to be used as the main basis for deduction. The interpretation results output the low-scoring evaluation dimensions, the street view elements that lead to the reduction of evaluation results, the road functional area to which the street view elements belong, the classification type of the street view elements, and the explanation of temporary interference.

[0013] Compared with the prior art, the beneficial effects that this invention can achieve are: 1. This invention acquires the current street view image of the target sampling point and the historical street view image corresponding to the target sampling point in space, forming a co-located street view image group. It then performs time-correlation on street view elements located in the same street space at multiple acquisition times to form a street view element temporal sequence. This makes the street view elements no longer just static recognition results in a single image, but have occurrence, spatial, and change states across acquisition time. After determining the temporal stable state based on the street view element temporal sequence, the street view elements can be divided into stable environmental elements, periodic management elements, and temporary interference elements. This distinguishes long-term street environment content, recurring management order content, and occasional temporary visual content, reducing the interference of a single acquisition state on the street view environment quality evaluation results and improving the stability and objectivity of the evaluation results.

[0014] 2. This invention generates streetscape environmental quality evaluation results based on stable environmental elements and periodic management elements, and suppresses the contribution of temporary interference elements to the evaluation results. This allows long-term existing streetscape elements with stable spatial performance to serve as the basis for long-term environmental quality evaluation, and streetscape elements that appear repeatedly and are concentrated in the same street functional area to serve as the basis for management order evaluation. Streetscape elements with low frequency of occurrence and unstable spatial performance are not used as the main basis for deduction. As a result, the output interpretation results can not only give the streetscape elements that lead to the reduction of evaluation results, but also give the corresponding classification type of the streetscape elements, so that the reasons for low scores can be attributed to stable environmental problems, periodic management problems, or temporary interference explanations, which facilitates subsequent street renewal, environmental quality diagnosis, and refined management. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a schematic diagram of the system composition of the present invention; Figure 3 This is a schematic diagram illustrating the street scene element attribution judgment and evaluation contribution control of the present invention. Detailed Implementation

[0016] To make the technical means, creative features, objectives, and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention.

[0017] Example: Figure 1 , Figure 2 and Figure 3 As shown, an intelligent evaluation method for interpreting street view environmental quality is proposed. Based on street view images of the same street location at multiple acquisition times, the method processes the street view elements in the images into a temporal sequence of street view elements with time states. The method then determines the classification of street view elements based on their persistence, location, and area states at multiple acquisition times, and controls the contribution of street view elements to the evaluation results of street view environmental quality based on their classification.

[0018] In this embodiment, the street view images all include the acquisition location, acquisition time, shooting direction, and camera parameters. The acquisition location is latitude and longitude coordinates or road mileage coordinates, the shooting direction is the azimuth angle corresponding to the street view image, and the camera parameters include image resolution, lens field of view, and image projection method. The road spatial data includes the road centerline, road boundary line, pedestrian space boundary, road side, and road mileage information. By matching the street view images with the road spatial data, street view images acquired at different times can be mapped to the same street spatial location.

[0019] First, the current street view image of the target sampling point is acquired. Based on the sampling position, shooting direction, and road spatial position of the target sampling point, historical street view images are retrieved from the historical street view image database. When retrieving historical street view images, the target sampling point is projected onto the road centerline to obtain the road number, road mileage position, and road side corresponding to the target sampling point. The acquisition position of the historical street view image is also projected onto the road centerline to obtain the road number, road mileage position, and road side corresponding to the historical street view image. Historical street view images that are located on the same road as the target sampling point, have a road mileage difference that meets the preset distance range, have the same road side, and have a shooting direction difference that meets the preset angle range are retained. The current street view image and the retained historical street view images form a co-located street view image group.

[0020] When forming a co-located street view image group, multiple similar street view images collected consecutively in the same month or the same collection batch are deduplicated. Specifically, the collection time interval, road mileage difference, and image similarity of adjacent historical street view images are calculated. When the collection time interval is small, the road mileage difference is small, and the image similarity is high, only the historical street view image with higher image clarity is retained. Through the above processing, the images in the co-located street view image group can cover different collection times, avoiding the bias caused by repeated image collection in a short period of time to subsequent time phase judgment.

[0021] Next, spatial alignment is performed on the street view images in the same street view image group. Spatial alignment includes field of view unification, road area localization, and image coordinate mapping. First, based on the shooting direction and camera field of view of the street view images, each street view image is cropped to a street observation field of view with the same orientation. Then, road boundary lines, building facade edge lines, and pedestrian space boundary lines are extracted from the street view images. These boundary lines are used as spatial alignment features. Finally, using the current street view image as the reference image, the spatial alignment features in the historical street view images are mapped to the image coordinates of the current street view image.

[0022] Specifically, in the spatial alignment process, static feature points are first extracted from the current street view image and the historical street view image. The static feature points are selected from road boundaries, building facades, fixed poles, fixed fences, curbstone boundaries, and road marking areas. For areas where motor vehicles, non-motor vehicles, pedestrians, and temporary objects are located, the weight is reduced or they are excluded during feature point matching to reduce the impact of dynamic targets on the alignment results. Subsequently, the static feature points in the current street view image and the historical street view image are matched, and mismatched points are removed by random sampling consistency method to obtain the spatial mapping relationship from the historical street view image to the current street view image. After completing the spatial mapping, the road area, pedestrian space area, building interface area, and green area in the historical street view image are transformed to the unified coordinate system of the current street view image.

[0023] After spatial alignment, street scene elements are identified for each street scene image. In this embodiment, a semantic segmentation model is used to output pixel-level category results, and street scene element regions are formed based on the semantic segmentation results. The input of the semantic segmentation model is the street scene image after the field of view is unified, and the output is a category label map with the same resolution as the street scene image. Each pixel in the category label map corresponds to a street scene category. Street scene categories include carriageway, sidewalk, green space, sky, building facade, wall, motor vehicle, non-motor vehicle, pedestrian, billboard, trash can, construction site fence, lighting facilities, and road obstacles.

[0024] After obtaining the category label map, connected component extraction is performed on the spatially connected pixel regions of the same category to generate candidate street view element regions. Candidate street view element regions with an area smaller than the preset noise area are deleted. Adjacent candidate street view element regions of the same category with a small boundary distance are merged. For candidate street view element regions that span multiple road functional areas, their main spatial location is determined based on the road functional area where the region's center of gravity is located. Thus, street view element information at each acquisition time is obtained. The street view element information includes street view element category, street view element region, street view element area ratio, center location of street view element region, and the road functional area to which the street view element belongs.

[0025] Among them, the proportion of street view element area is the ratio between the area of ​​street view element region and the area of ​​effective evaluation region of street view image. The effective evaluation region is the area in street view image used for street environment quality evaluation, excluding invalid areas at the sky edge, areas occluded by the collected vehicles, and areas with image edge distortion. The center position of the street view element region is the weighted center of the pixel coordinates within the street view element region. The weight can be determined according to the distance from the pixel within the street view element region to the region boundary, so that the influence of pixels inside the region on the center position is greater than that of edge noise pixels.

[0026] Subsequently, street view elements located in the same street space at multiple collection times are associated to form a street view element temporal sequence. Specifically, street view elements at each collection time are grouped according to category, and street view elements of the same category enter the same candidate association set. In the candidate association set, the spatial overlap, regional center distance, and consistency of the road functional area between street view element regions at different collection times are calculated. When the spatial overlap of two street view elements meets the overlap condition, or the regional center distance meets the distance condition and the road functional area is consistent, the two street view elements are determined to belong to the same street space location and are added to the same street view element temporal sequence.

[0027] During the association process, if multiple street view elements of the same category meet the association conditions at the same acquisition time, the street view element with the highest degree of overlap with the existing area in the time sequence is selected to be added to the time sequence. If the overlap of multiple candidate street view elements is close, the street view element with the smallest center distance is selected to be added to the time sequence. For street view elements that cannot be associated with the existing time sequence, a new street view element time sequence is established. Through the above processing, each street view element time sequence corresponds to the occurrence status of the same type of street view elements at the same street spatial location at multiple acquisition times.

[0028] After generating the temporal sequence of street view elements, the temporal stable state of the street view elements is calculated. The coverage rate of each street view element is calculated using the following formula: ; in: Indicates the first The coverage rate of the occurrence of a street view feature in a group of street view images at the same location, i.e., the coverage rate of the occurrence of the first street view feature in a group of street view images at the same location. The sufficiency of the appearance of each street view element at multiple acquisition times; Indicates the number of acquisition times in a group of street view images from the same location; Indicates the first Each street view element in the first The status at each collection time.

[0029] Suppose that the same street view image group contains The first collection time, the first Each street view element in the first The occurrence status at each collection time is recorded as follows: When the first Each street view element in the first When it occurs at a specific collection time, =1, when the first Each street view element in the first If it does not appear during a given collection time, =0.

[0030] No. The location stability of each street view element is calculated using the following formula: ; in: Indicates the first The positional stability of the first street scene element, i.e., the first Whether the street view elements are concentrated in the same street space location at multiple collection times; Indicates the first Each street view element in the first The center location of the region at each collection time; Indicates the first The average center position of a street view element at multiple acquisition times; This indicates the position offset adjustment parameter; This represents the smallest positive number that prevents the denominator from being zero.

[0031] No. The standard deviation of the area proportion of each street view element is calculated using the following formula: ; in: Indicates the first Standard deviation of the area proportion of each street view element at multiple collection times; Indicates the first Each street view element in the first Area percentage at each collection time; Indicates the first The average area percentage of each street view element across multiple data collection times.

[0032] No. The area fluctuation coefficient of each street view element is calculated according to the following formula: ; in: Indicates the first Area fluctuation coefficient of each street scene element; Indicates the first Standard deviation of the area proportion of each street view element at multiple collection times; Indicates the first The average area percentage of each street view element across multiple data collection times.

[0033] No. The area stability of each street scene element is calculated using the following formula: ; in: Indicates the first The area stability of the first street scene element, i.e., the first Whether the area of ​​each street view element is stable across multiple data collection times.

[0034] In addition, the first Several consecutive acquisition times of a street view element appearing in chronological order are merged into an occurrence cluster. The discrete repetition of each street view element is calculated using the following formula: ; in: Indicates the first The discrete repetition of the nth street view element, i.e., the nth Do the street scene elements exhibit the characteristic of intermittent repetition? Indicates the first The number of clusters of occurrence of each street scene element in chronological order; This indicates the number of reference clusters used to determine periodic repetition.

[0035] Next, based on the coverage rate Position stability Area stability and discrete repeatability The street scene elements are classified and categorized. Specifically, when the first... Coverage of street view elements Satisfying the continuity condition and position stability Satisfying the positional stability condition and area stability When the area stability condition is met, the first Street view elements are categorized into stable environmental elements; among them, the persistence condition refers to the high frequency of occurrence of street view elements across multiple acquisition times in the same street view image group; the location stability condition refers to the street view elements being mainly concentrated in the same street space across multiple acquisition times; and the area stability condition refers to the area proportion of street view elements remaining within an allowable range across multiple acquisition times. The street view elements did not meet the criteria for classifying stable environmental elements, but the discrete repetition was high. When the repetition condition is met and the occurrence location is within the same street functional area, the first... The street view elements are divided into periodically managed elements. The criteria for classifying stable environmental elements refer to the first... A street view element appears continuously at multiple acquisition times within the same street view image group, and is mainly located in the same street space at different acquisition times, while the area proportion changes within an allowable range. The repetition condition means that although the street view element does not exist continuously, it repeatedly appears at different acquisition times and is mainly concentrated in the same street functional area; when the... Coverage of street view elements Lower, and positional stability or area stability If the stability condition is not met, the first Each street view element is classified as a temporary disturbance element.

[0036] In a specific processing procedure, stable environmental elements include long-term existing road structures, greening structures, building interface structures, and the status of fixed facilities; periodic management elements include occupants, spillover objects, and temporary management objects that repeatedly appear in the same pedestrian space, vehicular space, or road edge space; and temporary interference elements include obstructions, pedestrians, temporarily parked vehicles, and short-term placed objects that only appear during a few collection times and whose locations are unstable. The above classification results are bound and stored with street view element categories, street view element areas, and street functional areas for subsequent evaluation contribution control and interpretation result generation.

[0037] In another embodiment, coverage will occur. Position stability Area stability and discrete repeatability The time-phase state vector is formed by combining the elements. The time-phase state vector is then input into a pre-trained classifier. The classifier outputs the category results of stable environmental elements, periodic management elements, and temporary disturbance elements. The training samples of the classifier are multi-temporal street scene element samples with labeled classifications. The input of the training samples is the time-phase state vector, and the output is the classification label. After training is completed, the classifier is used to automatically classify street scene elements in newly acquired street scene images.

[0038] Subsequently, street view environmental quality evaluation results are generated based on stable environmental factors and periodic management factors, and the contribution of temporary interference factors to the evaluation results is suppressed. Specifically, a basic influence quantity is configured for each street view factor. The basic influence quantity is used to represent the positive or negative impact of the street view factor on the evaluation dimension. Then, the evaluation contribution coefficient is determined according to the classification type of the street view factor, so that stable environmental factors participate in the evaluation according to the long-term environmental quality evaluation basis, periodic management factors participate in the evaluation according to the management order evaluation basis, and temporary interference factors are not used as the main basis for deduction.

[0039] No. The street view elements for the first The actual contribution of each evaluation dimension is calculated using the following formula: ; in: Indicates the first The street view elements for the first The actual contribution of each evaluation dimension; Indicates the first The evaluation contribution coefficient of each street view element; Indicates the first The street view elements for the first The basic influence of each evaluation dimension; When each street view element is a stable environmental element Take the first contribution value; the second When a street view element is a periodically managed element, Take the second contribution value; When a street view element is a temporary disturbance element Take the third contribution value; the first contribution value is greater than the second contribution value, the second contribution value is greater than the third contribution value, and the third contribution value is zero or close to zero.

[0040] The evaluation dimensions include ecological comfort, pedestrian friendliness, spatial order, visual safety, and street vitality. For each evaluation dimension, the actual contribution of the streetscape elements participating in that dimension is summarized to obtain the corresponding sub-dimensional score. Then, the streetscape environmental quality evaluation result is generated based on the sub-dimensional scores. Stable environmental elements are mainly used to reflect the impact of road space, green space, building interface, and fixed facilities on streetscape environmental quality. Periodic management elements are mainly used to reflect the impact of management order states such as repeated occupation, repeated spillover, and repeated occupancy on streetscape environmental quality. Temporary disturbance elements are used to generate disturbance prompts and are not used as the main basis for long-term environmental quality defects.

[0041] Finally, the street view environmental quality evaluation results and corresponding interpretation results are output. The interpretation results include low evaluation dimensions, street view elements that lead to the lower evaluation results, the road functional areas to which the street view elements belong, the classification type of the street view elements, and temporary interference descriptions. The classification types include stable environmental problems, periodic management problems, and temporary interference descriptions. For stable environmental problems, the interpretation results point to the long-term spatial environmental state. For periodic management problems, the interpretation results point to the recurring management order state. For temporary interference descriptions, the interpretation results are used to indicate the presence of occasional visual content in the current street view image, but this occasional visual content is not used as the main basis for deduction.

[0042] This embodiment also provides an intelligent evaluation system for interpretable street view environmental quality, used to execute the above-described intelligent evaluation method for interpretable street view environmental quality, including an image acquisition module, an alignment recognition module, a time correlation module, a state classification module, an evaluation processing module, and an interpretation output module.

[0043] The image acquisition module is used to acquire the current street view image of the target sampling point and the historical street view image corresponding to the target sampling point in space, forming a co-located street view image group. The image acquisition module reads the acquisition location, acquisition time, shooting direction and camera parameters of the street view image, and combines the road space data to filter the historical street view image so that the street view images in the co-located street view image group correspond to the same street space location.

[0044] The alignment and recognition module is used to perform spatial alignment and street scene element recognition on the same street scene image group, and obtain street scene element information at multiple acquisition times. The alignment and recognition module calculates the spatial mapping relationship based on road boundaries, building facade edges and static feature points, maps the historical street scene images to the unified coordinates of the current street scene image, generates a category label map through a semantic segmentation model, and then extracts the street scene element region and street scene element information through connected component extraction.

[0045] The time-related module is used to associate street scene elements located in the same street space at multiple acquisition times to form a street scene element time phase sequence. The time-related module matches street scene elements at different acquisition times according to the street scene element category, street scene element area, area center location and the road functional area to which it belongs, and adds the successfully matched street scene elements to the same street scene element time phase sequence.

[0046] The state classification module is used to determine the temporal stable state of street scene elements based on the temporal sequence of street scene elements, and classify street scene elements into stable environmental elements, periodic management elements, and temporary interference elements based on the temporal stable state. The state classification module calculates the occurrence coverage, location stability, area stability, and discrete repetition of street scene elements, and determines the classification type of street scene elements based on the calculation results.

[0047] The evaluation processing module is used to generate street view environmental quality evaluation results based on stable environmental factors and periodic management factors, and to suppress the evaluation contribution of temporary interference factors to the street view environmental quality evaluation results. The evaluation processing module configures the evaluation contribution coefficient according to the classification type of street view elements, and calculates the actual contribution of street view elements to each evaluation dimension based on the evaluation contribution coefficient and the basic impact quantity, thereby generating sub-dimension scores and street view environmental quality evaluation results.

[0048] The interpretation output module is used to output the street view environmental quality evaluation results and the corresponding interpretation results. The interpretation output module combines the low-scoring evaluation dimensions, the street view elements that lead to the reduction of the evaluation results, the road functional area to which the street view elements belong, and the corresponding classification type of the street view elements, and generates temporary interference descriptions for temporary interference elements.

[0049] Through the collaborative work of the above modules, the system can distinguish between long-term environmental conditions, periodic management conditions, and temporary disturbance conditions based on multi-temporal street view images, and output street view environmental quality evaluation results with assigned types.

[0050] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. An intelligent evaluation method for interpretable street view environmental quality, characterized in that, include: S1. Obtain the current street view image of the target sampling point, and obtain the historical street view image according to the street spatial location corresponding to the target sampling point. Combine the current street view image and the historical street view image into a co-located street view image group. S2. Spatial alignment and street scene element recognition are performed on the street scene images in the same street scene image group to obtain street scene element information at multiple acquisition times. S3. Based on street view element information, perform time correlation on street view elements located in the same street space at multiple collection times to form a street view element time sequence; S4. Determine the temporal stable state of street scene elements based on the temporal sequence of street scene elements, and classify street scene elements into stable environmental elements, periodic management elements, and temporary disturbance elements based on the temporal stable state. S5. Generate streetscape environmental quality assessment results based on stable environmental factors and periodic management factors, and suppress the contribution of temporary disturbance factors to the streetscape environmental quality assessment results. S6. Output the street view environmental quality evaluation results and the corresponding interpretation results.

2. The intelligent evaluation method for interpretable street view environmental quality according to claim 1, characterized in that, The process of acquiring historical street view images includes mapping the target sampling point to road space data to obtain the road number, road mileage location, road side and shooting direction corresponding to the target sampling point, and then selecting historical street view images from the historical street view image library that are in the same road space range as the target sampling point, have similar road mileage locations, consistent road sides and similar shooting directions, so that the historical street view images and the current street view images jointly reflect the same street space area.

3. The intelligent evaluation method for interpretable street view environmental quality according to claim 2, characterized in that, Before forming the corresponding street view image group, the selected historical street view images are deduplicated. This includes identifying street view images that were repeatedly collected within a short period of time based on the collection time, collection location, road mileage location, and image similarity of the historical street view images, and retaining street view images with higher image clarity among the repeatedly collected street view images.

4. The intelligent evaluation method for interpretable street view environmental quality according to claim 3, characterized in that, The spatial alignment includes taking the current street view image as the reference image, extracting static spatial features from the current street view image and historical street view images, and establishing a spatial mapping relationship from historical street view images to the current street view image based on the static spatial features, and converting the road area, pedestrian space area, building interface area and green area in the historical street view image to a unified image coordinate system. The street scene element recognition includes performing pixel-level category recognition on the spatially aligned street scene image to obtain a category label map, extracting connected components from the pixel regions in the category label map that are of the same category and spatially connected to obtain candidate street scene element regions, deleting candidate street scene element regions whose area is lower than the noise area condition, and merging candidate street scene element regions that are of the same category and spatially adjacent to each other to obtain street scene element information.

5. The intelligent evaluation method for interpretable street view environmental quality according to claim 4, characterized in that, The street view element information includes street view element category, street view element area, street view element area ratio, center location of street view element area, and road functional area to which the street view element belongs. The road functional area to which the street view element belongs is determined based on the positional relationship between the street view element area and the pedestrian space, vehicle space, road edge space, building interface space, or green space in the road space data.

6. The intelligent evaluation method for interpretable street view environmental quality according to claim 5, characterized in that, The time-related association includes taking street scene elements of the same category at different collection times as candidate association objects, matching them according to the degree of regional overlap, regional center distance and consistency of the functional area of ​​the road to which they belong, and adding the successfully matched candidate association objects to the time phase sequence of the same street scene element.

7. The intelligent evaluation method for interpretable street view environmental quality according to claim 6, characterized in that, The determination of the temporal stable state of street view elements includes: determining the occurrence and coverage state based on the frequency of occurrence of street view elements at multiple acquisition times; determining the positional stable state based on the changes in the regional center position of street view elements at multiple acquisition times; determining the area stable state based on the changes in the area proportion of street view elements at multiple acquisition times; and determining the discrete repetition state based on the intermittent repetition of street view elements at multiple acquisition times.

8. The intelligent evaluation method for interpretable street view environmental quality according to claim 7, characterized in that, The stable environmental elements refer to street scene elements that appear continuously at multiple collection times, are mainly concentrated in the same street space, and whose area proportion changes within an allowable range. The periodic management elements refer to street scene elements that do not meet the classification conditions of stable environmental elements, but appear repeatedly at multiple collection times and are mainly concentrated in the same street functional area. The temporary interference elements refer to street scene elements that appear with low frequency and whose location or area is unstable.

9. The intelligent evaluation method for interpretable street view environmental quality according to claim 8, characterized in that, The process of generating street view environmental quality evaluation results and outputting corresponding interpretation results includes: determining the contribution of street view elements to the evaluation dimensions based on their classification type; ensuring that stable environmental elements participate in the evaluation as a basis for long-term environmental quality evaluation; ensuring that periodic management elements participate in the evaluation as a basis for management order evaluation; ensuring that temporary interference elements are not used as the main basis for deduction; and outputting the low-scoring evaluation dimensions, the street view elements that lead to the reduction in evaluation results, the road functional area to which the street view elements belong, the classification type of the street view elements, and a description of temporary interference in the interpretation results.

10. A system for performing the intelligent evaluation method for interpretable street view environmental quality as described in claim 1, characterized in that, It includes an image acquisition module, an alignment recognition module, a time correlation module, a state classification module, an evaluation processing module, and an interpretation output module; The image acquisition module is used to perform street view image acquisition and the composition of a group of corresponding street view images; The alignment recognition module is used to perform spatial alignment of street view images and recognition of street view elements; The time correlation module is used to perform time correlation of street view elements to form a time phase sequence of street view elements; The state classification module is used to determine the temporal stable state of street scene elements and classify street scene elements according to the temporal stable state. The evaluation processing module is used to generate street view environmental quality evaluation results; The explanation output module is used to execute and output the explanation results.