A method and system for rapid road surface smoothness detection based on image recognition

Abstract

This invention discloses a rapid road surface smoothness detection method and system based on image recognition, belonging to the field of road engineering inspection technology. The method includes acquiring a continuous road image sequence and extracting the effective area image sequence of the continuous road; obtaining continuous road texture feature data through road texture feature extraction processing; establishing the dynamic change relationship of continuous road texture features during vehicle movement based on the continuous road texture feature data, and constructing a road texture inertial field; identifying abnormal change areas based on the road texture inertial field, and performing visual spectral perturbation analysis on the road image texture information in the abnormal change areas to obtain road visual perturbation feature data. By fusing continuous road texture dynamic change analysis with visual spectral change features, rapid road surface smoothness detection based on continuous image information is achieved.

Description

Technical Field

[0001] This invention relates to the field of road engineering inspection technology, and in particular to a method and system for rapid detection of road surface smoothness based on image recognition. Background Technology

[0002] With the development of road infrastructure construction and intelligent transportation technology, road maintenance and inspection are gradually moving towards automation and intelligence. Road surface smoothness, as an important indicator for evaluating road performance, reflects the surface structure and driving comfort of vehicles. Image recognition-based road surface detection technology uses vehicle-mounted camera equipment to collect road image sequences and combines image processing, computer vision, and feature recognition algorithms to analyze road conditions. It detects road anomalies by identifying visual features such as cracks, potholes, and ruts.

[0003] Existing image recognition-based road surface smoothness detection technologies mainly rely on single-frame images or local visual feature analysis. For areas with slight road surface undulations or continuous settlement, where the smoothness changes are small, it is difficult to fully utilize the dynamic texture change information in continuous road image sequences. Road texture features are easily affected by changes in lighting, road pollution, and image noise. Furthermore, existing technologies lack a comprehensive analysis of the correlation between spatial changes in road texture and spectral changes in texture features, making it difficult to achieve rapid smoothness evaluation of continuous road areas. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for rapid detection of road surface smoothness based on image recognition, which solves the problems of insufficient utilization of dynamic texture change information in continuous road image sequences and insufficient correlation analysis between road texture spatial changes and texture spectrum changes, resulting in insufficient accuracy of rapid detection of road surface smoothness.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a method for rapid detection of road surface smoothness based on image recognition, which includes: acquiring a continuous road image sequence, extracting a continuous road effective area image sequence, and obtaining continuous road texture feature data through road texture feature extraction processing; Based on continuous road texture feature data, the dynamic change relationship of continuous road texture features during vehicle movement is established, and a road texture inertial field is constructed. Based on the inertial field of road texture, abnormal change areas are identified, and visual spectral perturbation analysis is performed on the road image texture information in the abnormal change areas to obtain road visual perturbation feature data. Based on road visual disturbance feature data, the changes in texture features of road areas, changes in the texture spectrum of road images, and the continuous correlation features of road visual disturbance feature data are analyzed to obtain the road surface smoothness evaluation results. Based on the road surface smoothness evaluation results, and combined with the location data corresponding to the continuous road image sequence, a rapid road surface smoothness detection result is generated.

[0007] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the step of extracting the continuous road effective area image sequence includes, The system collects road image information during vehicle travel, arranges multiple road image frames according to the image acquisition time sequence to obtain a continuous road image sequence, and performs image region recognition processing on each road image frame in the continuous road image sequence to obtain road region location information. Based on the location information of the road area, road area cropping is performed on each road image frame in the continuous road image sequence to obtain a continuous effective road area image sequence.

[0008] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the continuous road texture feature data includes, Based on the continuous road effective area image sequence, image preprocessing is performed on each road image frame in the continuous road effective area image sequence to obtain the preprocessed continuous road effective area image sequence. Feature extraction processing is performed on the road surface texture information in the preprocessed continuous road effective area image sequence to obtain a road texture feature set; Based on the road texture feature set, the texture structure features, texture edge features, and texture distribution features in the road texture feature set are correlated to obtain continuous road texture feature data.

[0009] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the construction of the road texture inertial field includes, Based on continuous road texture feature data, according to the image acquisition time sequence corresponding to the continuous road texture feature data, temporal correlation processing is performed on the continuous road texture feature data corresponding to adjacent time to obtain the road texture feature change relationship at different time. The road texture features in the continuous road texture feature data are parsed to obtain the texture structure information, texture location coordinate information and texture neighborhood relationship information corresponding to each road texture feature in the continuous road texture feature data. Based on the texture location coordinate information, spatial location correlation analysis is performed on the road texture features in the continuous road texture feature data at different times, and the position coordinate offset between corresponding road texture features in adjacent image frames is calculated. Based on the location coordinate offset, determine the road texture features that are close in spatial location, combine the texture structure information to calculate the structural similarity between the road texture features corresponding to different times, and obtain the spatial correspondence of road texture features based on the location coordinate offset and structural similarity. Based on texture neighborhood information, the positional arrangement, structural distribution and mutual correlation of multiple associated road texture features around the road texture feature are jointly analyzed to construct the regional texture correlation relationship composed of road texture features. Based on the regional texture correlation relationship, the feature shift of a single road texture feature caused by lighting changes, road dust occlusion or image noise is corrected. Based on the texture structure association, the spatial correspondence of road texture features, and the regional texture association, the road texture change relationship in the continuous road texture feature data corresponding to different times is associated and described to obtain continuous road texture feature association data.

[0010] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the road visual disturbance feature data includes: Based on the road texture inertial field, the dynamic change state of road texture features in the road texture inertial field is traversed according to the spatial distribution, and the spatial connection relationship of road texture features is established according to the dynamic change state of road texture features with continuous spatial distribution. The preset spatial adjacency distance is determined based on the average spatial distance between adjacent road texture features in the continuous road texture feature data. Road texture features with a spatial distance not greater than the preset spatial adjacency distance and continuous dynamic change status are established to form a road texture feature spatial connection network. Based on the road texture feature spatial connection network, taking any road texture feature in the road texture feature spatial connection network as the starting node, and according to the dynamic change association parameters between adjacent road texture features in the road texture feature spatial connection network, the propagation path of the dynamic change state of the road texture feature in the road texture feature spatial connection network is calculated, and a road texture dynamic propagation network is formed according to the propagation path length, propagation direction change and propagation termination position. Based on the dynamic propagation network of road texture, the propagation continuity, propagation bifurcation and propagation termination relationships among the propagation paths in the dynamic propagation network of road texture are jointly analyzed. Based on the propagation continuity, propagation bifurcation and propagation termination relationships among the propagation paths, abnormal propagation regions are identified and abnormal change regions are obtained. Based on the abnormal change region, a spatial mapping relationship between the abnormal change region and the continuous road effective area image sequence is established in the continuous road effective area image sequence. According to the spatial mapping relationship and the image acquisition time order, the corresponding abnormal change region in the continuous road effective area image sequence is continuously tracked to obtain the road image texture information corresponding to the abnormal change region. Based on the texture information of road images, multiple texture analysis units are divided according to the spatial distribution direction of texture in the texture information of road images, and frequency domain transformation is performed on each unit. Based on the spatial arrangement relationship between the texture analysis units, a texture spectrum spatial distribution network is established to obtain texture spectrum feature data.

[0011] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the road surface smoothness evaluation results include: Based on road visual disturbance feature data, the dynamic change information of road texture, the spectral change information of road image texture, and the spatial correlation information in continuous road image sequences are analyzed to obtain the change amount of road texture features, the change amount of road image texture spectrum, and the continuous correlation features of road visual disturbance feature data, respectively. Based on the changes in road texture features, a spatial correlation between the changes in road texture features is established according to the spatial location corresponding to the road area. In addition, a collaborative change relationship between road texture change and texture spectrum change is established by combining the changes in road image texture spectrum, so as to obtain collaborative change data of road area. The continuous correlation features of road visual disturbance feature data are mapped to the corresponding road areas in road area collaborative change data, and the continuous evolution relationship between the continuous correlation features of road area collaborative change data and road visual disturbance feature data is established to obtain road area visual disturbance evaluation data. Based on the visual disturbance evaluation data of road areas, the road surface smoothness status is mapped according to the continuous evolution relationship between different road areas to obtain the road surface smoothness evaluation results.

[0012] As a preferred embodiment of the image recognition-based rapid road surface smoothness detection method of the present invention, the rapid road surface smoothness detection results include: Based on the road surface smoothness evaluation results, the road surface smoothness evaluation results are associated and matched with the location data corresponding to the continuous road image sequence to obtain the road area location evaluation data. The continuous spatial distribution relationship between the road area location evaluation data is established according to the location order corresponding to the continuous road image sequence to obtain the road smoothness spatial distribution data. Based on the spatial distribution data of road smoothness, the evaluation results of road surface smoothness between adjacent road areas are continuously integrated to form continuous road smoothness distribution data. Based on continuous road smoothness distribution data, the continuous road smoothness distribution data is continuously fused according to the positional order corresponding to the continuous road image sequence. The continuous road smoothness distribution data belonging to the same road segment are integrated, and the road surface smoothness rapid detection result is generated based on the integrated road segment smoothness distribution.

[0013] Secondly, the present invention provides a rapid road surface smoothness detection system based on image recognition, including an extraction module that acquires a continuous road image sequence and extracts a continuous road effective area image sequence, and obtains continuous road texture feature data through road texture feature extraction processing. The module constructs a dynamic relationship between continuous road texture features and vehicle movement, based on continuous road texture feature data, and builds a road texture inertial field. The disturbance module determines the abnormal change area based on the road texture inertial field, and performs visual spectral disturbance analysis on the road image texture information in the abnormal change area to obtain road visual disturbance feature data. The analysis module, based on road visual disturbance feature data, analyzes the changes in texture features of road areas, the changes in texture spectrum of road images, and the continuous correlation features of road visual disturbance feature data to obtain road surface smoothness evaluation results. The detection results module generates rapid road surface smoothness detection results based on the road surface smoothness evaluation results and the location data corresponding to the continuous road image sequence.

[0014] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein the computer program, when executed by the processor, implements any step of the image recognition-based rapid road surface smoothness detection method described in the first aspect of the present invention.

[0015] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the image recognition-based rapid road surface smoothness detection method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By acquiring continuous road image sequences and extracting effective road areas, continuous road texture feature data is obtained through road texture feature extraction processing; based on the continuous road texture feature data, the dynamic change relationship of road texture during vehicle movement is established, a road texture inertial field is constructed, and continuous road texture feature association data is established through road texture feature spatial correlation analysis, texture structure matching, and regional texture correlation, thereby realizing the continuous representation of road texture dynamic change information; based on the road texture inertial field, a road texture feature spatial connection network is constructed, and a road texture dynamic propagation network is formed according to the road texture dynamic change propagation relationship, identifying abnormal areas of road texture dynamic change, and performing visual spectrum perturbation analysis based on the road image texture information corresponding to the abnormal change areas to obtain road visual perturbation feature data; by fusing the continuous correlation features of road texture feature change, road image texture spectrum change, and road visual perturbation feature data, a road area visual perturbation evaluation relationship is established to obtain road surface smoothness evaluation results; combined with the location data corresponding to the continuous road image sequence, the road surface smoothness evaluation results are spatially correlated and road segment fused to generate rapid road surface smoothness detection results. By fusing continuous road texture dynamic change analysis with visual spectrum change features, rapid detection of road surface smoothness based on continuous image information was achieved. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of a method for rapid road surface smoothness detection based on image recognition.

[0019] Figure 2 This is a schematic diagram of a rapid road surface smoothness detection system based on image recognition. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0021] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0023] Reference Figures 1-2 This is one embodiment of the present invention, which provides a rapid road surface smoothness detection method based on image recognition, including the following steps: S1. Collect a continuous road image sequence and extract the effective area image sequence of the continuous road. Then, obtain the continuous road texture feature data through road texture feature extraction processing.

[0024] S1.1 Collect road image information during vehicle travel, arrange multiple road image frames according to the image acquisition time sequence to obtain a continuous road image sequence, perform image region recognition processing on each road image frame in the continuous road image sequence to obtain road region location information.

[0025] Furthermore, during vehicle operation, image acquisition devices installed on the vehicle continuously acquire road image information ahead of the vehicle. Multiple acquired road image frames are sorted according to their acquisition time sequence, and road image frames from different acquisition times are continuously correlated to obtain a continuous road image sequence. For each road image frame in the continuous road image sequence, image regions within the road image frame are identified. By analyzing the visual feature differences between road surface areas and non-road areas, the spatial distribution range of the road region within the road image frame is determined, obtaining the road region location information. The road region identification process can be implemented using existing image semantic segmentation methods. By analyzing pixel features, edge variation features, and regional texture features in the road image frame, image regions with continuous road surface features are identified, and the corresponding pixel positions are marked. Alternatively, it can be implemented using existing target region detection methods. By identifying road boundary features and the spatial structural relationships of road regions, the road region location information is determined.

[0026] S1.2. Based on the location information of the road area, perform road area cropping processing on each road image frame in the continuous road image sequence to obtain a continuous road effective area image sequence.

[0027] Furthermore, based on road area location information, spatial region localization is performed on each road image frame in the continuous road image sequence. The boundary range of the corresponding road area in the road image frame is determined according to the road area location information, and the road image frame is cropped according to the road area boundary range to remove image areas unrelated to the road surface and retain the effective road area image corresponding to the vehicle's driving direction. Based on the road area location information corresponding to different road image frames, the road areas in the continuous road image sequence are extracted synchronously to maintain the spatial consistency of the effective road areas at different times. Through road area cropping, the continuous road image sequence is converted into a continuous road effective area image sequence containing only road surface information. The pixel range corresponding to the road area is determined according to the road area location information, and the corresponding pixel area is extracted to form the effective road area image.

[0028] S1.3. Based on the continuous road effective area image sequence, perform image preprocessing on each road image frame in the continuous road effective area image sequence to obtain the preprocessed continuous road effective area image sequence.

[0029] Furthermore, by adjusting the image quality and visual feature distribution in road image frames, the recognizability of road surface texture information is improved. Image preprocessing can include image denoising, illumination equalization, and image enhancement. Denoising reduces the impact of random noise generated during road image frame acquisition; illumination equalization reduces brightness differences caused by different acquisition times, weather conditions, and road shadows; and image enhancement improves the expressive power of road surface texture edges and local structural features. For example, when some road areas in a continuous road image sequence experience brightness changes due to sunlight, illumination equalization adjusts the overall brightness distribution of the road image frames, ensuring comparability of road texture features across different time periods. When dust obscures local texture information, image enhancement enhances road texture edge information, enabling the road texture structure to be recognized by subsequent feature extraction processes. By uniformly preprocessing each road image frame in a continuous effective road area image sequence, a preprocessed continuous effective road area image sequence is obtained.

[0030] S1.4. Perform feature extraction processing on the road surface texture information in the preprocessed continuous road effective area image sequence to obtain a road texture feature set.

[0031] Furthermore, by analyzing the pixel variation relationships, texture structure variation relationships, and local visual feature variation relationships at different locations within the road surface area, feature description results corresponding to the road surface texture information are obtained, resulting in a road texture feature set. The gray-level co-occurrence matrix is ​​used to analyze the gray-level correlation relationships between different areas of the road surface, extracting texture features reflecting surface structure changes. Local binary mode is used to extract pixel variation relationships within local areas of the road surface, obtaining detailed road texture variation features. Directional filtering is used to analyze texture response changes in different directions on the road surface, obtaining road texture direction features. When the road surface remains continuously smooth, the texture features corresponding to adjacent areas in the road texture feature set exhibit continuous variation relationships. When potholes, bumps, or local damage exist on the road surface, the texture structure, texture direction, and local variation relationships of the corresponding areas in the road texture feature set will show abnormal changes. By uniformly extracting features from the road surface texture information in the preprocessed continuous effective road area image sequence, the road surface texture information in different road image frames is converted into a road texture feature set that can be correlated and analyzed.

[0032] S1.5. Based on the road texture feature set, correlate the texture structure features, texture edge features and texture distribution features in the road texture feature set to obtain continuous road texture feature data.

[0033] Furthermore, the texture structure features in the road texture feature set are analyzed to obtain the composition relationship of road surface textures and local structural changes; the texture edge features are analyzed to obtain the changes in road surface texture boundaries and the locations of local abrupt changes; combined with the texture distribution features, the positional relationship and distribution continuity between road texture features in different regions are analyzed to establish corresponding correlations between texture structure features, texture edge features, and texture distribution features. For example, when there are continuous flat areas on the road surface, the texture structure features, texture edge features, and texture distribution features maintain a stable correspondence; when there are local concave areas on the road surface, the texture edge features may undergo local changes, and the texture structure features and texture distribution features will change accordingly. Through correlation analysis, the comprehensive texture change information corresponding to this area can be obtained. By correlating the texture structure features, texture edge features, and texture distribution features in the road texture feature set, continuous road texture feature data can be obtained.

[0034] S2. Based on continuous road texture feature data, establish the dynamic change relationship of continuous road texture features during vehicle movement and construct the road texture inertial field.

[0035] S2.1 Based on continuous road texture feature data, according to the image acquisition time sequence corresponding to the continuous road texture feature data, perform time correlation processing on the continuous road texture feature data corresponding to adjacent time to obtain the road texture feature change relationship corresponding to different time.

[0036] Furthermore, the road texture features corresponding to each road image frame in the continuous road texture feature data are used as temporal correlation objects. Based on the temporal relationship between road image frames, corresponding analysis is performed on road texture features corresponding to adjacent times to determine whether the road texture features corresponding to adjacent times belong to continuous texture changes in the same road region. When the road texture features corresponding to adjacent times maintain a similar relationship in terms of spatial location, texture structure, and texture distribution, the road texture features corresponding to adjacent times are correlated, and the continuous change relationship of the road texture features over time is recorded. When the road texture features corresponding to adjacent times show changes in texture structure, edge changes, or texture distribution, the changes in road texture features are recorded to obtain the change state of the road texture features at the corresponding time. For example, when a vehicle passes through a smooth road area, the changes in road texture features corresponding to adjacent times usually remain in a continuous and stable state, and the change relationship of road texture features shows continuous change.

[0037] Specifically, when a vehicle passes through a road area with depressions, bumps, or localized damage, the road texture features at adjacent times may exhibit abrupt changes, resulting in abnormal changes in the relationship between road texture features. By associating continuous road texture feature data according to the order of image acquisition time, a continuous change relationship is formed between road texture features at different times, thus obtaining the relationship between road texture feature changes at different times.

[0038] S2.2 Perform feature parsing processing on the road texture features in the continuous road texture feature data to obtain the texture structure information, texture position coordinate information and texture neighborhood relationship information corresponding to each road texture feature in the continuous road texture feature data.

[0039] Furthermore, texture structure analysis is performed on the road texture features in the continuous road texture feature data. Based on the pixel variation relationship, texture direction variation relationship, and local region texture composition relationship in the road texture features, the structural features of the road texture features are analyzed to obtain texture structure information. For example, the road texture features corresponding to smooth road surface areas usually have a continuous and uniform texture structure, while the road texture features corresponding to pothole areas may show local texture breaks, texture direction changes, or texture density changes. Through texture structure analysis, the structural manifestations corresponding to different road texture features can be obtained. The spatial position of the road texture features in the continuous effective area image sequence is analyzed. Based on the pixel position and image coordinate relationship of the road texture features, the spatial position of the road texture features in the road image frame is determined, and texture position coordinate information is obtained.

[0040] Specifically, based on the texture location coordinates of each road texture feature, correlation analysis is performed on other road texture features within a certain spatial range around the road texture feature. By judging the positional relationship, structural similarity relationship, and distribution relationship between adjacent road texture features, the proximity relationship between road texture features is obtained, thus acquiring texture neighborhood relationship information. Through feature parsing processing of road texture features in continuous road texture feature data, texture structure information, texture location coordinate information, and texture neighborhood relationship information are obtained for each road texture feature.

[0041] It should be noted that by performing feature parsing on the road texture features in the continuous road texture feature data, the road texture features are decomposed into texture structure information, texture location coordinate information, and texture neighborhood relationship information. This enables the road texture features to simultaneously possess structural description capabilities, spatial positioning capabilities, and regional correlation capabilities. For example, when a single road texture feature changes, the texture structure information can be used to determine the change itself, the texture location coordinate information can be used to determine the location of the change, and the texture neighborhood relationship information can be used to analyze whether the surrounding road textures change synchronously. By obtaining the correlation information of the three types of road texture features, the continuous road texture feature data can reflect the dynamic changes in road textures during vehicle movement.

[0042] S2.3. Based on the texture position coordinate information, perform spatial position correlation analysis on the road texture features in the continuous road texture feature data at different times, and calculate the position coordinate offset between the corresponding road texture features in adjacent image frames.

[0043] Furthermore, based on the image acquisition time sequence corresponding to the continuous road texture feature data, the road texture features corresponding to adjacent image frames are obtained, and the texture position coordinate information corresponding to the road texture features in adjacent image frames is read respectively. Taking the road texture features in the current image frame as the association object, the road texture features corresponding to the current road texture features are searched in the next image frame according to the texture position coordinate information. By comparing the changes in texture position coordinates of the corresponding road texture features in the horizontal direction and the changes in texture position coordinates in the vertical direction in adjacent image frames, the spatial position change of the road texture features during vehicle movement is determined. The position coordinate offset is calculated by the difference in position coordinates of the corresponding road texture features in adjacent image frames. Through the calculation of position coordinate offset, the spatial position correspondence of road texture features corresponding to different times can be established. For example, when the vehicle passes through a flat road area, the position coordinate offset of the same road texture feature in adjacent image frames usually remains continuously changing; when the vehicle passes through an abnormal road surface area, the corresponding road texture feature may have a different position change trend than the normal area. The position coordinate offset can reflect the spatial change state of the road texture feature at different times, and the position coordinate offset between the corresponding road texture features in adjacent image frames can be obtained.

[0044] Specifically, by performing spatial location correlation analysis based on texture location coordinate information, road texture features corresponding to different times are matched accordingly. The location coordinate offset is used to describe the spatial changes of road texture features caused by vehicle movement. For example, the same road area will gradually change position in a continuous road image sequence as the vehicle moves. The location coordinate offset of the road texture features corresponding to the normal road area is continuous, while road surface depressions, bumps, or local damage areas will cause abnormal spatial changes in road texture features. The location coordinate offset can identify the differences in road texture changes in different areas. By using the location coordinate offset to perform spatial correlation analysis on the corresponding road texture features in adjacent image frames, the road texture features in continuous road texture feature data are transformed from single-frame static location information into spatial displacement information with temporal changes.

[0045] The expression for position coordinate offset is: ; in, For the first Time frame and the first The first time in the image frame The position coordinate offset of each road texture feature For the first The first time in the image frame The texture position coordinates of a road texture feature in the horizontal direction. For the first In the image frame at time 1 and 2 The horizontal coordinates of the road texture feature corresponding to each road texture feature. For the first In the image frame at time 1 and 2 The vertical coordinates of the road texture feature corresponding to each road texture feature. This is the acquisition time sequence number corresponding to the current image frame. This is the acquisition time sequence number corresponding to the next image frame adjacent to the current image frame. This refers to the road texture feature number in the continuous road texture feature data.

[0046] S2.4 Determine the road texture features that are close in spatial location based on the position coordinate offset, calculate the structural similarity between road texture features corresponding to different times based on the texture structure information, and obtain the spatial correspondence of road texture features based on the position coordinate offset and structural similarity.

[0047] Furthermore, based on the positional coordinate offset between road texture features in adjacent image frames, the spatial distance change between road texture features is judged, and road texture features whose positional coordinate offsets satisfy the spatial proximity relationship are selected as candidate association objects. For example, when there are road texture features with small positional changes in the image frame at time 1, it is judged that the two road texture features are likely to be spatially associated; when the positional coordinate offsets between different road texture features are large, the association probability between the two road texture features is reduced.

[0048] Based on the texture structure information corresponding to the road texture features in the candidate associated objects, the structural similarity between road texture features corresponding to different times is calculated. By comparing the texture structure composition relationship, texture direction change relationship, and local texture distribution relationship in the road texture features corresponding to different times, it is determined whether the road texture features corresponding to different times belong to the continuous change features in the same road area. For example, when a vehicle passes through a continuous flat road area, although the position coordinates of the road texture features corresponding to adjacent times change, the texture structure information remains highly consistent. When there are abnormal areas on the road surface, the corresponding road texture features may show changes in texture structure. The structural similarity can be used to determine whether the road texture features maintain a continuous correspondence.

[0049] Specifically, a joint analysis of positional coordinate offsets and structural similarity is performed to determine the spatial correspondence of road texture features based on their spatial proximity and the degree of preservation of texture structure. When road texture features simultaneously satisfy both spatial proximity and texture structure similarity, a correspondence is established between road texture features corresponding to different times. When road texture features only satisfy spatial proximity but have significant differences in texture structure, the correspondence between road texture features is reassessed to avoid incorrect matching caused by similar road texture regions, image noise, or local occlusion. Through the joint analysis of positional coordinate offsets and structural similarity, a stable spatial correspondence is established between road texture features corresponding to different times, thus obtaining the spatial correspondence of road texture features.

[0050] The structural similarity expression is:

[0051] in, The similarity of road texture features between adjacent image frames. For the first The texture structure feature vector corresponding to the road texture features at any given time. For the first The texture structure feature vector corresponding to the road texture features at any given time.

[0052] S2.5. Based on texture neighborhood relationship information, jointly analyze the positional arrangement, structural distribution, and interrelationship among multiple associated road texture features around the road texture feature, construct a regional texture association relationship composed of road texture features, and correct the feature shift of a single road texture feature caused by changes in lighting, road dust occlusion, or image noise according to the regional texture association relationship.

[0053] Furthermore, based on the texture neighborhood relationship information corresponding to the road texture features, multiple associated road texture features that have a spatial proximity relationship with a single road texture feature are obtained. The positional arrangement relationship between multiple associated road texture features is analyzed, and the positional coordinate relationship between multiple associated road texture features is obtained to determine whether the positional changes between adjacent road texture features remain continuous. In continuous smooth road areas, due to the continuous road surface structure, road texture features are usually formed by the continuous texture structure of the road surface. Therefore, adjacent road texture features have a stable positional arrangement relationship, the positional change direction of multiple associated road texture features is consistent, and there are no irregular positional jumps between road texture features. For example, when a vehicle is driving in a continuous smooth road area, multiple road texture features in the same area will synchronously shift their positions as the vehicle moves, but the arrangement direction and mutual distance relationship between adjacent road texture features still remain continuously changing.

[0054] Specifically, when potholes, cracks, or local damage exist on the road surface, the texture structure of the road surface changes, affecting the original spatial continuity between road texture features. This manifests as changes in the positional arrangement of multiple associated road texture features around the abnormal area. For example, the positional interval between adjacent road texture features changes, the orientation of road texture features deviates, or the connection between some road texture features and surrounding road texture features weakens. By analyzing the positional arrangement of multiple associated road texture features, it is possible to determine whether the road texture area maintains continuity and to identify changes in the continuity of the road texture based on changes in positional arrangement.

[0055] The structural distribution relationship among multiple associated road texture features is analyzed. Based on the texture structure information corresponding to the road texture features, the texture composition relationship, texture direction relationship, and texture change trend among adjacent road texture features are determined. When multiple associated road texture features are in a continuous and flat road area, the multiple road texture features usually maintain a similar structural distribution relationship. When there are abnormal changes in the road area, the road texture features corresponding to the abnormal area will cause structural distribution differences in the surrounding associated road texture features. The positional arrangement relationship and structural distribution relationship among road texture features are analyzed. Based on the common changes among multiple associated road texture features, the mutual correlation relationship among road texture features is established, and the regional texture correlation relationship composed of road texture features is constructed.

[0056] It should be noted that when a single road texture feature changes, but the positional arrangement, structural distribution, and interrelationships of multiple related road texture features remain stable, the change in the single road texture feature is likely caused by changes in illumination, road dust occlusion, or image noise. The correlation between the single road texture feature and the surrounding road texture features is then restored based on the regional texture correlation. For example, when a part of the road area is affected by changes in illumination, the brightness of a single road texture feature may change, but the surrounding multiple related road texture features still maintain a continuous arrangement. The regional texture correlation can reduce the feature shift caused by changes in illumination. When there is localized dust occlusion on the road surface, the single road texture feature may have structural defects, but the surrounding road texture features can still maintain their original regional correlation. The regional texture correlation is used to correct the single road texture feature, correcting the feature shift caused by changes in illumination, road dust occlusion, or image noise, thus obtaining the corrected road texture feature correlation.

[0057] S2.6. Based on the texture structure association, the spatial correspondence of road texture features, and the regional texture association, the road texture change relationship in the continuous road texture feature data corresponding to different times is associated and described to obtain continuous road texture feature association data.

[0058] Further, based on the correlation of texture structure, the changes in texture structure of road texture features at different times are analyzed. By comparing the changes in texture structure information in road texture features at adjacent times, it is determined whether the road texture features maintain a continuous structural relationship. When a vehicle passes through a continuous smooth road area, although the road texture features in the same area at different times change position due to the movement of the vehicle, the texture structure information maintains a continuous trend of change. When a vehicle passes through potholes, cracks, or locally damaged areas, the texture structure information in the corresponding road texture features changes. The process of road texture structure change is described through the correlation of texture structure.

[0059] Based on the spatial correspondence of road texture features, the positional changes of road texture features at different times are described in association. The correspondence between road texture features at different times is determined through the spatial correspondence of road texture features, and the spatial changes of road texture features in a continuous road image sequence are recorded. When the same road area appears in a continuous road image sequence, the position of the corresponding road texture feature shifts with vehicle movement, but the spatial correspondence of road texture features at different times is still maintained. When the road area undergoes abnormal changes, the spatial correspondence of the corresponding road texture features changes. The positional changes of road texture features are reflected through the spatial correspondence of road texture features.

[0060] Specifically, based on regional texture correlations, the overall change relationships among multiple associated road texture features are described. By analyzing the relationship between changes in a single road texture feature and changes in multiple surrounding associated road texture features, it is determined whether the road texture change is a local change or a regional continuous change. When a single road texture feature changes while multiple surrounding associated road texture features maintain a stable correlation, the road texture change is considered a local change. When multiple associated road texture features simultaneously undergo structural, positional, and distributional changes, the road texture change is considered a regional continuous change. A comprehensive correlation description is performed on texture structure correlations, spatial correspondences of road texture features, and regional texture correlations. The structural change process, spatial correspondence process, and regional correlation change process of road texture features at different times are integrated to form dynamic change relationships between continuous road texture features, thus obtaining continuous road texture feature correlation data.

[0061] S3. Determine the abnormal change area based on the road texture inertial field, and perform visual spectral perturbation analysis on the road image texture information in the abnormal change area to obtain road visual perturbation feature data.

[0062] S3.1 Based on the road texture inertial field, the dynamic change state of road texture features in the road texture inertial field is traversed according to the spatial distribution, and the spatial connection relationship of road texture features is established according to the dynamic change state of road texture features continuously distributed in spatial location.

[0063] The preset spatial adjacency distance is determined based on the average spatial spacing between adjacent road texture features in the continuous road texture feature data. Road texture features with a spatial spacing not greater than the preset spatial adjacency distance and continuous dynamic change status are established to form a road texture feature spatial connection network.

[0064] Furthermore, a preset spatial adjacency distance is determined based on the average spatial spacing between adjacent road texture features in the continuous road texture feature data. This preset spatial adjacency distance is used to determine whether there is a spatial proximity relationship between different road texture features. When the spatial spacing between two road texture features is not greater than the preset spatial adjacency distance, it is determined that the two road texture features are within an adjacent spatial range. The dynamic change states corresponding to the two road texture features are analyzed to see if they have a continuous relationship. For example, in a continuous smooth road area, the direction and trend of positional changes of adjacent road texture features as vehicles move remain continuous, and there is a stable dynamic change relationship between adjacent road texture features. When there are potholes, cracks, or local damage in the road area, the dynamic change state of road texture features near the abnormal area may show inconsistent change directions, interrupted change trends, or local abnormal changes.

[0065] Based on the spatial spacing relationship and the continuity of dynamic change states, the connection between road texture features is judged. When the spatial distance between two road texture features meets the preset spatial adjacency distance requirement and the dynamic change states corresponding to the two road texture features remain continuous, a spatial connection relationship between the two road texture features is established. When the spatial distance between two road texture features is close, but there are obvious differences in dynamic change states, a spatial connection relationship between the road texture features is not established to avoid incorrectly associating road texture features with different change states.

[0066] Specifically, a comprehensive analysis is performed on all road texture features in the road texture inertial field. Based on the spatial connectivity relationships established between multiple road texture features, a spatial connectivity network of road texture features is formed. In this network, road texture features serve as associated objects within the spatial connectivity relationships. These relationships describe the positional continuity and dynamic change continuity between road texture features, enabling the dynamic changes of road texture features in the road texture inertial field to form spatial relationships and thus creating the road texture feature spatial connectivity network.

[0067] S3.2 Based on the road texture feature spatial connection network, take any road texture feature in the road texture feature spatial connection network as the starting node, calculate the propagation path of the dynamic change state of the road texture feature in the road texture feature spatial connection network according to the dynamic change association parameters between adjacent road texture features in the road texture feature spatial connection network, and form a road texture dynamic propagation network according to the propagation path length, propagation direction change and propagation termination position.

[0068] Furthermore, the road texture features in the road texture feature spatial connection network are used as propagation nodes, and the dynamic change correlation parameters between adjacent road texture features are used as the propagation correlation basis. By analyzing the positional continuity relationship, texture change continuity relationship and dynamic change trend between adjacent road texture features, the propagation direction of the dynamic change state of the road texture features is determined. According to the dynamic change correlation parameters between adjacent road texture features in the road texture feature spatial connection network, the dynamic change state of the road texture features is continuously traversed. Starting from the starting node, adjacent road texture features with continuous correlation are searched and connected in sequence to form the propagation path of the dynamic change state of the road texture features.

[0069] Specifically, the propagation path of dynamic changes in road texture features is described based on the propagation path length, changes in propagation direction, and propagation termination position. The propagation path length reflects the continuous range of dynamic changes in road texture, the changes in propagation direction reflect the directional continuity during the dynamic changes, and the propagation termination position reflects the end point of the dynamic changes. When abnormal changes occur in the road area, the propagation path of dynamic changes in road texture may exhibit changes in propagation direction or reduced propagation continuity. Based on the relationships between the propagation range, propagation direction, and termination position of multiple dynamic change propagation paths of road texture features, correlation analysis is performed on different propagation paths to form a dynamic propagation network of road texture features. This transforms the dynamic change state of road texture features from a single road texture feature change into a continuous change relationship with spatial propagation relationships, thus forming a dynamic propagation network of road texture features.

[0070] The propagation path expression is: ; in, The propagation path of dynamic changes in road texture features. The number of road texture feature nodes included in the propagation path of the dynamic change state of road texture features. This serves as the node index in the propagation path of the dynamic changes in road texture features. The parameters are associated with the dynamic changes of texture features between adjacent roads in the propagation path.

[0071] S3.3 Based on the dynamic propagation network of road texture, the propagation continuity relationship, propagation bifurcation relationship and propagation termination relationship between each propagation path in the dynamic propagation network of road texture are jointly analyzed. Based on the propagation continuity relationship, propagation bifurcation relationship and propagation termination relationship between the propagation paths, abnormal propagation areas are identified and abnormal change areas are obtained.

[0072] Furthermore, a correlation analysis is performed on multiple propagation paths in the dynamic propagation network of road textures. Based on the connection nodes, propagation directions, and dynamic change correlation parameters between different propagation paths, the propagation continuity relationship between propagation paths is determined. When multiple propagation paths have a continuous connection relationship and the dynamic change state of road texture features remains consistent, it is determined that multiple propagation paths belong to a continuous road texture change area. When there are abnormal changes in the road area, the propagation direction may change or the propagation continuity relationship may be interrupted. The propagation bifurcation relationship between propagation paths is analyzed. Based on the changes in the number of propagation paths, the changes in propagation direction, and the road texture changes corresponding to the bifurcation positions, it is determined whether there is abnormal diffusion in the dynamic propagation of road textures. When a propagation bifurcation relationship appears in the dynamic propagation network of road textures that is inconsistent with the change state of the surrounding road textures, it is determined that there may be abnormal changes in the corresponding area.

[0073] Specifically, the propagation termination relationships between propagation paths are analyzed. Based on the end position of the propagation path and the corresponding road texture change state, it is determined whether there is an abnormal interruption in the dynamic propagation of road texture. When the dynamic propagation of road texture shows an abnormal termination or an abnormal change in the propagation range, it is determined that there may be abnormal changes in the corresponding area. The propagation continuity relationship, propagation bifurcation relationship, and propagation termination relationship are jointly analyzed, and the propagation abnormal area is identified based on the correlation change state between propagation paths. When the propagation path shows abnormal propagation direction, interruption of propagation continuity relationship, abnormal propagation bifurcation relationship, or abnormal propagation termination position, the corresponding area is determined as the propagation abnormal area, and the abnormal change area is obtained.

[0074] S3.4. Based on the abnormal change region, establish the spatial mapping relationship between the abnormal change region and the continuous road effective area image sequence in the continuous road effective area image sequence. According to the spatial mapping relationship and the image acquisition time sequence, perform continuous region tracking on the corresponding abnormal change region in the continuous road effective area image sequence to obtain the road image texture information corresponding to the abnormal change region.

[0075] Furthermore, based on the spatial correspondence of road texture features, the road texture features in the abnormal change area are matched with the road image frames in the continuous effective road area image sequence. Based on the spatial correspondence of road texture features, the corresponding positions of the road texture features at different times in the continuous effective road area image sequence are determined. The positions of multiple road texture features in the abnormal change area are mapped to the pixel areas in the corresponding road image frames, establishing a spatial mapping relationship between the abnormal change area and the continuous effective road area image sequence. For example, when the road texture dynamic propagation network identifies an abnormal change in a certain area, the abnormal change area is transformed into the road surface area in the corresponding road image frame based on the position coordinates of the road texture features in the abnormal change area, so that the abnormal change area can correspond to the specific road image texture position.

[0076] Specifically, based on spatial mapping relationships and the order of image acquisition time, continuous region tracking is performed on corresponding abnormal change regions in a continuous road effective area image sequence. The spatial position of abnormal change regions in different road image frames is updated based on the positional change relationships between adjacent image frames. By comparing the positional changes of road texture features corresponding to abnormal change regions in adjacent time-corresponding road image frames, the corresponding position of the abnormal change region in the next image frame is determined. When the abnormal change region maintains continuous spatial change in the continuous road effective area image sequence, the corresponding regions in adjacent image frames are continuously associated. When the abnormal change region is affected by changes in illumination, road occlusion, or image noise, causing changes in local texture information, the abnormal change region is continuously tracked based on the spatial correspondence of road texture features in adjacent time-corresponding frames and the regional texture association relationship. The tracked abnormal change regions obtained in road image frames at different times are associated, and the road surface texture information within the corresponding position range of the abnormal change region is extracted to obtain road image texture information containing the continuous change state of the abnormal change region.

[0077] S3.5 Based on the texture information of the road image, divide the texture into multiple texture analysis units according to the spatial distribution direction of the texture in the road image texture information, and perform frequency domain transformation on each unit. Based on the spatial arrangement relationship between the texture analysis units, establish a texture spectrum spatial distribution network to obtain texture spectrum feature data.

[0078] Furthermore, by analyzing the relationship between texture direction changes between adjacent regions in road image texture information, the main spatial distribution direction of road texture information is determined, enabling texture analysis units to be divided along the direction of road texture change. For example, road textures in continuous smooth road areas usually have continuous directional distribution characteristics, and texture analysis units are divided according to the extension direction of road textures; when there are potholes, cracks, or local damage on the road surface, the texture direction within the abnormal area may change, and texture analysis units are divided according to the spatial distribution direction of the texture.

[0079] Specifically, frequency domain transformation is performed on the multiple texture analysis units after division. The road image texture information in the texture analysis unit is transformed from the spatial domain to the frequency domain using existing frequency domain transformation methods. The texture spectrum information corresponding to each texture analysis unit is obtained. By analyzing the spectral response relationship corresponding to different spatial frequencies in the texture analysis unit, the texture frequency change in the road image texture information is obtained. For example, the texture analysis units in the smooth road area usually have a continuous and stable spectral distribution relationship. However, potholes, cracks, or locally damaged areas will cause changes in the spectral energy distribution, frequency change range, or spectral direction relationship in the corresponding texture analysis units due to changes in the road surface structure. The texture spectrum change information corresponding to different texture analysis units can be obtained through frequency domain transformation.

[0080] Based on the spatial positional relationships of multiple texture analysis units in the road image texture information, the spatial connectivity between adjacent texture analysis units is determined, and the corresponding texture spectrum information is associated. For example, when adjacent texture analysis units have a continuous spatial arrangement, the texture spectrum information corresponding to adjacent texture analysis units is associated, so that the texture spectrum changes corresponding to different spatial locations form a continuous distribution relationship. When there are local texture structure changes within an abnormal change area, the spectrum distribution relationship between corresponding texture analysis units may change. The texture spectrum spatial distribution network can reflect the texture spectrum changes corresponding to different locations within the abnormal area. Based on the texture spectrum information and spatial arrangement relationships of different texture analysis units in the texture spectrum spatial distribution network, the spectrum changes in the road image texture information are comprehensively described, and texture spectrum feature data is obtained.

[0081] S4. Based on road visual disturbance feature data, analyze the changes in texture features of road areas, changes in road image texture spectrum, and continuous correlation features of road visual disturbance feature data to obtain road surface smoothness evaluation results.

[0082] S4.1 Based on road visual disturbance feature data, analyze the dynamic change information of road texture, the change information of road image texture spectrum, and the spatial correlation information in the continuous road image sequence in the road visual disturbance feature data to obtain the change amount of road texture features, the change amount of road image texture spectrum, and the continuous correlation features of road visual disturbance feature data, respectively.

[0083] Furthermore, the dynamic changes in road texture are analyzed. Based on the changes in road texture features at different times in a continuous road image sequence, the degree of change in the spatial location, texture structure, and dynamic trend of road texture features is analyzed to obtain the amount of change in road texture features. For example, the changes in road texture features in a continuous and smooth road area usually remain continuous at different times, and the amount of change in road texture features is relatively stable. When there are potholes, cracks, or local damage in the road area, the changes in the road texture features in the corresponding area will show abnormal changes. The amount of change in road texture features can reflect the changes in the road texture structure.

[0084] Specifically, the information on the variation of road image texture spectrum is analyzed. Based on the spectral variation of different texture analysis units in the texture spectrum feature data, the relationship between the variation of road image texture information in the frequency domain is analyzed, and the amount of variation of road image texture spectrum is obtained. For example, in normal road areas, the texture spectrum variation between adjacent texture analysis units usually maintains a continuous relationship, while in abnormal change areas, due to changes in the road surface structure, the local texture spectrum distribution will change. The amount of variation of road image texture spectrum can reflect the changes in road texture details.

[0085] Spatial correlation information in continuous road image sequences is analyzed. Based on the positional changes of anomalous change areas in the continuous road image sequence, the spatial correspondence of road texture features, and the regional texture correlation, the continuous correlation state between corresponding road regions at different times is analyzed to obtain the continuous correlation features of road visual disturbance feature data. By separately analyzing the dynamic change information of road texture, the spectral change information of road image texture, and the spatial correlation information in the continuous road image sequence, different change information in the road visual disturbance feature data is described accordingly, and the continuous correlation features of road texture feature change amount, road image texture spectral change amount, and road visual disturbance feature data are obtained respectively.

[0086] S4.2 Based on the changes in road texture features, establish spatial correlations between the changes in road texture features according to the spatial location corresponding to the road area, and combine the changes in the road image texture spectrum to establish a collaborative change relationship between road texture changes and texture spectrum changes, thereby obtaining collaborative change data of the road area.

[0087] Furthermore, based on the texture location coordinates corresponding to the changes in road texture features, spatial matching is performed on the changes in road texture features at different locations, and the changing trends between adjacent road areas are analyzed. For example, in continuous and smooth road areas, the changes in road texture features corresponding to adjacent road areas usually maintain a continuous changing relationship. When there are local abnormal areas on the road, the changes in road texture features corresponding to the abnormal areas may differ from those of the surrounding road areas. Combining the changes in the road image texture spectrum, the correspondence between road texture changes and texture spectrum changes is analyzed. By comparing the changes in road texture features corresponding to the same road area and the changes in road image texture spectrum, it is determined whether there is a synergistic relationship between changes in road texture structure and changes in texture frequency. For example, potholes in the road usually cause changes in both road texture structure and texture spectrum at the same time. Through synergistic analysis, the visual change state inside the road area can be reflected. Based on the spatial correlation between changes in road texture features and the synergistic relationship between changes in road texture and changes in texture spectrum, multidimensional change information in the road area is associated and described to obtain synergistic change data of the road area.

[0088] S4.3 Map the continuous correlation features of road visual disturbance feature data to the corresponding road areas in the road area collaborative change data, establish the continuous evolution relationship between the continuous correlation features of road area collaborative change data and road visual disturbance feature data, and obtain road area visual disturbance evaluation data.

[0089] Furthermore, based on the spatial location information in the continuous correlation features of the road visual disturbance feature data, the continuous correlation features are matched to the corresponding road areas in the road area collaborative change data, and the changes of the corresponding road areas at different times are analyzed. Based on the change relationship between the continuous correlation features of the road area collaborative change data and the road visual disturbance feature data, a continuous evolution relationship is established. By analyzing the temporal correspondence between the road texture changes, texture spectrum changes, and spatial change states in the road area collaborative change data and the continuous correlation features, it is determined whether the visual changes in the road area continue to exist, gradually increase, or gradually disappear. For example, visual changes in normal road areas usually maintain continuous changes with vehicle movement, while abnormal road areas may show continuous changes or abnormally increasing trends. Based on the continuous evolution relationship between the continuous correlation features of the road area collaborative change data and the road visual disturbance feature data, road area visual disturbance evaluation data is formed.

[0090] S4.4 Based on the visual disturbance evaluation data of road areas, according to the continuous evolution relationship between different road areas, the visual disturbance evaluation data of road areas is mapped to the road surface smoothness state to obtain the road surface smoothness evaluation results.

[0091] Furthermore, based on the changes in road texture, texture spectrum, and continuous evolution status in the visual disturbance evaluation data of road areas, state analysis is performed on different road areas. When the visual disturbance evaluation data of a road area maintains a stable and continuous change, it is determined that the road surface smoothness status of the corresponding road area remains stable. When the visual disturbance evaluation data of a road area shows an abnormal change trend, it is determined that there is a change in the smoothness status of the road area. Based on the continuous evolution relationship between different road areas, the visual disturbance evaluation data of road areas are correlated and mapped, converting the visual change status of road areas into the corresponding road surface smoothness status evaluation results. For example, when the visual disturbance evaluation data of multiple road areas in a continuous road area maintains a consistent change relationship, the corresponding road surface smoothness status remains continuous; when the visual disturbance evaluation data of an abnormal area changes, the corresponding road surface smoothness status changes. Based on the visual disturbance evaluation data of the road area, the road surface smoothness status mapping is completed to obtain the road surface smoothness evaluation result.

[0092] S5. Based on the road surface smoothness evaluation results and combined with the location data corresponding to the continuous road image sequence, generate rapid road surface smoothness detection results.

[0093] S5.1 Based on the road surface smoothness evaluation results, the road surface smoothness evaluation results are associated and matched with the location data corresponding to the continuous road image sequence to obtain the road area location evaluation data. The continuous spatial distribution relationship between the road area location evaluation data is established according to the location order corresponding to the continuous road image sequence to obtain the road smoothness spatial distribution data.

[0094] Furthermore, based on the correlation between the road surface smoothness evaluation results and the corresponding location data in the continuous road image sequence, road area location evaluation data is obtained. This data includes the locational relationships of road areas and the road surface smoothness evaluation results at those locations. This transforms the road surface smoothness evaluation results from regional visual evaluation information into data with spatial locational correspondences. Following the positional order of the continuous road image sequence, the road area location evaluation data is arranged continuously in space. Spatial correlation analysis is performed on the road area location evaluation data corresponding to adjacent road areas to establish continuous spatial distribution relationships between different road locations. For example, when adjacent road image frames in a continuous road image sequence correspond to adjacent road areas, the corresponding road area location evaluation data maintains a continuous spatial relationship. When there are changes in the smoothness of a road area, the road area location evaluation data at the corresponding location changes. This continuous spatial distribution relationship reflects the distribution of road smoothness in road space. By establishing continuous spatial distribution relationships between road area location evaluation data, spatial distribution data of road smoothness is obtained.

[0095] S5.2 Based on the spatial distribution data of road smoothness, the evaluation results of road surface smoothness between adjacent road areas are continuously integrated to form continuous road smoothness distribution data.

[0096] Furthermore, based on the positional relationship between adjacent road areas in the road smoothness spatial distribution data, the spatial connection relationship between adjacent road areas is determined, and the changes in road smoothness evaluation results between adjacent road areas are compared. When the road smoothness evaluation results between adjacent road areas maintain a continuous change relationship, the adjacent road areas are continuously integrated to form a continuous road smoothness state description. For example, in a continuously smooth road area, the changes in the road smoothness evaluation results corresponding to adjacent road areas are small, and a stable road smoothness distribution can be formed through continuous integration. When there are local abnormal changes in a road area, the road smoothness evaluation results between adjacent road areas will change. Through spatial correlation, the change characteristics corresponding to the abnormal area can be preserved. Based on the continuous spatial distribution relationship in the road smoothness spatial distribution data, the road smoothness evaluation results corresponding to multiple adjacent road areas are merged to form continuous road smoothness distribution data that can reflect the overall state of the road section. By continuously integrating the road smoothness evaluation results between adjacent road areas, continuous road smoothness distribution data is formed.

[0097] S5.3 Based on continuous road smoothness distribution data, the continuous road smoothness distribution data is continuously fused according to the positional order corresponding to the continuous road image sequence. The continuous road smoothness distribution data belonging to the same road segment are integrated, and the road surface smoothness rapid detection result is generated based on the integrated road segment smoothness distribution.

[0098] Furthermore, the road segment range is determined based on the positional order corresponding to the continuous road image sequence. The evaluation results of multiple road areas belonging to the same road segment in the continuous road smoothness distribution data are correlated, and the continuous road smoothness distribution data within the same road segment is integrated. By analyzing the changes in pavement smoothness evaluation results at different locations within the road segment, an overall road segment smoothness distribution is formed. For example, when multiple locations within the same road segment maintain similar smoothness states, the corresponding continuous road smoothness distribution data are integrated to form a stable road segment smoothness distribution. When there are local abnormal areas within the road segment, the smoothness changes corresponding to the abnormal locations are retained, allowing the road segment smoothness distribution to reflect local changes. Based on the integrated road segment smoothness distribution, a rapid pavement smoothness detection result is generated. The corresponding positions in the continuous road image sequence, the road segment smoothness distribution, and the pavement smoothness evaluation results are correlated and output to form the rapid pavement smoothness detection result for the corresponding road segment.

[0099] This embodiment also provides a rapid road surface smoothness detection system based on image recognition, including: The extraction module acquires a continuous road image sequence and extracts the effective area image sequence of the continuous road. Through road texture feature extraction processing, it obtains continuous road texture feature data. The module constructs a dynamic relationship between continuous road texture features and vehicle movement, based on continuous road texture feature data, and builds a road texture inertial field. The disturbance module determines the abnormal change area based on the road texture inertial field, and performs visual spectral disturbance analysis on the road image texture information in the abnormal change area to obtain road visual disturbance feature data. The analysis module, based on road visual disturbance feature data, analyzes the changes in texture features of road areas, the changes in texture spectrum of road images, and the continuous correlation features of road visual disturbance feature data to obtain road surface smoothness evaluation results. The detection results module generates rapid road surface smoothness detection results based on the road surface smoothness evaluation results and the location data corresponding to the continuous road image sequence.

[0100] This embodiment also provides a computer device applicable to the image recognition-based rapid road surface smoothness detection method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the image recognition-based rapid road surface smoothness detection method proposed in the above embodiment.

[0101] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices 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 and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0102] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the image recognition-based rapid road surface smoothness detection method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0103] In summary, this invention acquires continuous road image sequences and extracts effective road areas, then combines this with road texture feature extraction processing to obtain continuous road texture feature data. Based on this continuous road texture feature data, it establishes the dynamic change relationship of road texture during vehicle movement, constructs a road texture inertial field, and establishes continuous road texture feature association data through road texture feature spatial correlation analysis, texture structure matching, and regional texture correlation, thus achieving continuous representation of road texture dynamic change information. Based on the road texture inertial field, it constructs a road texture feature spatial connection network and forms a road texture dynamic propagation network according to the road texture dynamic change propagation relationship, identifies abnormal areas of road texture dynamic change, and performs visual spectrum perturbation analysis based on the road image texture information corresponding to the abnormal change areas to obtain road visual perturbation feature data. It integrates the continuous correlation features of road texture feature change, road image texture spectrum change, and road visual perturbation feature data to establish a road area visual perturbation evaluation relationship and obtain road surface smoothness evaluation results. Combining the location data corresponding to the continuous road image sequence, it performs spatial correlation and road segment fusion on the road surface smoothness evaluation results to generate rapid road surface smoothness detection results. By fusing continuous road texture dynamic change analysis with visual spectrum change features, rapid detection of road surface smoothness based on continuous image information was achieved.

[0104] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A rapid method for detecting road surface smoothness based on image recognition, characterized in that: include, A sequence of continuous road images was acquired, and the effective area image sequence of the continuous road was extracted. Through road texture feature extraction processing, continuous road texture feature data was obtained. Based on continuous road texture feature data, the dynamic change relationship of continuous road texture features during vehicle movement is established, and a road texture inertial field is constructed. Based on the inertial field of road texture, abnormal change areas are identified, and visual spectral perturbation analysis is performed on the road image texture information in the abnormal change areas to obtain road visual perturbation feature data. Based on road visual disturbance feature data, the changes in texture features of road areas, changes in the texture spectrum of road images, and the continuous correlation features of road visual disturbance feature data are analyzed to obtain the road surface smoothness evaluation results. Based on the road surface smoothness evaluation results, and combined with the location data corresponding to the continuous road image sequence, a rapid road surface smoothness detection result is generated.

2. The method for rapid road surface smoothness detection based on image recognition as described in claim 1, characterized in that: The extraction of the continuous road effective area image sequence includes, The system collects road image information during vehicle travel, arranges multiple road image frames according to the image acquisition time sequence to obtain a continuous road image sequence, and performs image region recognition processing on each road image frame in the continuous road image sequence to obtain road region location information. Based on the location information of the road area, each road image frame in the continuous road image sequence is cropped to obtain a continuous effective road area image sequence.

3. The method for rapid road surface smoothness detection based on image recognition as described in claim 2, characterized in that: The continuous road texture feature data includes Based on the continuous road effective area image sequence, image preprocessing is performed on each road image frame in the continuous road effective area image sequence to obtain the preprocessed continuous road effective area image sequence. Feature extraction processing is performed on the road surface texture information in the preprocessed continuous road effective area image sequence to obtain a road texture feature set; Based on the road texture feature set, the texture structure features, texture edge features, and texture distribution features in the road texture feature set are correlated to obtain continuous road texture feature data.

4. The method for rapid road surface smoothness detection based on image recognition as described in claim 3, characterized in that: The construction of the road texture inertial field includes, Based on continuous road texture feature data, according to the image acquisition time sequence corresponding to the continuous road texture feature data, temporal correlation processing is performed on the continuous road texture feature data corresponding to adjacent time to obtain the road texture feature change relationship at different time. The road texture features in the continuous road texture feature data are parsed to obtain the texture structure information, texture location coordinate information and texture neighborhood relationship information corresponding to each road texture feature in the continuous road texture feature data. Based on the texture location coordinate information, spatial location correlation analysis is performed on the road texture features in the continuous road texture feature data at different times, and the position coordinate offset between corresponding road texture features in adjacent image frames is calculated. Based on the location coordinate offset, determine the road texture features that are close in spatial location, combine the texture structure information to calculate the structural similarity between the road texture features corresponding to different times, and obtain the spatial correspondence of road texture features based on the location coordinate offset and structural similarity. Based on texture neighborhood information, the positional arrangement, structural distribution and mutual correlation of multiple associated road texture features around the road texture feature are jointly analyzed to construct the regional texture correlation relationship composed of road texture features. Based on the regional texture correlation relationship, the feature shift of a single road texture feature caused by lighting changes, road dust occlusion or image noise is corrected. Based on the texture structure association, the spatial correspondence of road texture features, and the regional texture association, the road texture change relationship in the continuous road texture feature data corresponding to different times is associated and described to obtain continuous road texture feature association data.

5. The method for rapid road surface smoothness detection based on image recognition as described in claim 4, characterized in that: The road visual disturbance feature data includes, Based on the road texture inertial field, the dynamic change state of road texture features in the road texture inertial field is traversed according to the spatial distribution, and the spatial connection relationship of road texture features is established according to the dynamic change state of road texture features with continuous spatial distribution. The preset spatial adjacency distance is determined based on the average spatial distance between adjacent road texture features in the continuous road texture feature data. Road texture features with a spatial distance not greater than the preset spatial adjacency distance and continuous dynamic change status are established to form a road texture feature spatial connection network. Based on the road texture feature spatial connection network, taking any road texture feature in the road texture feature spatial connection network as the starting node, and according to the dynamic change association parameters between adjacent road texture features in the road texture feature spatial connection network, the propagation path of the dynamic change state of the road texture feature in the road texture feature spatial connection network is calculated, and a road texture dynamic propagation network is formed according to the propagation path length, propagation direction change and propagation termination position. Based on the dynamic propagation network of road texture, the propagation continuity, propagation bifurcation and propagation termination relationships among the propagation paths in the dynamic propagation network of road texture are jointly analyzed. Based on the propagation continuity, propagation bifurcation and propagation termination relationships among the propagation paths, abnormal propagation regions are identified and abnormal change regions are obtained. Based on the abnormal change region, a spatial mapping relationship between the abnormal change region and the continuous road effective area image sequence is established in the continuous road effective area image sequence. According to the spatial mapping relationship and the image acquisition time order, the corresponding abnormal change region in the continuous road effective area image sequence is continuously tracked to obtain the road image texture information corresponding to the abnormal change region. Based on the texture information of road images, multiple texture analysis units are divided according to the spatial distribution direction of texture in the texture information of road images, and frequency domain transformation is performed on each unit. Based on the spatial arrangement relationship between the texture analysis units, a texture spectrum spatial distribution network is established to obtain texture spectrum feature data.

6. The method for rapid road surface smoothness detection based on image recognition as described in claim 5, characterized in that: The road surface smoothness evaluation results include: Based on road visual disturbance feature data, the dynamic change information of road texture, the spectral change information of road image texture, and the spatial correlation information in continuous road image sequences are analyzed to obtain the change amount of road texture features, the change amount of road image texture spectrum, and the continuous correlation features of road visual disturbance feature data, respectively. Based on the changes in road texture features, a spatial correlation between the changes in road texture features is established according to the spatial location corresponding to the road area. In addition, a collaborative change relationship between road texture change and texture spectrum change is established by combining the changes in road image texture spectrum, so as to obtain collaborative change data of road area. The continuous correlation features of road visual disturbance feature data are mapped to the corresponding road areas in road area collaborative change data, and the continuous evolution relationship between the continuous correlation features of road area collaborative change data and road visual disturbance feature data is established to obtain road area visual disturbance evaluation data. Based on the visual disturbance evaluation data of road areas, the road surface smoothness status is mapped according to the continuous evolution relationship between different road areas to obtain the road surface smoothness evaluation results.

7. The method for rapid road surface smoothness detection based on image recognition as described in claim 6, characterized in that: The rapid road surface smoothness detection results include: Based on the road surface smoothness evaluation results, the road surface smoothness evaluation results are associated and matched with the location data corresponding to the continuous road image sequence to obtain the road area location evaluation data. The continuous spatial distribution relationship between the road area location evaluation data is established according to the location order corresponding to the continuous road image sequence to obtain the road smoothness spatial distribution data. Based on the spatial distribution data of road smoothness, the evaluation results of road surface smoothness between adjacent road areas are continuously integrated to form continuous road smoothness distribution data. Based on continuous road smoothness distribution data, the continuous road smoothness distribution data is continuously fused according to the positional order corresponding to the continuous road image sequence. The continuous road smoothness distribution data belonging to the same road segment are integrated, and the road surface smoothness rapid detection result is generated based on the integrated road segment smoothness distribution.

8. A rapid road surface smoothness detection system based on image recognition, based on the rapid road surface smoothness detection method based on image recognition as described in any one of claims 1 to 7, characterized in that: include, The extraction module acquires a continuous road image sequence and extracts the effective area image sequence of the continuous road. Through road texture feature extraction processing, it obtains continuous road texture feature data. The module constructs a dynamic relationship between continuous road texture features and vehicle movement, based on continuous road texture feature data, and builds a road texture inertial field. The disturbance module determines the abnormal change area based on the road texture inertial field, and performs visual spectral disturbance analysis on the road image texture information in the abnormal change area to obtain road visual disturbance feature data. The analysis module, based on road visual disturbance feature data, analyzes the changes in texture features of road areas, the changes in texture spectrum of road images, and the continuous correlation features of road visual disturbance feature data to obtain road surface smoothness evaluation results. The detection results module generates rapid road surface smoothness detection results based on the road surface smoothness evaluation results and the location data corresponding to the continuous road image sequence.

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 image recognition-based rapid road surface smoothness detection 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 the processor, it implements the steps of the image recognition-based rapid road surface smoothness detection method according to any one of claims 1 to 7.

Images