A method and system for identifying environmental anomalies in DC battery swapping stations

Abstract

This invention provides a method and system for identifying environmental anomalies within a DC battery swapping station. The method includes fusing multi-source images acquired from the DC battery swapping station; determining at least one suspected anomaly region based on the fused images; constructing a three-dimensional structural model of an object within the suspected anomaly region based on the reflection deformation information of structured light projected multiple times into the suspected anomaly region; determining the reflection change information of the object within the suspected anomaly region based on the intensity and time of the reflected echo from the surface of the object in the suspected anomaly region using acquired ultrasonic waves; and performing logical correlation analysis based on the three-dimensional structural model and the reflection change information to determine the target anomaly region within the DC battery swapping station to be identified. This method and system significantly improve the ability to identify small and concealed anomalies in DC battery swapping stations, effectively reduce missed detections, improve the accuracy and reliability of identifying environmental anomalies within DC battery swapping stations, and ensure the stable operation of the station.

Description

Technical Field

[0001] This invention relates to the field of environmental anomaly identification, and more specifically, to a method and system for identifying environmental anomalies within a DC battery swapping station. Background Technology

[0002] Current methods for identifying environmental anomalies in DC battery swapping stations primarily rely on vision-based algorithms. This involves analyzing images captured by cameras within the station and using trained algorithm models to identify unusual objects such as floating debris, suspended objects, bird nests, smoke or fire, open covers, unclosed enclosure doors, and water accumulation in cable trenches. While existing vision-based methods can detect some anomalies, they are insufficient for identifying small or concealed objects. Because these algorithms depend heavily on visual features, when anomalies are small (e.g., tiny floating objects) or located in concealed areas (e.g., areas obscured by equipment or water accumulation), the anomaly features may be subtle, making accurate extraction and matching difficult. This can lead to missed detections and potential safety hazards going unnoticed, impacting the stable operation of the DC battery swapping station. Summary of the Invention

[0003] To address the technical problem that existing technologies for identifying environmental anomalies in DC battery swapping stations primarily rely on visual features of images, leading to frequent missed detections and the inability to promptly identify potential safety hazards, this invention provides a method and system for identifying environmental anomalies within DC battery swapping stations.

[0004] According to one aspect of the present invention, the present invention provides a method for identifying environmental anomalies within a DC battery swapping station, comprising:

[0005] Coordinate mapping is performed on the initial images of each environmental area within the DC battery swapping station to be identified to obtain a fused image of each environmental area. The initial images include high-definition images captured by high-definition camera equipment and initial thermal images captured by thermal imaging camera equipment within the DC battery swapping station to be identified.

[0006] At least one suspected anomalous region is identified based on the fused image of each environmental region;

[0007] Based on the structured light projected multiple times into the suspected abnormal area by the high-definition camera device, the reflection deformation information of the structured light on the surface of the object in the suspected abnormal area is obtained, and a three-dimensional structural model of the object in the suspected abnormal area is constructed based on the reflection deformation information.

[0008] The suspected abnormal area is scanned using an ultrasonic scanning detector to obtain the reflected echo intensity and reflection time of ultrasonic waves on the surface of the object in the suspected abnormal area, and the reflection change information of the object in the suspected abnormal area is determined based on the reflected echo intensity and the reflection time.

[0009] Logical correlation analysis is performed based on the three-dimensional structural model and reflection change information of objects within each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified.

[0010] According to another aspect of the present invention, the present invention provides a system for identifying environmental anomalies within a DC battery swapping station, the system comprising:

[0011] The image fusion module is used to perform coordinate mapping on the initial images of each environmental area within the DC battery swapping station to be identified, and to obtain the fused image of each environmental area. The initial images include high-definition images captured by high-definition camera equipment and initial thermal imaging images captured by thermal imaging camera equipment within the DC battery swapping station to be identified.

[0012] The region filtering module is used to identify at least one suspected abnormal region based on the fused image of each environmental region.

[0013] The structural model construction module is used to obtain the reflection deformation information of the structured light on the surface of the object in the suspected abnormal area based on the structured light projected multiple times into the environment of the suspected abnormal area by the high-definition camera device, and to construct a three-dimensional structural model of the object in the suspected abnormal area based on the reflection deformation information.

[0014] An ultrasound-assisted detection module is used to scan the suspected abnormal area based on an ultrasonic scanning detector, obtain the reflected echo intensity and reflection echo time of the ultrasonic waves on the surface of the object in the suspected abnormal area, and determine the reflection change information of the object in the suspected abnormal area based on the reflected echo intensity and the reflection echo time.

[0015] The abnormal area identification module is used to perform logical correlation analysis based on the three-dimensional structural model and reflection change information of objects in each suspected abnormal area to determine the target abnormal area in the DC battery swapping station to be identified.

[0016] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the methods described in any of the above aspects of the present invention.

[0017] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method described in any of the preceding aspects of the present invention.

[0018] The present invention discloses a method and system for identifying environmental anomalies within a DC battery swapping station. The method includes: mapping the initial images of each environmental area within the DC battery swapping station to be identified to obtain a fused image of each environmental area; determining at least one suspected anomaly area based on the fused image of each environmental area; acquiring reflection deformation information of objects within the suspected anomaly area based on structured light projected multiple times into the suspected anomaly area by a high-definition camera, and constructing a three-dimensional structural model of the objects within the suspected anomaly area based on the reflection deformation information; scanning the suspected anomaly area using an ultrasonic scanning detector to acquire the reflected echo intensity and reflection echo time of ultrasonic waves on the surface of objects within the suspected anomaly area, and determining reflection change information of objects within the suspected anomaly area based on the reflected echo intensity and reflection echo time; and performing logical correlation analysis based on the three-dimensional structural model and reflection change information of objects within each suspected anomaly area to determine the target anomaly area within the DC battery swapping station to be identified. The method and system described herein utilize multi-source data acquisition and fusion from high-definition and thermal imaging equipment to complement each other, expanding the range of detectable abnormal features. Combined with structured light and ultrasonic scanning for auxiliary detection, it addresses the limitations of visual algorithms in spatial detection from the perspectives of three-dimensional structure and concealed space detection. Finally, logical correlation analysis is performed on the three-dimensional structural model and reflection change information, breaking through the bottleneck of a single visual algorithm and significantly improving the ability to identify small and concealed abnormal objects. This effectively reduces missed detections and solves the problem of insufficient identification capability of existing methods for small or concealed abnormal objects, improving the accuracy and reliability of environmental anomaly identification within DC battery swapping stations and ensuring the stable operation of the stations. Attached Figure Description

[0019] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures:

[0020] Figure 1 A flowchart of a method for identifying environmental anomalies within a DC battery swapping station according to a preferred embodiment of the present invention;

[0021] Figure 2 This is a schematic diagram of the structure of a system for identifying environmental anomalies within a DC battery swapping station according to a preferred embodiment of the present invention;

[0022] Figure 3This is a schematic diagram of the structure of an electronic device according to a preferred embodiment of the present invention. Detailed Implementation

[0023] Exemplary embodiments of the invention will now be described with reference to the accompanying drawings. However, the invention may be embodied in many different forms and is not limited to the embodiments described herein. These embodiments are provided to fully and completely disclose the invention and to fully convey its scope to those skilled in the art. The terminology used in the exemplary embodiments illustrated in the drawings is not intended to limit the invention. In the drawings, the same units / elements are referred to by the same reference numerals.

[0024] Unless otherwise stated, the terms used herein (including technical terms) have their common meaning as understood by one of ordinary skill in the art. Furthermore, it is understood that terms defined in commonly used dictionaries should be understood to have a meaning consistent with the context of their relevant field, and not to be interpreted as having an idealized or overly formal meaning.

[0025] Exemplary methods

[0026] Figure 1 This is a flowchart illustrating a method for identifying environmental anomalies within a DC battery swapping station according to a preferred embodiment of the present invention. Figure 1 As shown, the method for identifying environmental anomalies within a DC battery swapping station according to this preferred embodiment begins with step 101.

[0027] In step 101, coordinate mapping is performed on the initial images of each environmental area within the DC battery swapping station to be identified to obtain a fused image of each environmental area. The initial images include high-definition images acquired by high-definition camera equipment and initial thermal imaging images acquired by thermal imaging camera equipment within the DC battery swapping station to be identified.

[0028] Preferably, the step of performing coordinate mapping on the initial image of each environmental area within the DC battery swapping station to be identified, and obtaining the fused image of each environmental area, includes:

[0029] In step 1101, in each environmental region, the gradient magnitude around each pixel in the high-resolution image is calculated to determine the edge intensity change of each pixel, the texture complexity magnitude around each pixel is calculated to determine the texture complexity change of each pixel, and the temperature gradient magnitude of each pixel in the initial thermal imaging image is calculated.

[0030] Specifically, in each environmental region, high-definition images captured by high-definition camera equipment are analyzed. For each pixel in the high-definition image, gradient operators (such as the Sobel operator) can be used to calculate the gradient magnitude around the pixel to determine the edge intensity change. The texture complexity magnitude around the pixel is calculated using the gray-level co-occurrence matrix method to determine the texture complexity change.

[0031] In step 1102, pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule are designated as first target key pixels, and pixels in the initial thermal imaging image whose temperature gradient amplitude satisfies a custom second recognition rule are designated as second target key pixels.

[0032] Preferably, pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule are designated as first target key pixels, and pixels in the initial thermal imaging image whose temperature gradient magnitude satisfies a custom second recognition rule are designated as second target key pixels. The first recognition rule is to determine a pixel as a first target key pixel when its edge intensity change is greater than a custom intensity change threshold and its texture complexity change is greater than a custom complexity change threshold. The second recognition rule is to determine a pixel as a second target key pixel when its temperature gradient magnitude is greater than a custom gradient magnitude threshold.

[0033] Specifically, points with large gradient magnitudes and texture complexity magnitudes are usually located at object edges. Therefore, pixels in the high-definition image whose edge intensity changes are greater than a preset intensity change threshold and whose texture complexity changes are greater than a preset complexity change threshold are used as the first identification rule to determine the first target key pixel. The preset intensity change threshold and preset complexity change threshold are set according to actual conditions. For the initial thermal imaging image acquired by the thermal imaging camera, after calculating the temperature gradient magnitude of each pixel, since pixels with large temperature gradient magnitudes represent areas of drastic temperature changes, pixels in the initial thermal imaging image whose temperature gradient magnitude is greater than a preset gradient magnitude are used as the second identification rule to determine the second target key pixel. The preset gradient magnitude is set according to actual conditions.

[0034] In one embodiment, in the battery charging area of ​​a DC battery swapping station, at the edge of a battery box in a high-definition image, the edge intensity variation value of pixels in this area is calculated using the Sobel operator, and its texture complexity is also different from the surrounding area using the gray-level co-occurrence matrix. These pixels are identified as first target key pixels. In the initial thermal imaging image, the temperature gradient amplitude of pixels near the heat sink of the battery box is large because there is a significant temperature difference between the heat sink and the surrounding environment. These pixels are identified as second target key pixels.

[0035] In step 1103, the similarity between the first target key pixel and the second target key pixel at different scales is calculated, and the first target key pixel and the second target key pixel that satisfy the custom pixel matching rule are taken as the target pixel pair between the high-definition image and the initial thermal imaging image.

[0036] Specifically, similarity calculations are performed on the first and second target key pixels at different scales using the Scale Invariant Feature Transform (SIFT) algorithm. Therefore, for each first and second target key pixel, its feature descriptors (such as SIFT feature descriptors) at different scales are extracted, and the Euclidean distance between the feature descriptors is calculated. The minimum Euclidean distance is then used as the pixel matching rule to find the most similar pixel pair, thus obtaining the target pixel pair between the high-resolution image and the initial thermal image. In one embodiment, SIFT feature extraction is performed on the first target key pixels in the high-resolution image of the battery charging area and the second target key pixels in the initial thermal image at different scales. For example, if the SIFT feature descriptor of a key pixel in the high-resolution image at a certain scale is D1, and there are multiple second target key pixels in the initial thermal image, the Euclidean distance between the SIFT feature descriptors of these pixels at the corresponding scale and D1 is calculated. After calculation, if the feature descriptor of one pixel has the smallest Euclidean distance to D1, then these two pixels form a target pixel pair.

[0037] In step 1104, based on the position coordinates of each pixel in the target pixel pair, a mapping error function representing the coordinate mapping relationship between the high-definition image and the initial thermal imaging image is determined by affine transformation.

[0038] Specifically, affine transformation is used to construct the coordinate mapping relationship between target pixel pairs. Affine transformation is a linear transformation that preserves the "parallelism" and "straightness" of the image. Let the pixel coordinates in the high-resolution image be (x1, y1), and the matching pixel coordinates in the initial thermal image be (x2, y2). The affine transformation matrix is:

[0039]

[0040] Then we have:

[0041]

[0042] Therefore, the mapping error function E of the target pixel pair can be expressed as:

[0043]

[0044] Where N represents the number of target pixel pairs.

[0045] Using the coordinates of multiple sets of target pixel pairs, the parameter 'a' in the mapping error function E is solved using the least squares method. ij and t x t y This will give you the coordinate mapping relationship.

[0046] In step 1105, the initial thermal imaging image is resampled according to the mapping error function to obtain the target thermal imaging image.

[0047] Specifically, for each pixel in the target thermal image, its corresponding position in the initial thermal image is found using a mapping error function that characterizes the coordinate mapping relationship between the high-resolution image and the initial thermal image. Since coordinate mapping may result in non-integer coordinates, an interpolation algorithm (such as bilinear interpolation) is needed to determine the pixel value at that position. After resampling, the initial thermal image and the high-resolution image are aligned in coordinates to obtain the target thermal image.

[0048] In one embodiment, in the battery charging area, the coordinates of a pixel in the target thermal imaging image are (x... t ,y t ), and its corresponding coordinates in the initial thermal imaging image are calculated based on the coordinate mapping relationship as (x src ,y src ), where x src and y src It is a non-integer. Using the bilinear interpolation algorithm, based on (x... src ,y src The pixel value of the target thermal image is calculated by taking the pixel values ​​of the four integer coordinate points around it, and the pixel value is assigned to the target thermal image. The entire initial thermal image is then resampled point by point to obtain the target thermal image.

[0049] In step 1106, the pixel values ​​of each pixel in the high-definition image and the pixel values ​​of each pixel in the target thermal imaging image are fused to generate a fused image of each environmental region.

[0050] Specifically, the pixel values ​​of pixels in the high-resolution image are fused with the corresponding pixel values ​​in the target thermal image. A simple weighted average can be used for this fusion. For example, for a color high-resolution image, the pixel values ​​of each color channel are fused with the corresponding pixel values ​​in the target thermal image according to certain rules. For example, if the pixel value of the high-resolution image is I... h (x,y), the pixel value of the target thermal imaging image is I t (x,y), the merged pixel value I f The values ​​(x, y) can be determined based on features such as image edges and textures. When a point is located in a strong edge region in a high-resolution image, the pixel value of the high-resolution image is given greater weight. Conversely, if the temperature change characteristics of the point are obvious in the thermal imaging image, the pixel value of the thermal imaging image is used more and given greater weight, ultimately resulting in a fused image for each environmental region. Continuing with the above embodiment, during the fusion process in the battery charging area, for pixels at the edge of the battery box in the high-resolution image, due to their obvious edge features, the pixel value of the high-resolution image is given higher weight during fusion, allowing the fused image to more clearly display the edge shape of the battery box in that area. For pixels in the heat sink area of ​​the thermal imaging image where temperature changes drastically, the pixel value of the thermal imaging image is given higher weight during fusion, highlighting the temperature information of that area and generating a fused image containing rich visual and temperature information.

[0051] The embodiments of the present invention integrate high-definition image visual information and thermal imaging image temperature information into a fused image, which can comprehensively display the appearance, structure and temperature distribution of equipment in various environmental areas. This provides a richer and more comprehensive data foundation for subsequent accurate identification of abnormal areas, improves the accuracy and reliability of environmental anomaly identification in DC battery swapping stations, and ensures the stable operation of battery swapping stations.

[0052] In step 102, at least one suspected anomalous region is identified based on the fused image of each environmental region.

[0053] Specifically, by setting a series of normal image feature thresholds, including feature ranges such as color, texture, and temperature distribution, the actual feature values ​​of the fused image of a certain environmental area are compared with these thresholds. When the actual feature values ​​exceed the normal range, the environmental area is marked as a suspected abnormal area.

[0054] Preferably, determining at least one suspected anomalous region based on the fused image of each environmental region includes:

[0055] In step 1201, for each environmental region, local image features of the fused image at different scales are extracted, and the gradient changes of the local image features at different scales in different feature directions are calculated.

[0056] Specifically, for each environmental region, local image features of the fused image at different scales are extracted. For example, the fused image's I(x,y), under the variable s representing the scale, is centered at point (x,y) with a window size of w. s Then the extracted local image features F s (x,y) is:

[0057]

[0058] Among them, G s (i,j) is the Gaussian weighting function at scale s, used to highlight the features of the central region of the window, where G... s The formula for calculating (i,j) is as follows:

[0059]

[0060] Where, σ s It is the standard deviation corresponding to scale s, and the window size w. s Related, σ s =w s / 4.

[0061] Furthermore, after extracting local image features at different scales, the gradient changes of local image features at different scales in different feature directions can be determined. In one embodiment, at scale s, the feature F at point (x,y) is... s The gradient change of (x,y) on the variable θ representing the characteristic direction. for:

[0062]

[0063] in, and They are F s The partial derivatives of (x, y) in the x and y directions are calculated using the central difference method, for example...

[0064] In step 1202, the feature distance is calculated based on the gradient change of any two feature points of the local image features at each scale in each feature direction.

[0065] In step 1203, the local image features whose feature distances satisfy the custom clustering rules are clustered to generate several image feature classes.

[0066] Preferably, local image features whose feature distances satisfy a custom clustering rule are clustered to generate several image feature classes, wherein the clustering rule is as follows:

[0067] When taking any feature point (x1, y1) in the local image features as the center, with a radius of ∈ s Within the neighborhood of , there is at least MinPt s There are 10 feature points, that is, d. s,θ {(x1,y1),(x i ,y i )}≤∈ s The point (x) i ,y i The quantity is no less than MinPt s When determining feature point (x1, y1) and feature point (x... i ,y i ) belong to the same image feature cluster, where ∈ s and MinPt s It is a clustering parameter associated with the variable s representing the scale, d s,θ {(x1,y1),(x i ,y i )} is the feature point (x1, y1) and the feature point (x i ,y i The feature distance, defined by the variable s representing the scale and the variable θ representing the feature direction, is calculated using the following formula:

[0068]

[0069] In the formula, F s (x1,y1) and F s (x i ,y i ) are, respectively, the feature points (x1, y1) and (x2, y3) under the variable s representing the scale. i ,y i Local image features extracted centered on ) and The feature points (x1, y1) and (x2, y2) are respectively defined by the variable s representing the scale. i ,y i Local image features F at ) s (x1,y1) and F s (x i ,y i The gradient change on the variable θ that characterizes the feature direction.

[0070] Specifically, the clustering parameters ∈ s and MinPt s The value is determined based on experience.

[0071] In step 1204, the mean of the first gradient change and the mean of the first feature of the fused image at different scales and in different feature directions are calculated, and the mean of the second gradient change and the mean of the second feature of several image feature classes at different scales and in different feature directions are calculated respectively. The mean of the first gradient change is obtained by averaging the gradient magnitudes of all local image features at a certain scale and feature direction. The mean of the first feature is obtained by averaging each dimension of all local image feature vectors at the corresponding scale and feature direction. The mean of the second gradient change is obtained by averaging the gradient magnitudes of all feature points of an image feature class. The mean of the second feature is obtained by averaging the feature values ​​of all feature points of an image feature class.

[0072] Specifically, the gradient magnitude and eigenvector of all local image features at different scales and feature directions, as well as the solution of the gradient magnitude and eigenvalue of all feature points of an image feature class, are conventional techniques and will not be elaborated here.

[0073] In step 1205, the image feature class whose first gradient change mean and first feature mean of the fused image, as well as the image feature class whose second gradient change mean and second feature mean satisfy the custom first screening rule, is taken as the target cluster.

[0074] Preferably, the image feature class whose first gradient change mean and first feature mean, and second gradient change mean and second feature mean satisfy a custom first filtering rule are used as the target cluster, wherein the expression of the first filtering rule is as follows:

[0075]

[0076] In the formula, and These are the mean of the first gradient change and the mean of the first feature of the fused image, respectively. c and v c α and β are the mean of the second gradient change and the mean of the second feature for several image feature classes, respectively, and the threshold values ​​are set.

[0077] In one embodiment, the average gradient change of the battery storage area image is calculated at different scales and feature orientations. eigenvalues For a given image feature cluster, calculate the mean gradient change m of its internal feature points. c =2.0, characteristic mean v c ={1.3,1.5,1.1}. Given α=0.3 and β=0.2, calculate |2.0-1.5|=0.5>0.3. The cluster meets the screening criteria and is identified as the target cluster.

[0078] In step 1206, based on the mean of the first gradient change and the mean of the first feature of each abnormal image block, the neighborhood image is expanded according to a custom expansion rule to generate a target image block, wherein the abnormal image block is the region where the local image features are located in the target cluster.

[0079] Preferably, the step of expanding the neighborhood image of each abnormal image patch according to a custom expansion rule based on the mean of the first gradient change and the mean of the first feature, to generate a target image patch, wherein the expression of the expansion rule is:

[0080]

[0081] In the formula, m ab and v ab These are the mean of the first gradient change and the mean of the first feature for each anomalous image patch, respectively, m nb and v nb γ and ε are the mean of the second gradient change and the mean of the second feature of the proposed expanded neighborhood image, respectively, and the threshold values ​​are set.

[0082] In one embodiment, an anomalous image patch has been identified in the battery storage region, with a gradient change mean m. ab =2.0, characteristic mean v ab ={1.3,1.5,1.1}. Its neighborhood contains an image patch with a mean gradient change m. nb =1.8, characteristic mean v nb ={1.2,1.4,1.0}, γ=0.3 and δ=0.2, calculate |1.8-2.0|=0.2<0.3,

[0083] The neighboring image patch meets the expansion condition and is included in the expansion range of the target image patch.

[0084] In step 1207, a topological structure analysis is performed on the image region constructed based on the target image block to generate a topological structure analysis result, and a suspected abnormal region in the image region constructed based on the topological structure analysis result is determined.

[0085] Specifically, topological structure analysis is performed on the image regions constructed based on the target image blocks. Topological structure analysis can employ graph theory correlation methods, treating the target image blocks as nodes in a graph. If two target image blocks are adjacent or have a specific relationship, an edge is established between them. By analyzing the topological characteristics of the graph, such as connectivity and node degree, the normality of the image region's structure is determined. For each environmental region, anomaly analysis is performed based on the topological structure analysis results. If anomalies are found in the graph's topological structure (such as isolated nodes, abnormal connected components, etc.), the corresponding image region is identified as a suspected anomaly region. In one embodiment, in the battery energy storage region, in the constructed target image block graph, one target image block was found to have very few connecting edges with most other target image blocks, almost forming an isolated node. According to topological structure analysis, this situation belongs to an abnormal topological structure; therefore, the image region containing this target image block is identified as a suspected anomaly region, indicating that there is equipment failure or other abnormal conditions in this region.

[0086] In step 103, based on the structured light projected multiple times into the suspected abnormal area by the high-definition camera device, the reflection deformation information of the structured light on the surface of the object in the suspected abnormal area is obtained, and a three-dimensional structural model of the object in the suspected abnormal area is constructed based on the reflection deformation information.

[0087] Specifically, when a high-definition camera repeatedly projects structured light into a suspected anomaly area, the structured light will be reflected from the surface of objects within that area. Because the shapes of the object surfaces vary, the deformation of the reflected light also varies. Therefore, by analyzing the deformation information of the reflected light from the acquired high-definition images and using algorithms such as triangulation principles, the spatial coordinates of each point on the object's surface can be calculated, thus constructing a three-dimensional structural model of the object within the suspected anomaly area.

[0088] In one embodiment, a suspected abnormal area exists in the transformer area of ​​the DC battery swapping station to be identified. A high-definition camera is controlled to repeatedly project striped structured light into this area. The structured light is projected onto the transformer casing and surrounding objects. By comparing the acquired high-definition images, it is found that the stripes of the reflected light are distorted and displaced in certain locations. By analyzing these deformations of the reflected light, the distance and angle information of each point on the object surface relative to the high-definition camera are calculated, ultimately enabling the construction of a three-dimensional structural model of the transformer casing and surrounding objects.

[0089] Preferably, constructing a three-dimensional structural model of the object within the suspected abnormal region based on the reflection deformation information includes:

[0090] In step 1301, the corner change rate and surface curve change rate are calculated based on the reflection deformation information.

[0091] Specifically, the corner point change rate can be calculated by comparing the differences in corner point positions formed by structured light reflection at different times or from different viewpoints. The surface curve change rate can be calculated by analyzing the changes in curvature, shape, etc., of the curve formed by structured light reflection on the object's surface. The above calculations are all conventional techniques and will not be elaborated further here.

[0092] In step 1302, target feature points within the suspected abnormal region are extracted based on the corner change rate and the surface curve change rate.

[0093] Specifically, extracting target feature points within the suspected abnormal region based on the corner change rate and the surface curve change rate is also a conventional technique, which can employ classic feature detection and differential geometry methods, etc., and will not be elaborated here.

[0094] In step 1303, based on the intrinsic parameter matrix of the high-definition camera device and the projection direction of the structured light, the position information of the target feature in three-dimensional spatial coordinates is calculated using the principle of triangulation.

[0095] Specifically, after determining the target feature points, the position information of the target feature points in three-dimensional spatial coordinates is calculated using the intrinsic parameter matrix of the high-definition camera (including information such as focal length and principal point coordinates) and the projection direction of the structured light, based on the principle of triangulation. For example, if the intrinsic parameter matrix is ​​K and the structured light projection direction vector is... Given the known image coordinates p, the formula X = λK can be used. -1 p calculates the three-dimensional coordinates X, where λ is a scaling factor related to the direction of structured light projection.

[0096] In one embodiment, structured light reflection deformation information is analyzed in the bus connection area of ​​the DC battery swapping station to be identified. For a section of structured light reflection pattern at a bus connection, by comparing multiple images, it is found that the positions of certain corner points move significantly in different images. The corner point change rate of these corner points is calculated to be large. At the same time, the shape and curvature of some surface curves change significantly, and their surface curve change rate is also high. These points are identified as target feature points. The intrinsic parameter matrix K of the high-definition camera equipment and the structured light projection direction vector are known. The coordinates of these target feature points in three-dimensional space can be calculated using the above formula. For example, the coordinates of a certain target feature point in three-dimensional space are (x1, y1, z1).

[0097] In step 1304, for each target feature point, its corresponding first neighbor point is selected according to a custom neighbor point selection rule, and the normal vector of the first local surface constructed based on the first neighbor point is used as the first local surface normal vector of each target feature point in its first local surface. The first local surface is a local surface constructed by each target feature point based on the first neighbor point, and the neighbor point selection rule is determined according to a custom distance or angle.

[0098] Specifically, for each target feature point, a certain number of points are selected within its neighborhood to construct a local surface. The selection of neighborhood points can be determined based on conditions such as distance or angle. The first local surface can be constructed using methods such as least squares fitting of a plane. After determining the first local surface, the normal vector of each target feature point in its first local surface is obtained by calculating the normal vector of this plane. Let P be a point on the local surface. i (x i ,y i ,z i For i = 1, 2, ..., n, fit the plane equation ax + by + cz + d = 0 using the least squares method, and find the normal vector. This is the first local surface normal vector.

[0099] Continuing with the above embodiment, in the busbar connection region, for the target feature point with coordinates (x1, y1, z1), neighborhood points are selected around it with a certain radius. Using the least squares method, plane fitting is performed on these neighborhood points to obtain the plane equation 2x + 3y - 1z + 5 = 0. Therefore, the first local surface normal vector of the target feature point on its local surface is...

[0100] In step 1305, for each target feature point, a first neighboring point is selected according to a custom neighboring point determination condition, and the normal vector of the constructed second local surface is calculated based on the second neighboring point as the second local surface normal vector of each second neighboring point in its second local surface, wherein the second local surface is a local surface constructed by each first neighboring point based on the second neighboring point.

[0101] Specifically, the second local surface normal vector of the first neighboring point is also obtained by constructing the second local surface through its own neighboring points and calculating it, which will not be elaborated here.

[0102] In step 1306, the curvature of each target feature point on its first local surface is determined using a custom curvature calculation formula based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point.

[0103] Preferably, the curvature of each target feature point on its first local surface is determined using a custom curvature calculation formula based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point. The expression for the curvature calculation formula is:

[0104] k = sinθ / d

[0105] In the formula, θ and d are the angle and distance between the first local surface normal vector and the corresponding second local surface normal vector of the target feature point, respectively.

[0106] Continuing with the above embodiment, in the busbar connection region, for a target feature point with coordinates (x1, y1, z1), there is a neighboring point with coordinates (x2, y2, z2), and the second local surface normal vector of this neighboring point is... calculate and The included angle θ can be determined using the vector dot product formula. achievable And thus obtain The distance d between two points is calculated using the spatial distance formula. For example, if d = 0.5, the curvature k of the target feature point on its local surface can be obtained by substituting it into the curvature formula.

[0107] In step 1307, based on the curvature, a clustering algorithm is used to cluster the target feature points to generate several clusters.

[0108] Specifically, based on the curvature of each target feature point on its local surface, a clustering algorithm (such as K-Means++) is used to cluster the feature points. When selecting initial cluster centers, the K-Means++ algorithm prioritizes points farther from existing cluster centers to improve clustering performance. Target feature points with similar curvature are grouped into the same class, resulting in feature point clusters. During the clustering process, the distance from each feature point to each cluster center (e.g., Euclidean distance) is calculated, and the feature point is assigned to the class containing the nearest cluster center.

[0109] Continuing with the above embodiment, curvature values ​​of numerous target feature points are obtained in the bus connection region. The K-Means++ clustering algorithm is used, for example, setting the number of clusters to 3. First, a curvature value is selected as the first cluster center. Then, a second cluster center is selected based on its distance from the first cluster center, and so on, to determine the initial cluster centers. Next, the Euclidean distance from each target feature point to these three cluster centers is calculated, and the feature points are assigned to the cluster containing the nearest cluster center, ultimately resulting in three feature point clusters. For example, cluster 1 contains feature points with curvature within a certain range, and clusters 2 and 3 are obtained similarly.

[0110] In step 1308, each cluster is treated as a node, and a topological map of objects within the suspected abnormal region is constructed based on the connection relationships between the nodes.

[0111] Specifically, the topological structure of the object's surface is constructed based on different feature point clusters and their connections. Connections are determined by the spatial relationships between feature points; if two feature points are close to each other or located in the same local region, they are considered connected. Each feature point cluster is treated as a node, and the connections between nodes are considered as edges, thus constructing a topological graph.

[0112] Continuing with the above embodiment, in the bus connection region, there are three clusters of feature points. Analysis reveals that some feature points in cluster 1 are spatially close to some feature points in cluster 2, indicating a connection. Similar connections exist between cluster 2 and cluster 3. A topology graph is constructed using these three clusters as nodes and the connections between them as edges.

[0113] In step 1309, based on the three-dimensional coordinate information of the target feature points, a network generation algorithm is used to transform the topology map into a three-dimensional structural model of the object within the suspected abnormal region.

[0114] Specifically, by analyzing the topology map and combining the three-dimensional coordinate information of feature points, a three-dimensional structural model of objects in the suspected abnormal area can be constructed. Mesh generation algorithms can be used to transform the topology into a three-dimensional model.

[0115] Continuing with the above embodiments, after constructing the topology map, the topology map is transformed into a three-dimensional structural model of the object surface in the busbar connection area by combining the three-dimensional coordinate information of each feature point and using a mesh generation algorithm, clearly showing the shape, structure and other information of the busbar connection.

[0116] This invention enables the accurate construction of a 3D structural model of objects within a suspected anomaly area based on reflection deformation information. This allows for the extraction of key features from complex structured light reflection data in DC battery swapping station scenarios, identifying detailed surface features such as curvature variations. Through clustering and topology construction, the shape and structural relationships of the object's surface are fully presented. The resulting 3D structural model can be used for intuitive and accurate analysis of object surface conditions, improving the accuracy and reliability of anomaly identification within the DC battery swapping station and ensuring its stable operation.

[0117] In step 104, the suspected abnormal area is scanned using an ultrasonic scanning detector to obtain the reflected echo intensity and reflection time of ultrasonic waves on the surface of the object in the suspected abnormal area, and the reflection change information of the object in the suspected abnormal area is determined based on the reflected echo intensity and the reflection time.

[0118] Specifically, an ultrasonic scanning detector is used to scan the suspected abnormal area. Ultrasonic waves are reflected from the object's surface, and the intensity and duration of the reflected echoes carry information such as the object's surface material and distance. Therefore, the intensity and duration of the reflected echoes from the object's surface within the suspected abnormal area can be obtained. These reflected echo intensity and duration are then compared with standard values ​​under normal conditions to calculate the reflection change information, which can be used to determine whether there are defects or abnormalities on the object's surface.

[0119] In one embodiment, an ultrasonic scanning detector emits ultrasonic waves in a suspected abnormal area of ​​the bus connection region of a DC battery swapping station to be identified. When the ultrasonic waves encounter the metal surface at the bus connection, the reflected echo is received by the detector. If there are problems such as loosening or oxidation at the bus connection, the intensity of the reflected echo will be weakened, and the reflection time may also differ from that under normal connection conditions. Therefore, by comparing the actual received reflected echo intensity and time with standard values, the reflection change information of objects in the suspected abnormal area can be determined, and it can be judged whether there is a potential fault in the bus connection region.

[0120] Preferably, determining the reflection change information of objects within the suspected abnormal area based on the reflected echo intensity and the reflected echo time includes:

[0121] In step 1401, for the reflected echo time, a nonlinear transformation using chaotic mapping is used to obtain the transformed time series.

[0122] Specifically, for the reflected echo time data T={t1,t2,...,t n}, a nonlinear transformation is performed using chaotic mapping, where the nonlinear transformation formula is: Among them, t i It is the i-th time point in the original reflected echo time. This is the transformed time point. a, b, c, d, e are chaos parameters, with values ​​ranging from (0, 1).

[0123] In step 1402, the intensity of the reflected echo is analyzed by fractional Fourier transform to obtain the intensity spectrum after fractional Fourier transform.

[0124] Specifically, for the reflected echo intensity data I={i1,i2,...,i n This paper describes a spectral analysis based on the fractional Fourier transform (FRFT). FRFT analyzes signals in the time-frequency domain at a fractional order, capturing the frequency characteristics of signals with greater precision compared to the traditional Fourier transform. The formula for the fractional Fourier transform algorithm is as follows: Among them, I FRFT(u) is the result of the reflected echo intensity in the fractional Fourier transform domain, i(t) is the original reflected echo intensity signal, and K a (t,u) is the kernel function of the fractional Fourier transform, where a is a fractional transform with a range of (0,2).

[0125] In step 1403, based on the time series and the intensity spectrum, a custom time-frequency joint feature is calculated using a custom time-frequency joint feature expression after feature fusion based on the time series and the intensity spectrum.

[0126] Preferably, based on the time series and the intensity spectrum, a custom time-frequency joint feature expression is used to calculate the time-frequency joint feature after feature fusion based on the time series and the intensity spectrum, wherein the time-frequency joint feature expression is:

[0127]

[0128] In the formula, J(u,t) new ) is the joint time-frequency feature, θ is the parameter controlling the strength of the time correlation, and I FRFT (u) is the intensity spectrum, t new It is a time point in the time series, t ref It is a reference time point, and the median value of the time series can be selected.

[0129] In this preferred embodiment, The value is set based on experience.

[0130] In step 1404, a feature matrix is ​​constructed based on the time-frequency joint features, and singular value decomposition is performed on the feature matrix to obtain a singular value distribution. Then, a clustering analysis of dynamic density peaks is performed based on the singular value distribution to obtain a clustering result. The rows of the feature matrix are different time-frequency joint feature vectors.

[0131] Specifically, a feature matrix A = U∑V is constructed based on the aforementioned time-frequency joint features. T In this matrix, U and V are orthogonal matrices, and ∑ is a diagonal matrix. Singular value decomposition is performed on the characteristic matrix A, and the values ​​on the diagonal of the diagonal matrix are the singular values ​​σ. i Singular value distributions are obtained by calculating the density and distance relationships between singular values. For example, density can be defined as the ratio of the number of singular values ​​in a singular value's neighborhood to the size of the neighborhood, and distance can be calculated using Euclidean distance. Based on the singular value distribution, dynamic density peak clustering analysis is performed. In the singular value space, points with high density and relatively large distances from other high-density points are identified as cluster centers. Other singular values ​​are then assigned to corresponding clusters based on their distance from the cluster centers, resulting in the clustering results.

[0132] In one embodiment, the feature matrix A constructed in the bus connection region is an N*M matrix, and singular value decomposition is performed on it to obtain singular values ​​σ1, σ2, ..., σ r r = min(N, M). Given a neighborhood size of Δ, calculate each singular value σ. i density d i For example, in σ i Within a neighborhood of center Δ, there are n i If there are singular values, then d i =n i / Δ. Calculate the Euclidean distance between singular values. By analyzing density and distance, several dynamic density peak points are identified as cluster centers, and other singular values ​​are assigned to different clusters.

[0133] In step 1405, the reflection change information of objects in the suspected abnormal region is determined based on the change of the feature parameters of the clustering result over time, wherein the feature parameters include the center position, radius and shape factor of the cluster.

[0134] Specifically, the reflection changes of objects within suspected anomalous regions are determined based on the changes in the characteristic parameters of the clustering results over time. These characteristic parameters include the cluster center position, radius, and shape factor. The cluster center position reflects the central tendency of the joint time-frequency features, the radius represents the dispersion of the cluster, and the shape factor, obtained through geometric calculations, describes the shape of the cluster. For example, by comparing the changes in the cluster center position (ΔC), radius (ΔR), and shape factor (ΔS) at different times, the changes in the object's reflection characteristics are judged. If the changes in the center position (ΔC), radius (ΔR), and shape factor (ΔS) exceed a certain threshold, the object's reflection is considered to have changed significantly.

[0135] The embodiments of the present invention can extract rich and effective time-frequency joint features from the reflection echo time and intensity data, and mine potential patterns in the data through cluster analysis and other means. This enables the accurate capture of the time-varying information of the reflection characteristics of objects in suspected abnormal areas in DC battery swapping stations, which helps to accurately analyze the condition of object surfaces, detect minor defects, deformations and other abnormalities on the object surfaces, improve the accuracy and reliability of environmental anomaly identification in DC battery swapping stations, and ensure the stable operation of the battery swapping stations.

[0136] In step 105, logical correlation analysis is performed based on the three-dimensional structural model and reflection change information of objects within each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified.

[0137] Preferably, the step of performing logical correlation analysis based on the three-dimensional structural model and reflection change information of objects within each suspected anomaly area to determine the target anomaly area within the DC battery swapping station to be identified includes:

[0138] In step 1501, for each suspected abnormal region, the curvature of each triangular facet is calculated based on the coordinates of the three vertices of each triangular facet in the three-dimensional structural model and the curvature of each triangular facet in the two principal directions.

[0139] Specifically, for each suspected anomaly region, each triangular facet in the 3D structural model is obtained. Based on the coordinates of the three vertices of each facet and its curvature in the two principal directions, the curvature of each facet is calculated. For example, if the coordinates of the three vertices of a facet are P1(x1,y1,z1), P2(x2,y2,z2), and P3(x3,y3,z3), then the curvature H of that facet is calculated. m for:

[0140]

[0141] Where |(P2-P1)*(P3-P1)| represents the modulus of the cross product of vectors (P2-P1) and (P3-P1), which is twice the area of ​​the triangular facet; |P2-P1| and P3-P1| are the moduli of vectors (P2-P1) and (P3-P1) respectively; κ1 and κ2 are the curvatures of the triangular facet in the two principal directions, which are calculated by differential geometry.

[0142] In step 1502, the mean curvature of all triangular facets in the three-dimensional structural model is calculated to determine the curvature characteristic value of the three-dimensional structural model.

[0143] In step 1503, an association matrix is ​​constructed with the curvature feature value as the row and the rate of change feature value of the reflection change information as the column, wherein the rate of change feature value is calculated based on the rate of change of reflected echo intensity and the rate of change of reflected echo time.

[0144] In this preferred embodiment, the matrix elements in the correlation matrix represent the degree of correlation between curvature and rate of change.

[0145] In step 1504, the correlation matrix is ​​subjected to eigenvalue decomposition to determine the matrix eigenvector and the matrix eigenvalue corresponding to the matrix eigenvector, wherein the matrix eigenvector represents the feature pattern in the correlation matrix, and the matrix eigenvalue represents the importance of different feature patterns in the correlation matrix.

[0146] Specifically, the constructed correlation matrix W is subjected to eigenvalue decomposition, where the eigenvalue decomposition formula is: λ iThese are the eigenvalues ​​of the matrix. These are the corresponding matrix eigenvectors. The eigenvalues ​​λ are obtained by solving the characteristic equation |W-λI|=0. i Then substitute the eigenvalues ​​into The eigenvectors are obtained by solving. Where I represents the identity matrix. The eigenvalues ​​of the matrix are λ. i Eigenvectors represent the importance of different feature patterns in the correlation matrix; larger eigenvalues ​​correspond to more significant feature patterns. The characteristic patterns in the correlation matrix are characterized by their orientation, which reflects the linear combination relationship between curvature eigenvalues ​​and reflection rate eigenvalues.

[0147] In one embodiment, for the 3*2 correlation matrix W constructed for the suspected abnormal region of the transformer area, the eigenvalues ​​λ1 = 0.8 and λ2 = 0.3 (hypothetical values) are obtained by solving the characteristic equation |W-λI|=0. Substituting λ1 into... Solving for the corresponding eigenvectors yields the desired results. Similarly, we can obtain Here, λ1 = 0.8 indicates that its corresponding feature pattern is more important in the correlation matrix.

[0148] In step 1505, logical correlation analysis is performed based on the matrix feature vector and matrix feature value of each suspected abnormal region to determine the target abnormal region within the DC battery swapping station to be identified.

[0149] Preferably, the step of performing logical correlation analysis based on the matrix feature vector and matrix feature value of each suspected abnormal region to determine the target abnormal region within the DC battery swapping station to be identified includes:

[0150] In step 1551, for each suspected abnormal region, the matrix feature vector whose corresponding matrix feature values ​​satisfy the custom second screening rule is taken as the target feature vector.

[0151] Preferably, for each suspected abnormal region, the matrix feature vector whose corresponding matrix feature value satisfies the custom second filtering rule is taken as the target feature vector, wherein the second filtering rule is to determine the matrix feature vector as the target feature vector when the matrix feature value corresponding to a matrix feature vector is greater than the custom feature value threshold.

[0152] In step 1552, the target feature vector is combined to generate a feature descriptor for each suspected abnormal region.

[0153] Specifically, the selected target feature vectors are combined to obtain anomaly region feature descriptors. This feature combination can be achieved by concatenating the target feature vectors into a new vector in a specific order. For example, if there are n target feature vectors... And the dimensions of each target feature vector are d1, d2, ..., d. n Then the abnormal region feature descriptor It can be represented as Where, x T The transpose of the vector is used to represent the anomaly region feature descriptor, which integrates information from multiple important feature patterns and can more comprehensively describe the features of the suspected anomaly region.

[0154] In one embodiment, two target feature vectors have been identified in the suspected abnormal region of the battery energy storage area. and They are concatenated to obtain the feature descriptor of the abnormal region.

[0155] In step 1553, for any two suspected abnormal regions, differential analysis is performed based on their feature descriptors to obtain the difference value of the abnormal regions.

[0156] Specifically, for any two suspected anomalous regions, a difference analysis is performed based on their respective anomalous region feature descriptors to obtain the anomalous region difference value. The difference analysis can use metrics such as Euclidean distance, which will not be elaborated here.

[0157] In step 1554, the suspected abnormal areas whose abnormal area difference values ​​meet the custom third screening rules are identified as target abnormal areas within the DC battery swapping station to be identified.

[0158] Preferably, the suspected abnormal areas whose abnormal area difference values ​​satisfy the custom third screening rule are determined as target abnormal areas within the DC battery swapping station to be identified. The third screening rule is that when the abnormal area difference value between a suspected abnormal area and a custom number of other suspected abnormal areas is greater than a custom difference threshold, and the abnormal area difference value between a suspected abnormal area and all other suspected abnormal areas is greater than a custom average threshold, the suspected abnormal area is determined as a target abnormal area within the DC battery swapping station to be identified.

[0159] Specifically, set the custom quantity to N. th The difference threshold is D th The mean threshold is M. th For each suspected anomalous region, calculate its difference from other suspected anomalous regions. If a suspected anomalous region differs from N... th The difference in abnormal regions among other suspected abnormal regions is greater than D. th Furthermore, the mean difference between this suspected abnormal region and all other suspected abnormal regions is greater than M. thIf it is identified as an anomaly region, it is characterized as having a significant difference in feature patterns from most other regions.

[0160] In one embodiment, there are 5 suspected abnormal areas R1, R2, R3, R4, and R5 in a DC battery swapping station, with a preset number N. th =3, preset difference threshold D th =0.25, preset mean threshold M th =0.2. For the suspected anomalous region R3, the anomalous region differences between it and R1, R2, and R5 are 0.3, 0.28, and 0.35, respectively, all greater than 0.25. The mean anomalous region difference between R3 and the other four suspected anomalous regions is 0.26, which is greater than 0.2. Therefore, the anomaly identification system identifies R3 as the target anomalous region.

[0161] The preferred embodiment of this invention can accurately identify the target abnormal area within a DC battery swapping station from multiple suspected abnormal areas. By calculating the curvature feature value of the three-dimensional structural model, it can effectively capture the geometric shape change information of the object surface. By constructing an association matrix and performing feature decomposition, the potential relationship between curvature features and reflection change features is explored. Therefore, it can comprehensively utilize multi-source information to accurately determine which areas have abnormal situations that truly affect the normal operation of the DC battery swapping station, reduce false positives and false negatives, improve the accuracy and reliability of environmental anomaly identification within the DC battery swapping station, and ensure the stable operation of the battery swapping station.

[0162] The preferred embodiment of this method for identifying environmental anomalies within a DC battery swapping station utilizes multi-source data acquisition and fusion from high-definition and thermal imaging equipment. This allows high-definition and thermal imaging images to complement each other, expanding the range of detectable anomaly features. Furthermore, it incorporates structured light and ultrasonic scanning detection to assist in detection, addressing the limitations of visual algorithms in spatial detection from the perspectives of three-dimensional structure and concealed space detection. Finally, it performs logical correlation analysis on the three-dimensional structural model and reflection change information, overcoming the bottleneck of a single visual algorithm. This significantly improves the ability to identify small and concealed anomalies, effectively reducing missed detections and solving the problem of insufficient ability to identify small or concealed anomalies in existing methods. This enhances the accuracy and reliability of identifying environmental anomalies within the DC battery swapping station, ensuring the stable operation of the station.

[0163] Exemplary System

[0164] Figure 2 This is a schematic diagram of the structure of a system for identifying environmental anomalies within a DC battery swapping station according to a preferred embodiment of the present invention. Figure 2 As shown, the system 200 for identifying environmental anomalies within a DC battery swapping station according to this preferred embodiment includes:

[0165] The image fusion module 201 is used to perform coordinate mapping on the initial images of each environmental area in the DC battery swapping station to be identified, and to obtain the fused image of each environmental area. The initial images include high-definition images captured by high-definition camera equipment and initial thermal imaging images captured by thermal imaging camera equipment in the DC battery swapping station to be identified.

[0166] The region filtering module 202 is used to determine at least one suspected abnormal region based on the fused image of each environmental region;

[0167] The structural model construction module 203 is used to obtain the reflection deformation information of the structure light on the surface of the object in the suspected abnormal area based on the structure light projected multiple times into the environment of the suspected abnormal area by the high-definition camera device, and to construct a three-dimensional structural model of the object in the suspected abnormal area based on the reflection deformation information.

[0168] The ultrasonic-assisted detection module 204 is used to scan the suspected abnormal area based on an ultrasonic scanning detector, obtain the reflected echo intensity and reflection echo time of the ultrasonic waves on the surface of the object in the suspected abnormal area, and determine the reflection change information of the object in the suspected abnormal area based on the reflected echo intensity and the reflection echo time.

[0169] The abnormal area identification module 205 is used to perform logical correlation analysis based on the three-dimensional structural model and reflection change information of objects in each suspected abnormal area to determine the target abnormal area in the DC battery swapping station to be identified.

[0170] Preferably, the image fusion module 201 performs coordinate mapping on the initial images of each environmental area within the DC battery swapping station to be identified, and obtains a fused image of each environmental area, including:

[0171] In each environmental region, the gradient magnitude around each pixel in the high-resolution image is calculated to determine the edge intensity change of each pixel, the texture complexity magnitude around each pixel is calculated to determine the texture complexity change of each pixel, and the temperature gradient magnitude of each pixel in the initial thermal image is calculated.

[0172] The pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule are taken as first target key pixels, and the pixels in the initial thermal imaging image whose temperature gradient amplitude satisfies a custom second recognition rule are taken as second target key pixels.

[0173] Calculate the similarity between the first target key pixel and the second target key pixel at different scales, and take the first target key pixel and the second target key pixel that satisfy the custom pixel matching rule as the target pixel pair between the high-definition image and the initial thermal imaging image;

[0174] Based on the position coordinates of each pixel in the target pixel pair, an affine transformation is used to determine a mapping error function that characterizes the coordinate mapping relationship between the high-definition image and the initial thermal imaging image.

[0175] The initial thermal imaging image is resampled according to the mapping error function to obtain the target thermal imaging image;

[0176] The pixel values ​​of each pixel in the high-definition image and the pixel values ​​of each pixel in the target thermal image are fused to generate a fused image of each environmental region.

[0177] Preferably, the image fusion module 201 uses pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule as first target key pixels, and uses pixels in the initial thermal imaging image whose temperature gradient magnitude satisfies a custom second recognition rule as second target key pixels. The first recognition rule is to determine a pixel as a first target key pixel when the edge intensity change of a pixel is greater than a custom intensity change threshold and the texture complexity change is greater than a custom complexity change threshold. The second recognition rule is to determine a pixel as a second target key pixel when the temperature gradient magnitude of a pixel is greater than a custom gradient magnitude threshold.

[0178] Preferably, the region filtering module 202 determines at least one suspected abnormal region based on the fused image of each environmental region, including:

[0179] For each environmental region, local image features of the fused image at different scales are extracted, and the gradient changes of the local image features at different scales in different feature directions are calculated.

[0180] The feature distance is calculated based on the gradient change of any two feature points in each feature direction at each scale.

[0181] The local image features whose feature distances satisfy the custom clustering rules are clustered to generate several image feature classes;

[0182] The first gradient change mean and the first feature mean of the fused image are calculated at different scales and different feature directions. The second gradient change mean and the second feature mean of several image feature classes are calculated at different scales and different feature directions. The first gradient change mean is obtained by averaging the gradient magnitudes of all local image features at a certain scale and feature direction. The first feature mean is obtained by averaging each dimension of all local image feature vectors at the corresponding scale and feature direction. The second gradient change mean is obtained by averaging the gradient magnitudes of all feature points of an image feature class. The second feature mean is obtained by averaging the feature values ​​of all feature points of an image feature class.

[0183] The image feature class whose first gradient change mean and first feature mean of the fused image, as well as the second gradient change mean and second feature mean of the image feature class satisfy the custom first screening rule, is taken as the target cluster;

[0184] Based on the mean of the first gradient change and the mean of the first feature of each abnormal image patch, the neighborhood image is expanded according to a custom expansion rule to generate a target image patch, wherein the abnormal image patch is the region where the local image features are located in the target cluster;

[0185] A topological structure analysis is performed on the image region constructed based on the target image block to generate a topological structure analysis result, and suspected abnormal regions in the image region constructed based on the topological analysis result are determined.

[0186] Preferably, the region filtering module 202 clusters local image features whose feature distances satisfy a custom clustering rule to generate several image feature classes, wherein the clustering rule is as follows:

[0187] When taking any feature point (x1, y1) in the local image features as the center, with a radius of ∈ s Within the neighborhood of , there is at least MinPt s There are 10 feature points, that is, d. s,θ {(x1,y1),(x i ,y i )}≤∈ s The point (x) i ,y i The quantity is no less than MinPt s When determining feature point (x1, y1) and feature point (x... i ,y i ) belong to the same image feature cluster, where ∈ s and MinPt s It is a clustering parameter associated with the variable s representing the scale, d s,θ{(x1,y1),(x i ,y i )} is the feature point (x1, y1) and the feature point (x i ,y i The feature distance, defined by the variable s representing the scale and the variable θ representing the feature direction, is calculated using the following formula:

[0188]

[0189] In the formula, F s (x1,y1) and F s (x i ,y i ) are, respectively, the feature points (x1, y1) and (x2, y3) under the variable s representing the scale. i ,y i Local image features extracted centered on ) and The feature points (x1, y1) and (x2, y2) are respectively defined by the variable s representing the scale. i ,y i Local image features F at ) s (x1,y1) and F s (x i ,y i The gradient change on the variable θ that characterizes the feature direction.

[0190] Preferably, the region filtering module 202 uses image feature classes whose first gradient change mean and first feature mean, as well as second gradient change mean and second feature mean satisfy a custom first filtering rule, as target clusters. The expression for the first filtering rule is as follows:

[0191]

[0192] In the formula, and These are the mean of the first gradient change and the mean of the first feature of the fused image, respectively. c and v c α and β are the mean of the second gradient change and the mean of the second feature for several image feature classes, respectively, and the threshold values ​​are set.

[0193] Preferably, the region filtering module 202 expands the neighborhood image of each abnormal image patch according to a custom expansion rule based on the mean of the first gradient change and the mean of the first feature, thereby generating a target image patch. The expression for the expansion rule is:

[0194]

[0195] In the formula, mab and v ab These are the mean of the first gradient change and the mean of the first feature for each anomalous image patch, respectively, m nb and v nb γ and ε are the mean of the second gradient change and the mean of the second feature of the proposed expanded neighborhood image, respectively, and the threshold values ​​are set.

[0196] Preferably, the structural model construction module 203 constructs a three-dimensional structural model of the object within the suspected abnormal region based on the reflection deformation information, including:

[0197] Calculate the corner point change rate and the surface curve change rate based on the reflected deformation information;

[0198] Based on the corner change rate and the surface curve change rate, target feature points within the suspected abnormal region are extracted;

[0199] Based on the intrinsic parameter matrix of the high-definition camera and the projection direction of the structured light, the position information of the target feature in three-dimensional spatial coordinates is calculated using the principle of triangulation.

[0200] For each target feature point, its corresponding first neighbor point is selected according to a custom neighbor point selection rule, and the normal vector of the first local surface constructed based on the first neighbor point is used as the first local surface normal vector of each target feature point in its first local surface. The first local surface is a local surface constructed by each target feature point based on the first neighbor point. The neighbor point selection rule is determined according to a custom distance or angle.

[0201] For each target feature point, a first neighboring point is selected according to a custom neighboring point determination condition, and the normal vector of the constructed second local surface is calculated based on the second neighboring point as the second local surface normal vector of each second neighboring point in its second local surface. The second local surface is a local surface constructed by each first neighboring point based on the second neighboring point.

[0202] Based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point, the curvature of each target feature point on its first local surface is determined using a custom curvature calculation formula.

[0203] Based on the curvature, a clustering algorithm is used to cluster the target feature points to generate several clusters;

[0204] Each cluster is treated as a node, and a topological map of objects within the suspected abnormal region is constructed based on the connection relationships between the nodes.

[0205] Based on the three-dimensional coordinate information of the target feature points, a network generation algorithm is used to transform the topology map into a three-dimensional structural model of the objects within the suspected abnormal region.

[0206] Preferably, the structural model construction module 203 determines the curvature of each target feature point on its first local surface based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point, using a custom curvature calculation formula, wherein the expression of the curvature calculation formula is:

[0207] k = sinθ / d

[0208] In the formula, θ and d are the angle and distance between the first local surface normal vector and the corresponding second local surface normal vector of the target feature point, respectively.

[0209] Preferably, the ultrasound-assisted detection module 204 determines the reflection change information of objects within the suspected abnormal area based on the reflected echo intensity and the reflected echo time, including:

[0210] For the reflected echo time, a nonlinear transformation using chaotic mapping is used to obtain the transformed time series;

[0211] For the intensity of the reflected echo, a fractional Fourier transform is used for spectral analysis to obtain the intensity spectrum after the fractional Fourier transform.

[0212] Based on the time series and the intensity spectrum, the time-frequency joint features after feature fusion based on the time series and the intensity spectrum are calculated using a custom time-frequency joint feature expression;

[0213] A feature matrix is ​​constructed based on the time-frequency joint features, and singular value decomposition is performed on the feature matrix to obtain a singular value distribution. Then, a clustering analysis of dynamic density peaks is performed based on the singular value distribution to obtain a clustering result. The rows of the feature matrix are different time-frequency joint feature vectors.

[0214] The reflection change information of objects in the suspected abnormal region is determined based on the change of the feature parameters of the clustering results over time, wherein the feature parameters include the center position, radius and shape factor of the cluster.

[0215] Preferably, the ultrasound-assisted detection module 204 calculates the time-frequency joint features after feature fusion based on the time series and the intensity spectrum using a custom time-frequency joint feature expression, wherein the time-frequency joint feature expression is:

[0216]

[0217] In the formula, J(u,t) new This is a time-frequency joint feature. It is a parameter that controls the strength of time correlation, I FRFT (u) is the intensity spectrum, t new It is a time point in the time series, t ref It is a reference time point, and the median value of the time series can be selected.

[0218] Preferably, the abnormal area identification module 205 performs logical correlation analysis based on the three-dimensional structural model and reflection change information of objects within each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified, including:

[0219] For each suspected abnormal region, the curvature of each triangular facet is calculated based on the coordinates of the three vertices of each triangular facet in the three-dimensional structural model and the curvature of each triangular facet in the two principal directions.

[0220] The curvature characteristic value of the three-dimensional structural model is determined by calculating the mean curvature of all triangular facets in the three-dimensional structural model.

[0221] An association matrix is ​​constructed with the curvature feature value as the row and the rate of change feature value of the reflection change information as the column, wherein the rate of change feature value is calculated based on the rate of change of reflected echo intensity and the rate of change of reflected echo time.

[0222] The correlation matrix is ​​subjected to eigenvalue decomposition to determine the matrix eigenvectors and the corresponding matrix eigenvalues, wherein the matrix eigenvectors characterize the feature patterns in the correlation matrix, and the matrix eigenvalues ​​characterize the importance of different feature patterns in the correlation matrix;

[0223] Logical correlation analysis is performed on the matrix feature vector and matrix feature value of each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified.

[0224] Preferably, the abnormal area identification module 205 performs logical correlation analysis based on the matrix feature vector and matrix feature value of each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified, including:

[0225] For each suspected abnormal region, the matrix feature vector whose corresponding matrix feature values ​​satisfy the custom second screening rule is taken as the target feature vector;

[0226] The target feature vector is combined to generate a feature descriptor for each suspected abnormal region.

[0227] For any two suspected abnormal regions, perform difference analysis based on their feature descriptors to obtain the difference value of the abnormal regions.

[0228] The suspected abnormal areas whose abnormal area differences meet the custom third screening rule are identified as target abnormal areas within the DC battery swapping station to be identified.

[0229] Preferably, for each suspected abnormal region, the abnormal region identification module 205 uses the matrix feature vector whose corresponding matrix feature value satisfies the custom second filtering rule as the target feature vector, wherein the second filtering rule is to determine the matrix feature vector as the target feature vector when the matrix feature value corresponding to a matrix feature vector is greater than the custom feature value threshold.

[0230] Preferably, the abnormal area identification module 205 determines the suspected abnormal area whose abnormal area difference value meets the custom third screening rule as the target abnormal area within the DC battery swapping station to be identified. The third screening rule is that when the abnormal area difference value between a suspected abnormal area and a custom number of other suspected abnormal areas is greater than a custom difference threshold, and the abnormal area difference value between a suspected abnormal area and all other suspected abnormal areas is greater than a custom average threshold, the suspected abnormal area is determined to be the target abnormal area within the DC battery swapping station to be identified.

[0231] The preferred embodiment of this system and method for identifying environmental anomalies within a DC battery swapping station utilizes multi-source data acquisition and fusion from high-definition and thermal imaging cameras. Combined with structured light and ultrasonic scanning detection, the generated three-dimensional structural model and obtained reflection change information are logically correlated to determine the target anomaly area within the DC battery swapping station. This overcomes the bottleneck of single-vision algorithms, significantly improving the ability to identify small and concealed anomalies, effectively reducing missed detections, and solving the problem of insufficient ability to identify small or concealed anomalies in existing methods. This improves the accuracy and reliability of identifying environmental anomalies within the DC battery swapping station, ensuring the stable operation of the station.

[0232] Exemplary electronic devices

[0233] Figure 3 This is a schematic diagram of the structure of an electronic device according to a preferred embodiment of the present invention. Figure 3 As shown, the electronic device includes one or more processors 301 and memory 302.

[0234] The processor 301 may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0235] The memory 302 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 301 may execute the program instructions to implement the methods for identifying environmental anomalies within a DC battery swapping station as described in the various embodiments disclosed above, and / or other desired functions. In one example, the electronic device may also include an input device 303 and an output device 304, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0236] In addition, the input device 303 may also include, for example, a keyboard, a mouse, etc.

[0237] The output device 304 can output various information to the outside. The output device 304 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0238] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device relevant to this disclosure are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device may include any other suitable components depending on the specific application.

[0239] Exemplary computer program products and computer-readable storage media

[0240] In addition to the methods and devices described above, embodiments of this disclosure may also be computer program products comprising computer program instructions that, when executed by a processor, cause the processor to perform the steps of the methods for identifying environmental anomalies within a DC battery swapping station according to various embodiments of this disclosure as described in the "Exemplary Methods" section of this specification.

[0241] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of this disclosure. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0242] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the method for identifying environmental anomalies within a DC battery swapping station according to various embodiments of this disclosure as described in the "Exemplary Methods" section above.

[0243] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0244] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0245] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0246] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0247] The apparatus and methods of this disclosure may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of this disclosure are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, this disclosure may also be implemented as a program recorded on a recording medium, the program including machine-readable instructions for implementing the methods according to this disclosure. Thus, this disclosure also covers recording media storing programs for performing the methods according to this disclosure.

[0248] It should also be noted that in the apparatus, devices, and methods of this disclosure, the components or steps are decomposable and / or recombinable. Such decomposition and / or recombination should be considered equivalent to the present disclosure. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0249] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for identifying environmental anomalies within a DC battery swapping station, characterized in that, The method includes: Coordinate mapping is performed on the initial images of each environmental area within the DC battery swapping station to be identified to obtain a fused image of each environmental area. The initial images include high-definition images captured by high-definition camera equipment and initial thermal images captured by thermal imaging camera equipment within the DC battery swapping station to be identified. At least one suspected anomalous region is identified based on the fused image of each environmental region; Based on the structured light projected multiple times into the suspected abnormal area by the high-definition camera device, the reflection deformation information of the structured light on the surface of the object in the suspected abnormal area is obtained, and a three-dimensional structural model of the object in the suspected abnormal area is constructed based on the reflection deformation information. The suspected anomalous area is scanned using an ultrasonic scanning detector to obtain the reflected echo intensity and reflection time of ultrasonic waves from the surface of an object within the suspected anomalous area. Based on the reflected echo intensity and reflection time, information on the reflection changes of the object within the suspected anomalous area is determined. This determination includes: For the reflected echo time, a nonlinear transformation using chaotic mapping is used to obtain the transformed time series; For the intensity of the reflected echo, a fractional Fourier transform is used for spectral analysis to obtain the intensity spectrum after the fractional Fourier transform. Based on the time series and the intensity spectrum, the time-frequency joint features after feature fusion based on the time series and the intensity spectrum are calculated using a custom time-frequency joint feature expression; A feature matrix is ​​constructed based on the time-frequency joint features, and singular value decomposition is performed on the feature matrix to obtain a singular value distribution. Then, a clustering analysis of dynamic density peaks is performed based on the singular value distribution to obtain a clustering result. The rows of the feature matrix are different time-frequency joint feature vectors. The reflection change information of objects in the suspected abnormal region is determined based on the change of the feature parameters of the clustering results over time, wherein the feature parameters include the center position, radius and shape factor of the cluster; Logical correlation analysis is performed based on the three-dimensional structural model and reflection change information of objects within each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified.

2. The method according to claim 1, characterized in that, The process of mapping the initial images of each environmental area within the DC battery swapping station to be identified to obtain a fused image of each environmental area includes: In each environmental region, the gradient magnitude around each pixel in the high-resolution image is calculated to determine the edge intensity change of each pixel, the texture complexity magnitude around each pixel is calculated to determine the texture complexity change of each pixel, and the temperature gradient magnitude of each pixel in the initial thermal image is calculated. The pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule are taken as first target key pixels, and the pixels in the initial thermal imaging image whose temperature gradient amplitude satisfies a custom second recognition rule are taken as second target key pixels. Calculate the similarity between the first target key pixel and the second target key pixel at different scales, and take the first target key pixel and the second target key pixel that satisfy the custom pixel matching rule as the target pixel pair between the high-definition image and the initial thermal imaging image; Based on the position coordinates of each pixel in the target pixel pair, an affine transformation is used to determine a mapping error function that characterizes the coordinate mapping relationship between the high-definition image and the initial thermal imaging image. The initial thermal imaging image is resampled according to the mapping error function to obtain the target thermal imaging image; The pixel values ​​of each pixel in the high-definition image and the pixel values ​​of each pixel in the target thermal image are fused to generate a fused image of each environmental region.

3. The method according to claim 2, characterized in that, The pixels in the high-definition image whose edge intensity changes and texture complexity changes satisfy a custom first recognition rule are designated as first target key pixels, and the pixels in the initial thermal imaging image whose temperature gradient magnitude satisfies a custom second recognition rule are designated as second target key pixels. The first recognition rule is that when the edge intensity change of a pixel is greater than a custom intensity change threshold and the texture complexity change is greater than a custom complexity change threshold, the pixel is determined to be a first target key pixel. The second recognition rule is that when the temperature gradient magnitude of a pixel is greater than a custom gradient magnitude threshold, the pixel is determined to be a second target key pixel.

4. The method according to claim 1, characterized in that, Based on the fused image of each environmental region, at least one suspected anomalous region is identified, including: For each environmental region, local image features of the fused image at different scales are extracted, and the gradient changes of the local image features at different scales in different feature directions are calculated. The feature distance is calculated based on the gradient change of any two feature points in each feature direction at each scale. The local image features whose feature distances satisfy the custom clustering rules are clustered to generate several image feature classes; The first gradient change mean and the first feature mean of the fused image are calculated at different scales and different feature directions. The second gradient change mean and the second feature mean of several image feature classes are calculated at different scales and different feature directions. The first gradient change mean is obtained by averaging the gradient magnitudes of all local image features at a certain scale and feature direction. The first feature mean is obtained by averaging each dimension of all local image feature vectors at the corresponding scale and feature direction. The second gradient change mean is obtained by averaging the gradient magnitudes of all feature points of an image feature class. The second feature mean is obtained by averaging the feature values ​​of all feature points of an image feature class. The image feature class whose first gradient change mean and first feature mean of the fused image, as well as the second gradient change mean and second feature mean of the image feature class satisfy the custom first screening rule, is taken as the target cluster; Based on the mean of the first gradient change and the mean of the first feature of each abnormal image patch, the neighborhood image is expanded according to a custom expansion rule to generate a target image patch, wherein the abnormal image patch is the region where the local image features are located in the target cluster; A topological structure analysis is performed on the image region constructed based on the target image block to generate a topological structure analysis result, and suspected abnormal regions in the image region constructed based on the topological structure analysis result are determined.

5. The method according to claim 4, characterized in that, Local image features whose feature distances satisfy a custom clustering rule are clustered to generate several image feature classes, wherein the clustering rule is as follows: When taking any feature point in the local image features Centered on, with radius as Within the neighborhood of, containing at least 1 feature point, that is, satisfying point The quantity is no less than When determining feature points With feature points Belonging to the same image feature cluster, among which, and These are clustering parameters related to the variable s representing the scale. Feature points With feature points Variables on the characterization scale and variables representing the direction of characteristics The feature distance is calculated using the following formula: In the formula, and Variables at the representation scale Below, using feature points and feature points Local image features extracted from the center, and Variables at the representation scale Below, feature points and feature points Local image features at the location and Variables in the direction of characterization The gradient change on.

6. The method according to claim 4, characterized in that, Image feature classes whose first gradient change mean, first feature mean, second gradient change mean, and second feature mean satisfy a custom first filtering rule are used as target clusters, wherein the expression of the first filtering rule is as follows: In the formula, and These represent the mean of the first gradient change and the mean of the first feature of the fused image, respectively. and These represent the mean of the second gradient change and the mean of the second feature for several image feature classes, respectively. and The threshold value is set.

7. The method according to claim 4, characterized in that, The target image patch is generated by expanding its neighborhood image according to a custom expansion rule based on the mean of the first gradient change and the mean of the first feature of each abnormal image patch. The expression for the expansion rule is as follows: In the formula, and These represent the mean of the first gradient change and the mean of the first feature for each anomalous image patch. and These are the mean of the second gradient change and the mean of the second feature of the proposed expanded neighborhood image, respectively. and The threshold value is set.

8. The method according to claim 1, characterized in that, Based on the reflected deformation information, a three-dimensional structural model of the object within the suspected abnormal region is constructed, including: Calculate the corner point change rate and the surface curve change rate based on the reflected deformation information; Based on the corner change rate and the surface curve change rate, target feature points within the suspected abnormal region are extracted; Based on the intrinsic parameter matrix of the high-definition camera and the projection direction of the structured light, the position information of the target feature point in three-dimensional spatial coordinates is calculated using the principle of triangulation. For each target feature point, its corresponding first neighbor point is selected according to a custom neighbor point selection rule, and the normal vector of the first local surface constructed based on the first neighbor point is used as the first local surface normal vector of each target feature point in its first local surface. The first local surface is a local surface constructed by each target feature point based on the first neighbor point. The neighbor point selection rule is determined according to a custom distance or angle. For each target feature point, a first neighboring point is selected according to a custom neighboring point determination condition, and the normal vector of the constructed second local surface is calculated based on the second neighboring point as the second local surface normal vector of each second neighboring point in its second local surface. The second local surface is a local surface constructed by each first neighboring point based on the second neighboring point. Based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point, the curvature of each target feature point on its first local surface is determined using a custom curvature calculation formula. Based on the curvature, a clustering algorithm is used to cluster the target feature points to generate several clusters; Each cluster is treated as a node, and a topological map of objects within the suspected abnormal region is constructed based on the connection relationships between the nodes. Based on the three-dimensional coordinate information of the target feature points, a network generation algorithm is used to transform the topology map into a three-dimensional structural model of the objects within the suspected abnormal region.

9. The method according to claim 8, characterized in that, The curvature of each target feature point on its first local surface is determined using a custom curvature calculation formula based on the distance between the first local surface normal vector and the corresponding second local surface normal vector of each target feature point. The expression for the curvature calculation formula is as follows: In the formula, d and d are the angle and distance between the first local surface normal vector and the corresponding second local surface normal vector of the target feature point, respectively.

10. The method according to claim 1, characterized in that, The step involves calculating the time-frequency joint features after feature fusion based on the time series and the intensity spectrum using a custom time-frequency joint feature expression, wherein the time-frequency joint feature expression is: In the formula, It is a time-frequency joint feature. It is a parameter that controls the strength of time correlation. It is the intensity spectrum, These are the time points in the time series. It is a reference time point, and the median value of the time series can be selected.

11. The method according to claim 1, characterized in that, The logical correlation analysis based on the three-dimensional structural model and reflection change information of objects within each suspected anomaly area is used to determine the target anomaly area within the DC battery swapping station to be identified, including: For each suspected abnormal region, the curvature of each triangular facet is calculated based on the coordinates of the three vertices of each triangular facet in the three-dimensional structural model and the curvature of each triangular facet in the two principal directions. The curvature characteristic value of the three-dimensional structural model is determined by calculating the mean curvature of all triangular facets in the three-dimensional structural model. An association matrix is ​​constructed with the curvature feature value as the row and the rate of change feature value of the reflection change information as the column, wherein the rate of change feature value is calculated based on the rate of change of reflected echo intensity and the rate of change of reflected echo time. The correlation matrix is ​​subjected to eigenvalue decomposition to determine the matrix eigenvectors and the corresponding matrix eigenvalues, wherein the matrix eigenvectors characterize the feature patterns in the correlation matrix, and the matrix eigenvalues ​​characterize the importance of different feature patterns in the correlation matrix; Logical correlation analysis is performed on the matrix feature vector and matrix feature value of each suspected abnormal area to determine the target abnormal area within the DC battery swapping station to be identified.

12. The method according to claim 11, characterized in that, The step of performing logical correlation analysis based on the matrix feature vector and matrix feature value of each suspected abnormal region to determine the target abnormal region within the DC battery swapping station to be identified includes: For each suspected abnormal region, the matrix feature vector whose corresponding matrix feature values ​​satisfy the custom second screening rule is taken as the target feature vector; The target feature vector is combined to generate a feature descriptor for each suspected abnormal region. For any two suspected abnormal regions, perform difference analysis based on their feature descriptors to obtain the difference value of the abnormal regions. The suspected abnormal areas whose abnormal area differences meet the custom third screening rule are identified as target abnormal areas within the DC battery swapping station to be identified.

13. The method according to claim 12, characterized in that, For each suspected abnormal region, the matrix feature vector whose corresponding matrix feature value satisfies the custom second filtering rule is taken as the target feature vector. The second filtering rule is that when the matrix feature value corresponding to a matrix feature vector is greater than the custom feature value threshold, the matrix feature vector is determined to be the target feature vector.

14. The method according to claim 12, characterized in that, The suspected abnormal areas whose abnormal area difference values ​​meet the custom third screening rule are identified as target abnormal areas within the DC battery swapping station to be identified. The third screening rule is that when the abnormal area difference value between a suspected abnormal area and a custom number of other suspected abnormal areas is greater than a custom difference threshold, and the abnormal area difference value between a suspected abnormal area and all other suspected abnormal areas is greater than a custom average threshold, the suspected abnormal area is identified as a target abnormal area within the DC battery swapping station to be identified.

15. A system for identifying environmental anomalies within a DC battery swapping station, characterized in that, The system includes: The image fusion module is used to perform coordinate mapping on the initial images of each environmental area within the DC battery swapping station to be identified, and to obtain the fused image of each environmental area. The initial images include high-definition images captured by high-definition camera equipment and initial thermal imaging images captured by thermal imaging camera equipment within the DC battery swapping station to be identified. The region filtering module is used to identify at least one suspected abnormal region based on the fused image of each environmental region. The structural model construction module is used to obtain the reflection deformation information of the structured light on the surface of the object in the suspected abnormal area based on the structured light projected multiple times into the environment of the suspected abnormal area by the high-definition camera device, and to construct a three-dimensional structural model of the object in the suspected abnormal area based on the reflection deformation information. An ultrasound-assisted detection module is used to scan the suspected abnormal area using an ultrasonic scanning detector, acquire the reflected echo intensity and reflection echo time of ultrasonic waves on the surface of an object within the suspected abnormal area, and determine the reflection change information of the object within the suspected abnormal area based on the reflected echo intensity and reflection echo time. The determination of the reflection change information of the object within the suspected abnormal area based on the reflected echo intensity and reflection echo time includes: For the reflected echo time, a nonlinear transformation using chaotic mapping is used to obtain the transformed time series; For the intensity of the reflected echo, a fractional Fourier transform is used for spectral analysis to obtain the intensity spectrum after the fractional Fourier transform. Based on the time series and the intensity spectrum, the time-frequency joint features after feature fusion based on the time series and the intensity spectrum are calculated using a custom time-frequency joint feature expression; A feature matrix is ​​constructed based on the time-frequency joint features, and singular value decomposition is performed on the feature matrix to obtain a singular value distribution. Then, a clustering analysis of dynamic density peaks is performed based on the singular value distribution to obtain a clustering result. The rows of the feature matrix are different time-frequency joint feature vectors. The reflection change information of objects in the suspected abnormal region is determined based on the change of the feature parameters of the clustering results over time, wherein the feature parameters include the center position, radius and shape factor of the cluster; The abnormal area identification module is used to perform logical correlation analysis based on the three-dimensional structural model and reflection change information of objects in each suspected abnormal area to determine the target abnormal area in the DC battery swapping station to be identified.

16. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method as described in any one of claims 1-14.

17. An electronic device, characterized in that, include: The computer-readable storage medium as described in claim 16; as well as One or more processors for executing a program in the computer-readable storage medium.

Images