Method and system for detection of wrist and forearm positioning indicia based on image processing

Abstract

The application relates to the technical field of image processing, and provides a detection method and system for wrist arm positioning marks based on image processing, which solve the problems of high contour detection error rate caused by insufficient inhibition of environmental interference and false detection caused by parameter solidification. The method comprises the following steps: acquiring wrist arm operation parameters to generate a space constraint model; collecting visible light band and infrared band images of wrist arm positioning marks to generate multispectral enhanced images through fusion; based on the reflection characteristics of the positioning marks, combining weight distribution and feature aggregation of the space constraint model, acquiring an initial contour point set of the positioning marks and performing three-dimensional projection correction to generate a topological structure matched with the spatial posture of the wrist arm; and based on the verification result of the geometric distribution boundary of the topological structure and the space constraint model, generating the final detection coordinates of the positioning marks of the wrist arm. The application solves the problem of detection precision decline caused by complex environmental interference and mechanical deformation, and improves the robustness and accuracy of the positioning marks of the high-speed rail catenary wrist arm.

Description

Technical Field

[0001] This invention belongs to the field of image processing technology, and specifically relates to a method and system for detecting wrist and arm positioning markers based on image processing. Background Technology

[0002] With the rapid development of the high-speed rail network and the continuous increase in operating mileage, higher requirements are placed on the detection accuracy and efficiency of the contact wire cantilever positioning markers. To ensure the safe operation of high-speed rail, high-precision detection is required in complex environments, while dynamically adapting to the minute deformations of the cantilever caused by vibration and load, meeting the requirements of real-time performance and robustness.

[0003] Existing technical solutions include contour detection methods utilizing visible and infrared dual-band fusion. These methods simultaneously acquire images of the wrist and arm using visible and infrared cameras, fuse the images using fixed band registration parameters, extract the localization marker contours using a general edge detection algorithm, and filter false detection areas using static thresholds to output the final detection location. However, these existing solutions have several drawbacks. For example, the fixed band registration parameters result in insufficient suppression of environmental interference, leading to decreased contrast in the fused image and a high contour detection error rate. Furthermore, the general edge detection algorithm relies on fixed thresholds, which significantly increases the false detection rate in complex environments. Summary of the Invention

[0004] The purpose of this application is to provide a method and system for detecting wrist and arm positioning marks based on image processing, in order to solve the problems of high contour detection error rate due to insufficient suppression of environmental interference and false detection due to parameter fixation in the prior art.

[0005] To address the aforementioned technical problems, in a first aspect, this application provides a method for detecting wrist-arm positioning markers based on image processing, comprising:

[0006] The operating parameters of the cantilever arm in the high-speed railway catenary are obtained to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the positioner.

[0007] Visible light and infrared images of the positioning markers of the carpal tunnel are acquired within the geometric distribution boundary of the spatial constraint model. The visible light and infrared images are then fused to generate a multispectral enhanced image.

[0008] Based on the reflection features of the location marker in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, the initial contour point set of the location marker is obtained.

[0009] The initial contour point set is corrected by three-dimensional projection to generate a topology that matches the spatial pose of the carpal arm;

[0010] Based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model, the final detection coordinates of the positioning mark of the carpal arm are generated.

[0011] Optionally, based on the reflection features of the location marker in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, an initial contour point set of the location marker is obtained, including:

[0012] Texture gradient information, edge information, stability coefficient, and thermal radiation intensity information are extracted from the multispectral enhanced image;

[0013] Using a weight allocation mechanism, a dynamic attenuation factor and a boundary weight matrix are generated, and combined with the thermal radiation intensity information, a spatial attenuation gradient is generated.

[0014] The texture gradient information, edge information, stability coefficient, and thermal radiation intensity information are fused at multiple levels using a feature aggregation mechanism to generate a feature response map.

[0015] Based on the dynamic decay factor and the spatial decay gradient, the feature response map is adjusted to generate a set of candidate response points;

[0016] Based on the topology rules of the candidate response point set and the fracture repair rules of the edge information, a target connection path is generated, and an initial contour point set is generated by combining the candidate response point set.

[0017] Optionally, the feature response map is adjusted based on the dynamic decay factor and the spatial decay gradient to generate a candidate response point set, including:

[0018] The feature response map is adjusted based on the dynamic decay factor and the spatial decay gradient to generate an adjusted feature response map.

[0019] Based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information, a dynamic threshold map is generated.

[0020] The adjusted feature response map and the feature response map are subjected to feature extraction and convolution operations to generate a decayed feature response map;

[0021] The difference between the response value of each pixel in the attenuated feature response map and the dynamic threshold of the corresponding pixel in the dynamic threshold map is calculated to generate a difference value matrix. The pixels with difference values ​​greater than zero are selected from the difference value matrix as candidate response points to generate a candidate response point set.

[0022] Optionally, based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information, a dynamic threshold map is generated, including:

[0023] Based on the numerical distribution of the stability coefficient, a first threshold is generated, and based on the fluctuation characteristics of the stability coefficient, the first threshold is adjusted to generate a second threshold.

[0024] Based on the time-domain fluctuation characteristics of the thermal radiation intensity information, the thermal radiation intensity information is decomposed into thermal radiation energy distribution and energy fluctuation components.

[0025] Based on the thermal radiation energy distribution, a third threshold is generated. Based on the energy fluctuation component, the third threshold is scaled and modulated to generate a fourth threshold.

[0026] Based on the environmental type identifier of the wrist arm, the second threshold and the fourth threshold are subjected to multi-level filtering to obtain the initial threshold map;

[0027] Based on the attenuation coefficient of the adjusted feature response map, the initial threshold map is corrected to generate a dynamic threshold map.

[0028] Optionally, the initial contour point set is subjected to 3D projection correction to generate a topology that matches the spatial pose of the carpal arm, including:

[0029] The initial contour point set is decomposed into a planar projection point set and a height feature component;

[0030] Construct the target 3D mesh based on preset allowable deviation parameters;

[0031] Based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh, the planar projection point set is corrected to generate a corrected point set;

[0032] Based on the spatial distribution characteristics of the correction point set, a topology matching the spatial pose of the carpal arm is constructed.

[0033] Optionally, based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh, the planar projection point set is corrected to generate a corrected point set, including:

[0034] Based on the node spacing of the target 3D mesh, the planar projection point set is smoothed to generate a corrected point set;

[0035] Based on the thermal radiation intensity information corresponding to the height feature components, multiple target regions that conflict with the bending radius of the tube of the cantilever arm are identified, and a set of target region points is extracted from the target regions.

[0036] Adjust the height feature components of the target region point set according to the preset projection rules to generate the adjusted target region point set.

[0037] The corrected point set and the adjusted target region point set are spatially superimposed to obtain the superimposed point set. Isolated target points in the superimposed point set are then removed to generate the corrected point set.

[0038] Optionally, based on the verification results of the topology and the geometric distribution boundary of the spatial constraint model, the final detection coordinates of the carpal tunnel's positioning marker are generated, including:

[0039] Align the node coordinates of the topology with the geometric distribution boundary of the spatial constraint model to generate the aligned topology.

[0040] Calculate the overlap between the aligned topology and the geometric distribution boundary, and mark nodes with an overlap below a preset threshold as abnormal nodes;

[0041] The abnormal nodes are spatially corrected to obtain the corrected topology.

[0042] The overlap between the node coordinates of the modified topology and the geometric distribution boundary of the spatial constraint model is verified to generate verification results.

[0043] Based on the verification results, the final detection coordinates of the positioning marker of the wrist arm are generated.

[0044] Secondly, this application provides a detection system for wrist-arm positioning markers based on image processing, comprising:

[0045] The generation module is used to obtain the operating parameters of the cantilever arm in the high-speed railway catenary in order to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the positioner.

[0046] The acquisition module is used to acquire visible light and infrared images of the positioning marker of the carpal tunnel within the geometric distribution boundary of the spatial constraint model, and to fuse the visible light and infrared images to generate a multispectral enhanced image.

[0047] The acquisition module is used to acquire the initial contour point set of the positioning marker based on the reflection features of the positioning marker in the multispectral enhanced image, combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model;

[0048] The correction module is used to perform three-dimensional projection correction on the initial contour point set to generate a topology that matches the spatial pose of the carpal arm.

[0049] The output module is used to generate the final detection coordinates of the positioning marker of the carpal arm based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model.

[0050] Thirdly, this application provides a computing device including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to perform a method for detecting a wrist-arm positioning marker based on image processing as described in any of the first aspects.

[0051] Fourthly, this application provides a computer storage medium storing computer program instructions thereon, which, when executed by a processor, implement the method for detecting a wrist-arm positioning marker based on image processing as described in any one of the first aspects.

[0052] The beneficial effects of this application are:

[0053] This application provides a method for detecting carpal tunnel positioning markers based on image processing. By constructing a spatial constraint model based on carpal tunnel operational parameters, it can dynamically adapt to morphological changes caused by vibration and load, providing precise geometric boundary constraints for the detection process. Within this constraint boundary, dual-band images are acquired and fused to generate a multispectral enhanced image, improving the image's ability to suppress interference from complex environments and enhancing the distinction between the positioning marker and the background. An initial contour point set is extracted using the weight allocation and feature aggregation mechanisms of the spatial constraint model, optimizing the targeting of feature selection and reducing invalid feature interference. A topological structure matching the spatial posture of the carpal tunnel is generated through 3D projection correction, calibrating the consistency between the contour information and the actual spatial posture. Finally, detection coordinates are generated based on the topological structure and geometric boundary verification, comprehensively improving the accuracy and robustness of positioning marker detection and meeting the real-time detection requirements in complex environments.

[0054] Furthermore, based on the reflection characteristics of the positioning markers in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, dynamic attenuation factor, boundary weight matrix and spatial attenuation gradient are generated by extracting texture gradient information, edge information, stability coefficient and thermal radiation intensity information. After multi-level feature fusion, feature response map is obtained and adjusted to form candidate response point set. Then, according to topological rules and edge breakage repair rules, initial contour point set is generated to improve the accuracy and robustness of the contact wire cantilever positioning marker detection in complex environments, strengthen the resistance to various environmental interferences, optimize the distinguishability of effective features in the fused image, and dynamically adapt to the small deformation of the cantilever caused by vibration and load, so as to meet the real-time requirements of high-speed rail contact wire detection. Attached Figure Description

[0055] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0056] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 A flowchart illustrating a method for detecting wrist-arm positioning markers based on image processing, provided in an embodiment of this application;

[0058] Figure 2 A schematic diagram of the structure of a wrist-arm positioning marker detection system based on image processing provided in an embodiment of this application;

[0059] Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of this application. Detailed Implementation

[0060] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0061] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0063] Figure 1This application provides a flowchart of a method for detecting wrist-arm positioning markers based on image processing, as shown in the embodiments of this application. Figure 1 As shown, the method includes:

[0064] To address the challenges of environmental interference and mechanical deformation coupling in the inspection of overhead contact line cantilever arms in high-speed railways, this paper proposes several solutions. These solutions address shortcomings in existing technologies, such as insufficient image contrast due to fixed multispectral fusion parameters in rain, fog, or sudden changes in lighting, and the inability of static geometric models to adapt to dynamic cantilever arm deformation. A spatial constraint model is constructed using operational parameters to constrain the multispectral fusion range and drive boundary weight allocation. Robustly enhanced images are generated by combining visible light dynamic illumination compensation and infrared rain and fog suppression. Contour extraction accuracy is enhanced by utilizing cross-band reflection feature differences. Furthermore, the slope threshold and 3D projection correction are linked, and topology generation is constrained by mechanical deformation tolerance parameters. Ultimately, this achieves millimeter-level positioning accuracy and adaptive deformation detection in complex environments.

[0065] Based on this, this application provides a method for detecting wrist and arm positioning markers based on image processing, such as... Figure 1 ,include:

[0066] Step 101: Obtain the operating parameters of the cantilever arm in the high-speed rail catenary to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the locator.

[0067] In this step, the operational parameters refer to the dynamically changing mechanical parameters of the high-speed railway overhead contact line cantilever arm during operation. The spatial constraint model refers to the mathematical model generated based on the operational parameters. This model limits the physical feasible domain for image processing and contour correction by defining the geometric distribution boundaries of the cantilever arm positioning markers and the mechanical deformation tolerance rules. The slope threshold specifies the maximum allowable slope deviation value of the locator in the vertical direction, used to determine whether the cantilever arm installation conforms to engineering specifications, and serves as a constraint condition for three-dimensional projection correction.

[0068] In this embodiment of the application, according to the design specifications of the high-speed railway catenary, by analyzing the tilt state of the cantilever arm during actual installation, the installation angle of the cantilever arm in the working parameters of the cantilever arm in the high-speed railway catenary is converted into a spatial angle constraint of the detection area, so as to ensure that the tilt direction and range of the detection area are consistent with the actual installation posture of the cantilever arm, and avoid the detection range from deviating from the possible distribution area of ​​the cantilever arm positioning mark.

[0069] Next, the slope threshold of the locator is quantized and converted. Based on the specific value of the slope threshold, it is mapped to the allowable deviation range in the height direction. For example, the larger the slope threshold, the corresponding allowable deviation range in the height direction is adjusted accordingly to adapt to the installation requirements of the locator under different slopes. Finally, parametric modeling is used to integrate the converted spatial angle constraints and the allowable deviation range in the height direction to generate a spatial constraint model.

[0070] Step 102: Acquire visible light and infrared images of the positioning marker of the carpal tunnel within the geometric distribution boundary of the spatial constraint model, and fuse the visible light and infrared images to generate a multispectral enhanced image.

[0071] In this step, the positioning markers of the wrist arm refer to the categories of physical markers used for positioning on the wrist arm. They define the target features for image processing and directly affect the detection criteria such as reflectivity and texture details. The geometric distribution boundary refers to the boundary of the detection area defined by the spatial constraint model, characterizing the reasonable spatial distribution range of the wrist arm positioning markers.

[0072] In this embodiment, the visible light and infrared images of the wrist-arm positioning marker are acquired within the bounded range of the geometric distribution boundary of the spatial constraint model. First, dynamic illumination compensation processing is performed on the visible light image. Specifically, the histogram distribution of the image is analyzed, the gain is reduced in overexposed areas, and the brightness is increased in underexposed areas to generate a uniformly illuminated visible light image.

[0073] Next, using pre-calibrated band registration parameters, the illuminated visible light image and the infrared band image are precisely aligned to obtain the aligned image, ensuring the spatial consistency of the two band images. Subsequently, the aligned image is fused using a weighted fusion algorithm to finally generate a multispectral enhanced image, which not only fully preserves and enhances the detailed features of the visible light band image, but also suppresses rain and fog interference in the infrared band image.

[0074] Step 103: Based on the reflection features of the location marker in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, obtain the initial contour point set of the location marker.

[0075] In this step, the reflectivity features of the location marker refer to the differences in reflectivity of the location marker in different bands of the multispectral image. The weighting mechanism refers to the strategy of dynamically adjusting feature weights based on the distance between the geometric distribution boundary and the location marker. The feature aggregation mechanism refers to the technique of multi-scale fusion of visible light and infrared band features. The initial contour point set refers to the set of location marker contour points extracted from the multispectral enhanced image, which includes edge location and reflectivity feature information.

[0076] Step 104: Perform 3D projection correction on the initial contour point set to generate a topology that matches the spatial pose of the wrist arm.

[0077] In this step, the spatial orientation of the wrist arm refers to its three-dimensional spatial position and orientation during actual operation. The topology refers to the network of contour points connected by 3D projection correction, which satisfies mechanical constraints such as bolt spacing and bending radius.

[0078] Step 105: Based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model, generate the final detection coordinates of the positioning mark of the carpal arm.

[0079] In this step, the verification result refers to a quantitative indicator of the degree of matching between the topological structure and the geometric distribution boundary.

[0080] This application's embodiments construct a spatial constraint model through operational parameters to achieve robust multispectral fusion and interference suppression in complex environments; combined with mechanical deformation parameter-driven 3D projection correction and topology verification, it solves the contour offset and false detection problems caused by fixed parameters in traditional methods; it improves the positioning accuracy and environmental adaptability of high-speed rail catenary detection, meeting the requirements of high robustness and high real-time performance.

[0081] This application provides a specific embodiment. Step 103 involves obtaining the initial contour point set of the positioning marker based on the reflection features of the positioning marker in the multispectral enhanced image, combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model. This specifically includes the following steps:

[0082] Step 301: Extract texture gradient information, edge information, stability coefficient and thermal radiation intensity information from the multispectral enhanced image.

[0083] In this step, texture gradient information refers to parameters obtained by analyzing local pixel grayscale changes in visible light images and extracting minute structural features such as scratches and corrosion on the marked surface.

[0084] The stability coefficient is a parameter obtained by calculating the standard deviation of pixel values ​​in each region after suppressing rain and fog scattering in an infrared image. The smaller the standard deviation, the higher the corresponding value, which is used to characterize the region's anti-interference ability.

[0085] In this embodiment, the multispectral enhanced image is first separated into visible light and infrared images. By analyzing the grayscale changes of local pixels in the separated visible light image, the variation features of minute structures such as scratches and rust on the surface of the marker are extracted to generate texture gradient information.

[0086] Simultaneously, edge information is generated by identifying the edge positions of regions with abrupt changes in pixel grayscale in the visible light image, and then statistically analyzing the extension direction and continuity of these edge positions. For the separated infrared image, rain and fog scattering suppression processing is first performed. Specifically, by adjusting the brightness and contrast of the image, the blurring interference signals caused by rain and fog are filtered out, making the thermal radiation characteristics of the positioning markers in the infrared image clearer and reducing the impact of rain and fog on subsequent data extraction. Then, the standard deviation of pixel values ​​in each region after processing is calculated. The smaller the standard deviation, the more concentrated and stable the pixel values ​​in the region. Finally, a stability coefficient is generated based on this standard deviation, where the smaller the standard deviation, the higher the corresponding stability coefficient value.

[0087] The intensity values ​​of each pixel in the infrared image are extracted to generate thermal radiation intensity information that reflects the difference between the thermal characteristics of the mark and the background.

[0088] Step 302: Using the weight allocation mechanism, generate a dynamic attenuation factor and a boundary weight matrix, and combine the thermal radiation intensity information to generate a spatial attenuation gradient.

[0089] In this step, the dynamic attenuation factor refers to a parameter generated based on the distance relationship between the location marker and the geometric distribution boundary of the spatial constraint model, which is used to characterize the law of weight change with distance.

[0090] The boundary weight matrix is ​​a matrix constructed based on a dynamic decay factor, used to distinguish the weights of the target region and the background region.

[0091] The spatial attenuation gradient refers to the parameter obtained by multiplying the normalized thermal radiation intensity information with the dynamic attenuation factor, which is used to control the attenuation rate of the characteristic response in different regions.

[0092] In this embodiment, a weight allocation mechanism is used to generate a dynamic decay factor based on the distance between the positioning marker and the geometric distribution boundary of the spatial constraint model, and with reference to the rule that the closer the distance, the higher the weight, and the farther the distance, the faster the weight decays. This embodiment does not limit the value of the dynamic decay factor, and it can be set according to the actual situation.

[0093] Based on the dynamic attenuation factor, a boundary weight matrix is ​​constructed. In this boundary weight matrix, the region corresponding to the geometric distribution boundary is given a high weight, and the background region far from the boundary is given a low weight. The thermal radiation intensity information is normalized to adjust the values ​​to a uniform range. Then, the normalized thermal radiation intensity information is multiplied one by one with the dynamic attenuation factor to generate a spatial attenuation gradient.

[0094] Step 303: Use a feature aggregation mechanism to perform multi-level fusion of the texture gradient information, edge information, stability coefficient and thermal radiation intensity information to generate a feature response map.

[0095] In this step, the feature response map refers to the image obtained by multi-level fusion of texture gradient information, edge information, stability coefficient and thermal radiation intensity information through a feature aggregation mechanism. It is used to highlight the positioning and identification features and weaken environmental interference.

[0096] In this embodiment, the four types of information are processed using a hierarchical complementary fusion strategy of feature aggregation mechanism. Specifically, texture gradient information and thermal radiation intensity information are fused first. By leveraging the ability of texture gradient information to capture microstructures and the anti-interference characteristics of thermal radiation intensity information, the overall stability of the positioning marker contour is enhanced. Then, edge information and stability coefficient are fused. By leveraging the contour positioning ability of edge information and the anti-interference advantage of stability coefficient, the clarity of edge details is enhanced. Finally, the results of the two hierarchical fusions are integrated to comprehensively retain the advantages of contour stability and edge details, generating a feature response map.

[0097] Step 304: Based on the dynamic decay factor and the spatial decay gradient, adjust the feature response map to generate a candidate response point set.

[0098] In this step, the candidate response point set refers to the set of pixels obtained after adjusting, filtering and denoising the feature response map. This candidate response point set includes the contour points of potential localization markers.

[0099] Step 305: Based on the topology rules of the candidate response point set and the fracture repair rules of the edge information, generate the target connection path, and combine the candidate response point set to generate the initial contour point set.

[0100] In this step, topology rules refer to rules formulated based on the spatial distribution relationship of candidate response point sets, including adjacent point spacing and connection continuity. Fracture repair rules refer to edge connection strategies defined based on edge information, including maximum fracture tolerance distance and minimum radius of curvature. The target connection path refers to the optimized edge connection path generated through the topology rules and fracture repair rules.

[0101] In this embodiment, topological rules are formulated based on the candidate response point set, and fracture regions are identified according to the rules. Then, based on the directional features of the edge information, the reasonable connection direction of the fracture region is determined, and a target connection path that can repair the fracture contour is generated. Finally, the candidate response point set is used as the basic contour framework, and the fracture gaps are filled by the target connection path. The scattered candidate response points are connected coherently according to the reasonable path, while discrete noise points that do not match the path are removed, and finally integrated to form a complete and continuous initial contour point set.

[0102] This application's embodiments improve anti-interference capabilities in complex environments by leveraging dynamic weight allocation and multi-level feature fusion; and ensure contour integrity through topological rules and fracture repair mechanisms, thus solving the problems of poor environmental adaptability, large contour extraction deviation, and high false detection rate in existing solutions.

[0103] This application provides a specific embodiment. Step 304 involves adjusting the feature response map based on the dynamic decay factor and the spatial decay gradient to generate a candidate response point set, specifically including the following steps:

[0104] Step 311: Adjust the feature response map according to the dynamic decay factor and the spatial decay gradient to generate an adjusted feature response map.

[0105] In this step, the adjusted feature response map refers to the image obtained by performing pixel-by-pixel operations to adjust the dynamic attenuation factor, spatial attenuation gradient, and feature response map. It is used to enhance the target features of the positioning marker and suppress background interference.

[0106] In this embodiment, the response value of the feature response map is multiplied pixel by pixel with the dynamic decay factor and the spatial decay gradient to generate an adjusted feature response map. This map provides an optimized feature carrier for subsequent difference calculation to enhance the target and suppress interference.

[0107] Step 312: Generate a dynamic threshold map based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information.

[0108] In this step, the spatial distribution characteristics of the stability coefficient refer to the distribution pattern of the stability coefficient in different regions of the infrared image; high stability regions correspond to low noise interference, while low stability regions correspond to high noise interference. The temporal fluctuation characteristics of thermal radiation intensity information refer to the fluctuation pattern of thermal radiation intensity with time or environmental changes. The dynamic threshold map refers to a pixel-by-pixel adapted threshold image generated by combining the above two characteristics, used to accurately determine whether a pixel is a valid response point.

[0109] Step 313: Perform feature extraction and convolution operations on the adjusted feature response map and the feature response map to generate a decayed feature response map.

[0110] In this embodiment, local detail features and global contour features in the adjusted feature response map and the feature response map are extracted first. Then, a sliding window of fixed size is used to perform convolution operation on the two types of feature response maps. This convolution operation is performed by multiplying the two types of features by the weight coefficients in the sliding window and then summing them. Finally, the correlation between pixels in the window is calculated to smooth local abrupt changes and enhance feature coherence, thereby generating the attenuated feature response map.

[0111] Step 314: Calculate the difference between the response value of each pixel in the attenuated feature response map and the dynamic threshold of the corresponding pixel in the dynamic threshold map to generate a difference value matrix. Select the pixels with difference values ​​greater than zero from the difference value matrix as candidate response points to generate a candidate response point set.

[0112] In this step, the response value refers to the feature intensity value of each pixel in the attenuated feature response map, reflecting the probability that the pixel belongs to the localization marker contour. The dynamic threshold refers to the judgment value of the corresponding pixel in the dynamic threshold map, used to distinguish between valid features and noise. The difference matrix is ​​the matrix formed by calculating the difference between the response value of each pixel and the corresponding dynamic threshold.

[0113] In this embodiment, the difference between the response value of the attenuated feature response map and the dynamic threshold of the corresponding pixel in the dynamic threshold map is first calculated pixel by pixel, and a difference value matrix is ​​constructed based on these differences. Then, pixels with difference values ​​greater than zero are selected from the difference value matrix. These pixels represent valid points whose feature intensity exceeds the noise threshold. Finally, all valid points are integrated to form a candidate response point set.

[0114] The embodiments of this application improve the effectiveness and accuracy of response points, which not only solves the problem of poor adaptability of fixed parameters, but also reduces false detections caused by noise, providing a high-quality set of candidate response points for subsequent contour construction.

[0115] This application provides a specific embodiment. Step 312 involves generating a dynamic threshold map based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information. This specifically includes the following steps:

[0116] Step 321: Based on the numerical distribution of the stability coefficient, generate a first threshold, and adjust the first threshold according to the fluctuation characteristics of the stability coefficient to generate a second threshold.

[0117] In this step, the first threshold refers to the baseline threshold generated statistically based on the numerical distribution of the stability coefficient, which is used to initially adapt to the noise interference levels in different regions.

[0118] The second threshold is the threshold obtained by dynamically adjusting the first threshold based on the fluctuation characteristics of the stability coefficient, and is used to improve the threshold's adaptability to noise fluctuations.

[0119] In this embodiment, the numerical distribution of the stability coefficient is first statistically analyzed, and the median value of the concentrated distribution area is taken as the first threshold to ensure that the benchmark threshold matches the noise level of most areas. Then, the fluctuation characteristics of the stability coefficient are analyzed. If the stability coefficient of a certain area fluctuates greatly, it indicates that the noise interference is unstable. The first threshold corresponding to that area is then slightly adjusted to finally generate the second threshold.

[0120] Step 322: Based on the time-domain fluctuation characteristics of the thermal radiation intensity information, decompose the thermal radiation intensity information into thermal radiation energy distribution and energy fluctuation components.

[0121] In this step, thermal radiation energy distribution refers to the stable spatial distribution pattern of thermal radiation intensity information, reflecting the difference in thermal characteristics between the positioning marker and the background; energy fluctuation component refers to the fluctuation part of thermal radiation intensity information caused by changes in time or environment, which is used to dynamically adjust the threshold to offset the energy fluctuation impact caused by environmental interference.

[0122] In this embodiment, based on the temporal fluctuation characteristics of thermal radiation intensity information, thermal radiation energy distribution and energy fluctuation components are decomposed; specifically, by analyzing the variation law of thermal radiation intensity in continuous frames, a stable part that does not change with time and a variable part that fluctuates with the environment are separated, with the stable part being taken as thermal radiation energy distribution and the variable part being taken as energy fluctuation components.

[0123] Step 323: Based on the thermal radiation energy distribution, generate a third threshold, and perform scaling and modulation processing on the third threshold based on the energy fluctuation component to generate a fourth threshold.

[0124] In this step, the third threshold refers to a threshold generated based on the distribution of thermal radiation energy, used to match the thermal characteristic differences between the positioning marker and the background.

[0125] The fourth threshold is the threshold obtained by scaling and modulating the third threshold through energy fluctuation components, and is used to adapt to the dynamic fluctuations of thermal radiation energy.

[0126] In this embodiment, a third threshold is generated based on the gradient change of thermal radiation energy distribution. The larger the thermal radiation energy gradient, the higher the corresponding third threshold, so as to highlight the thermal difference between the positioning mark and the background. Then, the scaling ratio is determined based on the fluctuation amplitude of the energy fluctuation component. The third threshold is multiplied by the scaling ratio to generate a fourth threshold, so as to dynamically adapt to the fluctuation of thermal radiation energy.

[0127] Step 324: Based on the environment type identifier of the wrist arm, perform multi-level filtering on the second threshold and the fourth threshold to obtain the initial threshold map.

[0128] In this step, the environment type identifier of the cannula refers to the identifier that represents the type of scene in which the cannula is located, including scenes such as open air, tunnel, and rainstorm.

[0129] The initial threshold image refers to the threshold image obtained after multi-level filtering and fusion of the second and fourth thresholds based on the environment type identifier. It is used to integrate the advantages of the two types of thresholds and suppress environment-specific noise.

[0130] In this embodiment, the filtering strategy is first determined based on the environment type identifier of the arm. For example, if the environment in which the arm is located is a rainstorm environment, the filtering strategy focuses on suppressing rain and fog noise; if the environment in which the arm is located is a tunnel environment, the filtering strategy focuses on suppressing thermal radiation fluctuations. Then, the second threshold is low-pass filtered to retain wide-range adaptability, and the fourth threshold is high-pass filtered to enhance dynamic response. The two types of filtered thresholds are then fused according to the weights corresponding to the environment type identifier of the arm, and transient interference is suppressed through time-domain smoothing to obtain an initial threshold map.

[0131] Step 325: Based on the attenuation coefficient of the adjusted feature response map, the initial threshold map is corrected to generate a dynamic threshold map.

[0132] In this step, the attenuation coefficient of the adjusted feature response map refers to a parameter that reflects the degree of weight attenuation at different spatial locations in the adjusted feature response map. The farther away from the positioning marker, the larger the attenuation coefficient.

[0133] In this embodiment, the attenuation coefficient of the adjusted feature response map is first extracted, and deformation-sensitive regions with large attenuation coefficients are identified. These regions are easily affected by background interference. Then, neighborhood interpolation is performed on the threshold points of the initial threshold map of the deformation-sensitive regions to ensure the spatial continuity of the threshold and avoid abrupt threshold changes. At the same time, the threshold intensity is adjusted according to the magnitude of the attenuation coefficient. The larger the attenuation coefficient, the lower the threshold is appropriately to retain effective weak responses, and finally, a dynamic threshold map is generated.

[0134] The embodiments of this application solve the problem that fixed thresholds cannot adapt to complex environmental interference and energy fluctuations, improve the pertinence and adaptability of thresholds for different scenarios and regions, reduce missed detections and false detections caused by inappropriate thresholds, and provide a reliable guarantee for the accurate screening of subsequent candidate response points.

[0135] This application provides a specific embodiment. Step 104 involves performing three-dimensional projection correction on the initial contour point set to generate a topology that matches the spatial pose of the carpal arm. This specifically includes the following steps:

[0136] Step 401: Decompose the initial contour point set into a planar projection point set and a height feature component.

[0137] In this step, the planar projection point set refers to the coordinate set obtained by projecting the initial contour point set onto a two-dimensional plane along the vertical direction.

[0138] The height feature component refers to the parameter calculated based on the difference between the infrared thermal radiation intensity information and the visible light texture depth distribution, reflecting the spatial distribution characteristics of the positioning mark in the vertical direction.

[0139] In this embodiment, the three-dimensional coordinates of the initial contour point set are decomposed, the horizontal coordinates of each point are extracted, and the points are arranged in the order of the initial contour to form a planar projection point set, ensuring that the relative positions of the points in the plane are consistent with the initial contour. At the same time, by analyzing the spatial differences of the infrared thermal radiation intensity information and combining the estimated value of the texture depth distribution in the visible light band, the difference between the two is calculated point by point and smoothed and denoised to obtain the height feature component, so as to suppress the noise interference of a single sensor.

[0140] Step 402: Construct the target 3D mesh based on the preset allowable deviation parameters.

[0141] In this step, the preset allowable deviation parameter refers to the mechanical deformation tolerance range parameter defined according to the high-speed railway catenary design specification. This parameter includes the allowable height deviation value corresponding to the slope threshold, the projection density adjustment coefficient related to the installation angle, etc.

[0142] The target 3D mesh refers to a 3D mesh structure constructed based on preset allowable deviation parameters. Its node spacing is negatively correlated with the installation angle, and its height gradient is positively correlated with the slope threshold.

[0143] In this embodiment, the key parameters of the target 3D mesh are determined according to preset allowable deviation parameters. Specifically, the node spacing of the target 3D mesh is set to decrease proportionally to the cosine of the installation angle as the cantilever arm installation angle increases, ensuring higher mesh density in steep slope areas to improve correction accuracy. The height gradient of the target 3D mesh is linearly matched with the allowable height deviation range corresponding to the slope threshold, so that the height adaptability of the mesh meets the mechanical deformation requirements. Finally, the mesh node coordinates are aligned and calibrated point by point with the geometric distribution boundary of the spatial constraint model to construct the target 3D mesh covering the detection area, providing a spatial framework for the 3D correction of the planar projection point set.

[0144] Step 403: Based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh, the planar projection point set is corrected to generate a corrected point set.

[0145] In this step, the variation law of the height feature component refers to the distribution pattern of the height feature component in three-dimensional space. This variation law includes local abrupt changes and global trends. Local abrupt changes may be abnormal thermal radiation at the bolt connection, while global trends may be linear height shifts caused by slope changes.

[0146] The corrected point set refers to the set of three-dimensional spatial points obtained by height interpolation and conflict adjustment of the planar projection point set. Its height distribution conforms to the preset allowable deviation parameter constraints and reflects the spatial attitude of the corrected positioning mark.

[0147] In this embodiment, the variation law of the height feature component is first analyzed. If the height difference between adjacent points exceeds the preset allowable deviation, the target three-dimensional mesh is divided into sub-intervals based on the node spacing, and linear interpolation is performed in each interval to smooth the height change. If the thermal radiation intensity gradient corresponding to the height feature component conflicts with the bending radius of the cantilever pipe, the interpolation weight is adjusted according to the projection rule corresponding to the installation angle to correct the coordinates of the plane projection points in the conflict area. The adjusted plane coordinates are recombined with the height feature component, and abnormal points that exceed the mesh boundary are removed to generate a correction point set to ensure that the three-dimensional distribution of the point set conforms to the mechanical constraints.

[0148] Step 404: Based on the spatial distribution characteristics of the correction point set, construct a topology that matches the spatial pose of the wrist arm.

[0149] In this step, the spatial distribution characteristics of the corrected point set refer to the three-dimensional geometric properties of the corrected point set. These characteristics include the distance between adjacent points, the local radius of curvature, and spatial continuity. These characteristics must meet the mechanical rules constraints of the high-speed railway catenary.

[0150] In this embodiment, a topology is constructed based on the spatial distribution characteristics of the modified point set. Specifically: First, the spacing between adjacent points is compared with the allowable range of the standard bolt spacing to screen for effective connections, ensuring that the spacing does not exceed the standard constraints of bolt connections and avoiding contour breakage; then, the local radius of curvature is checked to eliminate abnormal points smaller than the deformation tolerance limit of the pipe, ensuring that the structure meets the mechanical strength requirements; finally, the effective points are connected in a reasonable order through spatial lines to form a topology that matches the spatial posture of the cantilever arm.

[0151] The embodiments of this application solve the problem that traditional static models cannot adapt to changes in the slope of the cantilever arm and mechanical deformation. By constraining mechanical rules, the physical rationality of the topological structure is guaranteed, and the accuracy and structural integrity of the three-dimensional detection of positioning markers in complex environments are improved.

[0152] This application provides a specific embodiment. Step 403 involves correcting the planar projection point set based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh to generate a corrected point set. This specifically includes the following steps:

[0153] Step 411: Based on the node spacing of the target 3D mesh, smooth the planar projection point set to generate a corrected point set.

[0154] In this embodiment, a sliding window is set based on the node spacing of the target 3D mesh. The window size is consistent with the node spacing to adapt to the mesh accuracy. The planar projection point set is input into the sliding window in sequence, and the average horizontal coordinate of all points in the sliding window is calculated point by point. The average value is used to replace the original coordinate of the center of the window to achieve smooth adjustment of the planar points. The sliding window is repeated to traverse all points to ensure that each point has undergone neighborhood smoothing processing to generate the corrected point set.

[0155] Step 412: Based on the thermal radiation intensity information corresponding to the height feature component, identify multiple target areas that conflict with the bending radius of the tube of the cantilever arm, and extract the target area point set from the target areas.

[0156] In this step, thermal radiation intensity information refers to the intensity values ​​of each pixel extracted from the infrared band image, reflecting the difference in thermal characteristics between the positioning mark and the background, and is used to help determine the rationality of the spatial point; the target area point set refers to the set of points extracted from the set of planar projection points that correspond to the area that has geometric conflict with the bending radius of the cantilever pipe, and the spatial distribution of these points does not conform to the mechanical deformation law of the pipe.

[0157] In this embodiment, each point in the corrected point set is associated with its corresponding height feature component, and the corresponding thermal radiation intensity information is inferred from the height feature component. The local curvature of each point is calculated based on the gradient change of the thermal radiation intensity information. The local curvature is compared with the bending radius of the cantilever tube. If the degree of bending corresponding to the local curvature exceeds the allowable deformation range of the tube, the area where the point is located is determined to be the target area. All points in the target area are extracted, grouped by region to form a target area point set, and the conflict points that need to be corrected are identified.

[0158] Step 413: Adjust the height feature components of the target region point set according to the preset projection rules to generate the adjusted target region point set.

[0159] In this step, the preset projection rules refer to the spatial projection adjustment rules based on the cantilever installation angle, slope threshold and pipe bending radius, which are used to regulate the height correction direction and magnitude of conflict points.

[0160] In this embodiment, a preset projection rule is invoked, the projection correction direction is determined based on the cantilever installation angle, the height adjustment ratio is calculated according to the slope threshold, and the upper limit of the correction range is set in combination with the pipe bending radius; for each point in the target area point set, the adjustment amount of the height feature component is calculated based on its corresponding thermal radiation intensity information and local curvature. The adjustment amount is the product of the original height feature component and the adjustment ratio, and does not exceed the upper limit of the correction range; the height feature component of the target area point set is corrected point by point according to the calculated adjustment amount to generate the adjusted target area point set.

[0161] Step 414: Spatially superimpose the corrected point set and the adjusted target region point set to obtain a superimposed point set. Remove isolated target points from the superimposed point set to generate a corrected point set.

[0162] In this step, isolated target points refer to anomalous points that are far from most points after spatial overlay and have no effective related points in their neighborhood. These points are mostly caused by noise interference or correction deviations, which will affect the integrity of the subsequent topology.

[0163] In this embodiment, the corrected point set and the adjusted target area point set are superimposed according to spatial coordinates. Points at the same spatial location retain the coordinates of the adjusted target area point set, while the remaining points retain the coordinates of the corrected point set, forming the superimposed point set. A neighborhood search range is set, and the number of neighboring points of each point in the superimposed point set within the search range is counted point by point. If the number of neighboring points is lower than the set standard, it is determined to be an isolated target point. All isolated target points are removed, and the remaining valid points are arranged in spatial order to generate a corrected point set, ensuring the spatial continuity and rationality of the point set.

[0164] The embodiments of this application solve the problems of mismatch between smoothing effect and mesh, lack of targeted adjustment of conflict points, and interference of isolated points with topology construction in traditional correction, thereby improving the mechanical rationality and spatial integrity of the correction point set.

[0165] This application provides a specific embodiment. Step 105, based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model, generates the final detection coordinates of the positioning mark of the carpal arm, specifically including the following steps:

[0166] Step 501: Align the node coordinates of the topology with the geometric distribution boundary of the spatial constraint model to generate the aligned topology.

[0167] In this step, the node coordinates of the topology refer to the three-dimensional spatial coordinates of each node that makes up the topology, which are used to accurately represent the spatial distribution of the positioning markers.

[0168] Aligned topology refers to the topology obtained by aligning the node coordinates of the topology with the geometric distribution boundary of the spatial constraint model, which is used to ensure that the topology is consistent with the preset spatial constraint benchmark.

[0169] In this embodiment, the origin and axis directions of the reference coordinates are first determined based on the geometric distribution boundary of the spatial constraint model. Then, the node coordinates of the topology are extracted point by point, and the positional deviation between each node coordinate and the corresponding reference point of the geometric distribution boundary is calculated. Next, the node coordinates are adjusted according to the deviation value. The specific adjustment process is as follows: when the deviation is positive, it is finely adjusted to the inside of the boundary; when the deviation is negative, it is finely adjusted to the outside of the boundary to ensure that all nodes fit within the constraint range of the geometric distribution boundary. After the above adjustment is completed, the node coordinates are integrated to generate the aligned topology.

[0170] Step 502: Calculate the overlap between the aligned topology and the geometric distribution boundary, and mark nodes with an overlap lower than a preset threshold as abnormal nodes.

[0171] In this step, the overlap refers to the degree of matching between the nodes of the aligned topology and the geometric distribution boundary of the spatial constraint model, which is quantified by the positional fit between the node coordinates and the boundary datum.

[0172] The preset threshold refers to the overlap qualification standard set based on the mechanical installation specifications of high-speed railway catenary, which is used to determine whether the node meets the spatial constraint requirements.

[0173] In this embodiment, the shortest distance from each node in the aligned topology to the geometric distribution boundary is calculated point by point. This distance is compared with the allowable deviation range of the boundary to obtain the overlap degree of each node. The smaller the distance from the node to the boundary, the higher the overlap degree. The overlap degree is highest when the distance is zero. The overlap degree of each node is compared with a preset threshold. If the overlap degree is lower than the preset threshold, it means that the node deviates from the constraint boundary beyond the allowable range and is marked as an abnormal node.

[0174] Step 503: Correct the spatial position of the abnormal node to obtain the corrected topology.

[0175] In this step, the corrected topology refers to the topology obtained after adjusting the spatial positions of the abnormal nodes, and the node coordinates all conform to the geometric distribution boundary requirements of the spatial constraint model.

[0176] In this embodiment, for each abnormal node, the coordinate distribution pattern of its adjacent valid nodes is analyzed, and the reasonable position range of the abnormal node is determined by combining the geometric distribution boundary characteristics of the spatial constraint model. Then, the average distance from the abnormal node to the adjacent valid nodes is calculated, and the corrected displacement of the node is determined according to the ratio of the average distance to the boundary datum. Next, the coordinates of the abnormal node are adjusted to the reasonable position range according to the corrected displacement to ensure that the overlap of the adjusted nodes is higher than a preset threshold. Finally, after all abnormal nodes are corrected, the coordinates of all corrected nodes are integrated to obtain the corrected topology.

[0177] Step 504: Verify the overlap between the node coordinates of the corrected topology and the geometric distribution boundary of the spatial constraint model to generate verification results.

[0178] In this step, the verification result refers to the judgment result obtained after checking the overlap between the node coordinates of the corrected topology and the geometric distribution boundary of the spatial constraint model. The result includes two cases: qualified and unqualified, which is used to confirm whether the topology meets the spatial constraint requirements.

[0179] In this embodiment of the application, the same calculation method as in step 502 is used to recalculate the overlap degree between each node in the corrected topology and the geometric distribution boundary point by point; then the overlap degree of all nodes is statistically analyzed. If the overlap degree of all nodes is higher than the preset threshold and the results of three consecutive detections are consistent, a qualified verification result is generated; if there are still nodes with an overlap degree lower than the preset threshold, an unqualified verification result is generated.

[0180] Step 505: Based on the verification results, generate the final detection coordinates of the positioning mark of the wrist arm.

[0181] In this step, the final detection coordinates refer to the three-dimensional spatial coordinates of the key nodes of the positioning marker extracted based on the qualified verification results, which are used to accurately characterize the actual installation position of the positioning marker on the cantilever arm of the high-speed rail catenary.

[0182] In this embodiment of the application, if the verification result is qualified, the core feature nodes of the positioning mark are selected from the modified topology. The core feature nodes include key position nodes such as contour vertices and bolt connection corresponding nodes. Then, the three-dimensional coordinates of these core feature nodes are extracted, sorted and arranged according to the structural order of the positioning mark, and duplicate coordinates and redundant edge coordinates are removed. Finally, they are integrated to form the final detection coordinates, providing data support for the operation and maintenance detection of the high-speed rail catenary.

[0183] The embodiments of this application solve the detection deviation problems caused by the lack of rigorous coordinate verification and the failure to correct abnormal nodes in a timely manner in traditional methods, improve the accuracy and reliability of the final detection coordinates, and make the position detection of the positioning mark meet the mechanical specifications of the high-speed rail catenary.

[0184] Figure 2 This application provides a schematic diagram of the structure of a wrist-arm positioning marker detection system based on image processing, as shown in the embodiment of the present application. Figure 2 As shown, the system includes:

[0185] The generation module 21 is used to obtain the operating parameters of the cantilever arm in the high-speed railway catenary in order to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the locator.

[0186] The acquisition module 22 is used to acquire visible light and infrared images of the positioning marker of the carpal tunnel within the geometric distribution boundary of the spatial constraint model, and fuse the visible light and infrared images to generate a multispectral enhanced image.

[0187] The acquisition module 23 is used to acquire the initial contour point set of the positioning marker based on the reflection features of the positioning marker in the multispectral enhanced image, combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model;

[0188] The correction module 24 is used to perform three-dimensional projection correction on the initial contour point set to generate a topology that matches the spatial posture of the carpal arm.

[0189] Output module 25 is used to generate the final detection coordinates of the positioning mark of the carpal arm based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model.

[0190] Figure 2 The image processing-based wrist-arm positioning marker detection system described above can perform... Figure 1 The implementation principle and technical effects of the image processing-based wrist-arm positioning marker detection method described in the illustrated embodiment will not be repeated here. The specific operation methods of each module and unit in the image processing-based wrist-arm positioning marker detection system described in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0191] In one possible design, Figure 2 The image processing-based wrist-arm positioning marker detection system of the embodiment shown can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0192] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0193] The processing component 32 is used to: acquire the operating parameters of the cantilever arm in the high-speed rail catenary to generate a spatial constraint model, the operating parameters including the installation angle of the cantilever arm and the slope threshold of the locator; acquire visible light and infrared images of the positioning markers of the cantilever arm within the geometric distribution boundary of the spatial constraint model, fuse the visible light and infrared images to generate a multispectral enhanced image; based on the reflection features of the positioning markers in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, acquire the initial contour point set of the positioning markers; perform three-dimensional projection correction on the initial contour point set to generate a topology matching the spatial posture of the cantilever arm; and generate the final detection coordinates of the positioning markers of the cantilever arm based on the verification results of the topology and the geometric distribution boundary of the spatial constraint model.

[0194] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits, digital signal processors, digital signal processing devices, programmable logic devices, field-programmable gate arrays, controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0195] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory, electrically erasable programmable read-only memory, erasable programmable read-only memory, programmable read-only memory, read-only memory, magnetic storage, flash memory, magnetic disk, or optical disk.

[0196] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0197] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0198] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0199] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0200] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The illustrated embodiment is a method for detecting wrist and arm positioning markers based on image processing.

[0201] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0202] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0203] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0204] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting wrist-arm positioning markers based on image processing, characterized in that, include: The operating parameters of the cantilever arm in the high-speed railway catenary are obtained to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the positioner. Visible light and infrared images of the positioning markers of the carpal tunnel are acquired within the geometric distribution boundary of the spatial constraint model. The visible light and infrared images are then fused to generate a multispectral enhanced image. Based on the reflection features of the location marker in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, the initial contour point set of the location marker is obtained. The initial contour point set is corrected by three-dimensional projection to generate a topology that matches the spatial pose of the carpal arm; Based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model, the final detection coordinates of the positioning mark of the carpal arm are generated.

2. The method for detecting wrist-arm positioning markers based on image processing according to claim 1, characterized in that, Based on the reflection features of the location marker in the multispectral enhanced image, and combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model, an initial contour point set of the location marker is obtained, including: Texture gradient information, edge information, stability coefficient, and thermal radiation intensity information are extracted from the multispectral enhanced image; Using a weight allocation mechanism, a dynamic attenuation factor and a boundary weight matrix are generated, and combined with the thermal radiation intensity information, a spatial attenuation gradient is generated. The texture gradient information, edge information, stability coefficient, and thermal radiation intensity information are fused at multiple levels using a feature aggregation mechanism to generate a feature response map. Based on the dynamic decay factor and the spatial decay gradient, the feature response map is adjusted to generate a set of candidate response points; Based on the topology rules of the candidate response point set and the fracture repair rules of the edge information, a target connection path is generated, and an initial contour point set is generated by combining the candidate response point set.

3. The method for detecting wrist-arm positioning markers based on image processing according to claim 2, characterized in that, Based on the dynamic decay factor and the spatial decay gradient, the feature response map is adjusted to generate a candidate response point set, including: The feature response map is adjusted based on the dynamic decay factor and the spatial decay gradient to generate an adjusted feature response map. Based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information, a dynamic threshold map is generated. The adjusted feature response map and the feature response map are subjected to feature extraction and convolution operations to generate a decayed feature response map; The difference between the response value of each pixel in the attenuated feature response map and the dynamic threshold of the corresponding pixel in the dynamic threshold map is calculated to generate a difference value matrix. The pixels with difference values ​​greater than zero are selected from the difference value matrix as candidate response points to generate a candidate response point set.

4. The method for detecting wrist-arm positioning markers based on image processing according to claim 3, characterized in that, Based on the spatial distribution characteristics of the stability coefficient and the temporal fluctuation characteristics of the thermal radiation intensity information, a dynamic threshold map is generated, including: Based on the numerical distribution of the stability coefficient, a first threshold is generated, and based on the fluctuation characteristics of the stability coefficient, the first threshold is adjusted to generate a second threshold. Based on the time-domain fluctuation characteristics of the thermal radiation intensity information, the thermal radiation intensity information is decomposed into thermal radiation energy distribution and energy fluctuation components. Based on the thermal radiation energy distribution, a third threshold is generated. Based on the energy fluctuation component, the third threshold is scaled and modulated to generate a fourth threshold. Based on the environmental type identifier of the wrist arm, the second threshold and the fourth threshold are subjected to multi-level filtering to obtain the initial threshold map; Based on the attenuation coefficient of the adjusted feature response map, the initial threshold map is corrected to generate a dynamic threshold map.

5. The method for detecting wrist-arm positioning markers based on image processing according to claim 1, characterized in that, The initial contour point set is subjected to 3D projection correction to generate a topology that matches the spatial pose of the carpal arm, including: The initial contour point set is decomposed into a planar projection point set and a height feature component; Construct the target 3D mesh based on preset allowable deviation parameters; Based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh, the planar projection point set is corrected to generate a corrected point set; Based on the spatial distribution characteristics of the correction point set, a topology matching the spatial pose of the carpal arm is constructed.

6. The method for detecting wrist-arm positioning markers based on image processing according to claim 5, characterized in that, Based on the variation law of the height feature components and the node spacing and height gradient of the target 3D mesh, the planar projection point set is corrected to generate a corrected point set, including: Based on the node spacing of the target 3D mesh, the planar projection point set is smoothed to generate a corrected point set; Based on the thermal radiation intensity information corresponding to the height feature components, multiple target regions that conflict with the bending radius of the tube of the cantilever arm are identified, and a set of target region points is extracted from the target regions. Adjust the height feature components of the target region point set according to the preset projection rules to generate the adjusted target region point set. The corrected point set and the adjusted target region point set are spatially superimposed to obtain the superimposed point set. Isolated target points in the superimposed point set are then removed to generate the corrected point set.

7. The method for detecting wrist-arm positioning markers based on image processing according to claim 1, characterized in that, Based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model, the final detection coordinates of the carpal tunnel's positioning marker are generated, including: Align the node coordinates of the topology with the geometric distribution boundary of the spatial constraint model to generate the aligned topology. Calculate the overlap between the aligned topology and the geometric distribution boundary, and mark nodes with an overlap below a preset threshold as abnormal nodes; The abnormal nodes are spatially corrected to obtain the corrected topology. The overlap between the node coordinates of the modified topology and the geometric distribution boundary of the spatial constraint model is verified to generate verification results. Based on the verification results, the final detection coordinates of the positioning marker of the wrist arm are generated.

8. A detection system for wrist-arm positioning markers based on image processing, characterized in that, include: The generation module is used to obtain the operating parameters of the cantilever arm in the high-speed railway catenary in order to generate a spatial constraint model. The operating parameters include the installation angle of the cantilever arm and the slope threshold of the positioner. The acquisition module is used to acquire visible light and infrared images of the positioning marker of the carpal tunnel within the geometric distribution boundary of the spatial constraint model, and to fuse the visible light and infrared images to generate a multispectral enhanced image. The acquisition module is used to acquire the initial contour point set of the positioning marker based on the reflection features of the positioning marker in the multispectral enhanced image, combined with the weight allocation mechanism and feature aggregation mechanism of the spatial constraint model; The correction module is used to perform three-dimensional projection correction on the initial contour point set to generate a topology that matches the spatial pose of the carpal arm. The output module is used to generate the final detection coordinates of the positioning marker of the carpal arm based on the verification results of the geometric distribution boundary of the topology and the spatial constraint model.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the image processing-based wrist and arm positioning marker detection method as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements a method for detecting a wrist-arm positioning marker based on image processing as described in any one of claims 1 to 7.

Images