Curtain wall thermal cycle detection method and system based on machine vision

Abstract

The invention discloses a curtain wall thermal cycle detection method and system based on machine vision. The method comprises the steps that sampling points are arranged at different parts of a curtain wall, and the inner side and the outer side are marked; collecting an image of the inner side of the glass at a thermal cycle test moment, extracting features, and obtaining sampling point information; inputting the sampling point information, the curtain wall thickness and the refractive index into a deformation degree prediction model, and outputting a deformation degree data set; integral deformation degree data of the curtain wall are obtained through deformation degree interpolation; and according to the deformation degree data, dividing deformation area grades, and adjusting sampling point density distribution of the next round of detection. The internal and external mark points are arranged at the curtain wall sampling points, the deformation degree is calculated through the image processing technology, the density of the sampling points is dynamically adjusted, fine monitoring of abnormal areas is achieved, and the detection efficiency and precision are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of machine vision detection, and in particular to a curtain wall thermal cycle detection method and system based on machine vision. Background Art

[0002] With the development of the construction industry, curtain walls, as an important building envelope structure, need to have good physical properties while ensuring the beauty of the building. Thermal cycle testing is one of the important means to evaluate the performance of curtain walls. It can simulate the ability of curtain walls to withstand temperature changes in actual environments, thereby evaluating their reliability in long-term use. At present, thermal cycle testing mainly relies on manual observation or traditional sensor technology to detect the deformation of curtain walls under thermal cycles, but these methods have certain limitations.

[0003] Traditional thermal cycle detection usually relies on manual observation by technicians, who observe the deformation of the curtain wall at different temperatures with the naked eye and record the relevant data. This method is simple and easy to use, but it is highly subjective and easily limited by the observer's experience and vision, resulting in low data accuracy. Another common method is to install sensors (such as displacement sensors) at key parts of the curtain wall to evaluate the thermal deformation of the curtain wall by recording the displacement changes at different temperatures. Although this method is more objective than manual observation, it cannot fully cover all areas of the curtain wall due to the limited number of sensors, and the installation and maintenance costs of the sensors are high. Summary of the invention

[0004] In order to solve at least one of the technical problems mentioned above, the present invention provides a curtain wall thermal cycle detection method and system based on machine vision.

[0005] In a first aspect, the present invention provides a curtain wall thermal cycle detection method based on machine vision, the method comprising:

[0006] Set multiple sampling points at different locations of the curtain wall specimen, and set marking points on both the inner and outer sides of the glass at each sampling point;

[0007] Collect images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain a plurality of curtain wall specimen images, perform feature extraction on the curtain wall specimen images to obtain a plurality of sampling point information, wherein the sampling point information includes an inner marking point position and an outer marking point position;

[0008] The sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images are sequentially input into a preset deformation prediction model, and a plurality of sampling point deformation data sets at different thermal cycle test times are correspondingly output;

[0009] According to the sampling point information and the deformation data set of the sampling points at different thermal cycle test times, deformation interpolation is performed on the curtain wall specimen image to obtain corresponding deformation data of multiple curtain wall specimens;

[0010] The deformation area of ​​the curtain wall specimens is graded according to the deformation data of the curtain wall specimens, and the sampling point density distribution of the next curtain wall thermal cycle detection process is adjusted according to the deformation area grade division results.

[0011] Preferably, the deformation prediction model is expressed as follows:

[0012]

[0013] Where D i is the deformation degree of the i-th sampling point, d is the curtain wall thickness of the curtain wall specimen, n1 is the refractive index of air, n2 is the curtain wall refractive index of the curtain wall specimen, θ1 is the angle between the shooting direction and the straight line perpendicular to the glass surface, θ2 is the angle between the refracted light and the straight line perpendicular to the surface of the curtain wall specimen, x in and in is the horizontal and vertical coordinates of the inner marking point of the i-th sampling point, x out and out are the horizontal and vertical coordinates of the external marker point of the i-th sampling point without refraction correction, x , out and o,ut are the horizontal and vertical coordinates of the external marker point of the i-th sampling point after refractive index calibration.

[0014] Preferably, the feature extraction of the curtain wall specimen image to obtain multiple sampling point information includes:

[0015] Grayscale the curtain wall specimen image;

[0016] Harris corner detection algorithm is used to detect the marking points of the curtain wall specimen image after grayscale processing to obtain multiple marking point data;

[0017] The K-Means algorithm is used to divide the marked point data into multiple classification clusters, where the number of classification clusters is equal to the number of sampling points;

[0018] The marker point data in different classification clusters are positioned to obtain the inner marker point positions and outer marker point positions corresponding to multiple sampling points.

[0019] Preferably, the step of classifying the deformation areas of the curtain wall specimens according to the deformation data of the curtain wall specimens, and adjusting the density distribution of sampling points of the next curtain wall thermal cycle detection process according to the deformation area classification result includes:

[0020] Based on the preset abnormal deformation threshold, the convex hull algorithm is used to divide the curtain wall specimen into normal area and abnormal area.

[0021] In the next curtain wall thermal cycle detection process, maintain the density distribution of sampling points in normal areas and increase the density of sampling points in abnormal areas.

[0022] Preferably, a plurality of sampling points are set at different positions of the curtain wall specimen, and marking points are set on both the inner and outer sides of the glass of each sampling point, including:

[0023] Multiple sampling points are evenly set on the curtain wall specimen, and marking points are set on both sides of the glass inside and outside of each sampling point, where the outer marking point is larger than the inner marking point.

[0024] In a second aspect, the present invention further provides a curtain wall thermal cycle detection system based on machine vision, the system comprising:

[0025] The sampling point setting module is used to set multiple sampling points at different parts of the curtain wall specimen, and marking points are set on both the inner and outer sides of the glass of each sampling point;

[0026] A sampling point feature acquisition module is used to collect images of the inner side of the curtain wall specimen glass at different thermal cycle test times to obtain multiple curtain wall specimen images, perform feature extraction on the curtain wall specimen images, and obtain multiple sampling point information, wherein the sampling point information includes an inner marking point position and an outer marking point position;

[0027] The deformation prediction module is used to sequentially input the sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images into a preset deformation prediction model, and correspondingly output a plurality of sampling point deformation data sets at different thermal cycle test times;

[0028] The deformation interpolation module is used to perform deformation interpolation on the curtain wall specimen image according to the sampling point information and the sampling point deformation data set at different thermal cycle test times, and obtain corresponding deformation data of multiple curtain wall specimens;

[0029] The sampling point density adjustment module is used to classify the deformation area of ​​the curtain wall specimen according to the deformation data of the curtain wall specimen, and adjust the sampling point density distribution of the next curtain wall thermal cycle detection process according to the deformation area classification result.

[0030] Preferably, the sampling point feature acquisition module includes:

[0031] A graphics acquisition unit, used to acquire images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain multiple curtain wall specimen images;

[0032] Graphics grayscale unit, used to grayscale the curtain wall specimen image;

[0033] A marking point extraction unit is used to perform marking point detection on the curtain wall specimen image after grayscale processing by using Harris corner point detection algorithm to obtain multiple marking point data;

[0034] The marker point classification unit is used to classify the marker point data into multiple classification clusters using the K-Means algorithm, where the number of classification clusters is equal to the number of sampling points;

[0035] The marker point positioning unit is used to locate the marker point data in different classification clusters to obtain the inner marker point positions and the outer marker point positions corresponding to the multiple sampling points.

[0036] Preferably, the sampling point density adjustment module includes:

[0037] A deformation area classification unit is used to divide the curtain wall specimen into a normal area and an abnormal area based on a preset abnormal deformation threshold using a convex hull algorithm;

[0038] The sampling point density adjustment unit is used to maintain the sampling point density distribution in the normal area and increase the sampling point density in the abnormal area in the next curtain wall thermal cycle detection process.

[0039] In a third aspect, the present invention further provides an electronic device comprising a processor and a memory, wherein the memory is used to store computer program code, and the computer program code comprises computer instructions. When the processor executes the computer instructions, the electronic device executes the method as described in the first aspect above and any possible implementation thereof.

[0040] In a fourth aspect, the present invention further provides a computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, wherein the computer program includes program instructions, and when the program instructions are executed by a processor of an electronic device, the processor executes the method as described in the first aspect above and any possible implementation thereof.

[0041] Compared with the prior art, the present invention has the following beneficial effects:

[0042] 1) The present invention sets inner and outer marking points for the sampling points of the glass curtain wall, so that the inner and outer marking points of the sampling points at different parts of the curtain wall can be captured in one image at the same time. By calculating the position offset of the inner and outer marking points of different sampling points, the deformation degree of each sampling point can be accurately calculated. Furthermore, according to the deformation degree of different parts, the distribution density of the sampling points of these parts is dynamically adjusted, so that the abnormal area with larger deformation can obtain more dense sampling data, thereby realizing more detailed data collection and analysis of the abnormal area. This embodiment not only improves the detection efficiency, but also significantly enhances the monitoring accuracy of potential problem areas.

[0043] 2) The deformation prediction model provided by the present invention takes into account the influence of the refractive index of the curtain wall. By correcting the position of the external marking point, it can more accurately reflect the actual position deviation of the internal and external marking points, thereby obtaining more accurate deformation data. Furthermore, the deformation prediction model does not require a complex large model, simplifies the calculation process, reduces the demand for computing resources, and makes the deformation calculation more efficient and quick. Through fast and accurate deformation calculation, data processing of a large number of sampling points can be completed in a short time, thereby improving the speed of the entire curtain wall thermal cycle detection process.

[0044] 3) The present invention achieves effective utilization of resources by classifying deformation areas and adjusting the density distribution of sampling points accordingly, especially increasing the detection density in key areas (abnormal areas) and enhancing the monitoring of potential problem areas.

[0045] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the background technology, the drawings required for use in the embodiments of the present invention or the background technology will be described below.

[0047] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0048] Figure 1 A schematic flow chart of a curtain wall thermal cycle detection method based on machine vision provided by an embodiment of the present invention;

[0049] Figure 2 A schematic diagram of the structure of a curtain wall thermal cycle detection system based on machine vision provided by an embodiment of the present invention;

[0050] Figure 3 The embodiment of the present invention provides Figure 2 A schematic diagram of the structure of the submodules of the sampling point feature acquisition module;

[0051] Figure 4 The embodiment of the present invention provides Figure 2 Schematic diagram of the structure of the submodules of the sampling point density adjustment module. DETAILED DESCRIPTION

[0052] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0054] Existing curtain wall thermal cycle detection technology mainly relies on manual observation and traditional sensor technology, which has problems such as low coverage, limited data accuracy, high cost and poor flexibility, making it difficult to achieve comprehensive, accurate and cost-effective deformation monitoring of curtain walls.

[0055] See also Figure 1 , Figure 1 The present invention provides a flow chart of a curtain wall thermal cycle detection method based on machine vision. Figure 1 As shown, the method includes:

[0056] S100, multiple sampling points are set at different parts of the curtain wall specimen, and marking points are set on both the inner and outer sides of the glass at each sampling point;

[0057] In this embodiment, multiple sampling points are set at different parts of the curtain wall specimen (which may be all parts of the curtain wall or preset key parts, such as corners, the middle, the top, etc.), and easily identifiable marking points are set on both the inner and outer sides of the glass of each sampling point (for example, a red circular mark is used as the inner mark, and a blue square mark is used as the outer mark). It should be noted that the marking points should have good contrast to facilitate subsequent image processing.

[0058] S200, collecting images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain a plurality of curtain wall specimen images, performing feature extraction on the curtain wall specimen images to obtain a plurality of sampling point information, wherein the sampling point information includes an inner marking point position and an outer marking point position;

[0059] In this embodiment, a high-resolution camera is used to capture images of the inner side of the curtain wall specimen glass at different thermal cycle test times to obtain multiple curtain wall specimen images. In order to ensure image quality, the background should be a pure color to reduce interference and facilitate image processing. Furthermore, consistent lighting conditions should be ensured during image acquisition to reduce the impact of environmental factors on image quality.

[0060] In a possible embodiment, a high-resolution camera may be used to capture images of the outer side of the curtain wall specimen glass at different thermal cycle test times to obtain multiple curtain wall specimen images.

[0061] In this embodiment, the curtain wall specimen image is input into a pre-trained marker point position prediction model based on a convolutional neural network to obtain sampling point information, wherein the sampling point information includes an inner marker point position and an outer marker point position.

[0062] It should be noted that the training process of the marker position prediction model is as follows:

[0063] A dataset of images of the inner side of the glass of curtain wall specimens at different thermal cycle test times is collected, and the marker points in each image are annotated (the inner marker points of the sampling points and the outer marker points of the sampling points), and the exact position of the marker points is recorded. Furthermore, the images are uniformly adjusted to the same size, and data enhancement is performed to increase the generalization ability of the model. Then, the Faster R-CNN model is selected as the marker point position prediction model, the cross entropy loss function is selected as the loss function, the model parameters are initialized using random initialization or other pre-trained weights, and the Adam optimizer is selected for model training. During the training process, the image is input into the model for forward propagation to obtain the predicted marker point position, and the loss is calculated based on the model's prediction results and the true label. The model parameters are updated through back propagation, and the training is iteratively performed on the training set and the validation set until the model reaches the expected performance level. Finally, the performance of the model is evaluated on the test set, and the best model weights are saved for subsequent use.

[0064] S300, sequentially inputting sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images into a preset deformation prediction model, and correspondingly outputting a plurality of sampling point deformation data sets at different thermal cycle test times;

[0065] In this embodiment, the sampling point information (including the inner marking point position and the outer marking point position) extracted from different curtain wall specimen images, the curtain wall thickness and the curtain wall refractive index are used as input, and the deformation prediction model trained in advance is used to obtain the sampling point deformation data set at different thermal cycle test times.

[0066] S400, performing deformation interpolation on the curtain wall specimen image according to the sampling point information and the sampling point deformation data set at different thermal cycle test times, and correspondingly obtaining a plurality of curtain wall specimen deformation data;

[0067] In this embodiment, based on the sampling point information and the sampling point deformation data set at different thermal cycle test times, bilinear interpolation is used to obtain the deformation data of the entire curtain wall specimen at different thermal cycle test times. It should be noted that the bilinear interpolation process can use a Gaussian smoothing filter to reduce the impact of noise and ensure a smooth transition of the deformation data.

[0068] S500: classify the deformation area of ​​the curtain wall specimen according to the deformation degree data of the curtain wall specimen, and adjust the sampling point density distribution of the next curtain wall thermal cycle detection process according to the deformation area classification result.

[0069] In this embodiment, cluster analysis or threshold segmentation is used to divide the deformation area levels of the curtain wall specimens according to the deformation data of the curtain wall specimens. For example, the area with a deformation greater than or equal to a preset threshold can be divided into a "high-risk area", and the area with a deformation less than the preset threshold can be divided into a "normal area". According to the result of the deformation area classification, the sampling point density of the "high-risk area" is increased in the next curtain wall thermal cycle detection process, and the sampling point density of the "normal area" is reduced. On the one hand, the detection accuracy of the "high-risk area" can be improved, and on the other hand, the detection efficiency of the "normal area" can be optimized and the amount of calculation can be reduced.

[0070] In this embodiment, by setting internal and external marking points for the sampling points of the glass curtain wall, the inner and outer marking points of the sampling points at different parts of the curtain wall can be captured in one image at the same time. By calculating the position offset of the internal and external marking points of different sampling points, the deformation degree of each sampling point can be accurately calculated. Furthermore, according to the deformation degree of different parts, the distribution density of the sampling points in these parts is dynamically adjusted, so that more dense sampling data can be obtained in abnormal areas with larger deformation, thereby realizing more detailed data collection and analysis of abnormal areas. This embodiment not only improves the detection efficiency, but also significantly enhances the monitoring accuracy of potential problem areas.

[0071] Preferably, the deformation prediction model is expressed as follows:

[0072]

[0073] Where D i is the deformation degree of the i-th sampling point, d is the curtain wall thickness of the curtain wall specimen, n1 is the refractive index of air, n2 is the curtain wall refractive index of the curtain wall specimen, θ1 is the angle between the shooting direction and the straight line perpendicular to the glass surface, θ2 is the angle between the refracted light and the straight line perpendicular to the surface of the curtain wall specimen, xin and in is the horizontal and vertical coordinates of the inner marking point of the i-th sampling point, x out and out are the horizontal and vertical coordinates of the external marker point of the i-th sampling point without refraction correction, x , out and o,ut are the horizontal and vertical coordinates of the external marker point of the i-th sampling point after refractive index calibration.

[0074] In this embodiment, the deformation prediction model takes into account the influence of the refractive index of the curtain wall. By correcting the position of the external marking point, it can more accurately reflect the actual position deviation of the internal and external marking points, thereby obtaining more accurate deformation data. Furthermore, the deformation prediction model does not require a complex large model, simplifies the calculation process, reduces the demand for computing resources, and makes the deformation calculation more efficient and quick. Through fast and accurate deformation calculation, data processing of a large number of sampling points can be completed in a short time, thereby improving the speed of the entire curtain wall thermal cycle detection process.

[0075] Preferably, the feature extraction of the curtain wall specimen image to obtain multiple sampling point information includes:

[0076] Grayscale the curtain wall specimen image;

[0077] Harris corner detection algorithm is used to detect the marking points of the curtain wall specimen image after grayscale processing to obtain multiple marking point data;

[0078] The K-Means algorithm is used to divide the marked point data into multiple classification clusters, where the number of classification clusters is equal to the number of sampling points;

[0079] The marker point data in different classification clusters are positioned to obtain the inner marker point positions and outer marker point positions corresponding to multiple sampling points.

[0080] In this embodiment, a weighted average method (for example, I=0.299R+0.587G+0.114B, where R, G and B represent the pixel values ​​of the red, green and blue channels of the original image, respectively, and I represents the grayscale value) is used to convert the color curtain wall specimen image into a grayscale image to reduce the computational complexity and highlight the contrast in the image. The Harris corner detection algorithm is used to calculate the gradient change around each pixel to identify the corner point, and a Harris response value is obtained. The larger the value, the more likely the pixel is a corner point. When the Harris response value is detected to be greater than a preset threshold, the corresponding pixel group is determined to be a marking point, and the position information of the marking point is recorded to obtain the marking point data. Furthermore, the number of classifications is set to the number of sampling points, and the data points are iteratively redistributed to the nearest centroid based on the K-Means algorithm, and finally converge to the best clustering result, so that the marking point data is divided into multiple classification clusters. For the marking point data in each classification cluster, the centroid position of each cluster is calculated to obtain the inner marking point position and the outer marking point position corresponding to the sampling point.

[0081] In this embodiment, the combination of grayscale processing and Harris corner detection algorithm can effectively reduce the false detection rate of marker points. Furthermore, the K-Means clustering algorithm is used to classify the marker point data, and the inner marker points and outer marker points corresponding to the sampling points are divided into a classification cluster, thereby accurately achieving rapid matching of marker points.

[0082] Preferably, the step of classifying the deformation areas of the curtain wall specimens according to the deformation data of the curtain wall specimens, and adjusting the density distribution of sampling points of the next curtain wall thermal cycle detection process according to the deformation area classification result includes:

[0083] Based on the preset abnormal deformation threshold, the convex hull algorithm is used to divide the curtain wall specimen into normal area and abnormal area.

[0084] In the next curtain wall thermal cycle detection process, maintain the density distribution of sampling points in normal areas and increase the density of sampling points in abnormal areas.

[0085] In this embodiment, according to the aforementioned feature extraction method (grayscale processing, Harris corner detection algorithm and K-Means algorithm, etc.), the inner mark point positions and the outer mark point positions corresponding to the plurality of sampling points are obtained, the position deviation (i.e., deformation data) of each sampling point is calculated based on the inner mark position and the outer mark position of the sampling point, a threshold T is defined for distinguishing normal deformation from abnormal deformation, all sampling points are projected onto a two-dimensional plane based on the threshold T, and two convex polygons containing all the sampling points are calculated using the convex hull algorithm, and it is determined whether the average deformation degree of the deformation data within the convex polygon is greater than or equal to the threshold T, if so, the corresponding convex polygon is determined to be an abnormal area, if not, the corresponding convex polygon is determined to be a normal area.

[0086] In the next curtain wall thermal cycle detection process, the density distribution of sampling points in the normal area is kept unchanged, and for the abnormal area, the density of sampling points is increased by increasing the number of sampling points. For example, new sampling points can be added in the abnormal area by using methods such as linear interpolation or quadratic interpolation.

[0087] In a possible implementation, multiple thresholds may be set to divide the curtain wall specimen into more refined deformation areas, thereby making a more refined adjustment to the sampling point distribution.

[0088] In this embodiment, by classifying the deformation areas and adjusting the density distribution of sampling points accordingly, effective utilization of resources is achieved, especially increasing the detection density in key areas (abnormal areas) and enhancing the monitoring of potential problem areas.

[0089] Preferably, a plurality of sampling points are set at different positions of the curtain wall specimen, and marking points are set on both the inner and outer sides of the glass of each sampling point, including:

[0090] Multiple sampling points are evenly set on the curtain wall specimen, and marking points are set on both sides of the glass inside and outside of each sampling point, where the outer marking point is larger than the inner marking point.

[0091] In this embodiment, multiple sampling points are evenly set at different positions of the curtain wall specimen to ensure the coverage of the entire curtain wall specimen surface. The number of sampling points is predetermined according to the size and shape of the curtain wall specimen and the expected detection accuracy. Marking points are set on both the inner and outer sides of the glass at each sampling point to facilitate subsequent deformation calculation. The size of the outer marking point is designed to be slightly larger than the inner marking point. The purpose of this is to ensure that the marking point can still be clearly identified when the outer light is more complex. In order to facilitate the detection of the image processing algorithm, the color of the marking point should form a clear contrast with the curtain wall background, such as using black or white marking points. Furthermore, the marking point should be made of a material that is easy to fix and will not cause damage to the curtain wall. At the same time, it is necessary to ensure that it will not fall off during the thermal cycle test.

[0092] In this embodiment, by setting larger outer marking points, good visibility can be maintained even under complex lighting conditions, thereby improving the robustness of the system.

[0093] In summary, the method provided in this embodiment can at least achieve the following effects:

[0094] 1) The present invention sets inner and outer marking points for the sampling points of the glass curtain wall, so that the inner and outer marking points of the sampling points at different parts of the curtain wall can be captured in one image at the same time. By calculating the position offset of the inner and outer marking points of different sampling points, the deformation degree of each sampling point can be accurately calculated. Furthermore, according to the deformation degree of different parts, the distribution density of the sampling points of these parts is dynamically adjusted, so that the abnormal area with larger deformation can obtain more dense sampling data, thereby realizing more detailed data collection and analysis of the abnormal area. This embodiment not only improves the detection efficiency, but also significantly enhances the monitoring accuracy of potential problem areas.

[0095] 2) The deformation prediction model provided by the present invention takes into account the influence of the refractive index of the curtain wall. By correcting the position of the external marking point, it can more accurately reflect the actual position deviation of the internal and external marking points, thereby obtaining more accurate deformation data. Furthermore, the deformation prediction model does not require a complex large model, simplifies the calculation process, reduces the demand for computing resources, and makes the deformation calculation more efficient and quick. Through fast and accurate deformation calculation, data processing of a large number of sampling points can be completed in a short time, thereby improving the speed of the entire curtain wall thermal cycle detection process.

[0096] 3) The present invention achieves effective utilization of resources by classifying deformation areas and adjusting the density distribution of sampling points accordingly, especially increasing the detection density in key areas (abnormal areas) and enhancing the monitoring of potential problem areas.

[0097] See also Figure 2 In one embodiment, a curtain wall thermal cycle detection system based on machine vision is also provided, the system comprising:

[0098] The sampling point setting module 100 is used to set multiple sampling points at different parts of the curtain wall specimen, and marking points are set on both sides of the glass inside and outside of each sampling point;

[0099] The sampling point feature acquisition module 200 is used to collect images of the inner side of the curtain wall specimen glass at different thermal cycle test times to obtain multiple curtain wall specimen images, perform feature extraction on the curtain wall specimen images, and obtain multiple sampling point information, wherein the sampling point information includes the inner marking point position and the outer marking point position;

[0100] The deformation prediction module 300 is used to sequentially input the sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images into a preset deformation prediction model, and correspondingly output a plurality of sampling point deformation data sets at different thermal cycle test times;

[0101] The deformation interpolation module 400 is used to perform deformation interpolation on the curtain wall specimen image according to the sampling point information and the sampling point deformation data set at different thermal cycle test times, and obtain corresponding deformation data of multiple curtain wall specimens;

[0102] The sampling point density adjustment module 500 is used to classify the deformation area of ​​the curtain wall specimen according to the deformation data of the curtain wall specimen, and adjust the sampling point density distribution of the next curtain wall thermal cycle detection process according to the deformation area classification result.

[0103] See also Figure 3 In one embodiment, the sampling point feature acquisition module 200 includes:

[0104] The image acquisition unit 201 is used to acquire images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain multiple curtain wall specimen images;

[0105] A graphic grayscale unit 202, used for grayscale processing of the curtain wall specimen image;

[0106] The marking point extraction unit 203 is used to perform marking point detection on the curtain wall specimen image after grayscale processing by using Harris corner point detection algorithm to obtain a plurality of marking point data;

[0107] A marker point classification unit 204 is used to classify the marker point data into a plurality of classification clusters using a K-Means algorithm, wherein the number of classification clusters is equal to the number of sampling points;

[0108] The marker point positioning unit 205 is used to position the marker point data in different classification clusters to obtain the inner marker point positions and the outer marker point positions corresponding to the multiple sampling points.

[0109] See also Figure 4 In one embodiment, the sampling point density adjustment module 500 includes:

[0110] The deformation region classification unit 501 is used to classify the curtain wall specimen into a normal region and an abnormal region by using a convex hull algorithm based on a preset abnormal deformation threshold;

[0111] The sampling point density adjustment unit 502 is used to maintain the sampling point density distribution in the normal area and increase the sampling point density in the abnormal area in the next curtain wall thermal cycle detection process.

[0112] It can be understood that the functions or modules included in the system provided in this embodiment can be used to execute the method described in the above method embodiment. Its specific implementation can refer to the description of the above method embodiment. For the sake of brevity, it will not be repeated here.

[0113] The present invention also provides an electronic device, including a processor and a memory, wherein the memory is used to store computer program code, and the computer program code includes computer instructions. When the processor executes the computer instructions, the electronic device executes a method as described in any possible implementation manner.

[0114] The present invention also provides a computer-readable storage medium, in which a computer program is stored. The computer program includes program instructions. When the program instructions are executed by a processor of an electronic device, the processor executes a method as described in any possible implementation manner.

[0115] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0116] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. Those skilled in the art can also clearly understand that the descriptions of the various embodiments of the present invention have different focuses. For the convenience and brevity of description, the same or similar parts may not be repeated in different embodiments. Therefore, for parts not described or not described in detail in a certain embodiment, refer to the records of other embodiments.

Claims

1. A curtain wall thermal cycle detection method based on machine vision, characterized in that: The method comprises: Set multiple sampling points at different locations of the curtain wall specimen, and set marking points on both the inner and outer sides of the glass at each sampling point; Collect images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain a plurality of curtain wall specimen images, perform feature extraction on the curtain wall specimen images to obtain a plurality of sampling point information, wherein the sampling point information includes an inner marking point position and an outer marking point position; The sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images are sequentially input into a preset deformation prediction model, and a plurality of sampling point deformation data sets at different thermal cycle test times are correspondingly output; According to the sampling point information and the deformation data set of the sampling points at different thermal cycle test times, deformation interpolation is performed on the curtain wall specimen image to obtain corresponding deformation data of multiple curtain wall specimens; The deformation area of ​​the curtain wall specimens is graded according to the deformation data of the curtain wall specimens, and the sampling point density distribution of the next curtain wall thermal cycle detection process is adjusted according to the deformation area grade division results.

2. The curtain wall thermal cycle detection method based on machine vision according to claim 1 is characterized in that: The deformation prediction model is expressed as follows: Where D i is the deformation degree of the i-th sampling point, d is the curtain wall thickness of the curtain wall specimen, n1 is the refractive index of air, n2 is the curtain wall refractive index of the curtain wall specimen, θ1 is the angle between the shooting direction and the straight line perpendicular to the glass surface, θ2 is the angle between the refracted light and the straight line perpendicular to the surface of the curtain wall specimen, x in and in is the horizontal and vertical coordinates of the inner marking point of the i-th sampling point, x out and out are the horizontal and vertical coordinates of the external marker point of the i-th sampling point without refraction correction, x , out and o,ut are the horizontal and vertical coordinates of the external marker point of the i-th sampling point after refractive index calibration.

3. The curtain wall thermal cycle detection method based on machine vision according to claim 1, characterized in that: The feature extraction of the curtain wall specimen image is performed to obtain multiple sampling point information, including: Grayscale the curtain wall specimen image; Harris corner detection algorithm is used to detect the marking points of the curtain wall specimen image after grayscale processing to obtain multiple marking point data; The K-Means algorithm is used to divide the marked point data into multiple classification clusters, where the number of classification clusters is equal to the number of sampling points; The marker point data in different classification clusters are positioned to obtain the inner marker point positions and outer marker point positions corresponding to multiple sampling points.

4. The curtain wall thermal cycle detection method based on machine vision according to claim 1, characterized in that: The method of classifying the deformation area of ​​the curtain wall specimen according to the deformation degree data of the curtain wall specimen, and adjusting the sampling point density distribution of the next curtain wall thermal cycle detection process according to the deformation area classification result, includes: Based on the preset abnormal deformation threshold, the convex hull algorithm is used to divide the curtain wall specimen into normal area and abnormal area. In the next curtain wall thermal cycle detection process, maintain the density distribution of sampling points in normal areas and increase the density of sampling points in abnormal areas.

5. The curtain wall thermal cycle detection method based on machine vision according to claim 1, characterized in that: The multiple sampling points are set at different positions of the curtain wall specimen, and marking points are set on both sides of the glass inside and outside of each sampling point, including: Multiple sampling points are evenly set on the curtain wall specimen, and marking points are set on both sides of the glass inside and outside of each sampling point, where the outer marking point is larger than the inner marking point.

6. A curtain wall thermal cycle detection system based on machine vision, characterized in that: The system comprises: The sampling point setting module is used to set multiple sampling points at different parts of the curtain wall specimen, and marking points are set on both the inner and outer sides of the glass of each sampling point; A sampling point feature acquisition module is used to collect images of the inner side of the curtain wall specimen glass at different thermal cycle test times to obtain multiple curtain wall specimen images, perform feature extraction on the curtain wall specimen images, and obtain multiple sampling point information, wherein the sampling point information includes an inner marking point position and an outer marking point position; The deformation prediction module is used to sequentially input the sampling point information, curtain wall thickness and curtain wall refractive index extracted from different curtain wall specimen images into a preset deformation prediction model, and correspondingly output a plurality of sampling point deformation data sets at different thermal cycle test times; The deformation interpolation module is used to perform deformation interpolation on the curtain wall specimen image according to the sampling point information and the sampling point deformation data set at different thermal cycle test times, and obtain corresponding deformation data of multiple curtain wall specimens; The sampling point density adjustment module is used to classify the deformation area of ​​the curtain wall specimen according to the deformation data of the curtain wall specimen, and adjust the sampling point density distribution of the next curtain wall thermal cycle detection process according to the deformation area classification result.

7. The curtain wall thermal cycle detection system based on machine vision according to claim 6, characterized in that: The sampling point feature acquisition module includes: A graphics acquisition unit, used to acquire images of the inner side of the glass of the curtain wall specimen at different thermal cycle test times to obtain multiple curtain wall specimen images; Graphics grayscale unit, used to grayscale the curtain wall specimen image; A marking point extraction unit is used to perform marking point detection on the curtain wall specimen image after grayscale processing by using Harris corner point detection algorithm to obtain multiple marking point data; The marker point classification unit is used to classify the marker point data into multiple classification clusters using the K-Means algorithm, where the number of classification clusters is equal to the number of sampling points; The marker point positioning unit is used to locate the marker point data in different classification clusters to obtain the inner marker point positions and the outer marker point positions corresponding to the multiple sampling points.

8. The curtain wall thermal cycle detection system based on machine vision according to claim 6, characterized in that: The sampling point density adjustment module comprises: A deformation area classification unit is used to divide the curtain wall specimen into a normal area and an abnormal area based on a preset abnormal deformation threshold using a convex hull algorithm; The sampling point density adjustment unit is used to maintain the sampling point density distribution in the normal area and increase the sampling point density in the abnormal area in the next curtain wall thermal cycle detection process.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory is used to store computer program codes, wherein the computer program codes include computer instructions, and when the processor executes the computer instructions, the electronic device executes the curtain wall thermal cycle detection method based on machine vision as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor of an electronic device, the processor executes the curtain wall thermal cycle detection method based on machine vision as described in any one of claims 1 to 5.

Images