Elevator performance detection method based on video analysis

Abstract

The invention discloses an elevator performance detection method based on video analysis, and relates to the technical field of elevator detection.The method comprises the steps that elevator polyurethane buffer construction information is obtained, and multi-view-angle sensing equipment is configured; performing an impact test on the buffer, and collecting synchronous test video data; multi-dimensional characteristic values such as deformation, stress strain and dynamic response are extracted based on characteristic engineering; in combination with pre-marked test data, constructing a machine learning judgment model taking the three types of features as input; and inputting the characteristic values to the model to realize automatic and accurate judgment on the quality condition of the buffer, and outputting a detection result containing physical performance and qualification judgment. The technical effects of improving the inspection accuracy, enhancing the inspection precision and improving the inspection efficiency are achieved by constructing a set of complete inspection process from video acquisition, preprocessing, feature value extraction, judgment, result output and report generation.

Description

Technical Field

[0001] The present invention relates to the technical field of elevator detection, and particularly relates to an elevator performance detection method based on video analysis. Background Art

[0002] In the field of elevator inspection, for the inspection of polyurethane buffers, traditional methods mainly rely on physical measurement and performance testing. For example, professional measuring tools are used to detect the dimensional specifications of the buffer to see if they meet the design requirements; by applying pressure to the buffer, parameters such as its deformation amount and recovery time are measured, and whether the performance of the buffer meets the standards is judged according to relevant standards. In terms of video processing, basic video playback and viewing technologies can be used to play back test videos, but they are only used as simple recording means in the inspection process, and there are technical problems such as insufficient inspection accuracy, incomplete data, and strong subjectivity. Summary of the Invention

[0003] The present invention provides an elevator performance detection method based on video analysis to solve the technical problems of insufficient inspection accuracy, incomplete data, and strong subjectivity in the prior art, and to achieve the technical effects of improving inspection accuracy, enhancing inspection precision, and improving inspection efficiency.

[0004] An elevator performance detection method based on video analysis provided by the present invention includes:

[0005] Obtain the structural information of the elevator polyurethane buffer, and configure multi-view perception devices according to the structural information.

[0006] Conduct an impact test on the buffer to be detected, and activate the multi-view perception devices to synchronously collect the test video data of the buffer to be detected.

[0007] Extract multi-dimensional feature values based on feature engineering from the test video data to obtain the deformation feature values, stress-strain feature values, and dynamic response feature values of the buffer to be detected.

[0008] Combine pre-labeled test data to construct a feature value judgment model based on machine learning, where the deformation features, stress-strain features, and dynamic response features are defined as the inputs of the feature value judgment model.

[0009] Input the deformation feature values, the stress-strain feature values, and the dynamic response feature values into the feature value judgment model to obtain a performance detection result, where the performance detection result includes physical performance and a qualified judgment result.

[0010] In a feasible implementation manner, configuring the multi-view perception devices according to the structural information includes:

[0011] Extract the external dimension information of the elevator polyurethane buffer based on the above-mentioned structure information.

[0012] Based on the external dimension information, deploy high-definition image acquisition devices in the front, side, and upper side of the buffer respectively.

[0013] In a feasible implementation, conduct an impact test on the buffer to be detected, and activate the multi-view perception device to synchronously collect the test video data of the buffer to be detected, including:

[0014] When the impact test starts, based on the preset acquisition parameters, synchronously obtain the entire process video of the test of the buffer to be detected through multiple high-definition image acquisition devices in the multi-view image acquisition device.

[0015] Use a filtering algorithm to perform denoising processing and image enhancement processing on the collected entire process video of the test, and output the processing result as the test video data.

[0016] In a feasible implementation, after collecting the test video data of the buffer to be detected, it further includes: performing frame rate unification processing on the test video data from different high-definition image acquisition devices.

[0017] In a feasible implementation, perform multi-dimensional eigenvalue extraction based on feature engineering on the test video data to obtain the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected, including:

[0018] Use image recognition and tracking algorithms to mark the key structure points of the buffer, extract the position information of the key structure points of the buffer at different times, and calculate the deformation eigenvalue, where the deformation eigenvalue at least includes the maximum deformation amount, the deformation rate of change, and the recovery time.

[0019] Based on the test video data, combine the material mechanics theoretical model and the finite element analysis method to inversely estimate the stress distribution and strain response eigenvalue during the test process.

[0020] Based on the position information, obtain the vibration frequency and amplitude through the time-frequency domain transformation method to form the dynamic response eigenvalue.

[0021] In a feasible implementation, combine the pre-labeled test data to construct a feature value judgment model based on machine learning, including:

[0022] Collect the test video data of multiple groups of sample buffers in different states, combine the corresponding measured physical properties and evaluation criteria, and establish a sample database by combining feature engineering.

[0023] Based on the sample database, the eigenvalue judgment model is constructed and supervised for training by combining machine learning algorithms. Among them, the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue in the sample database are defined as the sample inputs of the model, and the corresponding measured physical properties and qualified judgment results are defined as the model outputs.

[0024] In a feasible implementation, after obtaining the performance detection results, it includes:

[0025] Generate a visualization interface for the detection results, display the qualified judgment results in the performance detection results in a graphical manner, and distinguish the qualified and unqualified states by colors.

[0026] Combine the deformation eigenvalue, the stress-strain eigenvalue, the dynamic response eigenvalue, the physical properties in the performance detection results, and the corresponding evaluation criteria to construct an eigenvalue data report and synchronize it to the visualization interface.

[0027] In a feasible implementation, after obtaining the performance detection results, it further includes: automatically generating a detection report based on the performance detection results, where the detection report at least includes basic information of the buffer, test condition parameters, video acquisition time and location, the performance detection results, and improvement suggestions.

[0028] The present invention discloses an elevator performance detection method based on video analysis, including: obtaining the structural information of the elevator polyurethane buffer and configuring multi-view perception devices according to the structural information; conducting an impact test on the buffer to be detected, activating the multi-view perception devices to synchronously collect the test video data of the buffer to be detected; extracting multi-dimensional eigenvalues based on feature engineering from the test video data to obtain the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected; combining pre-labeled test data to construct an eigenvalue judgment model based on machine learning, where the deformation features, stress-strain features, and dynamic response features are defined as the inputs of the eigenvalue judgment model; inputting the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue into the eigenvalue judgment model to obtain the performance detection results, where the performance detection results include physical properties and qualified judgment results. The elevator performance detection method based on video analysis disclosed by the present invention solves the technical problems of insufficient inspection accuracy, incomplete data, and strong subjectivity, and achieves the technical effects of improving inspection accuracy, enhancing inspection precision, and improving inspection efficiency. Description of the Drawings

[0029] Figure 1 It is a schematic flowchart of an elevator performance detection method based on video analysis according to the present invention.

[0030] Figure 2Schematic diagram of the process for extracting multi-dimensional eigenvalue based on feature engineering in a method for detecting elevator performance based on video analysis according to the present invention. Detailed implementation mode

[0031] The following will combine the specification drawings and specific implementation modes to elaborate on the above technical solutions in detail to better understand the above technical solutions. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments of the present invention. It should be understood that the present invention is not limited to the example embodiments for explaining the present invention only. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention. In addition, it should be noted that for the convenience of description, only the parts related to the present invention are shown in the drawings rather than all.

[0032] Embodiment Figure 1 Schematic diagram of the process for a method for detecting elevator performance based on video analysis according to the present invention. Among them, the method for detecting elevator performance based on video analysis includes:

[0033] S100: Obtain the structural information of the elevator polyurethane buffer, and configure multi-view perception devices according to the structural information.

[0034] Specifically, the structural information refers to the physical structure and geometric parameters of the elevator polyurethane buffer, including external dimensions (such as length, width, height), positions of key components (such as vertices, bottom edge points), material properties (such as elastic modulus), etc. This structural information is the basis for subsequent video acquisition and feature extraction.

[0035] Specifically, the multi-view perception devices include multiple high-definition image acquisition devices (such as high-definition cameras), which are devices for taking pictures of the buffer from different angles. Exemplarily, they usually include cameras in the front, side, and side upper directions, so as to comprehensively record the state changes of the buffer during the test.

[0036] Through this step, reliable basic data is provided for subsequent eigenvalue extraction and judgment model construction, thus solving the problems of insufficient inspection accuracy, incomplete data, and strong subjectivity in the prior art.

[0037] In some embodiments, configuring the multi-view perception devices according to the structural information includes:

[0038] Based on the structural information, extract the external dimension information of the elevator polyurethane buffer; based on the external dimension information, deploy high-definition image acquisition devices in the front, side, and side upper directions of the buffer respectively.

[0039] Specifically, the external dimension information refers to the geometric dimension parameters of the buffer, such as the length, width, and height of the buffer, as well as the relative positions of key components.

[0040] Specifically, to configure the multi-view perception device according to the structural information, first, obtain the external dimension information of the buffer through a CAD model, design drawings, or actual measurement tools (such as a laser scanner). For example, the length of the buffer is 20 cm, the width is 15 cm, and the height is 10 cm. Then, based on the external dimension information, deploy high-definition cameras in the front, side, and side-upper positions of the buffer respectively. For example: deploy a camera in the front, about 50 cm away from the buffer surface, to ensure that the frontal deformation of the buffer can be clearly captured; deploy a camera on the side, about 30 cm away from the side of the buffer, to ensure that the lateral deformation and displacement of the buffer can be captured; deploy a camera in the side-upper position, about 40 cm away from the top of the buffer, to ensure that the top deformation and overall movement of the buffer can be recorded.

[0041] S200: Conduct an impact test on the buffer to be detected, and activate the multi-view perception device to synchronously collect the test video data of the buffer to be detected.

[0042] Specifically, the impact test simulates the working conditions when the elevator car or counterweight device impacts the bottom pit by applying an impact force to the buffer to test the performance of the buffer. This impact test involves a device for applying the impact force (such as an impact table or heavy object) and a system for controlling impact parameters (such as impact speed, impact mass). The test video data refers to the video data of the state changes such as deformation and recovery of the buffer collected by the multi-view perception device during the impact test, which is used for subsequent eigenvalue extraction and analysis.

[0043] In some embodiments, conducting an impact test on the buffer to be detected and activating the multi-view perception device to synchronously collect the test video data of the buffer to be detected includes:

[0044] When the impact test starts, based on the preset acquisition parameters, synchronously obtain the full-process video of the test of the buffer to be detected through multiple high-definition image acquisition devices in the multi-view image acquisition device; perform denoising processing and image enhancement processing on the collected full-process video of the test, and output the processing result as the test video data.

[0045] Specifically, an impact test is conducted on the buffer to be detected, and the multi-view perception device is activated to synchronously collect test video data. First, an impact test device is set up, including an impact object (such as a heavy object) and an impact platform. The mass and speed of the impact object can be adjusted according to the test standard. For example, the mass of the impact object is 500 kg and the impact speed is 1 m / s. Then, the impact test device is started, so that the impact object impacts the buffer at the set speed. At the same time as the impact starts, the multi-view perception device is activated through a trigger signal to ensure that all cameras start collecting test video data synchronously. Among them, each camera records the state changes such as deformation and recovery of the buffer during the impact process with preset acquisition parameters (such as a frame rate of 60 frames per second and a resolution of 1080p).

[0046] Furthermore, after obtaining the original test video data, a filtering algorithm (such as Gaussian filtering) is used to remove the noise generated by factors such as equipment noise and light interference, and the parameters such as the brightness, contrast, and color saturation of the video are adjusted for image enhancement processing to highlight the contour and detail features of the buffer.

[0047] Through the above process, the deformation, recovery, etc. of the buffer during the impact test are recorded from different angles, ensuring the comprehensiveness and accuracy of the data. Among them, the standardized setting of the impact test and the synchronous acquisition of the multi-view perception device provide high-quality video data for subsequent eigenvalue extraction, thereby improving the accuracy and reliability of the inspection. Through the filtering algorithm and image enhancement processing, the noise in the video data is removed, the clarity and stability of the video are improved, and a more reliable data basis is provided for subsequent eigenvalue extraction.

[0048] In some embodiments, after collecting the test video data of the buffer to be detected, it further includes: performing frame rate unification processing on the test video data from different high-definition image acquisition devices.

[0049] Specifically, the video frame rate collected by each high-definition image acquisition device is detected, and the video data collected by different cameras is adjusted to the same frame rate to ensure the consistency and accuracy of video analysis. Optionally, the frame rate unification method includes interpolation (increasing the frame rate) and frame dropping (decreasing the frame rate).

[0050] Exemplarily, if the frame rate of the front camera is 60 frames per second, the frame rate of the side camera is 30 frames per second, and the frame rate of the side upper camera is 45 frames per second; the video with a lower frame rate is interpolated to increase its frame rate to the target frame rate (such as 45 frames per second). For example, the 30-frame-per-second video of the side camera is interpolated to generate a 45-frame-per-second video; the video with a higher frame rate is frame-dropped to decrease its frame rate to the target frame rate (such as 45 frames per second). For example, the 60-frame-per-second video of the front camera is frame-dropped to generate a 45-frame-per-second video.

[0051] S300: Extract multi-dimensional eigenvalue based on feature engineering from the test video data to obtain the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected.

[0052] Specifically, feature engineering is a process of extracting useful information from raw data and converting it into eigenvalues, including feature selection, feature extraction, and feature construction. Among them, the deformation eigenvalue refers to the deformation characteristics of the buffer when it is impacted, including the maximum deformation amount, the change rate of the deformation amount, the recovery time, etc. The stress-strain eigenvalue refers to the stress and strain distribution characteristics generated inside the buffer when it is impacted, including the maximum stress, the strain concentration area, etc. The dynamic response eigenvalue refers to the vibration characteristics of the buffer after being impacted, including the vibration frequency, the amplitude, etc. By extracting the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue, the subtle changes and complex characteristics of the buffer during the test can be captured, which helps to improve the accuracy and reliability of the inspection.

[0053] In some embodiments, as Figure 2 shown, extracting multi-dimensional eigenvalue based on feature engineering from the test video data to obtain the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected includes:

[0054] Adopt an image recognition and tracking algorithm to mark the key structure points of the buffer, and extract the position information of the key structure points of the buffer at different times, and calculate the deformation eigenvalue. Among them, the deformation eigenvalue at least includes the maximum deformation amount, the change rate of the deformation amount, and the recovery time; based on the test video data, combine the material mechanics theoretical model and the finite element analysis method to inversely estimate the stress distribution and strain response eigenvalue during the test process; based on the position information, obtain the vibration frequency and amplitude through the time-frequency domain transformation method to form the dynamic response eigenvalue.

[0055] Specifically, to extract the deformation eigenvalue, first, use computer vision technology to identify and track the algorithm of specific objects or points in the video, such as the algorithm based on feature point detection (such as SIFT, SURF) or the object detection algorithm of deep learning (such as YOLO), to mark the key structure points of the buffer (such as the vertex, the bottom edge point); then, extract the position information of these key points at different times, and calculate the maximum deformation amount, the change rate of the deformation amount, and the recovery time of the buffer. For example, by calculating the displacement difference of the buffer vertex before and after the impact, the maximum deformation amount can be obtained; by analyzing the change curve of the vertex displacement with time, the deformation rate and the recovery time can be calculated. [[ID=*17]]

[0056] Specifically, to extract the stress and strain characteristic values, first, based on the deformation information of the buffer in the video, combined with the material mechanics theoretical model and the finite element analysis method, the stress distribution and strain response characteristic values during the test are reversely estimated. Among them, the material mechanics theoretical model refers to a theoretical model that describes the mechanical behavior of a material under stress, such as an elastic mechanics model, a plastic mechanics model, etc.; the material mechanics theoretical model is combined with parameters such as the deformation of the buffer and the elastic modulus of the material to calculate the stress magnitude and strain degree of the buffer at different parts.

[0057] Specifically, the dynamic response characteristic values are extracted, including image analysis of the tiny vibrations on the buffer surface, and the use of mathematical methods such as Fourier transform and wavelet transform to convert the vibration signal from the time domain to the frequency domain, thereby extracting the main components and outputting them as vibration frequency and amplitude.

[0058] Through the above process, multi-dimensional feature value extraction based on feature engineering is performed on the experimental video data, which can comprehensively and accurately reflect the performance of the buffer in multiple dimensions, solve the problem of single detection dimension in the existing technology, and provide rich input data for subsequent machine learning judgment models.

[0059] S400: In combination with the pre-labeled test data, a feature value judgment model based on machine learning is constructed, wherein deformation characteristics, stress-strain characteristics, and dynamic response characteristics are defined as inputs of the feature value judgment model.

[0060] Specifically, pre-labeled test data refers to test data that has been pre-labeled with buffer performance results (e.g., pass or fail), including eigenvalues (e.g., deformation characteristics, stress-strain characteristics, dynamic response characteristics) and corresponding physical performance results. This data serves as training data for the eigenvalue judgment model. The eigenvalue judgment model takes deformation characteristics, stress-strain characteristics, and dynamic response characteristics as input and outputs the corresponding buffer performance results.

[0061] In some embodiments, a feature value judgment model based on machine learning is constructed in combination with pre-labeled test data, including:

[0062] Collect multiple groups of sample buffer test video data in different states, combine the corresponding measured physical properties and evaluation standards, and combine feature engineering to establish a sample database; based on the sample database, combine the machine learning algorithm to construct and supervise the training of the eigenvalue judgment model, wherein the deformation eigenvalue, stress-strain eigenvalue and dynamic response eigenvalue in the sample database are defined as the sample input of the model, and the corresponding measured physical properties and qualified judgment results are defined as the model output.

[0063] Specifically, the sample database is a database that stores pre-labeled test data. Each set of sample data includes the characteristic values of the buffer (such as deformation characteristics, stress-strain characteristics, dynamic response characteristics) and the corresponding physical properties and qualified judgment results.

[0064] Specifically, a characteristic value judgment model based on machine learning is constructed in combination with the pre-labeled test data. First, collect test video data of multiple sets of sample buffers in different states, including qualified and unqualified buffers. Label each set of sample data and record its physical property results (such as maximum deformation, stress distribution, vibration frequency, comprehensive buffer performance retention rate, etc.) and qualified judgment results (qualified or unqualified); extract characteristic values from each set of sample data, including deformation characteristic values (such as maximum deformation, deformation rate of change, recovery time), stress-strain characteristic values (such as maximum stress, strain concentration area), and dynamic response characteristic values (such as vibration frequency, amplitude), and integrate the extracted characteristic values and the corresponding labeled results into a sample database, where the characteristic values are used as inputs and the labeled results are used as outputs.

[0065] Then, select a suitable machine learning algorithm (such as support vector machine, neural network), supervise and train the model based on the sample database, use an independent validation data set to verify the trained model, evaluate the accuracy and reliability of the model, and adjust the model parameters (such as learning rate, network structure) according to the verification results to optimize the model performance.

[0066] Through the above process, a characteristic value judgment model based on machine learning is constructed in combination with the pre-labeled test data. Among them, the machine learning model makes automatic judgments based on the characteristic values, reduces the influence of human subjective factors, and ensures the objectivity and consistency of the judgment results. By learning a large amount of sample data, the model can capture the complex relationship between the characteristic values and the buffer performance, improving the accuracy and reliability of the inspection. In summary, the judgment model based on machine learning provides a scientific and reliable method for the quality assessment of buffers, helps to standardize the elevator inspection work, and improves the overall safety level of the elevator industry.

[0067] Through this step, the present invention provides an efficient and accurate automated method for the performance detection of buffers, solving the problems of insufficient inspection accuracy, incomplete data, and strong subjectivity in the prior art.

[0068] S500: Input the deformation characteristic value, the stress-strain characteristic value, and the dynamic response characteristic value into the characteristic value judgment model to obtain a performance detection result, where the performance detection result includes physical properties and a qualified judgment result.

[0069] Specifically, the performance detection result refers to the buffer performance evaluation result output by the model, including physical performance indicators (such as maximum deformation amount, stress distribution, vibration frequency, comprehensive buffer performance retention rate, etc.) and pass / fail determination results (pass or fail).

[0070] Specifically, the process of inputting the eigenvalue into the eigenvalue judgment model and obtaining the performance detection result is as follows: First, the extracted deformation eigenvalues (such as maximum deformation amount, deformation rate of change, recovery time), stress-strain eigenvalues (such as maximum stress, strain concentration area), and dynamic response eigenvalues (such as vibration frequency, amplitude) are used as inputs and passed to the eigenvalue judgment model. Based on the patterns and relationships learned during the training process, the eigenvalue judgment model analyzes and calculates the input eigenvalues and outputs the performance detection result, including the physical performance indicators of the buffer and the pass / fail determination result. For example, the model may output that the maximum deformation amount of the buffer is 5 mm, the stress value in the stress concentration area is 120 MPa, the vibration frequency is 10 Hz, and it is determined that the buffer is qualified.

[0071] Through the eigenvalue judgment model of the above process, the input eigenvalues are automatically analyzed, reducing the workload and error of manual judgment, improving the detection efficiency. The model output not only includes the pass / fail determination result but also detailed physical performance indicators, providing comprehensive information for the performance evaluation of the buffer.

[0072] In some embodiments, after obtaining the performance detection result, it includes:

[0073] Generate a visualization interface for the detection result, display the pass / fail determination result in the performance detection result in a graphical manner, and distinguish the pass and fail states by color; combine the deformation eigenvalues, the stress-strain eigenvalues, the dynamic response eigenvalues, the physical performance in the performance detection result with the corresponding judgment criteria to construct an eigenvalue data report and synchronize it to the visualization interface.

[0074] Specifically, the visualization interface for the detection result refers to a user interface that displays the performance detection result in a graphical manner, including elements such as charts, color coding, text descriptions, etc., to intuitively present the detection result.

[0075] Specifically, the pass / fail determination result in the performance detection result is displayed in a graphical manner, and color coding is used to distinguish the pass and fail states. For example, the pass state is marked in green, the fail state is marked in red, and the warning state close to fail is marked in yellow. At the same time, the numerical values and trend charts of key eigenvalues (such as maximum deformation amount, stress concentration area, vibration frequency, etc.) are displayed on the interface.

[0076] Specifically, by combining the deformation eigenvalue, stress-strain eigenvalue, dynamic response eigenvalue, and physical property results, a detailed eigenvalue data report is constructed through the correlation relationships among the above data. The content of the report includes the basic information of the buffer (such as model, specification, manufacturer), test conditions (such as impact speed, environmental temperature), eigenvalue data (such as maximum deformation, stress distribution, vibration frequency), and pass / fail judgment results. Then, the report data is synchronized to the visualization interface to ensure that the information displayed on the interface is consistent with the content of the report.

[0077] Through the above process, with color coding and graphical display, inspectors can quickly identify the performance status of the buffer and help them deeply understand the performance of the buffer, providing a scientific basis for decision-making.

[0078] In some embodiments, after obtaining the performance detection results, it further includes: automatically generating a detection report based on the performance detection results, where the detection report at least includes the basic information of the buffer, test condition parameters, video acquisition time and location, the performance detection results, and improvement suggestions.

[0079] Optionally, the basic information of the buffer includes identification information such as the model, number, manufacturer, production batch, and factory date of the buffer, which is used to uniquely identify the object under test; the test condition parameters include the load level, impact mode, test speed, environmental temperature and humidity used in the test, which are used to clarify the basic boundary conditions of the detection process; the video acquisition time and location record the timestamp and geographical location information of the video data during the detection process, which is used to ensure the traceability and compliance of the detection process; the performance detection results may include key performance indicators such as the impact force response curve, deformation amount, energy absorption rate, recovery rate, and performance retention rate; the improvement suggestions are targeted technical improvement suggestions generated based on the evaluation of the buffer performance from the detection results, such as material improvement, replacement of the buffer, inspection of the elevator shaft environment, etc.

[0080] The above steps can automatically summarize and structure multi-source data during the detection process according to a preset template to generate a report, which can reduce manual input errors and improve the report compilation efficiency.

[0081] In summary, the elevator performance detection method based on video analysis provided by the present invention has the following technical effects:

[0082] By obtaining the structural information of the elevator polyurethane buffer and configuring multi - perspective perception devices according to the structural information; conducting an impact test on the buffer to be detected, activating the multi - perspective perception devices to synchronously collect the test video data of the buffer to be detected; extracting multi - dimensional eigenvalue from the test video data based on feature engineering to obtain the deformation eigenvalue, stress - strain eigenvalue and dynamic response eigenvalue of the buffer to be detected; combining with pre - labeled test data to construct a machine - learning - based eigenvalue judgment model, where the deformation feature, stress - strain feature and dynamic response feature are defined as the inputs of the eigenvalue judgment model; inputting the deformation eigenvalue, stress - strain eigenvalue and dynamic response eigenvalue into the eigenvalue judgment model to obtain the performance detection result, where the performance detection result includes physical performance and qualified judgment result, so as to achieve the technical effects of improving the inspection accuracy, enhancing the inspection precision and improving the inspection efficiency.

[0083] It should be understood that the disclosed embodiments of the present invention and the above descriptions enable those skilled in the art to implement the present invention using the present invention. At the same time, the present invention is not limited to the above - mentioned part of the embodiments. It should be understood that those of ordinary skill in the art can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. An elevator performance detection method based on video analysis, characterized in that, The method includes: Obtaining the structural information of the elevator polyurethane buffer, and configuring a multi - perspective perception device according to the structural information; Conducting an impact test on the buffer to be detected, and activating the multi - perspective perception device to synchronously collect the test video data of the buffer to be detected; Performing multi - dimensional eigenvalue extraction based on feature engineering on the test video data to obtain the deformation eigenvalue, stress - strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected; Combining pre - labeled test data to construct a machine - learning - based eigenvalue judgment model, where the deformation feature, stress - strain feature, and dynamic response feature are defined as the inputs of the eigenvalue judgment model; Inputting the deformation eigenvalue, the stress - strain eigenvalue, and the dynamic response eigenvalue into the eigenvalue judgment model to obtain a performance detection result, where the performance detection result includes physical performance and a pass - fail judgment result.

2. The elevator performance detection method based on video analysis according to claim 1, wherein Configuring a multi - perspective perception device according to the structural information, including: Based on the structural information, extracting the external dimension information of the elevator polyurethane buffer; Based on the external dimension information, arranging high - definition image acquisition devices in the front, side, and side - upper positions of the buffer respectively.

3. The elevator performance detection method based on video analysis according to claim 2, wherein, Conducting an impact test on the buffer to be detected, and activating the multi - perspective perception device to synchronously collect the test video data of the buffer to be detected, including: When the impact test starts, based on preset acquisition parameters, synchronously obtaining the entire - process test video of the buffer to be detected through multiple high - definition image acquisition devices in the multi - perspective image acquisition device; Using a filtering algorithm to perform denoising processing and image enhancement processing on the collected entire - process test video, and outputting the processing result as the test video data.

4. The elevator performance detection method based on video analysis according to claim 3, wherein After collecting the test video data of the buffer to be detected, it further includes: performing frame rate unification processing on the test video data from different high - definition image acquisition devices.

5. The elevator performance detection method based on video analysis according to claim 4, characterized in that, Performing multi - dimensional eigenvalue extraction based on feature engineering on the test video data to obtain the deformation eigenvalue, stress - strain eigenvalue, and dynamic response eigenvalue of the buffer to be detected, including: Using an image recognition and tracking algorithm to mark the key structural points of the buffer, and extracting the position information of the key structural points of the buffer at different times, and calculating the deformation eigenvalue, where the deformation eigenvalue at least includes the maximum deformation amount, the deformation rate of change, and the recovery time; Based on the test video data, combining the material mechanics theoretical model and the finite - element analysis method to inversely estimate the stress distribution and strain response eigenvalue during the test process; Based on the position information, obtaining the vibration frequency and amplitude through a time - frequency domain transformation method to form the dynamic response eigenvalue.

6. The elevator performance detection method based on video analysis according to claim 5, characterized in that Combining pre - labeled test data to construct a machine - learning - based eigenvalue judgment model, including: Collecting test video data of multiple groups of sample buffers in different states, combining the corresponding measured physical performance and evaluation criteria, and establishing a sample database by combining feature engineering; Based on the sample database, the eigenvalue judgment model is constructed and supervised for training by combining machine learning algorithms. Among them, the deformation eigenvalue, stress-strain eigenvalue, and dynamic response eigenvalue in the sample database are defined as the sample inputs of the model, and the corresponding measured physical properties and qualified judgment results are defined as the model outputs.

7. The elevator performance detection method based on video analysis according to claim 1, wherein Obtain the performance detection results. After that, it includes: Generate a visualization interface for the detection results, display the qualified judgment results in the performance detection results in a graphical manner, and distinguish the qualified and unqualified states by colors; Combine the deformation eigenvalue, the stress-strain eigenvalue, the dynamic response eigenvalue, the physical properties in the performance detection results, and the corresponding judgment criteria to construct an eigenvalue data report and synchronize it to the visualization interface.

8. The elevator performance detection method based on video analysis according to claim 1, characterized in that Obtain the performance detection results. After that, it also includes: automatically generating a detection report based on the performance detection results, where the detection report at least includes the basic information of the buffer, test condition parameters, video acquisition time and location, the performance detection results, and improvement suggestions.

Images