Traction type elevator steel wire rope damage detection method and device based on machine vision

Through a machine vision-based method, the convolutional neural network and multi-scale global feature extraction module combined with attention mechanism is used to realize the automation and real-time detection of traction elevator wire ropes, solving the low detection efficiency and strong subjectivity in the existing technology. The problem of this problem improves detection efficiency and safety, and provides a scientific basis for preventive maintenance.

CN118968483AInactive Publication Date: 2024-11-15贵州装备制造职业学院
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202411098443.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the prior art, the damage detection of traction elevator wire ropes relies on manual inspection, and there are problems such as low efficiency, great influence from ambient light, difficulty in detecting small damage, strong subjectivity, and inability to monitor in real time, and automation and real-time detection cannot be achieved.

Method used

Using a machine vision-based method, a damage detection model is constructed through a convolutional neural network, an image of the elevator wire rope is obtained by using the camera device, image preprocessing is performed and the model is input for detection, and damage detection results are output. A lightweight MobileNetV3 network and multi-scale global The feature extraction module combines attention mechanism to achieve automation and real-time detection.

Benefits of technology

Achieve efficient and accurate wire rope damage detection, overcome the limitations of traditional methods, improve detection efficiency and safety, can automatically and in real time detect wire rope damage, reduce maintenance costs, and provide data records and trend analysis to help predict. The service life of the wire rope.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118968483A_ABST
    Figure CN118968483A_ABST
Patent Text Reader

Abstract

The invention discloses a traction type elevator steel wire rope damage detection method and device based on machine vision, and belongs to the technical field of elevator safety. The method comprises the following steps: constructing a damage detection model based on a convolutional neural network, and training the damage detection model by using an elevator steel wire rope image sample set marked with damage; acquiring an elevator steel wire rope image through a camera device; carrying out image preprocessing on the acquired elevator steel wire rope image; and inputting the preprocessed elevator steel wire rope image into the trained damage detection model, carrying out elevator steel wire rope damage detection, and outputting an elevator steel wire rope damage detection result. By means of the method, the damage of the steel wire rope of the elevator can be efficiently and accurately detected, the limitation of a traditional detection method can be overcome, automatic and real-time detection can be achieved, and the working efficiency and safety of traction type elevator steel wire rope damage detection are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of elevator safety, and in particular to a method and device for detecting damage to a traction elevator wire rope based on machine vision. Background Art

[0002] An elevator is a vertical transportation device, mostly used to transport people or goods up and down in high-rise buildings; it usually consists of a car, guide rails, motors, control panels and safety systems. As an indispensable vertical transportation tool in modern buildings, the safety of traction elevators is directly related to the safety of passengers in daily use. Among them, the wire rope is one of the most critical components of a traction elevator. It not only bears the weight of the elevator car, but is also responsible for the lifting and lowering movement of the elevator. Therefore, the integrity of the wire rope is crucial to the overall safe operation of the elevator.

[0003] At present, the damage detection of traction elevator wire rope mainly adopts the method of regular manual inspection, which relies on the visual observation of professionals to determine whether there is obvious damage on the surface of the wire rope. However, this method has obvious limitations, such as low inspection efficiency, great influence by ambient light, difficulty in finding minor damage, and subjectivity, which is highly dependent on the operator's experience and attention; manual inspection also has problems such as high labor intensity and high risk; in addition, it is impossible to monitor the status of the wire rope in real time, and it is impossible to perform continuous inspection during elevator operation.

[0004] Therefore, how to automatically, efficiently, and accurately detect damage to traction elevator wire ropes in real time is a technical problem that urgently needs to be solved. Summary of the invention

[0005] In view of this, the present invention provides a method and device for detecting damage to a traction elevator wire rope based on machine vision, which at least solves some of the above-mentioned technical problems. The method and device can efficiently and accurately detect damage to the elevator wire rope, which can not only overcome the limitations of traditional detection methods, but also realize automated and real-time detection, thereby improving the work efficiency and safety of traction elevator wire rope damage detection.

[0006] To achieve the above object, the technical solution adopted by the present invention is:

[0007] In a first aspect, an embodiment of the present invention provides a method for detecting damage to a traction elevator wire rope based on machine vision, the method comprising the following steps:

[0008] S1. constructing a damage detection model based on a convolutional neural network, and training the damage detection model using a sample set of elevator wire rope images with marked damage;

[0009] S2, obtaining an image of the elevator wire rope through a camera device;

[0010] S3, performing image preprocessing on the acquired elevator wire rope image;

[0011] S4. Input the preprocessed elevator wire rope image into the trained damage detection model, perform elevator wire rope damage detection, and output the elevator wire rope damage detection result.

[0012] In an optional embodiment, in step S1, the damage detection model constructed includes: a backbone network, a multi-scale global feature extraction module, a different-scale attention enhancement module and a classification layer connected in sequence; wherein:

[0013] The backbone network is a lightweight MobileNetV3 network, which is used to extract features of elevator wire rope images;

[0014] The multi-scale global feature extraction module includes multiple branches for multi-scale information extraction, each branch uses dilated convolution to extract features at multiple scales, and extracts global information in a wider range by expanding the receptive field;

[0015] The different-scale attention enhancement module is used to generate attention of different-scale information on the output result of the multi-scale global feature extraction module, and combine the input feature map with the attention score to obtain feature maps with different-scale attention enhancement;

[0016] The classification layer is used to perform a dimensionality increase operation on the feature maps output by the attention enhancement modules of different scales, and then use the average pooling kernel to perform a global average pooling operation on the feature maps to compress the spatial information of each channel into a single numerical value; then, the pooled feature vector is passed to the fully connected layer to map the feature vector to the dimension of the number of damage severity categories; finally, the softmax activation function is used to convert the output into an elevator wire rope damage detection result.

[0017] In an optional embodiment, the lightweight MobileNetV3 network includes: a convolution module and multiple stacked bottleneck modules, wherein: the convolution module is used for downsampling; each bottleneck module contains point-by-point convolution for dimensionality increase, depth-wise separable convolution for feature extraction, and point-by-point convolution for dimensionality reduction.

[0018] In an optional embodiment, each branch adopts a dilation rate of r z The dilated convolution is used to extract features at multiple scales, and the global information is extracted by expanding the receptive field. The calculation formula of the effective receptive field is:

[0019] F = [(m-1)r z +1] 2 ;

[0020] Among them, F represents the effective receptive field, m represents the size of the dilated convolution kernel, and r z represents the expansion rate, and z represents the branch number.

[0021] In an optional embodiment, the damage detection model is trained using a Focal Loss loss function to measure the difference between the damage severity category predicted by the model and the actual severity category; the calculation formula of the Focal Loss loss function is:

[0022] ;

[0023] Among them, y n is the true damage severity category number corresponding to sample n, is the model for y n The predicted probability of is the importance weighting coefficient of samples of different injury severity categories; γ is the adjustment factor.

[0024] In an optional implementation, in step S2, the camera device is an infrared camera or an industrial CCD camera integrated with a lighting device.

[0025] In an optional implementation, in step S3, the image preprocessing includes: grayscale conversion, image scaling and histogram equalization.

[0026] In an optional implementation, in the histogram equalization process, the step change of pixels is detected point by point in a weighted manner to perform image enhancement, and the process includes:

[0027] The gray value of an image pixel is , , Respectively represent the total number of rows and columns of the image matrix, traverse all the pixels of the image, and correct the gray value of each pixel; the gray value of the image pixel after the image edge contour is enhanced is obtained by linear weighting; the formula is:

[0028]

[0029] In the formula, Represents the grayscale value of the enhanced image pixel; Represents the gray value of the corrected image pixel; is the set weight, and .

[0030] In a second aspect, an embodiment of the present invention further provides a traction elevator wire rope damage detection device based on machine vision, comprising: an image acquisition device, a memory and a processor, wherein:

[0031] The image acquisition device is used to collect and acquire the image of the elevator wire rope;

[0032] The memory stores computer instructions;

[0033] The processor is connected to the image acquisition device and the memory data. The processor executes the computer instructions to execute a machine vision-based traction elevator wire rope damage detection method to perform traction elevator wire rope damage detection.

[0034] In a third aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a processor, the computer instructions implement the above-mentioned method for detecting damage to traction elevator wire ropes based on machine vision.

[0035] Compared with the prior art, the present invention has at least the following beneficial effects:

[0036] 1. The present invention provides a method for detecting damage to a traction elevator wire rope based on machine vision, which can efficiently and accurately detect damage to the elevator wire rope. It can not only overcome the limitations of traditional manual detection methods, but also realize automatic and real-time detection, thereby improving the work efficiency and safety of traction elevator wire rope damage detection.

[0037] 2. The present invention adopts MobileNetV3 as the backbone network, and constructs a damage detection model based on multi-scale global feature extraction and attention mechanism. The model can automatically use regions of different scales in the elevator wire rope damage image to extract damage features and estimate the degree of damage, thereby improving the generalization ability of the model and can quickly, efficiently and accurately detect elevator wire rope damage.

[0038] 3. The present invention provides a method for detecting damage to traction elevator wire ropes based on machine vision. This method is a non-contact detection method that does not require direct contact with the wire rope, thus avoiding secondary damage that may be caused by physical contact. In addition, it is convenient for data recording and trend analysis: long-term data accumulation can help predict the service life of the wire rope and provide a scientific basis for preventive maintenance.

[0039] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.

[0040] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0043] Figure 1 A schematic flow chart of a method for detecting damage to a traction elevator wire rope based on machine vision is provided in an embodiment of the present invention.

[0044] Figure 2 A schematic diagram of the damage detection model structure provided in an embodiment of the present invention.

[0045] Figure 3 A schematic diagram of the structure of a multi-scale global feature extraction module provided in an embodiment of the present invention.

[0046] Figure 4 A schematic diagram of the structure of attention enhancement modules at different scales provided by an embodiment of the present invention.

[0047] Figure 5 A schematic structural diagram of a machine vision-based traction elevator wire rope damage detection device provided in an embodiment of the present invention.

[0048] Among them, 1-image acquisition device; 2-memory; 3-processor. DETAILED DESCRIPTION

[0049] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions 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.

[0050] In the description of the present invention, it should be noted that: in some processes described in the specification and drawings of this application, multiple operations appearing in a specific order are included, but it should be clearly understood that these operations may not be performed in the order in which they appear in this document or may be performed in parallel. In addition, various serial numbers are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0051] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. 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.

[0052] See also Figure 1 As shown, the present invention provides a traction elevator wire rope damage detection method based on machine vision, which is used to automatically detect the damage of the traction elevator wire rope. The method mainly includes:

[0053] S1. Build a damage detection model based on convolutional neural network and train the damage detection model using a sample set of elevator wire rope images with marked damage.

[0054] S2, obtaining an image of the elevator wire rope through a camera device;

[0055] S3, performing image preprocessing on the acquired elevator wire rope image;

[0056] S4. Input the preprocessed elevator wire rope image into the trained damage detection model, perform elevator wire rope damage detection, and output the elevator wire rope damage detection result.

[0057] The specific implementation methods of the method of the present invention are described in detail below:

[0058] 1. Building a damage detection model:

[0059] In the embodiment of the present invention, the lightweight MobileNetV3 is preferably used as the backbone network, and a damage detection model is constructed based on multi-scale global feature extraction and attention mechanism. Figure 2 As shown in the figure, the constructed damage detection model mainly includes: backbone network, multi-scale global feature extraction module, different scale attention enhancement module and classification layer connected in sequence; the above parts are introduced respectively as follows:

[0060] ①. Backbone network:

[0061] In a specific embodiment, a lightweight MobileNetV3 is used as the backbone network. While keeping the model lightweight, MobileNetV3 introduces an attention mechanism, which helps the model focus on more valuable features, thereby improving accuracy; while maintaining high precision, it also greatly reduces the amount of calculation and the number of parameters, thereby improving efficiency; the specific structure of the backbone network is as follows:

[0062] First, a 3×3 ordinary convolution is used for downsampling, and then 15 bottleneck modules are stacked. Each bottleneck module contains point-by-point convolution (1×1 convolution) for dimensionality increase, depth-separable convolution for feature extraction, optional SE attention module, and point-by-point convolution (1×1 convolution) for dimensionality reduction. At the same time, a residual connection structure is introduced to enhance the model's expressiveness and training stability. Among them, the SE attention module is selected to be added to the 4th to 6th and 11th to 15th bottleneck modules to enhance the model's representation ability. The MobileNetV3 network performs preliminary feature extraction of damage on the input elevator wire rope image.

[0063] ②. Multi-scale global feature extraction module:

[0064] like Figure 3 As shown, in the embodiment of the present invention, the feature map extracted by the backbone network is sent to the multi-scale global feature extraction module. The module contains three branches of multi-scale information extraction, each branch uses an expansion rate of r z The 3×3 dilated convolution is used to extract features at multiple scales, and a larger range of global information is extracted by expanding the receptive field. z The calculation formula of the effective receptive field of the dilated convolution kernel is as follows:

[0065] F = [(m-1)r z +1] 2 ;

[0066] Among them, F represents the effective receptive field, m represents the size of the dilated convolution kernel, and r z represents the expansion rate, and z represents the branch number.

[0067] In this module, 1×1 convolution is first used to increase the dimension of the feature map, and then three branches are used for multi-scale information extraction. Each branch uses a 3×3 dilated convolution with a different dilation rate. z = 1 for small local feature extraction; the second branch uses the dilation rate r z =2 is used to extract the damage features of a larger area, so as to obtain the damage features on a 5×5 receptive field; the third branch uses the expansion rate r z =3 to extract damage features in a large area. Dilated convolution helps capture global damage information by expanding the receptive field. The feature maps obtained from the three branches are concatenated using the “concatenate” operation.

[0068] ③. Attention enhancement modules at different scales:

[0069] like Figure 4As shown, in an embodiment of the present invention, the different scale attention enhancement modules generate attention to different scale information on the output results of the multi-scale global feature extraction module. First, for the feature map of size (C, H, W), the global average pooling operation and the global maximum pooling operation are used to compress the two-dimensional features of each channel into a real number, and the global average and global maximum values ​​of each channel are obtained. Then, a one-dimensional convolution kernel of an adaptive size is used to act on the feature map of size (C, 1, 1) to obtain the weight vector of each channel. The calculation formula for the size k of the one-dimensional convolution kernel is as follows:

[0070] ;

[0071] Among them, C represents the number of channels, λ is a scaling factor used to adjust the speed at which the one-dimensional convolution kernel size k changes with the number of channels C, b is a bias term, and the parameters λ=2 and b=1 are set in the model.

[0072] Subsequently, the weight vector is added element by element, and the weight of each channel is normalized using the sigmoid function to obtain the attention scores of feature maps of different scales. Finally, the input feature map of the module is multiplied by the attention score to obtain the feature maps of different scales with attention enhancement.

[0073] ④. Classification layer:

[0074] In an embodiment of the present invention, a 1×1 convolution is first used to perform a dimensionality increase operation on the feature maps output by the attention enhancement modules of different scales, and then a 7×7 average pooling kernel is used to perform a global average pooling operation on the feature maps to compress the spatial information of each channel into a single value. Then, the pooled feature vector is passed to two fully connected layers, and the feature vector is further mapped to the dimension of the number of severity categories. Finally, a softmax activation function is used to convert the output into an elevator wire rope damage detection result (the level of the damage severity category). Furthermore, when the detection result exceeds the threshold, an early warning can be issued to improve the safety of traction elevator operation.

[0075] 2. Model training:

[0076] A training set is constructed, and samples of the training set are input into the proposed damage detection model for training. Table 1 shows the elevator wire rope image sample set information with damage severity category labels used in the embodiment of the present invention.

[0077]

[0078] The data set is divided into a training set and a test set according to a data ratio of 8:2.

[0079] Furthermore, in the embodiment of the present invention, the Focal Loss loss function is used to measure the difference between the elevator wire rope damage severity category predicted by the model and the actual severity category, and the loss value is reduced to the minimum through multiple training iterations, thereby completing the model training. The calculation formula of the Focal Loss loss function is as follows:

[0080] ;

[0081] in, is the true damage severity category number corresponding to sample n, The model is The predicted probability of It is the importance weighting coefficient of samples of different damage severity categories, which can adjust the balance degree of different disease severity categories; is the adjustment factor, and This setting can reduce the loss contribution of easy-to-classify samples and increase the loss contribution of difficult-to-classify samples.

[0082] The damage detection model training cycle of the present invention is 50 epochs, and each epoch adopts small batch training with a batch size of 16. The input image size is 224×224 pixels. In addition, the Adam algorithm is used to optimize the model parameters, and the learning rate is set to 0.0001.

[0083] 3. Image acquisition:

[0084] In an embodiment of the present invention, a camera device is installed at a preset fixed position at multiple angles or a single angle in the elevator shaft to obtain an image of the elevator wire rope in real time; the camera device can be an infrared camera or an industrial CCD camera with an integrated lighting device; appropriate lighting helps to obtain high-quality images; LED lights or other types of light sources can be used to illuminate the wire rope, reduce shadows and reflections, and thus improve image quality; therefore, considering image quality, the present invention preferably uses an industrial CCD camera with an integrated LED lighting device to obtain the image of the elevator wire rope; and the collected images are stored to facilitate trend analysis of damage to the traction elevator wire rope. In specific implementation, one or more cameras are installed to ensure that they can cover the complete path of the elevator wire rope. The camera can be installed at the top, bottom or inside of the elevator shaft, depending on the actual monitoring environment; the camera is further calibrated to ensure the geometric accuracy of the image, including distortion correction and determination of internal and external parameter matrices; the camera acquisition frequency is set to ensure that continuous images of the wire rope can be captured; the image data is transmitted to the computing device by wire or wireless means; at the same time, encoders, laser rangefinders or other sensors are used to track the position and movement of the elevator car; the car position corresponding to each image is recorded, and the corresponding relationship between the image frame and the car position is established; finally, the absolute position of the damage on the wire rope can be calculated based on the damage features detected in the image and the position information of the car; the specific position changes of the damage can be tracked by analyzing the relative movement of the damage features in the continuous image frames. Long-term data accumulation can help predict the service life of the wire rope and provide a scientific basis for preventive maintenance.

[0085] 4. Image preprocessing:

[0086] Image preprocessing is an important step in computer vision and image analysis, the purpose of which is to improve image quality, enhance features of interest, reduce unnecessary noise, and thus simplify subsequent image analysis tasks. In a specific embodiment, image preprocessing preferably includes: grayscale conversion, image scaling, and histogram equalization. Among them, grayscale conversion can reduce the amount and complexity of data while maintaining the basic information of the image; image scaling can speed up processing and reduce storage requirements; at the same time, it also helps to standardize input for comparison or matching. Histogram equalization can improve image contrast, making the image look clearer and the features more obvious.

[0087] In an embodiment of the present invention, in the histogram equalization process, the step change of pixels is detected point by point in a weighted manner to perform image enhancement, and the process includes:

[0088] The gray value of an image pixel is , , Respectively represent the total number of rows and columns of the image matrix, traverse all the pixels of the image, and correct the gray value of each pixel; the gray value of the image pixel after the image edge contour is enhanced is obtained by linear weighting; the formula is:

[0089]

[0090] In the formula, Represents the grayscale value of the enhanced image pixel; Represents the gray value of the corrected image pixel; is the set weight, and .

[0091] V. Experiment:

[0092] In the embodiment of the present invention, the damage detection model is experimented on the Ubuntu 18.04.5 operating system with Python 3.7 installed. The hardware configuration includes an Intel(R) Xeon(R) CPU E5-2680 v4 @ 2.40GHz processor, 125GB memory and 4 NVIDIA TITAN V graphics processors (GPUs). In this experiment, the preprocessed elevator wire rope image is input into the trained damage detection model for traction elevator wire rope damage detection. After a large number of experimental verifications, the damage recognition accuracy is more than 95.80%, which has achieved a high damage recognition accuracy.

[0093] From the description of the above embodiments, those skilled in the art can know that the present invention proposes a method for detecting damage to traction elevator wire ropes based on machine vision, which includes: constructing a damage detection model based on a convolutional neural network, and training the damage detection model using a sample set of elevator wire rope images with marked damage; acquiring elevator wire rope images through a camera device; performing image preprocessing on the acquired elevator wire rope images; inputting the preprocessed elevator wire rope images into the trained damage detection model, performing elevator wire rope damage detection, and outputting elevator wire rope damage detection results. The specific advantages of the method of the present invention are as follows:

[0094] 1. The method of the present invention can efficiently and accurately detect elevator wire rope damage, which not only overcomes the limitations of traditional detection methods, but also realizes automatic and real-time detection, improves the work efficiency and safety of traction elevator wire rope damage detection, and reduces maintenance costs.

[0095] 2. The present invention adopts lightweight MobileNetV3 as the backbone network, and constructs a damage detection model based on multi-scale global feature extraction and attention mechanism. The model can automatically use regions of different scales in the elevator wire rope damage image to extract damage features and estimate the degree of damage, thereby improving the generalization ability of the model, and can quickly, efficiently and accurately detect elevator wire rope damage, thereby improving detection efficiency and accuracy.

[0096] 3. The present invention performs traction elevator wire rope damage detection based on machine vision, which is a non-contact detection: there is no need to directly touch the wire rope, thus avoiding secondary damage that may be caused by physical contact; and the use of the method of the present invention is also convenient for data recording, positioning of damage positions and trend analysis: long-term data accumulation can help predict the service life of the wire rope, provide a scientific basis for preventive maintenance, and has high practical value.

[0097] Furthermore, the present invention also provides a traction elevator wire rope damage detection system based on machine vision, which uses a traction elevator wire rope damage detection method based on machine vision in the above embodiment to perform traction elevator wire rope damage detection, and the system includes:

[0098] A model building module for building a damage detection model based on a convolutional neural network;

[0099] A model training module is used to train a damage detection model using a sample set of elevator wire rope images with marked damage;

[0100] An image acquisition module is used to acquire an image of the elevator wire rope through a camera device;

[0101] An image processing module, used for performing image preprocessing on the acquired elevator wire rope image;

[0102] The damage detection module is used to input the preprocessed elevator wire rope image into the trained damage detection model, perform elevator wire rope damage detection, and output the elevator wire rope damage detection result.

[0103] The implementation principle and technical effects of the system provided in the embodiment of the present invention are the same as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the system embodiment, reference can be made to the corresponding contents in the aforementioned method embodiment, which will not be repeated here.

[0104] In addition, the embodiment of the present invention also provides a traction elevator wire rope damage detection device based on machine vision, which is used to realize traction elevator wire rope damage detection. The device can be implemented by software and / or hardware. Figure 5As shown, the apparatus comprises an image acquisition device 1, a memory 2 and a processor 3. The image acquisition device 1, the memory 2 and the processor 3 may be connected via a bus or other means.

[0105] The image acquisition device 1 is used to collect and acquire the elevator wire rope image and send it to the processor 3. The processor 3 can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components and other chips, or a combination of the above chips.

[0106] The memory 2, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer executable programs and modules, such as programs or instructions corresponding to a machine vision-based traction elevator wire rope damage detection method.

[0107] The processor 3 executes various functional applications and data processing of the processor by running the non-transient software program or instructions stored in the memory 2, that is, implements the machine vision-based traction elevator wire rope damage detection method in the above method embodiment.

[0108] The memory 2 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and application programs required for at least one function; the data storage area may store data created by the processor 2 and the like.

[0109] In addition, the memory 2 may include a high-speed random access memory and may also include a non-volatile memory. More specific examples of the memory 2 (a non-exhaustive list) include: a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a static random access memory (SRAM), a portable compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanical encoding device, and any suitable combination of the above.

[0110] In some aspects, the memory 2 optionally includes memories remotely located relative to the processor 3 , and these remote memories may be connected to the processor 3 via a network.

[0111] Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, devices or computer program products, etc. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects.

[0112] It should be noted that the word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several distinct components, and by means of a suitably programmed computer.

[0113] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0114] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for detecting damage to a traction elevator wire rope based on machine vision, characterized in that: The method comprises the following steps: S1. constructing a damage detection model based on a convolutional neural network, and training the damage detection model using a sample set of elevator wire rope images with marked damage; S2, obtaining an image of the elevator wire rope through a camera device; S3, performing image preprocessing on the acquired elevator wire rope image; S4. Input the preprocessed elevator wire rope image into the trained damage detection model, perform elevator wire rope damage detection, and output the elevator wire rope damage detection result.

2. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 1, characterized in that: In step S1, the damage detection model constructed includes: a backbone network, a multi-scale global feature extraction module, a different-scale attention enhancement module and a classification layer connected in sequence; wherein: The backbone network is a lightweight MobileNetV3 network, which is used to extract features of elevator wire rope images; The multi-scale global feature extraction module includes multiple branches for multi-scale information extraction, each branch uses dilated convolution to extract features at multiple scales, and extracts global information in a wider range by expanding the receptive field; The different-scale attention enhancement module is used to generate attention of different-scale information on the output result of the multi-scale global feature extraction module, and combine the input feature map with the attention score to obtain feature maps with different-scale attention enhancement; The classification layer is used to perform a dimensionality increase operation on the feature maps output by the attention enhancement modules of different scales, and then use the average pooling kernel to perform a global average pooling operation on the feature maps to compress the spatial information of each channel into a single numerical value; then, the pooled feature vector is passed to the fully connected layer to map the feature vector to the dimension of the number of damage severity categories; finally, the softmax activation function is used to convert the output into an elevator wire rope damage detection result.

3. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 2, characterized in that: The lightweight MobileNetV3 network includes: a convolution module and multiple stacked bottleneck modules, wherein: the convolution module is used for downsampling; each bottleneck module contains point-by-point convolution for dimensionality increase, depth-separable convolution for feature extraction, and point-by-point convolution for dimensionality reduction.

4. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 2, characterized in that: Each branch adopts the expansion rate r z The dilated convolution is used to extract features at multiple scales, and the global information is extracted by expanding the receptive field. The calculation formula of the effective receptive field is: F=[(m-1)r z +1] 2 ; Among them, F represents the effective receptive field, m represents the size of the dilated convolution kernel, and r z represents the expansion rate, and z represents the branch number.

5. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 1, characterized in that: The damage detection model is trained using the Focal Loss loss function to measure the difference between the damage severity category predicted by the model and the actual severity category; the calculation formula of the Focal Loss loss function is: ; Among them, y n is the true damage severity category number corresponding to sample n, is the model for y n The predicted probability of is the importance weighting coefficient of samples of different injury severity categories; γ is the adjustment factor.

6. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 1, characterized in that: In the step S2, the camera device is an infrared camera or an industrial CCD camera integrated with a lighting device.

7. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 1, characterized in that: In the step S3, the image preprocessing includes: grayscale conversion, image scaling and histogram equalization.

8. The method for detecting damage to a traction elevator wire rope based on machine vision according to claim 7, characterized in that: In the process of histogram equalization, the step changes of pixels are detected point by point in a weighted form to perform image enhancement. The process includes: The gray value of an image pixel is , , Respectively represent the total number of rows and columns of the image matrix, traverse all the pixels of the image, and correct the gray value of each pixel; the gray value of the image pixel after the image edge contour is enhanced is obtained by linear weighting; the formula is: ; In the formula, Represents the grayscale value of the enhanced image pixel; Represents the gray value of the corrected image pixel; is the set weight, and .

9. A traction elevator wire rope damage detection device based on machine vision, characterized in that: include: An image acquisition device, a memory and a processor, wherein: The image acquisition device is used to collect and acquire the image of the elevator wire rope; The memory stores computer instructions; The processor is connected to the image acquisition device and the memory data. The processor executes the computer instructions to execute the machine vision-based traction elevator wire rope damage detection method as described in any one of claims 1 to 8 to perform traction elevator wire rope damage detection.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer instructions are executed by the processor, the method for detecting damage to a traction elevator wire rope based on machine vision as described in any one of claims 1 to 8 is implemented.

Citation Information

Cited By

  • Flaw detection method and system for alloy steel wire rope production

    CN119666929A

  • Characteristic pre-fusion ARV underwater oil pipe curvature detection method and system and storage medium

    CN120101698A

  • Fluid pyramid elevator foreign matter detection method based on multi-scale depth feature fusion

    CN121191098A