WO2021036670A1 - 多模输出的智能振动检测方法及装置 - Google Patents

多模输出的智能振动检测方法及装置 Download PDF

Info

Publication number
WO2021036670A1
WO2021036670A1 PCT/CN2020/105792 CN2020105792W WO2021036670A1 WO 2021036670 A1 WO2021036670 A1 WO 2021036670A1 CN 2020105792 W CN2020105792 W CN 2020105792W WO 2021036670 A1 WO2021036670 A1 WO 2021036670A1
Authority
WO
WIPO (PCT)
Prior art keywords
vibration
feature points
video
cluster
fault
Prior art date
Application number
PCT/CN2020/105792
Other languages
English (en)
French (fr)
Inventor
高风波
Original Assignee
深圳市广宁股份有限公司
深圳市豪视智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳市广宁股份有限公司, 深圳市豪视智能科技有限公司 filed Critical 深圳市广宁股份有限公司
Publication of WO2021036670A1 publication Critical patent/WO2021036670A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/20—Analysis of motion
    • G06T7/246—Analysis of motion using feature-based methods, e.g. the tracking of corners or segments
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
    • G01H9/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by using radiation-sensitive means, e.g. optical means
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M15/00—Testing of engines
    • G01M15/02—Details or accessories of testing apparatus
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/20—Administration of product repair or maintenance
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L67/00—Network arrangements or protocols for supporting network services or applications
    • H04L67/01—Protocols
    • H04L67/10—Protocols in which an application is distributed across nodes in the network
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10016—Video; Image sequence
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation

Definitions

  • This application relates to the field of Internet technology, and in particular to a method and device for intelligent vibration detection with multi-mode output.
  • the Internet belongs to the field of media, also known as the international network. It is a huge network connected between networks and networks. These networks are connected by a set of common protocols to form a logically single huge international network. This method of connecting computer networks to each other can be called “network interconnection”. On this basis, a global Internet that covers the whole world has been developed, called the Internet, which is a network structure connected to each other. "Internet+" is a new business form of Internet development under Innovation 2.0, and it is the evolution of the Internet form driven by Innovation 2.0 in the Knowledge Society and the new form of economic and social development that it has spawned.
  • Internet+ is a further practical result of Internet thinking, which promotes the continuous evolution of economic forms, thereby driving the vitality of social economic entities, and providing a broad network platform for reform, innovation, and development.
  • Internet + means “Internet + various traditional industries", but this is not a simple addition of the two, but the use of information and communication technology and Internet platforms to deepen the integration of the Internet and traditional industries to create new Develop ecology. It represents a new social form, that is, to give full play to the optimization and integration role of the Internet in the allocation of social resources, to deeply integrate the innovative achievements of the Internet into economic and social domains, to enhance the innovation and productivity of the entire society, and to form A broader new form of economic development using the Internet as an infrastructure and realization tool.
  • the traditional fault monitoring mechanism generally uses localized detection equipment, such as arranging laser Doppler vibrometer LDVs in a dedicated room, through which localized vibration detection and failure prediction are performed.
  • LDVs are expensive and use environment Restrictions (environmental influences such as temperature and light in the test environment will seriously deteriorate the measurement results), small test areas, and difficulty in remote monitoring, making it difficult to meet the needs of intelligent vibration detection in an increasing number of scenarios.
  • vibration detection is an important part.
  • Most of the existing vibration detection systems characterize the vibration of a vibrating object by separately acquiring vibration parameters or separately acquiring a vibration image, which results in that the vibration situation cannot be well observed and understood.
  • the purpose of the embodiments of the present application is to provide a multi-mode output intelligent vibration detection method and device.
  • the expression form of the vibration detection results is enriched and improved Accuracy and effectiveness of vibration detection.
  • the data transmission process in the vibration detection method disclosed in the embodiments of the present application can be based on Internet + technology to form a distributed intelligent vibration detection system of local + cloud or server.
  • the local collection device can perform accurate and original vibration detection. Image collection and preprocessing.
  • the cloud or server can be based on the distributed data obtained, combined with various special fault detection models obtained through statistical analysis of big data technology, to predict the fault of the detected target, and realize the Internet and traditional fault monitoring
  • the deep integration of the industry improves the intelligence and accuracy of fault monitoring, and meets the needs of intelligent vibration detection in an increasing number of scenarios.
  • the first aspect of the embodiments of the present application provides a vibration detection method, the method includes:
  • the transmitter in the process of locating the vibration is a vibrating object, and the target video corresponding to the vibrating object is acquired;
  • the stability data includes the vibration parameter and the amplified output video
  • the vibration condition of the vibrating object is determined according to the stability data.
  • a second aspect of the embodiments of the present application discloses a vibration detection device, which includes:
  • a receiving module configured to receive an instruction from a user to activate a device detection function, and present a vehicle vibration detection portal according to the instruction, the vehicle vibration detection portal provides vibration detection type options, and the vibration detection type options include vehicle type or engine model;
  • the selection module is used to receive the vibration detection type option selected by the user, and determine the engine model according to the vibration detection type option;
  • the prompt module is used to prompt the user to perform a vibration operation on the engine corresponding to the transmitter model, and the vibration operation includes a driving operation or an operation of stepping on the accelerator in neutral;
  • a positioning module configured to locate the transmitter as a vibrating object in the vibration process, and obtain a target video corresponding to the vibrating object;
  • An extraction module configured to extract parameters of the target video to obtain vibration parameters corresponding to the target video
  • An amplifying module configured to perform amplifying processing on the target video in parallel to obtain an amplified output video
  • the determining module is configured to output multiple modes of stability data in parallel, the stability data includes the vibration parameter and the amplified output video, and the vibration condition of the vibrating object is determined according to the stability data.
  • the third aspect of the embodiments of the present application discloses an electronic device, which is characterized by comprising a processor, a memory, a communication interface, and one or more programs, the one or more programs are stored in the memory, and It is configured to be executed by the processor, and the program includes instructions for executing the steps in the method of the first aspect.
  • the fourth aspect of the embodiments of the present application discloses a storage medium for storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute instructions of the steps in the corresponding method of the first aspect.
  • the embodiments of the present application disclose a vibration detection method and device, by receiving user instructions, obtaining the vehicle vibration detection entrance, and then receiving the engine model selected by the user, positioning the corresponding engine to obtain the target video corresponding to the vibrating object;
  • the target video is parameter extracted to obtain the vibration parameters corresponding to the target video;
  • the target video is amplified in parallel to obtain the amplified output video;
  • the stability data of multiple modes is output in parallel, and the stability data includes the vibration parameters and the amplified output video.
  • the stability data determines the vibration of the vibrating object.
  • the parameter extraction and amplification processing of the target video are performed in parallel, which can improve the efficiency of vibration detection.
  • the vibration parameters obtained by parameter extraction and the amplified output video obtained after the amplification processing are output in parallel, which enriches the manifestation of vibration detection results. , Improve the accuracy and effectiveness of vibration detection.
  • FIG. 1A is a structural block diagram of a vibration sensing device provided by an embodiment of the application.
  • FIG. 1B is a schematic flowchart of a vibration detection method provided by an embodiment of the application.
  • Fig. 1C is a schematic structural diagram of an engine provided by an embodiment of the application.
  • FIG. 1D is a schematic diagram of a setting interface of a sensor device according to an embodiment of the application.
  • FIG. 1E is a schematic diagram of a stability data display provided by an embodiment of the application.
  • Fig. 2 is a method for extracting parameters of a target video provided by an embodiment of the application.
  • FIG. 3 is a schematic flowchart of a method for verifying vibration conditions provided by an embodiment of the application.
  • FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • FIG. 5 is a schematic structural diagram of a vibration detection device provided by an embodiment of the application.
  • FIG. 1A is a structural block diagram 10 of a vibration sensing device provided by an embodiment of the present application, which is used to complete vibration detection.
  • the device includes an optical lens 11, an area image sensor device 12, a computing unit 13, a storage unit 14, and an interface unit 15.
  • the optical lens 11 is used to obtain a target image corresponding to a vibrating object, and adjust the focal length and aperture size of the optical lens 11, so that a clear image of the vibrating object can be formed on the optical sensing device.
  • the image of the vibrating object passing through the optical lens 11 is converted into electrical signal data by the area image sensor device 12 (the area image sensor device may be a CMOS sensor device or a CCD sensor device).
  • the electrical signal is stored in the storage unit 14, and the storage unit 14 is composed of a random access memory (RAM) and a read-only memory (ROM).
  • the calculation unit 13 is used to execute the corresponding algorithm.
  • the calculation unit 13 includes a CPU and a GPU, which can be used to alternately perform the vibration parameter extraction process and the video amplification process in the stability prediction model, and the acquired vibration parameters and the amplified output video form stability data.
  • it is connected to the user interface through the interface unit 15 and accepts the user to set the vibration sensing device.
  • the interface unit 15 is also used to output stability data.
  • the vibration parameters in the stability data include numerical values, vibration waveform graphs, modal graphs or spectrograms. These vibration parameters and the amplified output video are output in multi-mode, so that various parameters , Image and video contrast display, fully reflect the vibration of the vibrating object.
  • the interface unit 15 can be a USB interface, an RS232 interface, or other interfaces, which can also be used to set up a sensing device through the provided API interface.
  • FIG. 1B is a schematic flowchart of a vibration detection method provided by an embodiment of the application, as shown in FIG. As shown in 1B, the vibration detection method includes the following steps:
  • the vehicle vibration detection portal provides vibration detection type options, and the vibration detection type options include vehicle type or engine model.
  • Vibrating objects include objects that generate mechanical vibration through internal interaction, including engine vibration, transmitter vibration, or gear vibration, or the physics of mechanical vibration due to external forces, including wire vibration or bridge vibration.
  • the vibrating object will mechanically vibrate at a fixed frequency under normal conditions, and when the vibrating object fails, the vibration frequency will also change. Therefore, the target video corresponding to the vibrating object can be obtained and analyzed to determine the fault condition of the vibrating object.
  • the target video of the transmitter needs to be obtained.
  • the vibration detection sensor device is connected to the user interface through the interface. After the device detection function is activated, the user interface provides the user with the vibration detection type. If the user knows the engine model of his vehicle, he can directly select the vibration detection type option. If you know the vehicle engine model, you can select the vehicle type.
  • the vibration detection type option After the user selects the vibration detection type option, if the selected engine model is selected, the engine model can be directly determined. If the vehicle type is selected, the processor in the vibration detection sensing device can obtain the engine model corresponding to the vehicle type through the Internet, or The engine model corresponding to the vehicle type is directly extracted from the database.
  • the engine To perform vibration detection on the engine corresponding to the engine model, the engine needs to be vibrated. In the normal driving process of the vehicle, the engine is working, and it can be determined that the engine will vibrate. However, in some cases the vibration situation cannot be obtained during normal driving. For example, it is necessary to obtain the vibration of the engine at resonance or near resonance. It is difficult to achieve during normal driving, and there is a safety risk in resonance. This type of operation is not supported during normal driving. Therefore, the engine needs to be operated in neutral gear to obtain vibration. happening.
  • the transmitter in the process of locating the vibration is a vibrating object, and acquiring a target video corresponding to the vibrating object.
  • the engine in the positioning vibration process is a vibrating object, that is, if the engine does not vibrate, or the vibrating object is not an engine, it will not be used as a vibrating object for target video capture. This prevents the video of the engine in a non-vibrating state from being collected when there are multiple engines, or the video of other vibrating objects is collected as the target video. Improve the efficiency and accuracy of target video capture.
  • the obtaining the target video corresponding to the vibrating object further includes: obtaining a first video and a second video, where the first video and the second video are different source videos shot at the same time for the same target; Acquire a first frame image corresponding to the first video and a second frame image corresponding to the second video; overlap the first frame image and the second frame image, and compare the first frame image and the second frame image The pixels that cannot be overlapped in the second frame of image are removed; and the target video is obtained.
  • the same vibrating object when the same vibrating object is in video capture, it may be due to some external reasons, such as camera shaking, camera failure, etc., that the captured target video may be deviated. Then, use different cameras to capture the same vibrating object at the same time.
  • the two frame images correspond to Pixels should be completely overlapped, then remove the pixels that cannot overlap between the first frame of image and the second frame of image, that is, to remove the shooting deviation pixels in the two cameras, and the obtained frame image is less noisy.
  • Extracting the parameters of the target video can obtain the reflection of the vibration of the target video in the parameter changes.
  • the obtained vibration parameters can be numerical values, including vibration amplitude, vibration period, vibration frequency, etc., or related images, including vibration Waveform diagrams or spectrograms, etc., these vibration parameters can help to efficiently determine the relevant conclusions of the vibration situation.
  • N1 represents the number of regions containing key points
  • N2 represents the number of regions that do not contain key points
  • N N1+N2
  • R represents the preset feature point selection of the image frame Number
  • the actual number of initial feature points extracted from the N regions is: Screen the initial feature points to obtain stable multiple motion feature points; perform optical flow tracking on multiple motion feature points to obtain a time sequence of multiple motion feature points; filter the time sequence of multiple motion feature points Process to obtain the filtered signal; perform principal component analysis on the filtered signal to obtain the dimensionality reduction signal; perform parameter extraction on the dimensionality reduction signal to obtain the vibration parameters corresponding to the target video.
  • FIG. 1C is a schematic structural diagram of an engine provided by an embodiment of the application.
  • the engine includes an ignition coil 110, a cam
  • the valve, piston, ignition coil, and timing chain are always in working condition, and the vibration is caused by the work of these mechanisms.
  • the intermediate connecting structure such as a crank connecting rod mechanism or a crankshaft, will also be driven by the vibration of the connected mechanism, and the infrequent mechanism, such as the lubricating oil pan, has a relatively low vibration frequency.
  • the image frame can be divided into regions first, for example according to the functional structure, or evenly divided according to the grid.
  • the image frame in Figure 1C is divided into 5 regions by the dotted line 1, 2, 3, 4 .
  • the key points include valves, pistons, ignition coils, and timing chains, it can be seen that the area above the dotted line 1 contains the most key points, which is 2, followed by the area divided by the dotted line 1-2 and the dotted line 2-3.
  • the area allocates 90% of the initial feature points according to the proportion of the number of key points in each area, and allocates 10% of the initial feature points evenly for all areas that do not include key points. Because the number of initial feature points is a positive integer, the pair is based on the formula
  • the calculated number of feature points is selected and rounded to obtain the final initial number of feature points R', which is the sum of the initial feature points extracted for all regions including key points and regions not including key points.
  • the feature points of the preset type include at least one of the following: corner points, edges Spots, bright spots in dark areas, and dark spots in bright areas.
  • the corner point is the intersection between the contours. For the same scene, even if the viewing angle changes, the corner point usually has the characteristics of stability, and the pixels in the area near the point have large changes in the gradient direction and the gradient amplitude. .
  • the edge point is the intersection between the vibrating object and the background image.
  • the dark point in the bright area and the bright point in the dark area have high contrast with other pixels, and can be used as the initial feature point for better observation of the motion feature point.
  • the selected feature points are screened to obtain multiple stable motion feature points.
  • optical flow tracking is performed on the motion feature points to obtain the motion trajectory of the motion feature points on the time axis, that is, the time sequence.
  • filter the time series to filter out non-demand frequency motion trajectories.
  • the filtering can be time domain filtering or frequency domain filtering.
  • Principal component analysis (PCA) is performed on the filtered time series, and multiple indicators are synthesized into a few unrelated comprehensive indicators (ie principal components) to obtain a dimensionality reduction signal.
  • extract the parameters of the dimensionality reduction signal to obtain the vibration parameters corresponding to the target video, including the maximum vibration amplitude, the interval distribution of the vibration amplitude, and whether the resonance amplitude is reached or not.
  • filter the extracted initial feature points to obtain stable multiple motion feature points including: calculating the flow vector of the initial feature point based on the minimum difference square and SSD matching; calculating according to the flow vector corresponding to the initial feature point The offset distance of the initial feature point; cluster the multiple offset distances corresponding to the multiple initial feature points, and use the K-means clustering algorithm to cluster the multiple offset distances, where the k value is set to 4, Obtain 4 clusters; average the 4 clusters, and obtain the average of the first cluster, the average of the second cluster, the average of the third cluster, and the average of the fourth cluster, where , The average value of the first type cluster ⁇ the average value of the second type cluster ⁇ the average value of the third type cluster ⁇ the average value of the fourth type cluster; from the clusters corresponding to the average value of the second type cluster and the third type cluster average value Select multiple offset distances with variance less than the first variance threshold to form a fifth cluster cluster, and determine the initial feature points corresponding to the offset distances in the fifth cluster as multiple stable motion feature points.
  • Sum of Squared Differences (SSD) matching is used to calculate the flow vector of the initial feature points.
  • SSD Sum of Squared Differences
  • the motion trajectory of the initial feature point can be determined, and then the flow vector of the initial feature point can be obtained.
  • calculate the offset distance of the initial feature point according to the flow vector Find the modulus, get It is the offset distance of the initial feature point from point A to point B.
  • the K-means clustering algorithm uses the K-means clustering algorithm to cluster the multiple offset distances, set the k value to 4, and obtain 4 clusters Clusters, and then calculate the average of the 4 clusters respectively, and obtain the average of the first type of cluster, the average of the second type of cluster, the average of the third type of cluster, and the average of the fourth type of cluster, and the average of the first type of cluster
  • the value ⁇ the average value of the second type cluster ⁇ the average value of the third type cluster ⁇ the average value of the fourth type cluster, where the initial feature point corresponding to the cluster of the first type cluster average value can be determined as a stationary point (background point),
  • the initial feature points corresponding to the cluster averages of the four types of clusters can be determined as the points of violent motion (may be some irregularly moving points), and the clusters of the second type of cluster average and the third type of cluster average are corresponding to the clusters
  • the initial feature point has a moderate motion range, and it is most likely to be a stable feature
  • the offset distance variance of these moving feature points will also be a relatively stable value. Therefore, select the second type cluster average and the third type cluster average value. Multiple offset distances in the cluster cluster corresponding to the cluster average value are smaller than the first variance threshold to form the fifth cluster cluster, and the initial feature point corresponding to the offset distance in the fifth cluster is determined as the stable Of multiple motion feature points. This process can further improve the accuracy of the obtained stable multiple motion feature points.
  • performing optical flow tracking on multiple motion feature points to obtain a time sequence of multiple motion feature points including: determining the positions of the multiple motion feature points in the first frame of image; according to feature point matching and least squares method Determine the position of multiple motion feature points in the next adjacent frame of image; repeat the above to determine the position of multiple motion feature points in the next adjacent frame of image, until each frame of the multiple image frames is traversed; The sequence and position of the motion feature points determine the time sequence of multiple motion feature points.
  • the optical flow tracking of the motion feature point is to assume that the brightness and color of the motion feature point in different image frames do not change, and only the position changes. Therefore, it is necessary to track the position of the same motion feature point in different image frames.
  • it is tracked by the brightness or color matching of the feature point, and on the other hand, the location of the motion feature point is estimated by region division and position estimation. Locate the position of the same motion feature point in different image frames, and the position can be expressed by coordinates, and then a set of digital sequences of the motion feature point in time sequence can be obtained, that is, the time sequence of the motion feature point.
  • each image frame of the multiple image frames of the target video is divided into regions, and then the number of initial feature points selected corresponding to each region is calculated according to the key points contained in each region
  • the initial feature points of different regions can be allocated more accurately, which is helpful for the subsequent stable acquisition of multiple motion feature points, thereby improving the accuracy and effectiveness of the obtained vibration parameters;
  • by comparing the obtained initial feature points Perform offset distance calculation and offset distance clustering to complete the screening process, obtain multiple stable motion feature points, which can improve the accuracy of the obtained motion feature points; then perform optical flow tracking on the motion feature points to obtain the time series, and then Further filtering and vibration parameter extraction. The whole process improves the accuracy of vibration parameter extraction.
  • FIG. 1D is a schematic diagram of a sensor device setting interface provided by an embodiment of the application. As shown in Figure 1D, the sensor device setting interface provides users with two types of "add test point" and "delete test point".
  • the method further includes: receiving a user's operation of adding test points, and locating a plurality of addable test points selected by the user; and determining Multiple plane distances between the multiple test points that can be added; average the multiple plane distances to obtain an average plane distance; determine whether the average plane distance is less than a first distance threshold; if so, then Acquire multiple sets of short-distance addable test points corresponding to the smallest K distance values among the multiple plane distances; merge the multiple sets of short-distance addable test points into the same group to form multiple fusion addable test points; Take the remaining addable test points that have not been merged among the plurality of addable test points and the plurality of fused addable test points together as the target test point, and obtain the image frame of the target test point as the target video Image frame.
  • the user can add test points and delete test points on the sensor device setting interface.
  • the touch range is fixed, for example, each touch is It is a range of 1 cm 2 , but the actual range that the user wants to select may be 10 cm 2 , the user needs to operate multiple times, and the operating range may not completely cover the range that the user wants to select. Therefore, multiple plane distances between two addable test points are determined, and the multiple plane distances are averaged to obtain the average plane distance, where the plane distance refers to the corresponding distance on the screen of the setting interface.
  • the average plane distance is less than the first distance threshold, it means that the user selects test points within a very close range, and may want to select a larger area as the test point, then you can select the closest K test points for the same group fusion, that is Take the line between the two test points as the diameter to obtain a circle, and the area covered by the circle is the fusion to add test points. Finally, the fused fusion addable test point and the unfused addable test point are used as the target test point together, and the image frame of the target test point is obtained as the image frame of the target video for subsequent vibration parameter extraction, which can reduce the amount of data processing and improve The efficiency of vibration parameter extraction.
  • the parameter extraction process and the video amplification process need to be carried out in parallel, in the same processor, through two parallel processes, or through dual threads in the same process. Perform parallel processing.
  • performing amplification processing on the target video to obtain the amplified output video includes: performing spatial pyramid decomposition on a frame sequence composed of multiple frame images of the target video to obtain a pyramid structure composed of multiple sub-images with different spatial resolutions ; Perform time-domain bandpass filtering on each of the multiple sub-images in the pyramid structure to obtain the transformed signal corresponding to the target frequency band; Amplify the displacement corresponding to the transformed signal by A times to obtain the amplified signal, where A is taken The value range is (3, Amax), where the value of Amax is determined by the target frequency band and the displacement function of the transformed signal; combine the amplified signal and the pyramid structure to perform pyramid reconstruction to obtain the amplified output video.
  • the Euler motion amplification method can be used to enlarge the target video.
  • the pixels in the target video need to be converted into a function of time and space, that is, the multi-frame image of the target video is transformed by the image pyramid transformation.
  • the composed frame sequence is decomposed into multiple sub-images with different spatial resolutions and different scales to form a pyramid tower structure.
  • a Gaussian pyramid is used to decompose the multi-frame image of the target video, that is, a group of layers are layered in size.
  • the halved image sequence forms a pyramid structure, and each level of image in the sequence is the result of low-pass filtering of the previous level of image and interlaced sampling.
  • Pyramid decomposition is to perform spatial filtering on the frame sequence, and decompose to obtain different spatial frequencies
  • time-domain band-pass filtering can be performed on each frequency band to obtain the transformed signal of interest, that is, the transformed signal corresponding to the target frequency band, and only the transformation corresponding to the target frequency band The signal is amplified.
  • ideal band-pass filters, Butterworth band-pass filters, second-order infinite impulse response filters, etc. can be used.
  • ⁇ (t) represents the displacement signal
  • Amplify I(x, t) by ⁇ times, that is, amplify the displacement signal ⁇ (t), and the amplified signal is:
  • magnification is related to the spatial frequency and satisfies the following relationship:
  • the spatial frequency is ⁇
  • the spatial wavelength of the target frequency band is ⁇
  • ⁇ 2 ⁇ / ⁇
  • the maximum value of ⁇ can be determined by the displacement function of the target frequency band and the transformed signal. A max ⁇ .
  • the amplified signal After the amplified signal is obtained, it is recombined with the original frequency band, and then pyramid reconstruction, such as Laplace pyramid transform reconstruction, is used to obtain the amplified image, and then proceed to obtain the amplified output video.
  • pyramid reconstruction such as Laplace pyramid transform reconstruction
  • FIG. 1E is a schematic diagram of a stability data display provided by an embodiment of the application.
  • the vibration detection sensor device is connected to the user interface through the interface unit, and the interface unit is output in multi-mode on the user interface. Display the stability data of the zoomed-in video on the left area, and display the vibration parameters of the zoomed-in video corresponding to the target detection point in the right area.
  • the target detection point is a circular area corresponding to 110, and the right area That is, its corresponding vibration parameters.
  • the vibration parameter display area you can also choose to display frequency domain waveform graphs or frequency domain waveform graphs, and can also display specific parameter values such as the maximum vibration amplitude and the interval distribution of the vibration amplitude.
  • the vibration condition of the vibrating object can be determined through the multi-mode output result. Get conclusions about whether the vibrating object includes faults, fault types and names.
  • the method further includes: receiving the fault video and predicted fault name input by the user; matching the predicted fault name with the fault information in the fault list to determine whether the predicted fault name is included in the fault list, where the fault information includes the fault name And the stability data corresponding to the fault name; determine the target fault name that matches the predicted fault name, and obtain the vibration parameters corresponding to the fault video; match the vibration parameters corresponding to the fault video with the stability parameters corresponding to the target fault name; When the vibration parameters corresponding to the fault video and the stability parameters corresponding to the target fault name match successfully, it is determined that the predicted fault name is correct.
  • the vibration detection model for detection and obtains all the stability data
  • the predicted fault name is "bearing damage”.
  • the fault list After the fault list is successfully matched, it is determined that the fault name is also included in the fault list, which is the target fault name.
  • the fault name includes resonance
  • the method further includes updating the fault list, which specifically includes: acquiring the resonance speed corresponding to the vehicle engine model; acquiring the resonance target video corresponding to the engine model in the resonance speed range, where the resonance speed range is the difference between the resonance speed and the resonance speed.
  • the speed value range where the absolute value of the value is less than the first preset threshold; match the stability prediction model corresponding to the engine model, input the resonance target video into the stability prediction model, and obtain the resonance data of the engine model corresponding to the engine; compare the engine model and its corresponding
  • the resonance data is updated to the fault list as fault information.
  • resonance is a very typical vibration fault.
  • the transmitter has a resonance frequency, and different engines have different resonance frequencies. When the speed reaches a certain value, resonance problems may occur. Then when constructing the engine fault list, the resonance fault needs to be updated to it.
  • the resonance speed corresponding to the vehicle engine model which can be obtained directly according to the vehicle performance parameters, or can be obtained through independent experiments.
  • the rotation speed range refers to the rotation speed value range where the absolute value of the difference between the resonance speed and the resonance speed is less than the first preset threshold, for example, the resonance speed is 1200 rpm (revolutions per minute).
  • the first preset threshold is 100, and the rotation speed range is 1100 rpm to 1300 rpm.
  • resonance information may be displayed in the resonance target video corresponding to the engine. Therefore, the resonance target video is input into the stability prediction model, and the corresponding stability data is obtained as the resonance data. Finally, the engine model and its corresponding resonance data are updated to the fault list as fault information for subsequent fault prediction verification.
  • the vehicle vibration detection entrance is obtained by receiving user instructions, and then the engine model selected by the user is received, and the corresponding engine is located to obtain the target video corresponding to the vibrating object;
  • the target video is parameterized Extract to obtain the vibration parameters corresponding to the target video; parallelly amplify the target video to obtain the amplified output video; output multiple modes of stability data in parallel, and the stability data includes vibration parameters and amplified output video, which are determined according to the stability data
  • the vibration of the vibrating object the parameter extraction and amplification processing of the target video are performed in parallel, which can improve the efficiency of vibration detection.
  • the vibration parameters obtained by parameter extraction and the amplified output video obtained after the amplification processing are output in parallel, which enriches the manifestation of vibration detection results. , Improve the accuracy and effectiveness of vibration detection.
  • FIG. 2 is a method for extracting parameters of a target video according to an embodiment of the application. As shown in FIG. 2, the method includes the following steps:
  • Class cluster
  • each image frame of the multiple image frames of the target video is divided into regions, and then each image frame is calculated according to the key points contained in each region.
  • the area corresponds to the number of initial feature points selected, so that the initial feature points of different areas can be more accurately allocated, which is helpful for the subsequent stable acquisition of multiple motion feature points, thereby improving the accuracy and effectiveness of the obtained vibration parameters
  • the screening process is completed by calculating the offset distance and the offset distance clustering of the obtained initial feature points, and obtain a stable multiple motion feature points, which can improve the accuracy of the obtained motion feature points; then perform the motion feature points
  • Optical flow tracking, time series is obtained, and then further filtering and vibration parameter extraction are performed. The whole process improves the accuracy of vibration parameter extraction.
  • FIG. 3 is a schematic flowchart of a method for verifying vibration conditions provided by an embodiment of the application. As shown in FIG. 3, the method includes the following steps:
  • the resonance target video of the engine corresponding to the engine model in the resonance speed range and its corresponding resonance data are acquired, and the resonance data is updated to the fault list as fault information.
  • you need to verify whether the predicted fault name corresponding to the fault video is correct you can obtain the vibration parameters corresponding to the fault video, and determine whether the predicted fault name is correct according to the matching result of the vibration parameter and the stability parameter in the fault list, without the need to obtain the fault again
  • the motion zoom video corresponding to the video improves the efficiency of vibration verification on the one hand, and on the other hand, improves the accuracy of vibration verification through matching with the fault list.
  • FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present application.
  • the electronic device includes a processor, a memory, a communication interface, and one or more programs.
  • the above-mentioned one or more programs are stored in the above-mentioned memory and are configured to be executed by the above-mentioned processor, and the above-mentioned programs include instructions for executing the following steps:
  • the vehicle vibration detection portal Receiving an instruction from a user to activate a device detection function, and presenting a vehicle vibration detection portal according to the instruction, the vehicle vibration detection portal providing vibration detection type options, and the vibration detection type options include vehicle type or engine model;
  • the transmitter in the process of locating the vibration is a vibrating object, and the target video corresponding to the vibrating object is acquired;
  • the stability data includes the vibration parameter and the amplified output video
  • the vibration condition of the vibrating object is determined according to the stability data.
  • the electronic device obtains the vehicle vibration detection entrance by receiving user instructions, then receives the engine model selected by the user, locates the corresponding engine to obtain the target video corresponding to the vibrating object; extracts the parameters of the target video, Obtain the vibration parameters corresponding to the target video; amplify the target video in parallel to obtain the amplified output video; output multiple modes of stability data in parallel, the stability data includes vibration parameters and the amplified output video, and determine the vibrating object according to the stability data The vibration situation.
  • the parameter extraction and amplification processing of the target video are performed in parallel, which can improve the efficiency of vibration detection.
  • the vibration parameters obtained by parameter extraction and the amplified output video obtained after the amplification processing are output in parallel, which enriches the manifestation of vibration detection results. , Improve the accuracy and effectiveness of vibration detection.
  • the program stored in the electronic device 400 further includes other instructions for executing the various methods described in FIGS. 1B to 3.
  • FIG. 5 is a schematic structural diagram of a vibration detection device provided by an embodiment of the application. As shown in FIG. 5, the vibration detection device 500 includes:
  • the receiving module 501 is configured to receive an instruction from a user to activate a device detection function, and present a vehicle vibration detection portal according to the instruction, the vehicle vibration detection portal provides vibration detection type options, and the vibration detection type options include vehicle type or engine model;
  • the selection module 502 is configured to receive the vibration detection type option selected by the user and determine the engine model according to the vibration detection type option;
  • the prompt module 503 is configured to prompt the user to perform a vibration operation on the engine corresponding to the transmitter model.
  • the vibration operation includes a driving operation or an operation of stepping on the accelerator in neutral; a positioning module 504, used to locate the transmitter in the vibration process as a vibrating object, and obtain the target video corresponding to the vibrating object; an extraction module 505, used to The target video performs parameter extraction to obtain the vibration parameters corresponding to the target video; the amplifying module 506 is used to amplify the target video in parallel to obtain the amplified output video; the determining module 507 is used to output multiple modes in parallel Stability data, where the stability data includes the vibration parameter and the amplified output video, and the vibration condition of the vibrating object is determined according to the stability data.
  • the specific working process of the receiving module 501, the selection module 502, the prompt module 503, the positioning module 504, the extraction module 505, the amplification module 506, and the determination module 507 can be referred to the vibration detection described in steps 101-107 above. The corresponding description of the method will not be repeated here.
  • the extraction module 505 is specifically configured to:
  • the initial feature points are selected according to the number of key points contained in each area, and the number of selected points is:
  • N1 represents the number of regions containing key points
  • N2 represents the number of regions that do not contain key points
  • N N1+N2
  • R represents the preset feature point selection of the image frame Number
  • the initial feature points extracted in each image frame are obtained according to the number of feature points Ti selected in each region, where the number of initial feature points is:
  • Parameter extraction is performed on the dimensionality reduction signal to obtain vibration parameters corresponding to the target video.
  • the extraction module 505 is also specifically configured to:
  • Clustering multiple offset distances corresponding to multiple initial feature points clustering the multiple offset distances using the K-means clustering algorithm, where the k value is set to 4 to obtain 4 clustering classes cluster;
  • Calculate the average of the 4 clusters and obtain the average of the first type of cluster, the average of the second type of cluster, the average of the third type of cluster, and the average of the fourth type of cluster, where the average of the first type of cluster Value ⁇ Average value of the second type cluster ⁇ Average value of the third type cluster ⁇ Average value of the fourth type cluster;
  • the extraction module 505 is further specifically configured to:
  • the time sequence of the plurality of motion feature points is determined according to the sequence and position of determining the plurality of motion feature points.
  • the amplifying module 506 is specifically configured to:
  • Pyramid reconstruction is performed by combining the amplified signal and the pyramid structure to obtain an amplified output video.
  • the vibration detection device 500 further includes a verification module 508, which is specifically configured to:
  • the fault name includes resonance
  • the verification module 508 is also used to update the fault list, specifically:
  • the resonance rotation speed interval is a rotation speed value interval in which an absolute value of a difference between the resonance rotation speed and the resonance rotation speed is less than a first preset threshold
  • the engine model and its corresponding resonance data are updated to the fault list as fault information.
  • the vibration detection device disclosed in the embodiment of the present application obtains the vehicle vibration detection entrance by receiving user instructions, and then receives the engine model selected by the user, locates the corresponding engine to obtain the target video corresponding to the vibrating object; parameterizes the target video Extract to obtain the vibration parameters corresponding to the target video; parallelly amplify the target video to obtain the amplified output video; output multiple modes of stability data in parallel, and the stability data includes vibration parameters and amplified output video, which are determined according to the stability data
  • the vibration of the vibrating object the parameter extraction and amplification processing of the target video are performed in parallel, which can improve the efficiency of vibration detection.
  • the vibration parameters obtained by parameter extraction and the amplified output video obtained after the amplification processing are output in parallel, which enriches the manifestation of vibration detection results. , Improve the accuracy and effectiveness of vibration detection.
  • a storage medium for storing a computer program for electronic data exchange, wherein the computer program causes a computer to execute instructions of the steps in any of the above methods.
  • the disclosed method can be implemented in other ways.
  • the method embodiments described above are only illustrative, for example, the division of the units is only a logical function division, and there may be other divisions in actual implementation, for example, multiple units or components may be combined or may be Integrate into another system, or some features can be ignored or not implemented.
  • the displayed or discussed mutual coupling or direct coupling or communication connection may be indirect coupling or communication connection through some interfaces, methods or units, and may be in electrical or other forms.
  • the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the objectives of the solutions of the embodiments.
  • the functional units in the various embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units may be integrated into one unit.
  • the above-mentioned integrated unit can be realized in the form of hardware or software program module.
  • the integrated unit is implemented in the form of a software program module and sold or used as an independent product, it can be stored in a computer readable memory.
  • the technical solution of the present application essentially or the part that contributes to the existing technology or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a memory.
  • a number of instructions are included to enable a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application.
  • the aforementioned memory includes: U disk, Read-Only Memory (ROM, Read-Only Memory), Random Access Memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk and other media that can store program codes.
  • the program can be stored in a computer-readable memory, and the memory can include: a flash disk , Read-only memory (English: Read-Only Memory, abbreviation: ROM), random access device (English: Random Access Memory, abbreviation: RAM), magnetic disk or optical disk, etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Quality & Reliability (AREA)
  • Strategic Management (AREA)
  • Tourism & Hospitality (AREA)
  • Operations Research (AREA)
  • General Business, Economics & Management (AREA)
  • Marketing (AREA)
  • Economics (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)

Abstract

本申请实施例提供一种多模输出的智能振动检测方法及装置，方法包括：获取振动物体对应的目标视频；对目标视频进行参数提取，获得目标视频对应的振动参数；并行地对目标视频进行放大处理，获得放大输出视频；并行输出多种模式的稳定性数据，稳定性数据包括振动参数和放大输出视频，根据稳定性数据确定振动物体的振动情况。采用本申请实施例，通过并行地获取和输出多种模式的稳定性数据，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。

Description

多模输出的智能振动检测方法及装置 技术领域
本申请涉及互联网技术领域，具体涉及一种多模输出的智能振动检测方法及装置。
背景技术
互联网Internet属于传媒领域，又称国际网络，是网络与网络之间所串连成的庞大网络，这些网络以一组通用的协议相连，形成逻辑上的单一巨大国际网络。这种将计算机网络互相联接在一起的方法可称作“网络互联”，在这基础上发展出覆盖全世界的全球性互联网络称互联网，即是互相连接一起的网络结构。“互联网+”是创新2.0下的互联网发展的新业态，是知识社会创新2.0推动下的互联网形态演进及其催生的经济社会发展新形态。“互联网+”是互联网思维的进一步实践成果，推动经济形态不断地发生演变，从而带动社会经济实体的生命力，为改革、创新、发展提供广阔的网络平台。通俗的说，“互联网+”就是“互联网+各个传统行业”，但这并不是简单的两者相加，而是利用信息通信技术以及互联网平台，让互联网与传统行业进行深度融合，创造新的发展生态。它代表一种新的社会形态，即充分发挥互联网在社会资源配置中的优化和集成作用，将互联网的创新成果深度融合于经济、社会各域之中，提升全社会的创新力和生产力，形成更广泛的以互联网为基础设施和实现工具的经济发展新形态。
传统的故障监测机制一般是采用本地化检测设备，如在专用房间布置激光多普勒测振仪LDVs，通过该设备进行本地化的振动检测，以及故障预测等，但是LDVs存在价格昂贵，使用环境受限(测试环境的温度、光照等环境影响会使测量结果严重变坏)，测试区域小、难以实现远程监控等缺点，难以满足日益增多的各类场景中的智能化的振动检测需求。
所有的机械和运动系统都会产生各种各样的振动，其中一些振动反映的是系统的正常运动状态，而另外一些则反映了系统的异常运动状态(内部故障、轴连接不平衡等)。所以，要预测性维护系统设备，振动检测都是重要的一环。现有的振动检测系统大多通过单独获取振动参数，或者单独获取振动图像来表征振动物体的振动，这导致振动情况不能够被很好地观察和理解。
发明内容
有鉴于此，本申请实施例的目的在于提供一种多模输出的智能振动检测方法及装置，通过并行地获取和输出多种模式的稳定性数据，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。
具体的，本申请实施例所公开的振动检测方法中的数据传输流程可以基于互联网+技术，形成本地+云端或服务器的分布式智能化振动检测系统，一方面本地可以通过采集装置进行精确的原始影像采集和预处理，另一方面云端或服务器可以基于获取到的分布式数据，结合通过大数据技术统计分析得到的各类专用故障检测模型，预测被检测目标的故障，实现互联网与传统故障监测行业的深度融合，提高故障监测的智能性和准确度，满足日益增多的各类场景中的智能化的振动检测需求。
为了解决上述技术问题，本申请实施例第一方面提供了一种振动检测方法，所述方法包括：
接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项；
接收用户选择的发动机型号，或接收用户选择的车辆类型，并根据所述车辆类型定位发动机型号；
提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；
定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；
对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；
并行地对所述目标视频进行放大处理，获得放大输出视频；
并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
本申请实施例第二方面公开一种振动检测装置，所述装置包括：
接收模块，用于接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号；
选择模块，用于接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号；
提示模块，用于提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；
定位模块，用于定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；
提取模块，用于对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；
放大模块，用于并行地对所述目标视频进行放大处理，获得放大输出视频；
确定模块，用于并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
本申请实施例第三方面公开了一种电子装置，其特征在于，包括处理器、存储器、通信接口，以及一个或多个程序，所述一个或多个程序被存储在所述存储器中，并且被配置由所述处理器执行，所述程序包括用于执行第一方面所述的方法中的步骤的指令。
本申请实施例第四方面公开一种存储介质，用于存储用于电子数据交换的计算机程序，其中，所述计算机程序使得计算机执行第一方面对应方法所述的步骤的指令。
可以看出，本申请实施例公开了一种振动检测方法和装置，通过接收用户指令，获取车辆振动检测入口，再接收用户选择的发动机型号，定位对应的发动机获得振动物体对应的目标视频；对目标视频进行参数提取，获得目标视频对应的振动参数；并行地对目标视频进行放大处理，获得放大输出视频；并行输出多种模式的稳定性数据，稳定性数据包括振动参数和放大输出视频，根据稳定性数据确定振动物体的振动情况。在这个过程中，目标视频的参数提取和放大处理并行地进行，可以提升振动检测效率，并行地输出参数提取 获得的振动参数和放大处理后获得的放大输出视频，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。
附图说明
为了更清楚地说明本申请实施例中的技术方案，下面将对实施例描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图是本申请的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动的前提下，还可以根据这些附图获得其他的附图。
图1A为本申请实施例提供的一种振动传感装置结构框图。
图1B为本申请实施例提供的一种振动检测方法流程示意图。
图1C为本申请实施例提供的一种发动机的结构示意图。
图1D为本申请实施例提供的一种传感装置设置界面示意图。
图1E为本申请实施例提供的一种稳定性数据展示示意图。
图2为本申请实施例提供的一种对目标视频进行参数提取的方法。
图3为本申请实施例提供的一种振动情况验证方法的流程示意图。
图4是本申请实施例提供的一种电子装置的结构示意图。
图5为本申请实施例提供的一种振动检测装置的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本申请方案，下面将结合本申请实施例中的附图，对本申请实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本申请一部分实施例，而不是全部的实施例。基于本申请中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都属于本申请保护的范围。
在本文中提及“实施例”意味着，结合实施例描述的特定特征、结构或特性可以包含在本申请的至少一个实施例中。在说明书中的各个位置出现该短语并不一定均是指相同的实施例，也不是与其它实施例互斥的独立的或备选的实施例。本领域技术人员显式地和隐式地理解的是，本文所描述的实施例可以与其它实施例相结合。
首先介绍本申请实施例涉及的振动传感装置，请参阅图1A，图1A为本申请实施例提供的一种振动传感装置结构框图10，用于完成振动检测。该装置包括光学镜头11，面阵图像传感装置12，计算单元13，存储单元14，接口单元15。光学镜头11用于获取振动物体对应的目标图像，调整光学镜头11的焦距和光圈大小，使得可以在光学传感装置上形成振动物体的清晰影像。经过光学镜头11的振动物体影相经过面阵图像传感装置12(面阵图像传感装置可以是CMOS传感装置，或者CCD传感装置)转换为电信号数据。电信号储存在存储单元14，存储单元14由随机存取存储器(Random Access Memory，RAM)和只读存储器(Read-Only Memory，ROM)组成。计算单元13用来进行执行相应的算法。其中计算单元13包括CPU和GPU，可以用于交替进行稳定性预测模型中的振动参数提取过程和视频放大处理过程，获取到的振动参数和放大输出视频后组成稳定性数据。最后经过接口单元15与用户界面进行连接，并接受用户对振动传感装置进行设置。接口单元15还用于对稳定性数据进行输出，稳定性数据中的振动参数包括数值、振动波形图、模态图或频 谱图，这些振动参数与放大输出视频进行多模输出，使得各种参数、图像和视频对比展示，全方面体现振动物体的振动情况。接口单元15可以为USB接口、RS232接口，或者其他接口，用于另外也可以通过提供的API接口设置传感装置。
其次，本申请实施例提供一种振动检测方法，应用于如图1A所示的振动传感装置，请参阅图1B，图1B为本申请实施例提供的一种振动检测方法流程示意图，如图1B所示，振动检测方法包括如下步骤：
101、接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号。
振动物体包括内部相互作用产生机械振动的物体，包括引擎振动、发送机振动或者齿轮振动等，或者由于外力作用产生机械振动的物理，包括电线振动或桥梁振动等。振动物体在正常情况下会按照固定的频率机械振动，而当振动物体产生故障时，振动频率也会发生变化。因此，可以通过获取振动物体对应的目标视频，并进行分析确定振动物体的故障情况。
在振动物体为车辆发动机时，需要获取发送机的目标视频。振动检测传感装置与用户界面通过接口进行连接，在启动设备检测功能后，在用户界面为用户提供振动检测类型，如果用户知道自己车辆的发动机型号，则可以直接选择振动检测类型选项，如果不知道车辆发动机型号，则可以选择车辆类型。
102、接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号。
在用户选择了振动检测类型选项后，如果选择的是发动机型号，则可以直接确定发动机型号，如果选择了车辆类型，振动检测传感装置中的处理器可以联网获取车辆类型对应的发动机型号，或者直接从数据库中提取车辆类型对应的发动机型号。
103、提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作。
要对发动机型号对应的发动机进行振动检测，那么需要使发动机振动。在车辆正常驾驶过程中，发动机处于工作状态，可以确定发动机会振动。但是在一些情况下的振动情况，正常驾驶过程中无法获得。例如需要获得发动机在共振或接近共振情况下的振动情况，正常驾驶时很难达到，并且共振情况下有安全风险，正常驾驶途中不支持该种操作，因此需要对发动机进行空挡踩油门操作获得振动情况。
104、定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频。
定位振动过程中的发动机为振动物体，即如果发动机没有振动，或者振动的物体不为发动机，都不会作为振动物体进行目标视频采集。这样可以防止在发动机有多个的情况下，采集了非振动状态的发动机的视频，或者采集了其他振动物体的视频作为目标视频。提升了目标视频采集的效率和准确性。
可选的，所述获取振动物体对应的目标视频还包括：获取第一视频和第二视频，其中所述第一视频和第二视频为针对同一目标拍摄在相同时间内拍摄的不同源视频；获取述第一视频对应的第一帧图像和所述第二视频对应的第二帧图像；对所述第一帧图像和所述第 二帧图像进行重叠，对所述第一帧图像和所述第二帧图像不能重叠的像素点进行清除；获得所述目标视频。
具体地，同一个振动物体在进行视频采集时，可能因为一些外界原因，例如摄像头晃动、摄像头故障等，使得采集的目标视频存在偏差，那么，用不同的摄像头在同一时间内拍摄同一振动物体的不同源视频，获得第一视频和第二视频，并且对第一视频对应的第一帧图像和第二视频对应的第二帧图像进行重叠，在摄像头正常的情况下，两个帧图像对应的像素点应该完全重叠，那么清除掉第一帧图像和第二帧图像不能重叠的像素点，即为清除两个摄像头中的拍摄偏差像素点，获得的即为噪声更少的帧图像。同样的，还可以对该振动物体拍摄更多的同一时间段内的不同源视频，进一步减少视频噪声。
105、对所述目标视频进行参数提取，获得所述目标视频对应的振动参数。
对目标视频进行参数提取，可以获得目标视频的振动情况在参数变化上的体现，例如获取到的振动参数可以是数值，包括振动幅度、振动周期、振动频率等，也可以是相关图像，包括振动波形图或频谱图等，这些振动参数可以有助于高效确定振动情况的相关结论。
可选的，获取目标视频中发动机的多个图像帧，并对图像帧进行区域划分，获得N个区域；确定N个区域中每个区域包含的关键点个数Mi,i∈[1,N]，关键点为振动产生位置；按照每个区域包含的关键点个数进行特征点选取，选取个数为：
Figure PCTCN2020105792-appb-000001
其中Ti表示第i个区域选取的特征点个数，N1表示包含关键点的区域个数，N2表示不包含关键点的区域个数，N＝N1+N2，R表示图像帧预设特征点选取个数，
Figure PCTCN2020105792-appb-000002
表示保留
Figure PCTCN2020105792-appb-000003
的整数值，
Figure PCTCN2020105792-appb-000004
表示保留
Figure PCTCN2020105792-appb-000005
的整数值；
获得N个区域的实际提取的初始特征点个数为：
Figure PCTCN2020105792-appb-000006
对所述初始特征点进行筛选，获得稳定的多个运动特征点；对多个运动特征点进行光流跟踪，获得多个运动特征点的时间序列；对多个运动特征点的时间序列进行滤波处理，获得滤波后信号；对滤波后信号进行主成分分析，获得降维信号；对降维信号进行参数提取，获得目标视频对应的振动参数。
具体地，在本申请实施例中采用光流跟踪法进行振动参数提取。首先可以按照一定的周期获取目标视频对应的多帧图像，请参阅图1C，图1C为本申请实施例提供的一种发动机的结构示意图，如图1C所示，发动机中包括点火线圈110，凸轮机构111，气门112，活塞113，曲柄连杆机构114，曲轴115，润滑油底壳116和正时链条117，在发动机工作时，每个机构的振动情况不同。例如气门、活塞、点火线圈和正时链条一直处于工作状态， 振动正是由于这些机构的工作引起振动。而中间连接结构如曲柄连杆机构或曲轴等，也会由于所连接机构的振动被带动，而不常工作的机构，例如润滑油底壳，振动频率相对较低。
根据上述内容，可以首先对图像帧进行区域划分，例如按照功能结构进行区域划分，也可以按照网格进行平均划分，例如图1C中图像帧由虚线1,2,3,4划分成5个区域。在划分区域后，确定每个区域中包含的关键点。假设关键点包括气门、活塞、点火线圈和正时链条，那么可知虚线1上方区域包含的关键点最多，为2个，其次是虚线1-2和虚线2-3分割的区域为1个，那么可以知道包括关键点的区域N1＝3，不包括关键点的区域N2＝2，假设预设期望提取的初始特征点个数为R，R可以为100,95等数值，为所有的包括关键点的区域按照每个区域的关键点个数比例分配90％的初始特征点，为所有不包括关键点的区域平均分配10％的初始特征点，因为初始特征点个数为正整数，因此对根据公式计算出的特征点选取个数进行取整，获得最终的初始特征点个数R’，为所有包括关键点的区域和不包括关键点的区域提取的初始特征点之和。
确定每个区域的初始特征点提取个数后，从图像帧的各个区域提取对应个数预设类型的特征点作为初始特征点，预设类型的特征点包括以下至少一种：角点、边缘点、暗区的亮点以及亮区的暗点。角点是轮廓之间的交点，对于同一场景，即使视角发生变化，角点通常具备稳定性质的特征，并且该点附近区域的像素点无论在梯度方向上还是其梯度幅值上有着较大变化。边缘点为振动物体与背景图像之间的交点，亮区的暗点和暗区的亮点与其他像素点之间的对比度高，都可以作为初始特征点，以便更好地进行运动特征点观测。
然后对选取的特征点进行筛选，获得稳定的多个运动特征点。再对运动特征点进行光流跟踪，获得运动特征点在时间轴上的运动轨迹，即时间序列。然后对时间序列进行滤波，滤除非需求频率的运动轨迹。其中，滤波可以是时域滤波或频域滤波。对滤波后的时间序列进行主成分分析(principal component analysis，PCA)，把多个指标合成为少数几个相互无关的综合指标(即主成分)，获得降维信号。最后对降维信号进行参数提取，获得目标视频对应的振动参数，包括振动最大幅度、振动幅度取区间分布，是否达到共振振幅等。
可选的，对提取的初始特征点进行筛选，获得稳定的多个运动特征点，包括：基于最小差值平方和SSD匹配，计算初始特征点的流向量；根据初始特征点对应的流向量计算初始特征点的偏移距离；对多个初始特征点对应的多个偏移距离进行聚类，采用K-means聚类算法对多个偏移距离进行聚类，其中，设置k值为4，获得4个聚类类簇；对4个聚类类簇求平均值，并获得第一类簇平均值，第二类簇平均值、第三类簇平均值和第四类簇平均值，其中，第一类簇平均值<第二类簇平均值<第三类簇平均值<第四类簇平均值；从第二类簇平均值和第三类簇平均值对应的聚类类簇中挑选方差小于第一方差阈值的多个偏移距离组成第五聚类类簇，并确定第五类簇中的偏移距离对应的初始特征点作为稳定的多个运动特征点。
具体地，选择初始特征点后，采用最小差值平方和(Sum of Squared Differences,SSD)匹配，计算初始特征点的流向量。SSD的值越小，说明特征点之间的相似度越大，根据这 一原则可以确定初始特征点的运动轨迹，进而获得初始特征点的流向量。再根据流向量计算初始特征点的偏移距离，例如对
Figure PCTCN2020105792-appb-000007
求模，获得
Figure PCTCN2020105792-appb-000008
即为初始特征点从点A到点B的偏移距离。
获得多个初始特征点的偏移距离后，对这些值进行聚类，采用K-means聚类算法对所述多个偏移距离进行聚类，将k值设置为4，获得4个聚类类簇，再分别计算4个类簇的平均值，并获得第一类簇平均值，第二类簇平均值、第三类簇平均值和第四类簇平均值，且第一类簇平均值<第二类簇平均值<第三类簇平均值<第四类簇平均值，其中第一类簇平均值的类簇对应的初始特征点即可确定为静止点(背景点)，第四类簇平均值的类簇对应的初始特征点即可确定为剧烈运动的点(可能是一些不规律运动的点)，第二类簇平均值和第三类簇平均值的类簇对应的初始特征点运动幅度适中，最有可能为稳定的运动特征点。进一步地，机械振动的物体，其运动特征点进行一定幅度的往返运动，那么这些运动特征点的偏移距离方差也会是一个较稳定的值，因此，挑选第二类簇平均值和第三类簇平均值对应的聚类类簇中方差小于第一方差阈值的多个偏移距离组成第五聚类类簇，并确定第五类簇中的偏移距离对应的初始特征点作为稳定的多个运动特征点。该过程可以进一步提升获得的稳定的多个运动特征点的准确性。
可选的，对多个运动特征点进行光流跟踪，获得多个运动特征点的时间序列，包括：确定多个运动特征点在第一帧图像中的位置；根据特征点匹配和最小二乘法确定多个运动特征点在下一相邻帧图像中的位置；重复上述确定多个运动特征点在下一相邻帧图像的位置，直到遍历完多个图像帧中的每一帧图像；根据确定多个运动特征点的顺序和位置确定多个运动特征点的时间序列。
具体地，对运动特征点进行光流跟踪，就是假设在运动特征点在不同的图像帧中亮度色彩不发生变化，只有位置发生变化。因此，需要追踪相同的运动特征点在不同图像帧中的位置，一方面通过特征点的亮度或色彩匹配进行追踪，另一方面是通过区域划分和位置估算定位运动特征点的位置。定位到同一个运动特征点在不同的图像帧中的位置，位置可以通过坐标表示，就可以获得该运动特征点按时间顺序的一组数字序列，即为该运动特征点的时间序列。
可见，在本申请实施例中，通过对目标视频的多个图像帧中的每个图像帧进行区域划分，然后根据每个区域中包含的关键点计算每个区域对应选取的初始特征点个数，这样可以对不同区域的初始特征点进行更精确地分配，有助于后续稳定的多个运动特征点的获取，进而提升获得的振动参数的准确性和有效性；通过对获得的初始特征点进行偏移距离计算和偏移距离聚类完成筛选过程，获得稳定的多个运动特征点，可以提升获得的运动特征点的准确性；再对运动特征点进行光流跟踪，获得时间序列，再进一步进行滤波和振动参数提取。整个过程提升了振动参数提取的准确性。
另外，在某些情况下，可能只是想对发动机指定部位进行振动检测，那么对于发动机上的振动检测部位可以通过用户进行手动定位和选择。用户可以通过操作振动检测传感装置接口处的交互界面进行传感装置设置，进而确定进行振动检测的区域或部位。请参阅图1D，图1D为本申请实施例提供的一种传感装置设置界面示意图，如图1D所示，传感装置 设置界面为用户提供“添加测试点”和“删除测试点”两种操作，当用户选择“添加测试点”后，可以通过触控或者参数输入定位到传感装置设置界面展示的振动物体的对应位置上，添加对应位置作为目标测试点，例如图1D中的1,2,3三个点。同样的，当用户选择“删除测试点”后，也可以通过触控或参数输入定位到振动物体的对应位置，并删除已选择的目标测试点。另外，还可以在传感装置设置界面进行播放、停止、循环、亮度调节等基本设置，便于用户进行观察和测试点选择。
可选的，在对提取的初始特征点进行筛选，获得稳定的多个运动特征点之前，所述方法还包括：接收用户的添加测试点操作，定位用户选择的多个可添加测试点；确定所述多个可添加测试点的之间的多个平面距离；对所述多个平面距离求取平均值，获得平均平面距离；确定所述平均平面距离是否小于第一距离阈值；若是，则获取所述多个平面距离中最小的K个距离值对应的多组近距离可添加测试点；将所述多组近距离可添加测试点进行同组融合，形成多个融合可添加测试点；将所述多个可添加测试点中未进行融合的剩余可添加测试点和所述多个融合可添加测试点共同作为目标测试点，获取所述目标测试点的图像帧作为所述目标视频的图像帧。
具体地，根据上述内容可知，用户可以在传感装置设置界面添加测试点和删除测试点，在添加测试点的时候，如果通过触控实现，触控范围是固定的，比如每次触控都是1cm 2的范围，但是用户实际想选择的范围可能是10cm 2，用户需要多次操作，且可能操作范围不能完全覆盖想选择的范围。因此，确定多个可添加测试点两两之间的多个平面距离，并对多个平面距离求取平均值，获得平均平面距离，其中平面距离是指在设置界面的画面上对应的距离。如果平均平面距离小于第一距离阈值，说明用户在很近的范围内选择测试点，可能是想选择一个较大区域作为测试点，那么可以选择距离最近的K个测试点进行同组融合，即以两个测试点之间的连线作为直径获得圆，圆所覆盖的区域即为融合可添加测试点。最后将融合后的融合可添加测试点和未融合的可添加测试点共同作为目标测试点，获取目标测试点的图像帧作为目标视频的图像帧进行后续振动参数提取，可以减少数据处理量，提升振动参数提取的效率。
106、并行地对所述目标视频进行放大处理，获得放大输出视频。
为了提升稳定性预测模型的运行效率，参数提取过程和视频放大处理过程需要并行进行，在同一个处理器中，通过并行的两个进程，或者通过同一个进程中的双线程对上述两个过程进行并行处理。
可选的，对目标视频进行放大处理，获得放大输出视频，包括：将目标视频的多帧图像组成的帧序列进行空域金字塔分解，得到由多个不同空间分辨率的子图像组成的金字塔形结构；对金字塔型结构中的多个子图像中每个子图像进行时域带通滤波处理，得到目标频带对应的变换信号；对变换信号对应的位移进行A倍放大，获得放大后信号，其中A的取值范围为(3，Amax),其中Amax的值由目标频带和变换信号的位移函数确定；结合放大后信号和金字塔形结构进行金字塔重构，得到放大输出视频。
具体地，对目标视频进行放大处理可以采用欧拉运动放大方法对目标视频进行放大，首先需要将目标视频中的像素转换成时间和空间的函数，即通过图像金字塔变换将目标视频的多帧图像组成的帧序列分解成多个不同空间分辨率、不同尺度大小的子图像以构成一 个金字塔塔型结构，例如采用高斯金字塔对目标视频的多帧图像进行分解，即由一组在尺寸上逐层减半的图像序列组成金字塔结构，序列中的每一级图像均为其前一级图像低通滤波并隔行隔列采样的结果。
进行金字塔分解即对帧序列进行空域滤波，分解得到不同空间频率的
频带，并对这些频带分别进行放大。因为处于不同空间频率的频带对应的信噪比不同，空间频率越低，图像噪声越少，信噪比越高，因此每层空间频率的频带可以设置不同的放大系数。例如可以使用一个线性可变的放大倍数来放大不同频率的频带。金字塔结构中，从顶层到底层，放大倍数依次降低。
通过金字塔处理得到不同空间频率的频带后，还可以对每个频带进行时域的带通滤波处理，以得到感兴趣的变换信号，即目标频带对应的变换信号，并只对目标频带对应的变换信号进行放大处理。在进行带通滤波处理时，可以采用理想带通滤波器，Butterworth带通滤波器，二阶无限脉冲响应滤波器等。
获得目标频带对应的变换信号后，令I(x，t)为点x在时刻t的灰度值，且初始值为f(x)，则：
Figure PCTCN2020105792-appb-000009
其中δ(t)表示位移信号。
对I(x，t)放大α倍，即对位移信号δ(t)进行放大，且放大后的信号为：
Figure PCTCN2020105792-appb-000010
因为微小运动进行放大时，倍数太小是没有意义的，因此A的最小取值大于3。另外，放大倍数与空间频率相关，且满足如下关系：
Figure PCTCN2020105792-appb-000011
其中，空间频率为ω，目标频带的空间波长为λ，且λ＝2π/ω，则可通过目标频带和变换信号的位移函数确定α的最大值。A max≤α。
获得放大后信号之后，将其重新与原本的频带相结合，再通过金字塔重构，例如拉普拉斯金字塔变换重构，得到放大后的图像，进行得到放大输出视频。
可见，在本申请实施例中，通过并行地获取振动参数和放大输出视频作为振动物体对应的稳定性参数，可以将两种类型的参数进行对应查看，进而更准确地反应振动物体的振动情况，提升了振动检测的精确度和有效性。
107、并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
具体地，通过上述并行的参数提取过程和放大处理过程，获得多种模式的稳定性数据，包括振动参数和放大输出视频，然后对这些稳定性数据并行地输出，可以对照地进行振动情况的观察和研究，丰富了振动情况的表现形式，进而提升了振动检测结果的准确性。
在获得振动物体对应的稳定性数据后，可以根据稳定性数据查看振动物体的振动情况。请参阅图1E，图1E为本申请实施例提供的一种稳定性数据展示示意图，如图1E所示，振动检测传感装置经过接口单元连接到用户界面，在用户界面对接口单元多模输出的稳定性数据进行展示，在左边区域进行放大视频的展示，在右边区域进行放大视频对应目标检测点的振动参数的展示，例如在图1E中，目标检测点为110对应的圆形区域，右边即为其对应的振动参数。在振动参数展示区域还可以选择展示频域波形图或频域波形图，还可以展示振动最大幅度、振动幅度取区间分布等具体参数值，通过多模输出结果确定所述振动物体的振动情况，得到振动物体是否包括故障，故障类型名称等结论。
可选的，方法还包括：接收用户输入的故障视频和预测故障名称；将预测故障名称与故障列表中的故障信息进行匹配，确定故障列表中是否包括预测故障名称，其中，故障信息包括故障名称和与故障名称对应的稳定性数据；确定与预测故障名称匹配的目标故障名称，获取故障视频对应的振动参数；将故障视频对应的振动参数与目标故障名称对应的稳定性参数进行匹配；当确定故障视频对应的振动参数与目标故障名称对应的稳定性参数匹配成功时，确定预测故障名称正确。
具体地，在对振动物体进行振动检测时，如果用户已经对振动情况有了一定的预判断，再将振动物体输入到振动检测模型中进行检测并获得全部的稳定性数据，就将耗费大量不必要的时间。可以获取用户预判断的预测故障名称和对应的故障视频，然后将预测故障名称与故障列表进行匹配，确定预测故障名称是否在故障列表中，如果在，则可以直接获取故障视频对应的振动参数，即不再获取故障视频对应的运动放大视频，然后将故障视频对应的振动参数与故障列表中对应的振动参数进行匹配。例如预测故障名称为“轴承损坏”，与故障列表匹配成功后，确定故障列表中也包括该故障名称，即为目标故障名称。获取故障视频对应的振动参数，并且将之与故障列表中“轴承损坏”这一目标故障名称对应的稳定性数据中的振动参数进行匹配，包括振动图像匹配或者参数区间匹配，振动图像匹配即时域波形图或者频域波形图的匹配，参数区间匹配即最大振幅值，振幅平均值等的匹配，图像匹配需要达到第一相似度即确定匹配成功，例如90％等，参数值匹配可以是差值小于第一预设差值即确定匹配成功，例如0.5等。确定振动参数匹配成功后，可以确定预测故障名称正确。
可选的，故障名称包括共振，方法还包括更新故障列表，具体包括：获取车辆发动机型号对应的共振转速；获取发动机型号在共振转速区间对应的共振目标视频，共振转速区间为与共振转速的差值绝对值小于第一预设阈值的转速值区间；匹配发动机型号对应的稳定性预测模型，将共振目标视频输入稳定性预测模型，获取发动机型号对应发动机的共振数据；将发动机型号及其对应的共振数据作为故障信息更新到故障列表中。
具体地，共振是一个很典型的振动故障，发送机都有共振频率，不同的发动机共振频率不同。当转速达到一定数值时，就可能产生共振问题。那么在构建发动机的故障列表时，就需要将共振这一故障更新到其中。首先需要获取车辆发动机型号对应的共振转速，这可以根据车辆性能参数直接获取，也可以进行自主实验获取。然后获取发送机型号在共振转速区间对应的共振目标视频，转速区间是指与共振转速的差值绝对值小于第一预设阈值的转速值区间，例如共振转速为1200rpm(转/分)，第一预设阈值为100，那么转速区间为 1100rpm～1300rpm。在这个区间内，都可能在发动机对应的共振目标视频中显示出共振信息。因此将共振目标视频输入到稳定性预测模型中，获取其对应的稳定性数据作为共振数据。最后将发动机型号及其对应的共振数据作为故障信息更新到所述故障列表中，用于后续的故障预测验证。
可见，在本申请实施例公开的振动检测方法中，通过接收用户指令，获取车辆振动检测入口，再接收用户选择的发动机型号，定位对应的发动机获得振动物体对应的目标视频；对目标视频进行参数提取，获得目标视频对应的振动参数；并行地对目标视频进行放大处理，获得放大输出视频；并行输出多种模式的稳定性数据，稳定性数据包括振动参数和放大输出视频，根据稳定性数据确定振动物体的振动情况。在这个过程中，目标视频的参数提取和放大处理并行地进行，可以提升振动检测效率，并行地输出参数提取获得的振动参数和放大处理后获得的放大输出视频，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。
请参阅图2，图2为本申请实施例提供的一种对目标视频进行参数提取的方法，如图2所示，该方法包括如下步骤：
201、获取目标视频中发动机的多个图像帧，并对所述多个图像帧中的每个图像帧进行区域划分，获得N个区域；
202、确定N个区域中每个区域包含的关键点个数，并根据关键点个数确定每个区域选取的特征点个数；
203、按照每个区域选取的特征点个数获得每个图像帧中提取的初始特征点；
204、基于最小差值平方和SSD匹配，计算所述初始特征点的流向量；
205、根据所述初始特征点对应的流向量计算所述初始特征点的偏移距离；
206、对多个初始特征点对应的多个偏移距离进行聚类，采用K-means聚类算法对所述多个偏移距离进行聚类，其中，设置k值为4，获得4个聚类类簇；
207、对所述4个聚类类簇求平均值，并获得第一类簇平均值，第二类簇平均值、第三类簇平均值和第四类簇平均值，其中，第一类簇平均值<第二类簇平均值<第三类簇平均值<第四类簇平均值；
208、从第二类簇平均值和第三类簇平均值对应的聚类类簇中挑选方差小于第一方差阈值的多个偏移距离组成第五聚类类簇，并确定所述第五类簇中的偏移距离对应的初始特征点作为所述稳定的多个运动特征点；
209、确定所述多个运动特征点在第一帧图像中的位置；
210、根据特征点匹配和最小二乘法确定所述多个运动特征点在下一相邻帧图像中的位置；
211、重复上述确定多个运动特征点在下一相邻帧图像的位置，直到遍历完所述多个图像帧中的每一帧图像；
212、根据确定所述多个运动特征点的顺序和位置确定所述多个运动特征点的时间序列；
213、对所述多个运动特征点的时间序列进行滤波处理，获得滤波后信号，并对所述滤波后信号进行主成分分析，获得降维信号；
214、对所述降维信号进行参数提取，获得所述目标视频对应的振动参数。
其中，上述步骤201-步骤214的具体描述可以参照步骤101-107所描述的振动检测方法的相应描述，在此不再赘述。
可见，本申请实施例公开的对目标视频进行参数提取的方法中，通过对目标视频的多个图像帧中的每个图像帧进行区域划分，然后根据每个区域中包含的关键点计算每个区域对应选取的初始特征点个数，这样可以对不同区域的初始特征点进行更精确地分配，有助于后续稳定的多个运动特征点的获取，进而提升获得的振动参数的准确性和有效性；通过对获得的初始特征点进行偏移距离计算和偏移距离聚类完成筛选过程，获得稳定的多个运动特征点，可以提升获得的运动特征点的准确性；再对运动特征点进行光流跟踪，获得时间序列，再进一步进行滤波和振动参数提取。整个过程提升了振动参数提取的准确性。
请参阅图3，图3为本申请实施例提供的一种振动情况验证方法的流程示意图，如图3所示，该方法包括如下步骤：
301、获取车辆发动机型号对应的共振转速；
302、获取所述发动机型号在共振转速区间对应的共振目标视频，所述共振转速区间为与所述共振转速的差值绝对值小于第一预设阈值的转速值区间；
303、匹配所述发动机型号对应的稳定性预测模型，将所述共振目标视频输入所述稳定性预测模型，获取所述发动机型号对应发动机的共振数据；
304、将所述发动机型号及其对应的共振数据作为故障信息更新到所述故障列表中；
305、接收用户输入的故障视频和预测故障名称；
306、将所述预测故障名称与故障列表中的故障信息进行匹配，其中，所述故障信息包括故障名称和与所述故障名称对应的稳定性数据；
307、确定与所述预测故障名称匹配的目标故障名称，获取所述故障视频对应的振动参数；
308、将所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数进行匹配；
309、当确定所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数匹配成功时，确定所述预测故障名称正确。
其中，上述步骤301-步骤309的具体描述可以参照步骤101-步骤107所描述的振动检测方法的相应描述，在此不再赘述。
在本申请实施例中，通过获取车辆的发动机型号，在获取发动机型号对应的发动机在共振转速区间的共振目标视频以及其对应的共振数据，并将共振数据作为故障信息更新到故障列表中。在需要验证故障视频对应的预测故障名称是否正确时，可以获得故障视频对应的振动参数，并根据振动参数与故障列表中的稳定性参数匹配结果判定预测故障名称是否正确，而不需要再获取故障视频对应的运动放大视频，这一方面提升了振动情况验证的效率，另一方面通过与故障列表的匹配提升了振动情况验证的准确率。
如上述一致地，请参阅图4，图4是本申请实施例提供的一种电子装置的结构示意图，如图4所示，该电子装置包括处理器、存储器、通信接口以及一个或多个程序，其中，上 述一个或多个程序被存储在上述存储器中，并且被配置由上述处理器执行，上述程序包括用于执行以下步骤的指令：
接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号；
接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号；
提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；
定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；
对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；
并行地对所述目标视频进行放大处理，获得放大输出视频；
并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
可以看出，本申请实施例中，电子装置通过接收用户指令，获取车辆振动检测入口，再接收用户选择的发动机型号，定位对应的发动机获得振动物体对应的目标视频；对目标视频进行参数提取，获得目标视频对应的振动参数；并行地对目标视频进行放大处理，获得放大输出视频；并行输出多种模式的稳定性数据，稳定性数据包括振动参数和放大输出视频，根据稳定性数据确定振动物体的振动情况。在这个过程中，目标视频的参数提取和放大处理并行地进行，可以提升振动检测效率，并行地输出参数提取获得的振动参数和放大处理后获得的放大输出视频，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。
可选情况下，电子装置400中存储的程序还包括用于执行图1B～图3中描述的各类方法中的其他指令。
请参阅图5，图5为本申请实施例提供的一种振动检测装置的结构示意图，如图5所示，所述振动检测装置500包括：
接收模块501，用于接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号；选择模块502，用于接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号；提示模块503，用于提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；定位模块504，用于定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；提取模块505，用于对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；放大模块506，用于并行地对所述目标视频进行放大处理，获得放大输出视频；确定模块507，用于并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
在此需要说明的是，上述接收模块501、选择模块502、提示模块503、定位模块504、提取模块505、放大模块506和确定模块507的具体工作过程参见上述步骤101-107所描述的振动检测方法的相应描述，在此不再赘述。
在可选情况下，在所述对所述目标视频进行参数提取，获得所述目标视频对应的振动参数方面，所述提取模块505具体用于：
获取目标视频中发动机的多个图像帧，并对所述多个图像帧中的每个图像帧进行区域划分，获得N个区域；
确定N个区域中每个区域包含的关键点个数Mi,i∈[1,N]，关键点为振动产生位置；
按照每个区域包含的关键点个数进行初始特征点选取，选取个数为：
Figure PCTCN2020105792-appb-000012
其中Ti表示第i个区域选取的特征点个数，N1表示包含关键点的区域个数，N2表示不包含关键点的区域个数，N＝N1+N2，R表示图像帧预设特征点选取个数，
Figure PCTCN2020105792-appb-000013
表示保留
Figure PCTCN2020105792-appb-000014
的整数值，
Figure PCTCN2020105792-appb-000015
表示保留
Figure PCTCN2020105792-appb-000016
的整数值；
按照每个区域选取的特征点个数Ti获得每个图像帧中提取的初始特征点，其中初始特征点个数为：
Figure PCTCN2020105792-appb-000017
对所述初始特征点进行筛选，获得稳定的多个运动特征点；
对所述多个运动特征点进行光流跟踪，获得所述多个运动特征点的时间序列；
对所述多个运动特征点的时间序列进行滤波处理，获得滤波后信号；
对所述滤波后信号进行主成分分析，获得降维信号；
对所述降维信号进行参数提取，获得所述目标视频对应的振动参数。
在可选情况下，在对提取的初始特征点进行筛选，获得稳定的多个运动特征点方面，提取模块505还具体用于：
基于最小差值平方和SSD匹配，计算所述初始特征点的流向量；
根据所述初始特征点对应的流向量计算所述初始特征点的偏移距离；
对多个初始特征点对应的多个偏移距离进行聚类，采用K-means聚类算法对所述多个偏移距离进行聚类，其中，设置k值为4，获得4个聚类类簇；
对所述4个聚类类簇求平均值，并获得第一类簇平均值，第二类簇平均值、第三类簇平均值和第四类簇平均值，其中，第一类簇平均值<第二类簇平均值<第三类簇平均值<第四类簇平均值；
从第二类簇平均值和第三类簇平均值对应的聚类类簇中挑选方差小于第一方差阈值的多个偏移距离组成第五聚类类簇，并确定所述第五类簇中的偏移距离对应的初始特征点作为所述稳定的多个运动特征点。
在可选情况下，在对所述多个运动特征点进行光流跟踪，获得所述多个运动特征点的时间序列方面，所述提取模块505还具体用于：
确定所述多个运动特征点在第一帧图像中的位置；
根据特征点匹配和最小二乘法确定所述多个运动特征点在下一相邻帧图像中的位置；
重复上述确定多个运动特征点在下一相邻帧图像的位置，直到遍历完所述多个图像帧中的每一帧图像；
根据确定所述多个运动特征点的顺序和位置确定所述多个运动特征点的时间序列。
在可选情况下，在所述对所述目标视频进行放大处理，获得放大输出视频方面，所述放大模块506具体用于：
将所述目标视频的多帧图像组成的帧序列进行空域金字塔分解，得到由多个不同空间分辨率的子图像组成的金字塔形结构；
对所述金字塔型结构中的多个子图像中每个子图像进行时域带通滤波处理，得到目标频带对应的变换信号；
对所述变换信号对应的位移进行A倍放大，获得放大后信号，其中A的取值范围为(3，Amax),其中Amax的值由目标频带和变换信号的位移函数确定；
结合所述放大后信号和所述金字塔形结构进行金字塔重构，得到放大输出视频。
可选情况下，所述振动检测装置500还包括验证模块508，具体用于：
接收用户输入的故障视频和预测故障名称；
将所述预测故障名称与故障列表中的故障信息进行匹配，其中，所述故障信息包括故障名称和与所述故障名称对应的稳定性数据；
确定与所述预测故障名称匹配的目标故障名称，获取所述故障视频对应的振动参数；
将所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数进行匹配；
当确定所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数匹配成功时，确定所述预测故障名称正确。
可选情况下，所述故障名称包括共振，所述验证模块508还用于更新故障列表，具体用于：
获取车辆发动机型号对应的共振转速；
获取所述发动机型号在共振转速区间对应的共振目标视频，所述共振转速区间为与所述共振转速的差值绝对值小于第一预设阈值的转速值区间；
匹配所述发动机型号对应的稳定性预测模型，将所述共振目标视频输入所述稳定性预测模型，获取所述发动机型号对应发动机的共振数据；
将所述发动机型号及其对应的共振数据作为故障信息更新到所述故障列表中。
可以看出，本申请实施例公开的振动检测装置，通过接收用户指令，获取车辆振动检测入口，再接收用户选择的发动机型号，定位对应的发动机获得振动物体对应的目标视频；对目标视频进行参数提取，获得目标视频对应的振动参数；并行地对目标视频进行放大处理，获得放大输出视频；并行输出多种模式的稳定性数据，稳定性数据包括振动参数和放大输出视频，根据稳定性数据确定振动物体的振动情况。在这个过程中，目标视频的参数提取和放大处理并行地进行，可以提升振动检测效率，并行地输出参数提取获得的振动参数和放大处理后获得的放大输出视频，丰富了振动检测结果的表现形式，提升了振动检测的精确度和有效性。
在一些实施例里，提供一种存储介质，用于存储用于电子数据交换的计算机程序，其中，所述计算机程序使得计算机执行上述任一方法所述的步骤的指令。
需要说明的是，对于前述的各方法实施例，为了简单描述，故将其都表述为一系列的动作组合，但是本领域技术人员应该知悉，本申请并不受所描述的动作顺序的限制，因为依据本申请，某些步骤可以采用其他顺序或者同时进行。其次，本领域技术人员也应该知悉，说明书中所描述的实施例均属于可选实施例，所涉及的动作和模块并不一定是本申请所必须的。
在上述实施例中，对各个实施例的描述都各有侧重，某个实施例中没有详述的部分，可以参见其他实施例的相关描述。
在本申请所提供的几个实施例中，应该理解到，所揭露的方法，可通过其它的方式实现。例如，以上所描述的方法实施例仅仅是示意性的，例如所述单元的划分，仅仅为一种逻辑功能划分，实际实现时可以有另外的划分方式，例如多个单元或组件可以结合或者可以集成到另一个系统，或一些特征可以忽略，或不执行。另一点，所显示或讨论的相互之间的耦合或直接耦合或通信连接可以是通过一些接口，方法或单元的间接耦合或通信连接，可以是电性或其它的形式。
所述作为分离部件说明的单元可以是或者也可以不是物理上分开的，作为单元显示的部件可以是或者也可以不是物理单元，即可以位于一个地方，或者也可以分布到多个网络单元上。可以根据实际的需要选择其中的部分或者全部单元来实现本实施例方案的目的。
另外，在本申请各个实施例中的各功能单元可以集成在一个处理单元中，也可以是各个单元单独物理存在，也可以两个或两个以上单元集成在一个单元中。上述集成的单元既可以采用硬件的形式实现，也可以采用软件程序模块的形式实现。
集成的单元如果以软件程序模块的形式实现并作为独立的产品销售或使用时，可以存储在一个计算机可读取存储器中。基于这样的理解，本申请的技术方案本质上或者说对现有技术做出贡献的部分或者该技术方案的全部或部分可以以软件产品的形式体现出来，该计算机软件产品存储在一个存储器中，包括若干指令用以使得一台计算机设备(可为个人计算机、服务器或者网络设备等)执行本申请各个实施例所述方法的全部或部分步骤。而前述的存储器包括：U盘、只读存储器(ROM，Read-Only Memory)、随机存取存储器(RAM，Random Access Memory)、移动硬盘、磁碟或者光盘等各种可以存储程序代码的介质。
本领域普通技术人员可以理解上述实施例的各种方法中的全部或部分步骤是可以通过程序来指令相关的硬件来完成，该程序可以存储于一计算机可读存储器中，存储器可以包括：闪存盘、只读存储器(英文：Read-Only Memory，简称：ROM)、随机存取器(英文：Random Access Memory，简称：RAM)、磁盘或光盘等。
以上对本申请实施例进行了详细介绍，本文中应用了具体个例对本申请的原理及实施方式进行了阐述，以上实施例的说明只是用于帮助理解本申请的方法及其核心思想；同时，对于本领域的一般技术人员，依据本申请的思想，在具体实施方式及应用范围上均会有改变之处，综上所述，本说明书内容不应理解为对本申请的限制。

Claims (10)

  1. 一种振动检测方法，其特征在于，应用于振动检测传感装置，所述方法包括：
    接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号；
    接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号；
    提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；
    定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；
    对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；
    并行地对所述目标视频进行放大处理，获得放大输出视频；
    并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
  2. 根据权利要求1所述的方法，其特征在于，所述对所述目标视频进行参数提取，获得所述目标视频对应的振动参数，包括：
    获取目标视频中发动机的多个图像帧，并对所述多个图像帧中的每个图像帧进行区域划分，获得N个区域；
    确定N个区域中每个区域包含的关键点个数Mi,i∈[1,N]，关键点为振动产生位置；
    按照每个区域包含的关键点个数进行初始特征点选取，选取个数为：
    Figure PCTCN2020105792-appb-100001
    其中Ti表示第i个区域选取的特征点个数，N1表示包含关键点的区域个数，N2表示不包含关键点的区域个数，N＝N1+N2，R表示图像帧预设特征点选取个数，
    Figure PCTCN2020105792-appb-100002
    表示保留
    Figure PCTCN2020105792-appb-100003
    的整数值，
    Figure PCTCN2020105792-appb-100004
    表示保留
    Figure PCTCN2020105792-appb-100005
    的整数值；
    按照每个区域选取的特征点个数Ti获得每个图像帧中提取的初始特征点，其中初始特征点个数为：
    Figure PCTCN2020105792-appb-100006
    对所述初始特征点进行筛选，获得稳定的多个运动特征点；
    对所述多个运动特征点进行光流跟踪，获得所述多个运动特征点的时间序列；
    对所述多个运动特征点的时间序列进行滤波处理，获得滤波后信号；
    对所述滤波后信号进行主成分分析，获得降维信号；
    对所述降维信号进行参数提取，获得所述目标视频对应的振动参数。
  3. 根据权利要求2所述的方法，其特征在于，所述对提取的初始特征点进行筛选，获得稳定的多个运动特征点，包括：
    基于最小差值平方和SSD匹配，计算所述初始特征点的流向量；
    根据所述初始特征点对应的流向量计算所述初始特征点的偏移距离；
    对多个初始特征点对应的多个偏移距离进行聚类，采用K-means聚类算法对所述多个偏移距离进行聚类，其中，设置k值为4，获得4个聚类类簇；
    对所述4个聚类类簇求平均值，并获得第一类簇平均值，第二类簇平均值、第三类簇平均值和第四类簇平均值，其中，第一类簇平均值<第二类簇平均值<第三类簇平均值<第四类簇平均值；
    从第二类簇平均值和第三类簇平均值对应的聚类类簇中挑选方差小于第一方差阈值的多个偏移距离组成第五聚类类簇，并确定所述第五类簇中的偏移距离对应的初始特征点作为所述稳定的多个运动特征点。
  4. 根据权利要求2或3所述的方法，其特征在于，对所述多个运动特征点进行光流跟踪，获得所述多个运动特征点的时间序列，包括：
    确定所述多个运动特征点在第一帧图像中的位置；
    根据特征点匹配和最小二乘法确定所述多个运动特征点在下一相邻帧图像中的位置；
    重复上述确定多个运动特征点在下一相邻帧图像的位置，直到遍历完所述多个图像帧中的每一帧图像；
    根据确定所述多个运动特征点的顺序和位置确定所述多个运动特征点的时间序列。
  5. 根据权利要求1-4任一项所述的方法，其特征在于，所述对所述目标视频进行放大处理，获得放大输出视频，包括：
    将所述目标视频的多帧图像组成的帧序列进行空域金字塔分解，得到由多个不同空间分辨率的子图像组成的金字塔形结构；
    对所述金字塔型结构中的多个子图像中每个子图像进行时域带通滤波处理，得到目标频带对应的变换信号；
    对所述变换信号对应的位移进行A倍放大，获得放大后信号，其中A的取值范围为(3，Amax),其中Amax的值由目标频带和变换信号的位移函数确定；
    结合所述放大后信号和所述金字塔形结构进行金字塔重构，得到放大输出视频。
  6. 根据权利要求5所述的方法，其特征在于，所述方法还包括振动情况验证，具体包括：
    接收用户输入的故障视频和预测故障名称；
    将所述预测故障名称与故障列表中的故障信息进行匹配，其中，所述故障信息包括故障名称和与所述故障名称对应的稳定性数据；
    确定与所述预测故障名称匹配的目标故障名称，获取所述故障视频对应的振动参数；
    将所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数进行匹配；
    当确定所述故障视频对应的振动参数与所述目标故障名称对应的稳定性参数匹配成功时，确定所述预测故障名称正确。
  7. 根据权利要求6所述的方法，其特征在于，所述故障名称包括共振，所述方法还包括更新故障列表，具体包括：
    获取车辆发动机型号对应的共振转速；
    获取所述发动机型号在共振转速区间对应的共振目标视频，所述共振转速区间为与所述共振转速的差值绝对值小于第一预设阈值的转速值区间；
    匹配所述发动机型号对应的稳定性预测模型，将所述共振目标视频输入所述稳定性预测模型，获取所述发动机型号对应发动机的共振数据；
    将所述发动机型号及其对应的共振数据作为故障信息更新到所述故障列表中。
  8. 一种振动检测装置，其特征在于，所述装置包括：
    接收模块，用于接收用户启动设备检测功能的指令，根据所述指令呈现车辆振动检测入口，所述车辆振动检测入口提供振动检测类型选项，所述振动检测类型选项包括车辆类型或发动机型号；
    选择模块，用于接收用户选择振动检测类型选项，并根据所述振动检测类型选项确定发动机型号；
    提示模块，用于提示用户对所述发送机型号对应的发动机进行振动操作，所述振动操作包括驾驶操作，或空挡踩油门操作；
    定位模块，用于定位振动过程中的所述发送机为振动物体，获取所述振动物体对应的目标视频；
    提取模块，用于对所述目标视频进行参数提取，获得所述目标视频对应的振动参数；
    放大模块，用于并行地对所述目标视频进行放大处理，获得放大输出视频；
    确定模块，用于并行输出多种模式的稳定性数据，所述稳定性数据包括所述振动参数和所述放大输出视频，根据所述稳定性数据确定所述振动物体的振动情况。
  9. 一种电子装置，其特征在于，包括处理器、存储器、通信接口，以及一个或多个程序，所述一个或多个程序被存储在所述存储器中，并且被配置由所述处理器执行，所述程序包括用于执行如权利要求1-7任一项所述的方法中的步骤的指令。
  10. 一种计算机可读存储介质，其特征在于，存储用于电子数据交换的计算机程序，其中，所述计算机程序使得计算机执行如权利要求1-7任一项所述的方法。
PCT/CN2020/105792 2019-08-28 2020-07-30 多模输出的智能振动检测方法及装置 WO2021036670A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910800074.6A CN110956645B (zh) 2019-08-28 2019-08-28 多模输出的智能振动检测方法及装置
CN201910800074.6 2019-08-28

Publications (1)

Publication Number Publication Date
WO2021036670A1 true WO2021036670A1 (zh) 2021-03-04

Family

ID=69975475

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/105792 WO2021036670A1 (zh) 2019-08-28 2020-07-30 多模输出的智能振动检测方法及装置

Country Status (2)

Country Link
CN (1) CN110956645B (zh)
WO (1) WO2021036670A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114581252A (zh) * 2022-03-03 2022-06-03 平安科技（深圳）有限公司 目标案件的预测方法和装置、电子设备、存储介质
CN117593330A (zh) * 2023-11-23 2024-02-23 四川省寰宇众恒科技有限公司 一种可视化的实时振动监测方法

Families Citing this family (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110956645B (zh) * 2019-08-28 2023-10-31 深圳市广宁股份有限公司 多模输出的智能振动检测方法及装置
CN113848700A (zh) * 2020-06-28 2021-12-28 北京配天技术有限公司 一种机器人抖动抑制方法、装置、设备及可读存储介质
CN113158785B (zh) * 2021-03-11 2022-11-15 复旦大学 一种振荡信号模态参数的识别方法
CN113405644B (zh) * 2021-05-17 2022-04-08 中山大学 非接触的振动测量方法、装置和存储介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101354315A (zh) * 2008-09-05 2009-01-28 爱立迈科（宁波）计测仪器有限公司 一种基于振动信号的发动机状态跟踪检测装置及方法
JP2013027215A (ja) * 2011-07-25 2013-02-04 Canon Inc モータ制御装置および光学機器
CN106225681A (zh) * 2016-07-25 2016-12-14 肖锐 一种大跨度桥梁健康状态监测装置
CN108317997A (zh) * 2017-01-17 2018-07-24 大连因特视智能传感科技有限公司 基于视频分析的轨道车辆在隧道内的振动检测系统
CN108414240A (zh) * 2018-03-15 2018-08-17 广东工业大学 一种检测机器异常振动的方法与装置
CN110956645A (zh) * 2019-08-28 2020-04-03 深圳市广宁股份有限公司 多模输出的智能振动检测方法及装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW201120812A (en) * 2009-12-04 2011-06-16 Huper Lab Co Ltd Stabilization method for vibrating video frames
ES2686470T3 (es) * 2015-12-10 2018-10-18 Airbus Defence and Space GmbH Dispositivo modular para un análisis de vibraciones de vídeo a alta velocidad
CN110068388A (zh) * 2019-03-29 2019-07-30 南京航空航天大学 一种基于视觉和盲源分离的振动检测方法

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101354315A (zh) * 2008-09-05 2009-01-28 爱立迈科（宁波）计测仪器有限公司 一种基于振动信号的发动机状态跟踪检测装置及方法
JP2013027215A (ja) * 2011-07-25 2013-02-04 Canon Inc モータ制御装置および光学機器
CN106225681A (zh) * 2016-07-25 2016-12-14 肖锐 一种大跨度桥梁健康状态监测装置
CN108317997A (zh) * 2017-01-17 2018-07-24 大连因特视智能传感科技有限公司 基于视频分析的轨道车辆在隧道内的振动检测系统
CN108414240A (zh) * 2018-03-15 2018-08-17 广东工业大学 一种检测机器异常振动的方法与装置
CN110956645A (zh) * 2019-08-28 2020-04-03 深圳市广宁股份有限公司 多模输出的智能振动检测方法及装置

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114581252A (zh) * 2022-03-03 2022-06-03 平安科技（深圳）有限公司 目标案件的预测方法和装置、电子设备、存储介质
CN114581252B (zh) * 2022-03-03 2024-04-05 平安科技（深圳）有限公司 目标案件的预测方法和装置、电子设备、存储介质
CN117593330A (zh) * 2023-11-23 2024-02-23 四川省寰宇众恒科技有限公司 一种可视化的实时振动监测方法

Also Published As

Publication number Publication date
CN110956645B (zh) 2023-10-31
CN110956645A (zh) 2020-04-03

Similar Documents

Publication Publication Date Title
WO2021036670A1 (zh) 多模输出的智能振动检测方法及装置
WO2021052025A1 (zh) 电风扇振动故障检测方法及装置
WO2021017998A1 (zh) 文本位置定位方法和系统以及模型训练方法和系统
US10255686B2 (en) Estimating depth from a single image
JP6437514B2 (ja) 特徴の空間局所化を利用することによる物体検出時間の減少
CN111612760A (zh) 用于检测障碍物的方法和装置
JP2019514123A (ja) 地理的領域におけるコンテナ内に格納された量の遠隔決定
JP2018524678A (ja) 画像からの事業発見
JP2016515431A (ja) 視線追跡のための方法
CN103778643B (zh) 一种实时生成目标深度信息的方法及其装置
CN107767358B (zh) 一种图像中物体模糊度确定方法和装置
JP2018077644A (ja) 情報処理システム及びプログラム
US10042899B2 (en) Automatic registration
US20230040550A1 (en) Method, apparatus, system, and storage medium for 3d reconstruction
JPWO2014203687A1 (ja) 画像処理方法、画像処理装置、および画像処理プログラム
CN110276321A (zh) 一种遥感视频目标跟踪方法及系统
CN115661336A (zh) 一种三维重建方法及相关装置
WO2021052024A1 (zh) 洗衣机振动故障检测方法及装置
CN105574844B (zh) 辐射响应函数估计方法和装置
CN113160210A (zh) 基于深度相机的排水管道缺陷检测方法及装置
US11080286B2 (en) Method and system for merging multiple point cloud scans
CN110675427A (zh) 智能振动检测方法及相关产品
CN116109962A (zh) 一种基于视频相位的振动频率测量方法、装置、电子设备及存储介质
CN105809159A (zh) 一种基于成像的视觉权重图提取方法
Peng et al. Full-field visual vibration measurement of rotating machine under complex conditions via unsupervised retinex model

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20857488

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20857488

Country of ref document: EP

Kind code of ref document: A1